RFC 7015






Internet Engineering Task Force (IETF)                       B. Trammell
Request for Comments: 7015                                    ETH Zurich
Category: Standards Track                                      A. Wagner
ISSN: 2070-1721                                              Consecom AG
                                                               B. Claise
                                                     Cisco Systems, Inc.
                                                          September 2013


  Flow Aggregation for the IP Flow Information Export (IPFIX) Protocol

Abstract



   This document provides a common implementation-independent basis for
   the interoperable application of the IP Flow Information Export
   (IPFIX) protocol to the handling of Aggregated Flows, which are IPFIX
   Flows representing packets from multiple Original Flows sharing some
   set of common properties.  It does this through a detailed
   terminology and a descriptive Intermediate Aggregation Process
   architecture, including a specification of methods for Original Flow
   counting and counter distribution across intervals.

Status of This Memo



   This is an Internet Standards Track document.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Further information on
   Internet Standards is available in Section 2 of RFC 5741.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   http://www.rfc-editor.org/info/rfc7015.
















Trammell, et al.             Standards Track                    [Page 1]

RFC 7015                    IPFIX Aggregation             September 2013


Copyright Notice



   Copyright (c) 2013 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (http://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.

Table of Contents
   1. Introduction ....................................................3
      1.1. IPFIX Protocol Overview ....................................4
      1.2. IPFIX Documents Overview ...................................5
   2. Terminology .....................................................5
   3. Use Cases for IPFIX Aggregation .................................7
   4. Architecture for Flow Aggregation ...............................8
      4.1. Aggregation within the IPFIX Architecture ..................8
      4.2. Intermediate Aggregation Process Architecture .............12
           4.2.1. Correlation and Normalization ......................14
   5. IP Flow Aggregation Operations .................................15
      5.1. Temporal Aggregation through Interval Distribution ........15
           5.1.1. Distributing Values across Intervals ...............16
           5.1.2. Time Composition ...................................18
           5.1.3. External Interval Distribution .....................19
      5.2. Spatial Aggregation of Flow Keys ..........................19
           5.2.1. Counting Original Flows ............................21
           5.2.1. Counting Distinct Key Values .......................22
      5.3. Spatial Aggregation of Non-key Fields .....................22
           5.3.1. Counter Statistics .................................22
           5.3.2. Derivation of New Values from Flow Keys and
                  Non-key fields .....................................23
      5.4. Aggregation Combination ...................................23
   6. Additional Considerations and Special Cases in Flow
      Aggregation ....................................................24
      6.1. Exact versus Approximate Counting during Aggregation ......24
      6.2. Delay and Loss Introduced by the IAP ......................24
      6.3. Considerations for Aggregation of Sampled Flows ...........24
      6.4. Considerations for Aggregation of Heterogeneous Flows .....25
   7. Export of Aggregated IP Flows Using IPFIX ......................25
      7.1. Time Interval Export ......................................25
      7.2. Flow Count Export .........................................25
           7.2.1. originalFlowsPresent ...............................26



Trammell, et al.             Standards Track                    [Page 2]

RFC 7015                    IPFIX Aggregation             September 2013


           7.2.2. originalFlowsInitiated .............................26
           7.2.3. originalFlowsCompleted .............................26
           7.2.4. deltaFlowCount .....................................26
      7.3. Distinct Host Export ......................................27
           7.3.1. distinctCountOfSourceIPAddress .....................27
           7.3.2. distinctCountOfDestinationIPAddress ................27
           7.3.3. distinctCountOfSourceIPv4Address ...................27
           7.3.4. distinctCountOfDestinationIPv4Address ..............28
           7.3.5. distinctCountOfSourceIPv6Address ...................28
           7.3.6. distinctCountOfDestinationIPv6Address ..............28
      7.4. Aggregate Counter Distribution Export .....................28
           7.4.1. Aggregate Counter Distribution Options Template ....29
           7.4.2. valueDistributionMethod Information Element ........29
   8. Examples .......................................................31
      8.1. Traffic Time Series per Source ............................32
      8.2. Core Traffic Matrix .......................................37
      8.3. Distinct Source Count per Destination Endpoint ............42
      8.4. Traffic Time Series per Source with Counter Distribution ..44
   9. Security Considerations ........................................46
   10. IANA Considerations ...........................................46
   11. Acknowledgments ...............................................46
   12. References ....................................................47
      12.1. Normative References .....................................47
      12.2. Informative References ...................................47

1.  Introduction



   The assembly of packet data into Flows serves a variety of different
   purposes, as noted in the requirements [RFC3917] and applicability
   statement [RFC5472] for the IP Flow Information Export (IPFIX)
   protocol [RFC7011].  Aggregation beyond the Flow level, into records
   representing multiple Flows, is a common analysis and data reduction
   technique as well, with applicability to large-scale network data
   analysis, archiving, and inter-organization exchange.  This
   applicability in large-scale situations, in particular, led to the
   inclusion of aggregation as part of the IPFIX Mediation Problem
   Statement [RFC5982], and the definition of an Intermediate
   Aggregation Process in the Mediator framework [RFC6183].

   Aggregation is used for analysis and data reduction in a wide variety
   of applications, for example, in traffic matrix calculation,
   generation of time series data for visualizations or anomaly
   detection, or data reduction for long-term trending and storage.
   Depending on the keys used for aggregation, it may additionally have
   an anonymizing effect on the data: for example, aggregation
   operations that eliminate IP addresses make it impossible to later
   directly identify nodes using those addresses.




Trammell, et al.             Standards Track                    [Page 3]

RFC 7015                    IPFIX Aggregation             September 2013


   Aggregation, as defined and described in this document, covers the
   applications defined in [RFC5982], including Sections 5.1 "Adjusting
   Flow Granularity", 5.4 "Time Composition", and 5.5 "Spatial
   Composition".  However, Section 4.2 of this document specifies a more
   flexible architecture for an Intermediate Aggregation Process than
   that envisioned by the original Mediator work [RFC5982].  Instead of
   a focus on these specific limited use cases, the Intermediate
   Aggregation Process is specified to cover any activity commonly
   described as "Flow aggregation".  This architecture is intended to
   describe any such activity without reference to the specific
   implementation of aggregation.

   An Intermediate Aggregation Process may be applied to data collected
   from multiple Observation Points, as it is natural to use aggregation
   for data reduction when concentrating measurement data.  This
   document specifically does not address the protocol issues that arise
   when combining IPFIX data from multiple Observation Points and
   exporting from a single Mediator, as these issues are general to
   IPFIX Mediation; they are therefore treated in detail in the
   Mediation Protocol document [IPFIX-MED-PROTO].

   Since Aggregated Flows as defined in the following section are
   essentially Flows, the IPFIX protocol [RFC7011] can be used to
   export, and the IPFIX File Format [RFC5655] can be used to store,
   aggregated data "as is"; there are no changes necessary to the
   protocol.  This document provides a common basis for the application
   of IPFIX to the handling of aggregated data, through a detailed
   terminology, Intermediate Aggregation Process architecture, and
   methods for Original Flow counting and counter distribution across
   intervals.  Note that Sections 5, 6, and 7 of this document are
   normative.

1.1.  IPFIX Protocol Overview



   In the IPFIX protocol, { type, length, value } tuples are expressed
   in Templates containing { type, length } pairs, specifying which
   { value } fields are present in data records conforming to the
   Template, giving great flexibility as to what data is transmitted.
   Since Templates are sent very infrequently compared with Data
   Records, this results in significant bandwidth savings.  Various
   different data formats may be transmitted simply by sending new
   Templates specifying the { type, length } pairs for the new data
   format.  See [RFC7011] for more information.

   The IPFIX Information Element Registry [IANA-IPFIX] defines a large
   number of standard Information Elements that provide the necessary {
   type } information for Templates.  The use of standard elements
   enables interoperability among different vendors' implementations.



Trammell, et al.             Standards Track                    [Page 4]

RFC 7015                    IPFIX Aggregation             September 2013


   Additionally, non-standard enterprise-specific elements may be
   defined for private use.

1.2.  IPFIX Documents Overview



   "Specification of the IP Flow Information Export (IPFIX) Protocol for
   the Exchange of Flow Information" [RFC7011] and its associated
   documents define the IPFIX protocol, which provides network engineers
   and administrators with access to IP traffic Flow information.

   IPFIX has a formal description of IPFIX Information Elements, their
   names, types, and additional semantic information, as specified in
   the IPFIX Information Model [RFC7012].  The IPFIX Information Element
   registry [IANA-IPFIX] is maintained by IANA.  New Information Element
   definitions can be added to this registry subject to an Expert Review
   [RFC5226], with additional process considerations described in
   [RFC7013].

   "Architecture for IP Flow Information Export" [RFC5470] defines the
   architecture for the export of measured IP Flow information out of an
   IPFIX Exporting Process to an IPFIX Collecting Process and the basic
   terminology used to describe the elements of this architecture, per
   the requirements defined in "Requirements for IP Flow Information
   Export" [RFC3917].  The IPFIX protocol document [RFC7011] covers the
   details of the method for transporting IPFIX Data Records and
   Templates via a congestion-aware transport protocol from an IPFIX
   Exporting Process to an IPFIX Collecting Process.

   "IP Flow Information Export (IPFIX) Mediation: Problem Statement"
   [RFC5982] introduces the concept of IPFIX Mediators, and defines the
   use cases for which they were designed; "IP Flow Information Export
   (IPFIX) Mediation: Framework" [RFC6183] then provides an
   architectural framework for Mediators.  Protocol-level issues (e.g.,
   Template and Observation Domain handling across Mediators) are
   covered by "Operation of the IP Flow Information Export (IPFIX)
   Protocol on IPFIX Mediators" [IPFIX-MED-PROTO].

   This document specifies an Intermediate Process for Flow aggregation
   that may be applied at an IPFIX Mediator, as well as at an original
   Observation Point prior to export, or for analysis and data reduction
   purposes after receipt at a Collecting Process.

2.  Terminology



   Terms used in this document that are defined in the Terminology
   section of the IPFIX protocol document [RFC7011] are to be
   interpreted as defined there.




Trammell, et al.             Standards Track                    [Page 5]

RFC 7015                    IPFIX Aggregation             September 2013


   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
   document are to be interpreted as described in [RFC2119].

   In addition, this document defines the following terms:

   Aggregated Flow:  A Flow, as defined by [RFC7011], derived from a set
      of zero or more Original Flows within a defined Aggregation
      Interval.  Note that an Aggregated Flow is defined in the context
      of an Intermediate Aggregation Process only.  Once an Aggregated
      Flow is exported, it is essentially a Flow as in [RFC7011] and can
      be treated as such.

   Intermediate Aggregation Process:  an Intermediate Aggregation
      Process (IAP), as in [RFC6183], that aggregates records, based
      upon a set of Flow Keys or functions applied to fields from the
      record.

   Aggregation Interval:  A time interval imposed upon an Aggregated
      Flow.  Intermediate Aggregation Processes may use a regular
      Aggregation Interval (e.g., "every five minutes", "every calendar
      month"), though regularity is not necessary.  Aggregation
      intervals may also be derived from the time intervals of the
      Original Flows being aggregated.

   Partially Aggregated Flow:  A Flow during processing within an
      Intermediate Aggregation Process; refers to an intermediate data
      structure during aggregation within the Intermediate Aggregation
      Process architecture detailed in Section 4.2.

   Original Flow:  A Flow given as input to an Intermediate Aggregation
      Process in order to generate Aggregated Flows.

   Contributing Flow:  An Original Flow that is partially or completely
      represented within an Aggregated Flow.  Each Aggregated Flow is
      made up of zero or more Contributing Flows, and an Original Flow
      may contribute to zero or more Aggregated Flows.

   Original Exporter:  The Exporter from which the Original Flows are
      received; meaningful only when an IAP is deployed at a Mediator.

   The terminology presented herein improves the precision of, but does
   not supersede or contradict the terms related to, Mediation and
   aggregation defined in the Mediation Problem Statement [RFC5982] and
   the Mediation Framework [RFC6183] documents.  Within this document,
   the terminology defined in this section is to be considered
   normative.




Trammell, et al.             Standards Track                    [Page 6]

RFC 7015                    IPFIX Aggregation             September 2013


3.  Use Cases for IPFIX Aggregation



   Aggregation, as a common data reduction method used in traffic data
   analysis, has many applications.  When used with a regular
   Aggregation Interval and Original Flows containing timing
   information, it generates time series data from a collection of Flows
   with discrete intervals, as in the example in Section 8.1.  This time
   series data is itself useful for a wide variety of analysis tasks,
   such as generating input for network anomaly detection systems or
   driving visualizations of volume per time for traffic with specific
   characteristics.  As a second example, traffic matrix calculation
   from Flow data, as shown in Section 8.2 is inherently an aggregation
   action, by spatially aggregating the Flow Key down to input or output
   interface, address prefix, or autonomous system (AS).

   Irregular or data-dependent Aggregation Intervals and key aggregation
   operations can also be used to provide adaptive aggregation of
   network Flow data.  Here, full Flow Records can be kept for Flows of
   interest, while Flows deemed "less interesting" to a given
   application can be aggregated.  For example, in an IPFIX Mediator
   equipped with traffic classification capabilities for security
   purposes, potentially malicious Flows could be exported directly,
   while known-good or probably-good Flows (e.g., normal web browsing)
   could be exported simply as time series volumes per web server.

   Aggregation can also be applied to final analysis of stored Flow
   data, as shown in the example in Section 8.3.  All such aggregation
   applications in which timing information is not available or not
   important can be treated as if an infinite Aggregation Interval
   applies.

   Note that an Intermediate Aggregation Process that removes
   potentially sensitive information as identified in [RFC6235] may tend
   to have an anonymizing effect on the Aggregated Flows as well;
   however, any application of aggregation as part of a data protection
   scheme should ensure that all the issues raised in [RFC6235] are
   addressed, specifically Sections 4 ("Anonymization of IP Flow Data"),
   7.2 ("IPFIX-Specific Anonymization Guidelines"), and 9 ("Security
   Considerations").

   While much of the discussion in this document, and all of the
   examples, apply to the common case that the Original Flows to be
   aggregated are all of the same underlying type (i.e., are represented
   with identical Templates or compatible Templates containing a core
   set Information Elements that can be freely converted to one
   another), and that each packet observed by the Metering Process
   associated with the Original Exporter is represented, this is not a
   necessary assumption.  Aggregation can also be applied as part of a



Trammell, et al.             Standards Track                    [Page 7]

RFC 7015                    IPFIX Aggregation             September 2013


   technique using both aggregation and correlation to pull together
   multiple views of the same traffic from different Observation Points
   using different Templates.  For example, consider a set of
   applications running at different Observation Points for different
   purposes -- one generating Flows with round-trip times for passive
   performance measurement, and one generating billing records.  Once
   correlated, these Flows could be used to produce Aggregated Flows
   containing both volume and performance information together.  The
   correlation and normalization operation described in Section 4.2.1
   handles this specific case of correlation.  Flow correlation in the
   general case is outside the scope of this document.

4.  Architecture for Flow Aggregation



   This section specifies the architecture of the Intermediate
   Aggregation Process and how it fits into the IPFIX architecture.

4.1.  Aggregation within the IPFIX Architecture



   An Intermediate Aggregation Process could be deployed at any of three
   places within the IPFIX architecture.  While aggregation is most
   commonly done within a Mediator that collects Original Flows from an
   Original Exporter and exports Aggregated Flows, aggregation can also
   occur before initial export, or after final collection, as shown in
   Figure 1.  The presence of an IAP at any of these points is, of
   course, optional.

























Trammell, et al.             Standards Track                    [Page 8]

RFC 7015                    IPFIX Aggregation             September 2013


   +===========================================+
   |  IPFIX Exporter        +----------------+ |
   |                        | Metering Proc. | |
   | +-----------------+    +----------------+ |
   | | Metering Proc.  | or |      IAP       | |
   | +-----------------+----+----------------+ |
   | |           Exporting Process           | |
   | +-|----------------------------------|--+ |
   +===|==================================|====+
       |                                  |
   +===|===========================+      |
   |   |  Aggregating Mediator     |      |
   + +-V-------------------+       |      |
   | | Collecting Process  |       |      |
   + +---------------------+       |      |
   | |         IAP         |       |      |
   + +---------------------+       |      |
   | |  Exporting Process  |       |      |
   + +-|-------------------+       |      |
   +===|===========================+      |
       |                                  |
   +===|==================================|=====+
   |   | Collector                        |     |
   | +-V----------------------------------V-+   |
   | |         Collecting Process           |   |
   | +------------------+-------------------+   |
   |                    |        IAP        |   |
   |                    +-------------------+   |
   |  (Aggregation      |   File Writer     |   |
       for Storage)     +-----------|-------+   |
   +================================|===========+
                                    |
                             +------V-----------+
                             |    IPFIX File    |
                             +------------------+

                 Figure 1: Potential Aggregation Locations

   The Mediator use case is further shown in Figures A and B in
   [RFC6183].

   Aggregation can be applied for either intermediate or final analytic
   purposes.  In certain circumstances, it may make sense to export
   Aggregated Flows directly after metering, for example, if the
   Exporting Process is applied to drive a time series visualization, or
   when Flow data export bandwidth is restricted and Flow or packet
   sampling is not an option.  Note that this case, where the
   Aggregation Process is essentially integrated into the Metering



Trammell, et al.             Standards Track                    [Page 9]

RFC 7015                    IPFIX Aggregation             September 2013


   Process, is basically covered by the IPFIX architecture [RFC5470]:
   the Flow Keys used are simply a subset of those that would normally
   be used, and time intervals may be chosen other than those available
   from the cache policies customarily offered by the Metering Process.
   A Metering Process in this arrangement MAY choose to simulate the
   generation of larger Flows in order to generate Original Flow counts,
   if the application calls for compatibility with an Intermediate
   Aggregation Process deployed in a separate location.

   In the specific case that an Intermediate Aggregation Process is
   employed for data reduction for storage purposes, it can take
   Original Flows from a Collecting Process or File Reader and pass
   Aggregated Flows to a File Writer for storage.

   Deployment of an Intermediate Aggregation Process within a Mediator
   [RFC5982] is a much more flexible arrangement.  Here, the Mediator
   consumes Original Flows and produces Aggregated Flows; this
   arrangement is suited to any of the use cases detailed in Section 3.
   In a Mediator, Original Flows from multiple sources can also be
   aggregated into a single stream of Aggregated Flows.  The
   architectural specifics of this arrangement are not addressed in this
   document, which is concerned only with the aggregation operation
   itself.  See [IPFIX-MED-PROTO] for details.

   The data paths into and out of an Intermediate Aggregation Process
   are shown in Figure 2.

























Trammell, et al.             Standards Track                   [Page 10]

RFC 7015                    IPFIX Aggregation             September 2013


   packets --+               IPFIX Messages      IPFIX Files
             |                     |                  |
             V                     V                  V
   +==================+ +====================+ +=============+
   | Metering Process | | Collecting Process | | File Reader |
   |                  | +====================+ +=============+
   | (Original Flows  |            |                  |
   |    or direct     |            |  Original Flows  |
   |   aggregation)   |            V                  V
   + - - - - - - - - -+======================================+
   |           Intermediate Aggregation Process (IAP)        |
   +=========================================================+
             | Aggregated                  Aggregated |
             | Flows                            Flows |
             V                                        V
   +===================+                       +=============+
   | Exporting Process |                       | File Writer |
   +===================+                       +=============+
             |                                        |
             V                                        V
       IPFIX Messages                            IPFIX Files

           Figure 2: Data Paths through the Aggregation Process

   Note that as Aggregated Flows are IPFIX Flows, an Intermediate
   Aggregation Process may aggregate already Aggregated Flows from an
   upstream IAP as well as Original Flows from an upstream Original
   Exporter or Metering Process.

   Aggregation may also need to correlate Original Flows from multiple
   Metering Processes, each according to a different Template with
   different Flow Keys and values.  This arrangement is shown in Figure
   3; in this case, the correlation and normalization operation
   described in Section 4.2.1 handles merging the Original Flows before
   aggregation.
















Trammell, et al.             Standards Track                   [Page 11]

RFC 7015                    IPFIX Aggregation             September 2013


   packets --+---------------------+------------------+
             |                     |                  |
             V                     V                  V
   +====================+ +====================+ +====================+
   | Metering Process 1 | | Metering Process 2 | | Metering Process n |
   +====================+ +====================+ +====================+
             |                     |  Original Flows  |
             V                     V                  V
   +==================================================================+
   | Intermediate Aggregation Process  +  correlation / normalization |
   +==================================================================+
             | Aggregated                  Aggregated |
             | Flows                            Flows |
             V                                        V
   +===================+                       +=============+
   | Exporting Process |                       | File Writer |
   +===================+                       +=============+
             |                                        |
             +------------> IPFIX Messages <----------+

   Figure 3: Aggregating Original Flows from Multiple Metering Processes

4.2.  Intermediate Aggregation Process Architecture



   Within this document, an Intermediate Aggregation Process can be seen
   as hosting a function composed of four types of operations on
   Partially Aggregated Flows, as illustrated in Figure 4: interval
   distribution (temporal), key aggregation (spatial), value aggregation
   (spatial), and aggregate combination.  "Partially Aggregated Flows",
   as defined in Section 2, are essentially the intermediate results of
   aggregation, internal to the Intermediate Aggregation Process.




















Trammell, et al.             Standards Track                   [Page 12]

RFC 7015                    IPFIX Aggregation             September 2013


           Original Flows  /   Original Flows requiring correlation
   +=============|===================|===================|=============+
   |             |   Intermediate    |    Aggregation    |   Process   |
   |             |                   V                   V             |
   |             |   +-----------------------------------------------+ |
   |             |   |   (optional) correlation and normalization    | |
   |             |   +-----------------------------------------------+ |
   |             |                          |                          |
   |             V                          V                          |
   |  +--------------------------------------------------------------+ |
   |  |                interval distribution (temporal)              | |
   |  +--------------------------------------------------------------+ |
   |           | ^                         | ^                |        |
   |           | |  Partially Aggregated   | |                |        |
   |           V |         Flows           V |                |        |
   |  +-------------------+       +--------------------+      |        |
   |  |  key aggregation  |<------|  value aggregation |      |        |
   |  |     (spatial)     |------>|      (spatial)     |      |        |
   |  +-------------------+       +--------------------+      |        |
   |            |                          |                  |        |
   |            |   Partially Aggregated   |                  |        |
   |            V          Flows           V                  V        |
   |  +--------------------------------------------------------------+ |
   |  |                     aggregate combination                    | |
   |  +--------------------------------------------------------------+ |
   |                                       |                           |
   +=======================================|===========================+
                                           V
                                   Aggregated Flows

    Figure 4: Conceptual Model of Aggregation Operations within an IAP

   Interval distribution:  a temporal aggregation operation that imposes
      an Aggregation Interval on the Partially Aggregated Flow.  This
      Aggregation Interval may be regular, irregular, or derived from
      the timing of the Original Flows themselves.  Interval
      distribution is discussed in detail in Section 5.1.

   Key aggregation:  a spatial aggregation operation that results in the
      addition, modification, or deletion of Flow Key fields in the
      Partially Aggregated Flows.  New Flow Keys may be derived from
      existing Flow Keys (e.g., looking up an AS number (ASN) for an IP
      address), or "promoted" from specific non-key fields (e.g., when
      aggregating Flows by packet count per Flow).  Key aggregation can
      also add new non-key fields derived from Flow Keys that are
      deleted during key aggregation: mainly counters of unique reduced
      keys.  Key aggregation is discussed in detail in Section 5.2.




Trammell, et al.             Standards Track                   [Page 13]

RFC 7015                    IPFIX Aggregation             September 2013


   Value aggregation:  a spatial aggregation operation that results in
      the addition, modification, or deletion of non-key fields in the
      Partially Aggregated Flows.  These non-key fields may be "demoted"
      from existing key fields, or derived from existing key or non-key
      fields.  Value aggregation is discussed in detail in Section 5.3.

   Aggregate combination:  an operation combining multiple Partially
      Aggregated Flows having undergone interval distribution, key
      aggregation, and value aggregation that share Flow Keys and
      Aggregation Intervals into a single Aggregated Flow per set of
      Flow Key values and Aggregation Interval.  Aggregate combination
      is discussed in detail in Section 5.4.

   Correlation and normalization:  an optional operation that applies
      when accepting Original Flows from Metering Processes that export
      different views of essentially the same Flows before aggregation.
      The details of correlation and normalization are specified in
      Section 4.2.1, below.

   The first three of these operations may be carried out any number of
   times in any order, either on Original Flows or on the results of one
   of the operations above, with one caveat: since Flows carry their own
   interval data, any spatial aggregation operation implies a temporal
   aggregation operation, so at least one interval distribution step,
   even if implicit, is required by this architecture.  This is shown as
   the first step for the sake of simplicity in the diagram above.  Once
   all aggregation operations are complete, aggregate combination
   ensures that for a given Aggregation Interval, set of Flow Key
   values, and Observation Domain, only one Flow is produced by the
   Intermediate Aggregation Process.

   This model describes the operations within a single Intermediate
   Aggregation Process, and it is anticipated that most aggregation will
   be applied within a single process.  However, as the steps in the
   model may be applied in any order and aggregate combination is
   idempotent, any number of Intermediate Aggregation Processes
   operating in series can be modeled as a single process.  This allows
   aggregation operations to be flexibly distributed across any number
   of processes, should application or deployment considerations so
   dictate.

4.2.1.  Correlation and Normalization



   When accepting Original Flows from multiple Metering Processes, each
   of which provides a different view of the Original Flow as seen from
   the point of view of the IAP, an optional correlation and
   normalization operation combines each of these single Flow Records




Trammell, et al.             Standards Track                   [Page 14]

RFC 7015                    IPFIX Aggregation             September 2013


   into a set of unified Partially Aggregated Flows before applying
   interval distribution.  These unified Flows appear as if they had
   been measured at a single Metering Process that used the union of the
   set of Flow Keys and non-key fields of all Metering Processes sending
   Original Flows to the IAP.

   Since, due to export errors or other slight irregularities in Flow
   metering, the multiple views may not be completely consistent;
   normalization involves applying a set of corrections that are
   specific to the aggregation application in order to ensure
   consistency in the unified Flows.

   In general, correlation and normalization should take multiple views
   of essentially the same Flow, as determined by the configuration of
   the operation itself, and render them into a single unified Flow.
   Flows that are essentially different should not be unified by the
   correlation and normalization operation.  This operation therefore
   requires enough information about the configuration and deployment of
   Metering Processes from which it correlates Original Flows in order
   to make this distinction correctly and consistently.

   The exact steps performed to correlate and normalize Flows in this
   step are application, implementation, and deployment specific, and
   will not be further specified in this document.

5.  IP Flow Aggregation Operations



   As stated in Section 2, an Aggregated Flow is simply an IPFIX Flow
   generated from Original Flows by an Intermediate Aggregation Process.
   Here, we detail the operations by which this is achieved within an
   Intermediate Aggregation Process.

5.1.  Temporal Aggregation through Interval Distribution



   Interval distribution imposes a time interval on the resulting
   Aggregated Flows.  The selection of an interval is specific to the
   given aggregation application.  Intervals may be derived from the
   Original Flows themselves (e.g., an interval may be selected to cover
   the entire time containing the set of all Flows sharing a given Key,
   as in Time Composition, described in Section 5.1.2) or externally
   imposed; in the latter case the externally imposed interval may be
   regular (e.g., every five minutes) or irregular (e.g., to allow for
   different time resolutions at different times of day, under different
   network conditions, or indeed for different sets of Original Flows).

   The length of the imposed interval itself has trade-offs.  Shorter
   intervals allow higher-resolution aggregated data and, in streaming
   applications, faster reaction time.  Longer intervals generally lead



Trammell, et al.             Standards Track                   [Page 15]

RFC 7015                    IPFIX Aggregation             September 2013


   to greater data reduction and simplified counter distribution.
   Specifically, counter distribution is greatly simplified by the
   choice of an interval longer than the duration of longest Original
   Flow, itself generally determined by the Original Flow's Metering
   Process active timeout; in this case, an Original Flow can contribute
   to at most two Aggregated Flows, and the more complex value
   distribution methods become inapplicable.

   |                |                |                |
   | |<--Flow A-->| |                |                |
   |        |<--Flow B-->|           |                |
   |          |<-------------Flow C-------------->|   |
   |                |                |                |
   |   interval 0   |   interval 1   |   interval 2   |

              Figure 5: Illustration of Interval Distribution

   In Figure 5, we illustrate three common possibilities for interval
   distribution as applies with regular intervals to a set of three
   Original Flows.  For Flow A, the start and end times lie within the
   boundaries of a single interval 0; therefore, Flow A contributes to
   only one Aggregated Flow.  Flow B, by contrast, has the same duration
   but crosses the boundary between intervals 0 and 1; therefore, it
   will contribute to two Aggregated Flows, and its counters must be
   distributed among these Flows; though, in the two-interval case, this
   can be simplified somewhat simply by picking one of the two intervals
   or proportionally distributing between them.  Only Flows like Flow A
   and Flow B will be produced when the interval is chosen to be longer
   than the duration of longest Original Flow, as above.  More
   complicated is the case of Flow C, which contributes to more than two
   Aggregated Flows and must have its counters distributed according to
   some policy as in Section 5.1.1.

5.1.1.  Distributing Values across Intervals



   In general, counters in Aggregated Flows are treated the same as in
   any Flow.  Each counter is independently calculated as if it were
   derived from the set of packets in the Original Flow.  For example,
   delta counters are summed, the most recent total count for each
   Original Flow taken then summed across Flows, and so on.

   When the Aggregation Interval is guaranteed to be longer than the
   longest Original Flow, a Flow can cross at most one Interval
   boundary, and will therefore contribute to at most two Aggregated
   Flows.  Most common in this case is to arbitrarily but consistently
   choose to account the Original Flow's counters either to the first or
   to the last Aggregated Flow to which it could contribute.




Trammell, et al.             Standards Track                   [Page 16]

RFC 7015                    IPFIX Aggregation             September 2013


   However, this becomes more complicated when the Aggregation Interval
   is shorter than the longest Original Flow in the source data.  In
   such cases, each Original Flow can incompletely cover one or more
   time intervals, and apply to one or more Aggregated Flows.  In this
   case, the Intermediate Aggregation Process must distribute the
   counters in the Original Flows across one or more resulting
   Aggregated Flows.  There are several methods for doing this, listed
   here in roughly increasing order of complexity and accuracy; most of
   these are necessary only in specialized cases.

   End Interval:  The counters for an Original Flow are added to the
      counters of the appropriate Aggregated Flow containing the end
      time of the Original Flow.

   Start Interval:  The counters for an Original Flow are added to the
      counters of the appropriate Aggregated Flow containing the start
      time of the Original Flow.

   Mid Interval:  The counters for an Original Flow are added to the
      counters of a single appropriate Aggregated Flow containing some
      timestamp between start and end time of the Original Flow.

   Simple Uniform Distribution:  Each counter for an Original Flow is
      divided by the number of time intervals the Original Flow covers
      (i.e., of appropriate Aggregated Flows sharing the same Flow
      Keys), and this number is added to each corresponding counter in
      each Aggregated Flow.

   Proportional Uniform Distribution:  This is like simple uniform
      distribution, but accounts for the fractional portions of a time
      interval covered by an Original Flow in the first and last time
      interval.  Each counter for an Original Flow is divided by the
      number of time _units_ the Original Flow covers, to derive a mean
      count rate.  This rate is then multiplied by the number of time
      units in the intersection of the duration of the Original Flow and
      the time interval of each Aggregated Flow.

   Simulated Process:  Each counter of the Original Flow is distributed
      among the intervals of the Aggregated Flows according to some
      function the Intermediate Aggregation Process uses based upon
      properties of Flows presumed to be like the Original Flow.  For
      example, Flow Records representing bulk transfer might follow a
      more or less proportional uniform distribution, while interactive
      processes are far more bursty.

   Direct:  The Intermediate Aggregation Process has access to the
      original packet timings from the packets making up the Original
      Flow, and uses these to distribute or recalculate the counters.



Trammell, et al.             Standards Track                   [Page 17]

RFC 7015                    IPFIX Aggregation             September 2013


   A method for exporting the distribution of counters across multiple
   Aggregated Flows is detailed in Section 7.4.  In any case, counters
   MUST be distributed across the multiple Aggregated Flows in such a
   way that the total count is preserved, within the limits of accuracy
   of the implementation.  This property allows data to be aggregated
   and re-aggregated with negligible loss of original count information.
   To avoid confusion in interpretation of the aggregated data, all the
   counters in a given Aggregated Flow MUST be distributed via the same
   method.

   More complex counter distribution methods generally require that the
   interval distribution process track multiple "current" time intervals
   at once.  This may introduce some delay into the aggregation
   operation, as an interval should only expire and be available for
   export when no additional Original Flows applying to the interval are
   expected to arrive at the Intermediate Aggregation Process.

   Note, however, that since there is no guarantee that Flows from the
   Original Exporter will arrive in any given order, whether for
   transport-specific reasons (i.e., UDP reordering) or reasons specific
   to the implementation of the Metering Process or Exporting Process,
   even simpler distribution methods may need to deal with Flows
   arriving in an order other than start time or end time.  Therefore,
   the use of larger intervals does not obviate the need to buffer
   Partially Aggregated Flows within "current" time intervals, to ensure
   the IAP can accept Flow time intervals in any arrival order.  More
   generally, the interval distribution process SHOULD accept Flow start
   and end times in the Original Flows in any reasonable order.  The
   expiration of intervals in interval distribution operations is
   dependent on implementation and deployment requirements, and it MUST
   be made configurable in contexts in which "reasonable order" is not
   obvious at implementation time.  This operation may lead to delay and
   loss introduced by the IAP, as detailed in Section 6.2.

5.1.2.  Time Composition



   Time Composition, as in Section 5.4 of [RFC5982] (or interval
   combination), is a special case of aggregation, where interval
   distribution imposes longer intervals on Flows with matching keys and
   "chained" start and end times, without any key reduction, in order to
   join long-lived Flows that may have been split (e.g., due to an
   active timeout shorter than the actual duration of the Flow).  Here,
   no Key aggregation is applied, and the Aggregation Interval is chosen
   on a per-Flow basis to cover the interval spanned by the set of
   Aggregated Flows.  This may be applied alone in order to normalize
   split Flows, or it may be applied in combination with other
   aggregation functions in order to obtain more accurate Original Flow
   counts.



Trammell, et al.             Standards Track                   [Page 18]

RFC 7015                    IPFIX Aggregation             September 2013


5.1.3.  External Interval Distribution



   Note that much of the difficulty of interval distribution at an IAP
   can be avoided simply by configuring the original Exporters to
   synchronize the time intervals in the Original Flows with the desired
   aggregation interval.  The resulting Original Flows would then be
   split to align perfectly with the time intervals imposed during
   interval imposition, as shown in Figure 6, though this may reduce
   their usefulness for non-aggregation purposes.  This approach allows
   the Intermediate Aggregation Process to use Start Interval or End
   Interval distribution, while having equivalent information to that
   available to direct interval distribution.

   |                |                |                |
   |<----Flow D---->|<----Flow E---->|<----Flow F---->|
   |                |                |                |
   |   interval 0   |   interval 1   |   interval 2   |

         Figure 6: Illustration of External Interval Distribution

5.2.  Spatial Aggregation of Flow Keys



   Key aggregation generates a new set of Flow Key values for the
   Aggregated Flows from the Original Flow Key and non-key fields in the
   Original Flows or from correlation of the Original Flow information
   with some external source.  There are two basic operations here.
   First, Aggregated Flow Keys may be derived directly from Original
   Flow Keys through reduction, or they may be derived by the dropping
   of fields or precision in the Original Flow Keys.  Second, Aggregated
   Flow Keys may be derived through replacement, e.g., by removing one
   or more fields from the Original Flow and replacing them with fields
   derived from the removed fields.  Replacement may refer to external
   information (e.g., IP to AS number mappings).  Replacement may apply
   to Flow Keys as well as non-key fields.  For example, consider an
   application that aggregates Original Flows by packet count (i.e.,
   generating an Aggregated Flow for all one-packet Flows, one for all
   two-packet Flows, and so on).  This application would promote the
   packet count to a Flow Key.

   Key aggregation may also result in the addition of new non-key fields
   to the Aggregated Flows, namely, Original Flow counters and unique
   reduced key counters.  These are treated in more detail in Sections
   5.2.1 and 5.2.2, respectively.

   In any key aggregation operation, reduction and/or replacement may be
   applied any number of times in any order.  Which of these operations
   are supported by a given implementation is implementation and
   application dependent.



Trammell, et al.             Standards Track                   [Page 19]

RFC 7015                    IPFIX Aggregation             September 2013


   Original Flow Keys

   +---------+---------+----------+----------+-------+-----+
   | src ip4 | dst ip4 | src port | dst port | proto | tos |
   +---------+---------+----------+----------+-------+-----+
        |         |         |          |         |      |
     retain   mask /24      X          X         X      X
        |         |
        V         V
   +---------+-------------+
   | src ip4 | dst ip4 /24 |
   +---------+-------------+

   Aggregated Flow Keys (by source address and destination /24 network)

          Figure 7: Illustration of Key Aggregation by Reduction

   Figure 7 illustrates an example reduction operation, aggregation by
   source address and destination /24 network.  Here, the port,
   protocol, and type-of-service information is removed from the Flow
   Key, the source address is retained, and the destination address is
   masked by dropping the lower 8 bits.

   Original Flow Keys

   +---------+---------+----------+----------+-------+-----+
   | src ip4 | dst ip4 | src port | dst port | proto | tos |
   +---------+---------+----------+----------+-------+-----+
        |         |         |          |         |      |
        V         V         |          |         |      |
   +-------------------+    X          X         X      X
   | ASN lookup table  |
   +-------------------+
        |         |
        V         V
   +---------+---------+
   | src asn | dst asn |
   +---------+---------+

   Aggregated Flow Keys (by source and destination ASN)

                 Figure 8: Illustration of Key Aggregation
                       by Reduction and Replacement

   Figure 8 illustrates an example reduction and replacement operation,
   aggregation by source and destination Border Gateway Protocol (BGP)
   Autonomous System Number (ASN) without ASN information available in
   the Original Flow.  Here, the port, protocol, and type-of-service



Trammell, et al.             Standards Track                   [Page 20]

RFC 7015                    IPFIX Aggregation             September 2013


   information is removed from the Flow Keys, while the source and
   destination addresses are run though an IP address to ASN lookup
   table, and the Aggregated Flow Keys are made up of the resulting
   source and destination ASNs.

5.2.1.  Counting Original Flows



   When aggregating multiple Original Flows into an Aggregated Flow, it
   is often useful to know how many Original Flows are present in the
   Aggregated Flow. Section 7.2 introduces four new Information Elements
   to export these counters.

   There are two possible ways to count Original Flows, which we call
   conservative and non-conservative.  Conservative Flow counting has
   the property that each Original Flow contributes exactly one to the
   total Flow count within a set of Aggregated Flows.  In other words,
   conservative Flow counters are distributed just as any other counter
   during interval distribution, except each Original Flow is assumed to
   have a Flow count of one.  When a count for an Original Flow must be
   distributed across a set of Aggregated Flows, and a distribution
   method is used that does not account for that Original Flow
   completely within a single Aggregated Flow, conservative Flow
   counting requires a fractional representation.

   By contrast, non-conservative Flow counting is used to count how many
   Contributing Flows are represented in an Aggregated Flow.  Flow
   counters are not distributed in this case.  An Original Flow that is
   present within N Aggregated Flows would add N to the sum of non-
   conservative Flow counts, one to each Aggregated Flow.  In other
   words, the sum of conservative Flow counts over a set of Aggregated
   Flows is always equal to the number of Original Flows, while the sum
   of non-conservative Flow counts is strictly greater than or equal to
   the number of Original Flows.

   For example, consider Flows A, B, and C as illustrated in Figure 5.
   Assume that the key aggregation step aggregates the keys of these
   three Flows to the same aggregated Flow Key, and that start interval
   counter distribution is in effect.  The conservative Flow count for
   interval 0 is 3 (since Flows A, B, and C all begin in this interval),
   and for the other two intervals is 0.  The non-conservative Flow
   count for interval 0 is also 3 (due to the presence of Flows A, B,
   and C), for interval 1 is 2 (Flows B and C), and for interval 2 is 1
   (Flow C).  The sum of the conservative counts 3 + 0 + 0 = 3, the
   number of Original Flows; while the sum of the non-conservative
   counts 3 + 2 + 1 = 6.






Trammell, et al.             Standards Track                   [Page 21]

RFC 7015                    IPFIX Aggregation             September 2013


   Note that the active and inactive timeouts used to generate Original
   Flows, as well as the cache policy used to generate those Flows, have
   an effect on how meaningful either the conservative or non-
   conservative Flow count will be during aggregation.  In general,
   Original Exporters using the IPFIX Configuration Model SHOULD be
   configured to export Flows with equal or similar activeTimeout and
   inactiveTimeout configuration values, and the same cacheMode, as
   defined in [RFC6728].  Original Exporters not using the IPFIX
   Configuration Model SHOULD be configured equivalently.

5.2.2.  Counting Distinct Key Values



   One common case in aggregation is counting distinct key values that
   were reduced away during key aggregation.  The most common use case
   for this is counting distinct hosts per Flow Key; for example, in
   host characterization or anomaly detection, distinct sources per
   destination or distinct destinations per source are common metrics.
   These new non-key fields are added during key aggregation.

   For such applications, Information Elements for distinct counts of
   IPv4 and IPv6 addresses are defined in Section 7.3.  These are named
   distinctCountOf(KeyName).  Additional such Information Elements
   should be registered with IANA on an as-needed basis.

5.3.  Spatial Aggregation of Non-key Fields



   Aggregation operations may also lead to the addition of value fields
   that are demoted from key fields or are derived from other value
   fields in the Original Flows.  Specific cases of this are treated in
   the subsections below.

5.3.1.  Counter Statistics



   Some applications of aggregation may benefit from computing different
   statistics than those native to each non-key field (e.g., flags are
   natively combined via union and delta counters by summing).  For
   example, minimum and maximum packet counts per Flow, mean bytes per
   packet per Contributing Flow, and so on.  Certain Information
   Elements for these applications are already provided in the IANA
   IPFIX Information Elements registry [IANA-IPFIX] (e.g.,
   minimumIpTotalLength).

   A complete specification of additional aggregate counter statistics
   is outside the scope of this document, and should be added in the
   future to the IANA IPFIX Information Elements registry on a per-
   application, as-needed basis.





Trammell, et al.             Standards Track                   [Page 22]

RFC 7015                    IPFIX Aggregation             September 2013


5.3.2.  Derivation of New Values from Flow Keys and Non-key fields



   More complex operations may lead to other derived fields being
   generated from the set of values or Flow Keys reduced away during
   aggregation.  A prime example of this is sample entropy calculation.
   This counts distinct values and frequency, so it is similar to
   distinct key counting as in Section 5.2.2; however, it may be applied
   to the distribution of values for any Flow field.

   Sample entropy calculation provides a one-number normalized
   representation of the value spread and is useful for anomaly
   detection.  The behavior of entropy statistics is such that a small
   number of keys showing up very often drives the entropy value down
   towards zero, while a large number of keys, each showing up with
   lower frequency, drives the entropy value up.

   Entropy statistics are generally useful for identifier keys, such as
   IP addresses, port numbers, AS numbers, etc.  They can also be
   calculated on Flow length, Flow duration fields, and the like, even
   if this generally yields less distinct value shifts when the traffic
   mix changes.

   As a practical example, one host scanning a lot of other hosts will
   drive source IP entropy down and target IP entropy up.  A similar
   effect can be observed for ports.  This pattern can also be caused by
   the scan-traffic of a fast Internet worm.  A second example would be
   a Distributed Denial of Service (DDoS) flooding attack against a
   single target (or small number of targets) that drives source IP
   entropy up and target IP entropy down.

   A complete specification of additional derived values or entropy
   Information Elements is outside the scope of this document.  Any such
   Information Elements should be added in the future to the IANA IPFIX
   Information Elements registry on a per-application, as-needed basis.

5.4.  Aggregation Combination



   Interval distribution and key aggregation together may generate
   multiple Partially Aggregated Flows covering the same time interval
   with the same set of Flow Key values.  The process of combining these
   Partially Aggregated Flows into a single Aggregated Flow is called
   aggregation combination.  In general, non-Key values from multiple
   Contributing Flows are combined using the same operation by which
   values are combined from packets to form Flows for each Information
   Element.  Delta counters are summed, flags are unioned, and so on.






Trammell, et al.             Standards Track                   [Page 23]

RFC 7015                    IPFIX Aggregation             September 2013


6.  Additional Considerations and Special Cases in Flow Aggregation



6.1.  Exact versus Approximate Counting during Aggregation



   In certain circumstances, particularly involving aggregation by
   devices with limited resources, and in situations where exact
   aggregated counts are less important than relative magnitudes (e.g.,
   driving graphical displays), counter distribution during key
   aggregation may be performed by approximate counting means (e.g.,
   Bloom filters).  The choice to use approximate counting is
   implementation and application dependent.

6.2.  Delay and Loss Introduced by the IAP



   When accepting Original Flows in export order from traffic captured
   live, the Intermediate Aggregation Process waits for all Original
   Flows that may contribute to a given interval during interval
   distribution.  This is generally dominated by the active timeout of
   the Metering Process measuring the Original Flows.  For example, with
   Metering Processes configured with a five-minute active timeout, the
   Intermediate Aggregation Process introduces a delay of at least five
   minutes to all exported Aggregated Flows to ensure it has received
   all Original Flows.  Note that when aggregating Flows from multiple
   Metering Processes with different active timeouts, the delay is
   determined by the maximum active timeout.

   In certain circumstances, additional delay at the original Exporter
   may cause an IAP to close an interval before the last Original
   Flow(s) accountable to the interval arrives.  In this case, the IAP
   MAY drop the late Original Flow(s).  Accounting of Flows lost at an
   Intermediate Process due to such issues is covered in
   [IPFIX-MED-PROTO].

6.3.  Considerations for Aggregation of Sampled Flows



   The accuracy of Aggregated Flows may also be affected by sampling of
   the Original Flows, or sampling of packets making up the Original
   Flows.  At the time of writing, the effect of sampling on Flow
   aggregation is still an open research question.  However, to maximize
   the comparability of Aggregated Flows, aggregation of sampled Flows
   should only be applied to Original Flows sampled using the same
   sampling rate and sampling algorithm, Flows created from packets
   sampled using the same sampling rate and sampling algorithm, or
   Original Flows that have been normalized as if they had the same
   sampling rate and algorithm before aggregation.  For more on packet
   sampling within IPFIX, see [RFC5476].  For more on Flow sampling
   within the IPFIX Mediator framework, see [RFC7014].




Trammell, et al.             Standards Track                   [Page 24]

RFC 7015                    IPFIX Aggregation             September 2013


6.4.  Considerations for Aggregation of Heterogeneous Flows



   Aggregation may be applied to Original Flows from different sources
   and of different types (i.e., represented using different, perhaps
   wildly different Templates).  When the goal is to separate the
   heterogeneous Original Flows and aggregate them into heterogeneous
   Aggregated Flows, each aggregation should be done at its own
   Intermediate Aggregation Process.  The Observation Domain ID on the
   Messages containing the output Aggregated Flows can be used to
   identify the different Processes and to segregate the output.

   However, when the goal is to aggregate these Flows into a single
   stream of Aggregated Flows representing one type of data, and if the
   Original Flows may represent the same original packet at two
   different Observation Points, the Original Flows should be correlated
   by the correlation and normalization operation within the IAP to
   ensure that each packet is only represented in a single Aggregated
   Flow or set of Aggregated Flows differing only by aggregation
   interval.

7.  Export of Aggregated IP Flows Using IPFIX



   In general, Aggregated Flows are exported in IPFIX as any other Flow.
   However, certain aspects of Aggregated Flow export benefit from
   additional guidelines or new Information Elements to represent
   aggregation metadata or information generated during aggregation.
   These are detailed in the following subsections.

7.1.  Time Interval Export



   Since an Aggregated Flow is simply a Flow, the existing timestamp
   Information Elements in the IPFIX Information Model (e.g.,
   flowStartMilliseconds, flowEndNanoseconds) are sufficient to specify
   the time interval for aggregation.  Therefore, no new aggregation-
   specific Information Elements for exporting time interval information
   are necessary.

   Each Aggregated Flow carrying timing information SHOULD contain both
   an interval start and interval end timestamp.

7.2.  Flow Count Export



   The following four Information Elements are defined to count Original
   Flows as discussed in Section 5.2.1.







Trammell, et al.             Standards Track                   [Page 25]

RFC 7015                    IPFIX Aggregation             September 2013


7.2.1.  originalFlowsPresent



   Description:  The non-conservative count of Original Flows
      contributing to this Aggregated Flow.  Non-conservative counts
      need not sum to the original count on re-aggregation.

   Abstract Data Type:  unsigned64

   Data Type Semantics:  deltaCounter

   ElementID:  375

7.2.2.  originalFlowsInitiated



   Description:  The conservative count of Original Flows whose first
      packet is represented within this Aggregated Flow.  Conservative
      counts must sum to the original count on re-aggregation.

   Abstract Data Type:  unsigned64

   Data Type Semantics:  deltaCounter

   ElementID:  376

7.2.3.  originalFlowsCompleted



   Description:  The conservative count of Original Flows whose last
      packet is represented within this Aggregated Flow.  Conservative
      counts must sum to the original count on re-aggregation.

   Abstract Data Type:  unsigned64

   Data Type Semantics:  deltaCounter

   ElementID:  377

7.2.4.  deltaFlowCount



   Description:  The conservative count of Original Flows contributing
      to this Aggregated Flow; may be distributed via any of the methods
      expressed by the valueDistributionMethod Information Element.

   Abstract Data Type:  unsigned64

   Data Type Semantics:  deltaCounter

   ElementID:  3




Trammell, et al.             Standards Track                   [Page 26]

RFC 7015                    IPFIX Aggregation             September 2013


7.3.  Distinct Host Export



   The following six Information Elements represent the distinct counts
   of source and destination network-layer addresses used to export
   distinct host counts reduced away during key aggregation.

7.3.1.  distinctCountOfSourceIPAddress



   Description:  The count of distinct source IP address values for
      Original Flows contributing to this Aggregated Flow, without
      regard to IP version.  This Information Element is preferred to
      the IP-version-specific counters, unless it is important to
      separate the counts by version.

   Abstract Data Type:  unsigned64

   Data Type Semantics:  totalCounter

   ElementID:  378

7.3.2.  distinctCountOfDestinationIPAddress



   Description:  The count of distinct destination IP address values for
      Original Flows contributing to this Aggregated Flow, without
      regard to IP version.  This Information Element is preferred to
      the version-specific counters below, unless it is important to
      separate the counts by version.

   Abstract Data Type:  unsigned64

   Data Type Semantics:  totalCounter

   ElementID:  379

7.3.3.  distinctCountOfSourceIPv4Address



   Description:  The count of distinct source IPv4 address values for
      Original Flows contributing to this Aggregated Flow.

   Abstract Data Type:  unsigned32

   Data Type Semantics:  totalCounter

   ElementID:  380







Trammell, et al.             Standards Track                   [Page 27]

RFC 7015                    IPFIX Aggregation             September 2013


7.3.4.  distinctCountOfDestinationIPv4Address



   Description:  The count of distinct destination IPv4 address values
      for Original Flows contributing to this Aggregated Flow.

   Abstract Data Type:  unsigned32

   Data Type Semantics:  totalCounter

   ElementID:  381

7.3.5.  distinctCountOfSourceIPv6Address



   Description:  The count of distinct source IPv6 address values for
      Original Flows contributing to this Aggregated Flow.

   Abstract Data Type:  unsigned64

   Data Type Semantics:  totalCounter

   ElementID:  382

7.3.6.  distinctCountOfDestinationIPv6Address



   Description:  The count of distinct destination IPv6 address values
      for Original Flows contributing to this Aggregated Flow.

   Abstract Data Type:  unsigned64

   Data Type Semantics:  totalCounter

   ElementID:  383

7.4.  Aggregate Counter Distribution Export



   When exporting counters distributed among Aggregated Flows, as
   described in Section 5.1.1, the Exporting Process MAY export an
   Aggregate Counter Distribution Option Record for each Template
   describing Aggregated Flow records; this Options Template is
   described below.  It uses the valueDistributionMethod Information
   Element, also defined below.  Since, in many cases, distribution is
   simple, accounting the counters from Contributing Flows to the first
   Interval to which they contribute, this is the default situation, for
   which no Aggregate Counter Distribution Record is necessary;
   Aggregate Counter Distribution Records are only applicable in more
   exotic situations, such as using an Aggregation Interval smaller than
   the durations of Original Flows.




Trammell, et al.             Standards Track                   [Page 28]

RFC 7015                    IPFIX Aggregation             September 2013


7.4.1.  Aggregate Counter Distribution Options Template



   This Options Template defines the Aggregate Counter Distribution
   Record, which allows the binding of a value distribution method to a
   Template ID.  The scope is the Template ID, whose uniqueness, per
   [RFC7011], is local to the Transport Session and Observation Domain
   that generated the Template ID.  This is used to signal to the
   Collecting Process how the counters were distributed.  The fields are
   as below:

   +-----------------------------+-------------------------------------+
   | IE                          | Description                         |
   +-----------------------------+-------------------------------------+
   | templateId [scope]          | The Template ID of the Template     |
   |                             | defining the Aggregated Flows to    |
   |                             | which this distribution option      |
   |                             | applies.  This Information Element |
   |                             | MUST be defined as a Scope field.   |
   | valueDistributionMethod     | The method used to distribute the   |
   |                             | counters for the Aggregated Flows   |
   |                             | defined by the associated Template. |
   +-----------------------------+-------------------------------------+

7.4.2.  valueDistributionMethod Information Element



   Description:  A description of the method used to distribute the
      counters from Contributing Flows into the Aggregated Flow records
      described by an associated scope, generally a Template.  The
      method is deemed to apply to all the non-Key Information Elements
      in the referenced scope for which value distribution is a valid
      operation; if the originalFlowsInitiated and/or
      originalFlowsCompleted Information Elements appear in the
      Template, they are not subject to this distribution method, as
      they each infer their own distribution method.  This is intended
      to be a complete set of possible value distribution methods; it is
      encoded as follows:















Trammell, et al.             Standards Track                   [Page 29]

RFC 7015                    IPFIX Aggregation             September 2013


   +-------+-----------------------------------------------------------+
   | Value | Description                                               |
   +-------+-----------------------------------------------------------+
   | 0     | Unspecified: The counters for an Original Flow are        |
   |       | explicitly not distributed according to any other method  |
   |       | defined for this Information Element; use for arbitrary   |
   |       | distribution, or distribution algorithms not described by |
   |       | any other codepoint.                                      |
   |       | --------------------------------------------------------- |
   |       |                                                           |
   | 1     | Start Interval: The counters for an Original Flow are     |
   |       | added to the counters of the appropriate Aggregated Flow  |
   |       | containing the start time of the Original Flow.  This     |
   |       | should be assumed the default if value distribution       |
   |       | information is not available at a Collecting Process for  |
   |       | an Aggregated Flow.                                       |
   |       | --------------------------------------------------------- |
   |       |                                                           |
   | 2     | End Interval: The counters for an Original Flow are added |
   |       | to the counters of the appropriate Aggregated Flow        |
   |       | containing the end time of the Original Flow.             |
   |       | --------------------------------------------------------- |
   |       |                                                           |
   | 3     | Mid Interval: The counters for an Original Flow are added |
   |       | to the counters of a single appropriate Aggregated Flow   |
   |       | containing some timestamp between start and end time of   |
   |       | the Original Flow.                                        |
   |       | --------------------------------------------------------- |
   |       |                                                           |
   | 4     | Simple Uniform Distribution: Each counter for an Original |
   |       | Flow is divided by the number of time intervals the       |
   |       | Original Flow covers (i.e., of appropriate Aggregated     |
   |       | Flows sharing the same Flow Key), and this number is      |
   |       | added to each corresponding counter in each Aggregated    |
   |       | Flow.                                                     |
   |       | --------------------------------------------------------- |
   |       |                                                           |
   | 5     | Proportional Uniform Distribution: Each counter for an    |
   |       | Original Flow is divided by the number of time units the  |
   |       | Original Flow covers, to derive a mean count rate.  This  |
   |       | mean count rate is then multiplied by the number of time  |
   |       | units in the intersection of the duration of the Original |
   |       | Flow and the time interval of each Aggregated Flow.       |
   |       |  This is like simple uniform distribution, but accounts   |
   |       | for the fractional portions of a time interval covered by |
   |       | an Original Flow in the first and last time interval.     |
   |       | --------------------------------------------------------- |




Trammell, et al.             Standards Track                   [Page 30]

RFC 7015                    IPFIX Aggregation             September 2013


   |       | --------------------------------------------------------- |
   | 6     | Simulated Process: Each counter of the Original Flow is   |
   |       | distributed among the intervals of the Aggregated Flows   |
   |       | according to some function the Intermediate Aggregation   |
   |       | Process uses based upon properties of Flows presumed to   |
   |       | be like the Original Flow.  This is essentially an        |
   |       | assertion that the Intermediate Aggregation Process has   |
   |       | no direct packet timing information but is nevertheless   |
   |       | not using one of the other simpler distribution methods.  |
   |       | The Intermediate Aggregation Process specifically makes   |
   |       | no assertion as to the correctness of the simulation.     |
   |       | --------------------------------------------------------- |
   |       |                                                           |
   | 7     | Direct: The Intermediate Aggregation Process has access   |
   |       | to the original packet timings from the packets making up |
   |       | the Original Flow, and uses these to distribute or        |
   |       | recalculate the counters.                                 |
   +-------+-----------------------------------------------------------+

   Abstract Data Type:  unsigned8

   ElementID:  384

8.  Examples



   In these examples, the same data, described by the same Template,
   will be aggregated multiple different ways; this illustrates the
   various different functions that could be implemented by Intermediate
   Aggregation Processes.  Templates are shown in IESpec format as
   introduced in [RFC7013].  The source data format is a simplified
   Flow: timestamps, traditional 5-tuple, and octet count; the Flow Key
   fields are the 5-tuple.  The Template is shown in Figure 9.

   flowStartMilliseconds(152)[8]
   flowEndMilliseconds(153)[8]
   sourceIPv4Address(8)[4]{key}
   destinationIPv4Address(12)[4]{key}
   sourceTransportPort(7)[2]{key}
   destinationTransportPort(11)[2]{key}
   protocolIdentifier(4)[1]{key}
   octetDeltaCount(1)[8]

                   Figure 9: Input Template for Examples

   The data records given as input to the examples in this section are
   shown below; timestamps are given in H:MM:SS.sss format.  In this and
   subsequent figures, flowStartMilliseconds is shown in H:MM:SS.sss
   format as 'start time', flowEndMilliseconds is shown in H:MM:SS.sss



Trammell, et al.             Standards Track                   [Page 31]

RFC 7015                    IPFIX Aggregation             September 2013


   format as 'end time', sourceIPv4Address is shown as 'source ip4' with
   the following 'port' representing sourceTransportPort,
   destinationIPv4Address is shown as 'dest ip4' with the following
   'port' representing destinationTransportPort, protocolIdentifier is
   shown as 'pt', and octetDeltaCount as 'oct'.

  start time |end time   |source ip4 |port |dest ip4      |port|pt|  oct
  9:00:00.138 9:00:00.138 192.0.2.2   47113 192.0.2.131    53   17   119
  9:00:03.246 9:00:03.246 192.0.2.2   22153 192.0.2.131    53   17    83
  9:00:00.478 9:00:03.486 192.0.2.2   52420 198.51.100.2   443  6   1637
  9:00:07.172 9:00:07.172 192.0.2.3   56047 192.0.2.131    53   17   111
  9:00:07.309 9:00:14.861 192.0.2.3   41183 198.51.100.67  80   6  16838
  9:00:03.556 9:00:19.876 192.0.2.2   17606 198.51.100.68  80   6  11538
  9:00:25.210 9:00:25.210 192.0.2.3   47113 192.0.2.131    53   17   119
  9:00:26.358 9:00:30.198 192.0.2.3   48458 198.51.100.133 80   6   2973
  9:00:29.213 9:01:00.061 192.0.2.4   61295 198.51.100.2   443  6   8350
  9:04:00.207 9:04:04.431 203.0.113.3 41256 198.51.100.133 80   6    778
  9:03:59.624 9:04:06.984 203.0.113.3 51662 198.51.100.3   80   6    883
  9:00:30.532 9:06:15.402 192.0.2.2   37581 198.51.100.2   80   6  15420
  9:06:56.813 9:06:59.821 203.0.113.3 52572 198.51.100.2   443  6   1637
  9:06:30.565 9:07:00.261 203.0.113.3 49914 198.51.100.133 80   6    561
  9:06:55.160 9:07:05.208 192.0.2.2   50824 198.51.100.2   443  6   1899
  9:06:49.322 9:07:05.322 192.0.2.3   34597 198.51.100.3   80   6   1284
  9:07:05.849 9:07:09.625 203.0.113.3 58907 198.51.100.4   80   6   2670
  9:10:45.161 9:10:45.161 192.0.2.4   22478 192.0.2.131    53   17    75
  9:10:45.209 9:11:01.465 192.0.2.4   49513 198.51.100.68  80   6   3374
  9:10:57.094 9:11:00.614 192.0.2.4   64832 198.51.100.67  80   6    138
  9:10:59.770 9:11:02.842 192.0.2.3   60833 198.51.100.69  443  6   2325
  9:02:18.390 9:13:46.598 203.0.113.3 39586 198.51.100.17  80   6  11200
  9:13:53.933 9:14:06.605 192.0.2.2   19638 198.51.100.3   80   6   2869
  9:13:02.864 9:14:08.720 192.0.2.3   40429 198.51.100.4   80   6  18289

                    Figure 10: Input Data for Examples

8.1.  Traffic Time Series per Source



   Aggregating Flows by source IP address in time series (i.e., with a
   regular interval) can be used in subsequent heavy-hitter analysis and
   as a source parameter for statistical anomaly detection techniques.
   Here, the Intermediate Aggregation Process imposes an interval,
   aggregates the key to remove all key fields other than the source IP
   address, then combines the result into a stream of Aggregated Flows.
   The imposed interval of five minutes is longer than the majority of
   Flows; for those Flows crossing interval boundaries, the entire Flow
   is accounted to the interval containing the start time of the Flow.






Trammell, et al.             Standards Track                   [Page 32]

RFC 7015                    IPFIX Aggregation             September 2013


   In this example, the Partially Aggregated Flows after each conceptual
   operation in the Intermediate Aggregation Process are shown.  These
   are meant to be illustrative of the conceptual operations only, and
   not to suggest an implementation (indeed, the example shown here
   would not necessarily be the most efficient method for performing
   these operations).  Subsequent examples will omit the Partially
   Aggregated Flows for brevity.

   The input to this process could be any Flow Record containing a
   source IP address and octet counter; consider for this example the
   Template and data from the introduction.  The Intermediate
   Aggregation Process would then output records containing just
   timestamps, source IP, and octetDeltaCount, as in Figure 11.

   flowStartMilliseconds(152)[8]
   flowEndMilliseconds(153)[8]
   sourceIPv4Address(8)[4]
   octetDeltaCount(1)[8]

           Figure 11: Output Template for Time Series per Source































Trammell, et al.             Standards Track                   [Page 33]

RFC 7015                    IPFIX Aggregation             September 2013


   Assume the goal is to get 5-minute (300 s) time series of octet
   counts per source IP address.  The aggregation operations would then
   be arranged as in Figure 12.

                    Original Flows
                          |
                          V
              +-----------------------+
              | interval distribution |
              |  * impose uniform     |
              |    300s time interval |
              +-----------------------+
                  |
                  | Partially Aggregated Flows
                  V
   +------------------------+
   |  key aggregation       |
   |   * reduce key to only |
   |     sourceIPv4Address  |
   +------------------------+
                  |
                  | Partially Aggregated Flows
                  V
             +-------------------------+
             |  aggregate combination  |
             |   * sum octetDeltaCount |
             +-------------------------+
                          |
                          V
                  Aggregated Flows

       Figure 12: Aggregation Operations for Time Series per Source

   After applying the interval distribution step to the source data in
   Figure 10, only the time intervals have changed; the Partially
   Aggregated Flows are shown in Figure 13.  Note that interval
   distribution follows the default Start Interval policy; that is, the
   entire Flow is accounted to the interval containing the Flow's start
   time.












Trammell, et al.             Standards Track                   [Page 34]

RFC 7015                    IPFIX Aggregation             September 2013


  start time |end time   |source ip4 |port |dest ip4      |port|pt|  oct
  9:00:00.000 9:05:00.000 192.0.2.2   47113 192.0.2.131    53   17   119
  9:00:00.000 9:05:00.000 192.0.2.2   22153 192.0.2.131    53   17    83
  9:00:00.000 9:05:00.000 192.0.2.2   52420 198.51.100.2   443  6   1637
  9:00:00.000 9:05:00.000 192.0.2.3   56047 192.0.2.131    53   17   111
  9:00:00.000 9:05:00.000 192.0.2.3   41183 198.51.100.67  80   6  16838
  9:00:00.000 9:05:00.000 192.0.2.2   17606 198.51.100.68  80   6  11538
  9:00:00.000 9:05:00.000 192.0.2.3   47113 192.0.2.131    53   17   119
  9:00:00.000 9:05:00.000 192.0.2.3   48458 198.51.100.133 80   6   2973
  9:00:00.000 9:05:00.000 192.0.2.4   61295 198.51.100.2   443  6   8350
  9:00:00.000 9:05:00.000 203.0.113.3 41256 198.51.100.133 80   6    778
  9:00:00.000 9:05:00.000 203.0.113.3 51662 198.51.100.3   80   6    883
  9:00:00.000 9:05:00.000 192.0.2.2   37581 198.51.100.2   80   6  15420
  9:00:00.000 9:05:00.000 203.0.113.3 39586 198.51.100.17  80   6  11200
  9:05:00.000 9:10:00.000 203.0.113.3 52572 198.51.100.2   443  6   1637
  9:05:00.000 9:10:00.000 203.0.113.3 49914 197.51.100.133 80   6    561
  9:05:00.000 9:10:00.000 192.0.2.2   50824 198.51.100.2   443  6   1899
  9:05:00.000 9:10:00.000 192.0.2.3   34597 198.51.100.3   80   6   1284
  9:05:00.000 9:10:00.000 203.0.113.3 58907 198.51.100.4   80   6   2670
  9:10:00.000 9:15:00.000 192.0.2.4   22478 192.0.2.131    53   17    75
  9:10:00.000 9:15:00.000 192.0.2.4   49513 198.51.100.68  80   6   3374
  9:10:00.000 9:15:00.000 192.0.2.4   64832 198.51.100.67  80   6    138
  9:10:00.000 9:15:00.000 192.0.2.3   60833 198.51.100.69  443  6   2325
  9:10:00.000 9:15:00.000 192.0.2.2   19638 198.51.100.3   80   6   2869
  9:10:00.000 9:15:00.000 192.0.2.3   40429 198.51.100.4   80   6  18289

         Figure 13: Interval Imposition for Time Series per Source

   After the key aggregation step, all Flow Keys except the source IP
   address have been discarded, as shown in Figure 14.  This leaves
   duplicate Partially Aggregated Flows to be combined in the final
   operation.



















Trammell, et al.             Standards Track                   [Page 35]

RFC 7015                    IPFIX Aggregation             September 2013


   start time |end time   |source ip4 |octets
   9:00:00.000 9:05:00.000 192.0.2.2      119
   9:00:00.000 9:05:00.000 192.0.2.2       83
   9:00:00.000 9:05:00.000 192.0.2.2     1637
   9:00:00.000 9:05:00.000 192.0.2.3      111
   9:00:00.000 9:05:00.000 192.0.2.3    16838
   9:00:00.000 9:05:00.000 192.0.2.2    11538
   9:00:00.000 9:05:00.000 192.0.2.3      119
   9:00:00.000 9:05:00.000 192.0.2.3     2973
   9:00:00.000 9:05:00.000 192.0.2.4     8350
   9:00:00.000 9:05:00.000 203.0.113.3    778
   9:00:00.000 9:05:00.000 203.0.113.3    883
   9:00:00.000 9:05:00.000 192.0.2.2    15420
   9:00:00.000 9:05:00.000 203.0.113.3  11200
   9:05:00.000 9:10:00.000 203.0.113.3   1637
   9:05:00.000 9:10:00.000 203.0.113.3    561
   9:05:00.000 9:10:00.000 192.0.2.2     1899
   9:05:00.000 9:10:00.000 192.0.2.3     1284
   9:05:00.000 9:10:00.000 203.0.113.3   2670
   9:10:00.000 9:15:00.000 192.0.2.4       75
   9:10:00.000 9:15:00.000 192.0.2.4     3374
   9:10:00.000 9:15:00.000 192.0.2.4      138
   9:10:00.000 9:15:00.000 192.0.2.3     2325
   9:10:00.000 9:15:00.000 192.0.2.2     2869
   9:10:00.000 9:15:00.000 192.0.2.3    18289

           Figure 14: Key Aggregation for Time Series per Source

   Aggregate combination sums the counters per key and interval; the
   summations of the first two keys and intervals are shown in detail in
   Figure 15.




















Trammell, et al.             Standards Track                   [Page 36]

RFC 7015                    IPFIX Aggregation             September 2013


     start time |end time   |source ip4 |octets
     9:00:00.000 9:05:00.000 192.0.2.2      119
     9:00:00.000 9:05:00.000 192.0.2.2       83
     9:00:00.000 9:05:00.000 192.0.2.2     1637
     9:00:00.000 9:05:00.000 192.0.2.2    11538
   + 9:00:00.000 9:05:00.000 192.0.2.2    15420
                                          -----
   = 9:00:00.000 9:05:00.000 192.0.2.2    28797

     9:00:00.000 9:05:00.000 192.0.2.3      111
     9:00:00.000 9:05:00.000 192.0.2.3    16838
     9:00:00.000 9:05:00.000 192.0.2.3      119
   + 9:00:00.000 9:05:00.000 192.0.2.3     2973
                                          -----
   = 9:00:00.000 9:05:00.000 192.0.2.3    20041

             Figure 15: Summation during Aggregate Combination

   This can be applied to each set of Partially Aggregated Flows to
   produce the final Aggregated Flows that are shown in Figure 16, as
   exported by the Template in Figure 11.

   start time |end time   |source ip4 |octets
   9:00:00.000 9:05:00.000 192.0.2.2    28797
   9:00:00.000 9:05:00.000 192.0.2.3    20041
   9:00:00.000 9:05:00.000 192.0.2.4     8350
   9:00:00.000 9:05:00.000 203.0.113.3  12861
   9:05:00.000 9:10:00.000 192.0.2.2     1899
   9:05:00.000 9:10:00.000 192.0.2.3     1284
   9:05:00.000 9:10:00.000 203.0.113.3   4868
   9:10:00.000 9:15:00.000 192.0.2.2     2869
   9:10:00.000 9:15:00.000 192.0.2.3    20614
   9:10:00.000 9:15:00.000 192.0.2.4     3587

          Figure 16: Aggregated Flows for Time Series per Source

8.2.  Core Traffic Matrix



   Aggregating Flows by source and destination ASN in time series is
   used to generate core traffic matrices.  The core traffic matrix
   provides a view of the state of the routes within a network, and it
   can be used for long-term planning of changes to network design based
   on traffic demand.  Here, imposed time intervals are generally much
   longer than active Flow timeouts.  The traffic matrix is reported in
   terms of octets, packets, and flows, as each of these values may have
   a subtly different effect on capacity planning.





Trammell, et al.             Standards Track                   [Page 37]

RFC 7015                    IPFIX Aggregation             September 2013


   This example demonstrates key aggregation using derived keys and
   Original Flow counting.  While some Original Flows may be generated
   by Exporting Processes on forwarding devices, and therefore contain
   the bgpSourceAsNumber and bgpDestinationAsNumber Information
   Elements, Original Flows from Exporting Processes on dedicated
   measurement devices without routing data contain only a
   destinationIPv[46]Address.  For these Flows, the Mediator must look
   up a next-hop AS from an IP-to-AS table, replacing source and
   destination addresses with ASNs.  The table used in this example is
   shown in Figure 17.  (Note that due to limited example address space,
   in this example we ignore the common practice of routing only blocks
   of /24 or larger.)

   prefix           |ASN
   192.0.2.0/25      64496
   192.0.2.128/25    64497
   198.51.100/24     64498
   203.0.113.0/24    64499

                        Figure 17: Example ASN Map

   The Template for Aggregated Flows produced by this example is shown
   in Figure 18.

   flowStartMilliseconds(152)[8]
   flowEndMilliseconds(153)[8]
   bgpSourceAsNumber(16)[4]
   bgpDestinationAsNumber(17)[4]
   octetDeltaCount(1)[8]

               Figure 18: Output Template for Traffic Matrix

   Assume the goal is to get 60-minute time series of octet counts per
   source/destination ASN pair.  The aggregation operations would then
   be arranged as in Figure 19.
















Trammell, et al.             Standards Track                   [Page 38]

RFC 7015                    IPFIX Aggregation             September 2013


                    Original Flows
                          |
                          V
              +-----------------------+
              | interval distribution |
              |  * impose uniform     |
              |    3600s time interval|
              +-----------------------+
                  |
                  | Partially Aggregated Flows
                  V
   +------------------------+
   |  key aggregation       |
   |  * reduce key to only  |
   |    sourceIPv4Address + |
   |    destIPv4Address     |
   +------------------------+
                  |
                  V
   +------------------------+
   |  key aggregation       |
   |  * replace addresses   |
   |    with ASN from map   |
   +------------------------+
                  |
                  | Partially Aggregated Flows
                  V
             +-------------------------+
             |  aggregate combination  |
             |   * sum octetDeltaCount |
             +-------------------------+
                          |
                          V
                  Aggregated Flows

           Figure 19: Aggregation Operations for Traffic Matrix

   After applying the interval distribution step to the source data in
   Figure 10, the Partially Aggregated Flows are shown in Figure 20.
   Note that the Flows are identical to those in the interval
   distribution step in the previous example, except the chosen interval
   (1 hour, 3600 seconds) is different; therefore, all the Flows fit
   into a single interval.








Trammell, et al.             Standards Track                   [Page 39]

RFC 7015                    IPFIX Aggregation             September 2013


   start time |end time |source ip4 |port |dest ip4      |port|pt|  oct
   9:00:00     10:00:00  192.0.2.2   47113 192.0.2.131    53   17   119
   9:00:00     10:00:00  192.0.2.2   22153 192.0.2.131    53   17    83
   9:00:00     10:00:00  192.0.2.2   52420 198.51.100.2   443  6   1637
   9:00:00     10:00:00  192.0.2.3   56047 192.0.2.131    53   17   111
   9:00:00     10:00:00  192.0.2.3   41183 198.51.100.67  80   6  16838
   9:00:00     10:00:00  192.0.2.2   17606 198.51.100.68  80   6  11538
   9:00:00     10:00:00  192.0.2.3   47113 192.0.2.131    53   17   119
   9:00:00     10:00:00  192.0.2.3   48458 198.51.100.133 80   6   2973
   9:00:00     10:00:00  192.0.2.4   61295 198.51.100.2   443  6   8350
   9:00:00     10:00:00  203.0.113.3 41256 198.51.100.133 80   6    778
   9:00:00     10:00:00  203.0.113.3 51662 198.51.100.3   80   6    883
   9:00:00     10:00:00  192.0.2.2   37581 198.51.100.2   80   6  15420
   9:00:00     10:00:00  203.0.113.3 52572 198.51.100.2   443  6   1637
   9:00:00     10:00:00  203.0.113.3 49914 197.51.100.133 80   6    561
   9:00:00     10:00:00  192.0.2.2   50824 198.51.100.2   443  6   1899
   9:00:00     10:00:00  192.0.2.3   34597 198.51.100.3   80   6   1284
   9:00:00     10:00:00  203.0.113.3 58907 198.51.100.4   80   6   2670
   9:00:00     10:00:00  192.0.2.4   22478 192.0.2.131    53   17    75
   9:00:00     10:00:00  192.0.2.4   49513 198.51.100.68  80   6   3374
   9:00:00     10:00:00  192.0.2.4   64832 198.51.100.67  80   6    138
   9:00:00     10:00:00  192.0.2.3   60833 198.51.100.69  443  6   2325
   9:00:00     10:00:00  203.0.113.3 39586 198.51.100.17  80   6  11200
   9:00:00     10:00:00  192.0.2.2   19638 198.51.100.3   80   6   2869
   9:00:00     10:00:00  192.0.2.3   40429 198.51.100.4   80   6  18289

             Figure 20: Interval Imposition for Traffic Matrix

   The next steps are to discard irrelevant key fields and to replace
   the source and destination addresses with source and destination ASNs
   in the map; the results of these key aggregation steps are shown in
   Figure 21.



















Trammell, et al.             Standards Track                   [Page 40]

RFC 7015                    IPFIX Aggregation             September 2013


   start time |end time |source ASN |dest ASN |octets
   9:00:00     10:00:00  AS64496     AS64497      119
   9:00:00     10:00:00  AS64496     AS64497       83
   9:00:00     10:00:00  AS64496     AS64498     1637
   9:00:00     10:00:00  AS64496     AS64497      111
   9:00:00     10:00:00  AS64496     AS64498    16838
   9:00:00     10:00:00  AS64496     AS64498    11538
   9:00:00     10:00:00  AS64496     AS64497      119
   9:00:00     10:00:00  AS64496     AS64498     2973
   9:00:00     10:00:00  AS64496     AS64498     8350
   9:00:00     10:00:00  AS64499     AS64498      778
   9:00:00     10:00:00  AS64499     AS64498      883
   9:00:00     10:00:00  AS64496     AS64498    15420
   9:00:00     10:00:00  AS64499     AS64498     1637
   9:00:00     10:00:00  AS64499     AS64498      561
   9:00:00     10:00:00  AS64496     AS64498     1899
   9:00:00     10:00:00  AS64496     AS64498     1284
   9:00:00     10:00:00  AS64499     AS64498     2670
   9:00:00     10:00:00  AS64496     AS64497       75
   9:00:00     10:00:00  AS64496     AS64498     3374
   9:00:00     10:00:00  AS64496     AS64498      138
   9:00:00     10:00:00  AS64496     AS64498     2325
   9:00:00     10:00:00  AS64499     AS64498    11200
   9:00:00     10:00:00  AS64496     AS64498     2869
   9:00:00     10:00:00  AS64496     AS64498    18289

              Figure 21: Key Aggregation for Traffic Matrix:
                         Reduction and Replacement

   Finally, aggregate combination sums the counters per key and
   interval.  The resulting Aggregated Flows containing the traffic
   matrix, shown in Figure 22, are then exported using the Template in
   Figure 18.  Note that these Aggregated Flows represent a sparse
   matrix: AS pairs for which no traffic was received have no
   corresponding record in the output.

   start time  end time  source ASN  dest ASN  octets
   9:00:00     10:00:00  AS64496     AS64497      507
   9:00:00     10:00:00  AS64496     AS64498    86934
   9:00:00     10:00:00  AS64499     AS64498    17729

              Figure 22: Aggregated Flows for Traffic Matrix

   The output of this operation is suitable for re-aggregation: that is,
   traffic matrices from single links or Observation Points can be
   aggregated through the same interval imposition and aggregate
   combination steps in order to build a traffic matrix for an entire
   network.



Trammell, et al.             Standards Track                   [Page 41]

RFC 7015                    IPFIX Aggregation             September 2013


8.3.  Distinct Source Count per Destination Endpoint



   Aggregating Flows by destination address and port, and counting
   distinct sources aggregated away, can be used as part of passive
   service inventory and host characterization.  This example shows
   aggregation as an analysis technique, performed on source data stored
   in an IPFIX File.  As the Transport Session in this File is bounded,
   removal of all timestamp information allows summarization of the
   entire time interval contained within the interval.  Removal of
   timing information during interval imposition is equivalent to an
   infinitely long imposed time interval.  This demonstrates both how
   infinite intervals work, and how unique counters work.  The
   aggregation operations are summarized in Figure 23.






































Trammell, et al.             Standards Track                   [Page 42]

RFC 7015                    IPFIX Aggregation             September 2013


                    Original Flows
                          |
                          V
              +-----------------------+
              | interval distribution |
              |  * discard timestamps |
              +-----------------------+
                  |
                  | Partially Aggregated Flows
                  V
   +----------------------------+
   |  value aggregation         |
   |  * discard octetDeltaCount |
   +----------------------------+
                  |
                  | Partially Aggregated Flows
                  V
   +----------------------------+
   |  key aggregation           |
   |   * reduce key to only     |
   |     destIPv4Address +      |
   |     destTransportPort,     |
   |   * count distinct sources |
   +----------------------------+
                  |
                  | Partially Aggregated Flows
                  V
       +----------------------------------------------+
       |  aggregate combination                       |
       |   * no-op (distinct sources already counted) |
       +----------------------------------------------+
                          |
                          V
                  Aggregated Flows

            Figure 23: Aggregation Operations for Source Count

   The Template for Aggregated Flows produced by this example is shown
   in Figure 24.

   destinationIPv4Address(12)[4]
   destinationTransportPort(11)[2]
   distinctCountOfSourceIPAddress(378)[8]

                Figure 24: Output Template for Source Count






Trammell, et al.             Standards Track                   [Page 43]

RFC 7015                    IPFIX Aggregation             September 2013


   Interval distribution, in this case, merely discards the timestamp
   information from the Original Flows in Figure 10, and as such is not
   shown.  Likewise, the value aggregation step simply discards the
   octetDeltaCount value field.  The key aggregation step reduces the
   key to the destinationIPv4Address and destinationTransportPort,
   counting the distinct source addresses.  Since this is essentially
   the output of this aggregation function, the aggregate combination
   operation is a no-op; the resulting Aggregated Flows are shown in
   Figure 25.

   dest ip4      |port |dist src
   192.0.2.131    53           3
   198.51.100.2   80           1
   198.51.100.2   443          3
   198.51.100.67  80           2
   198.51.100.68  80           2
   198.51.100.133 80           2
   198.51.100.3   80           3
   198.51.100.4   80           2
   198.51.100.17  80           1
   198.51.100.69  443          1

               Figure 25: Aggregated Flows for Source Count

8.4.  Traffic Time Series per Source with Counter Distribution



   Returning to the example in Section 8.1, note that our source data
   contains some Flows with durations longer than the imposed interval
   of five minutes.  The default method for dealing with such Flows is
   to account them to the interval containing the Flow's start time.

   In this example, the same data is aggregated using the same
   arrangement of operations and the same output Template as in
   Section 8.1, but using a different counter distribution policy,
   Simple Uniform Distribution, as described in Section 5.1.1.  In order
   to do this, the Exporting Process first exports the Aggregate Counter
   Distribution Options Template, as in Figure 26.

   templateId(12)[2]{scope}
   valueDistributionMethod(384)[1]

        Figure 26: Aggregate Counter Distribution Options Template

   This Template is followed by an Aggregate Counter Distribution Record
   described by this Template; assuming the output Template in Figure 11
   has ID 257, this record would appear as in Figure 27.





Trammell, et al.             Standards Track                   [Page 44]

RFC 7015                    IPFIX Aggregation             September 2013


   template ID | value distribution method
           257   4 (simple uniform)

             Figure 27: Aggregate Counter Distribution Record

   Following metadata export, the aggregation steps follow as before.
   However, two long Flows are distributed across multiple intervals in
   the interval imposition step, as indicated with "*" in Figure 28.
   Note the uneven distribution of the three-interval, 11200-octet Flow
   into three Partially Aggregated Flows of 3733, 3733, and 3734 octets;
   this ensures no cumulative error is injected by the interval
   distribution step.

 start time |end time   |source ip4 |port |dest ip4      |port|pt|  oct
 9:00:00.000 9:05:00.000 192.0.2.2   47113 192.0.2.131    53   17   119
 9:00:00.000 9:05:00.000 192.0.2.2   22153 192.0.2.131    53   17    83
 9:00:00.000 9:05:00.000 192.0.2.2   52420 198.51.100.2   443  6   1637
 9:00:00.000 9:05:00.000 192.0.2.3   56047 192.0.2.131    53   17   111
 9:00:00.000 9:05:00.000 192.0.2.3   41183 198.51.100.67  80   6  16838
 9:00:00.000 9:05:00.000 192.0.2.2   17606 198.51.100.68  80   6  11538
 9:00:00.000 9:05:00.000 192.0.2.3   47113 192.0.2.131    53   17   119
 9:00:00.000 9:05:00.000 192.0.2.3   48458 198.51.100.133 80   6   2973
 9:00:00.000 9:05:00.000 192.0.2.4   61295 198.51.100.2   443  6   8350
 9:00:00.000 9:05:00.000 203.0.113.3 41256 198.51.100.133 80   6    778
 9:00:00.000 9:05:00.000 203.0.113.3 51662 198.51.100.3   80   6    883
 9:00:00.000 9:05:00.000 192.0.2.2   37581 198.51.100.2   80   6   7710*
 9:00:00.000 9:05:00.000 203.0.113.3 39586 198.51.100.17  80   6   3733*
 9:05:00.000 9:10:00.000 203.0.113.3 52572 198.51.100.2   443  6   1637
 9:05:00.000 9:10:00.000 203.0.113.3 49914 197.51.100.133 80   6    561
 9:05:00.000 9:10:00.000 192.0.2.2   50824 198.51.100.2   443  6   1899
 9:05:00.000 9:10:00.000 192.0.2.3   34597 198.51.100.3   80   6   1284
 9:05:00.000 9:10:00.000 203.0.113.3 58907 198.51.100.4   80   6   2670
 9:05:00.000 9:10:00.000 192.0.2.2   37581 198.51.100.2   80   6   7710*
 9:05:00.000 9:10:00.000 203.0.113.3 39586 198.51.100.17  80   6   3733*
 9:10:00.000 9:15:00.000 192.0.2.4   22478 192.0.2.131    53   17    75
 9:10:00.000 9:15:00.000 192.0.2.4   49513 198.51.100.68  80   6   3374
 9:10:00.000 9:15:00.000 192.0.2.4   64832 198.51.100.67  80   6    138
 9:10:00.000 9:15:00.000 192.0.2.3   60833 198.51.100.69  443  6   2325
 9:10:00.000 9:15:00.000 192.0.2.2   19638 198.51.100.3   80   6   2869
 9:10:00.000 9:15:00.000 192.0.2.3   40429 198.51.100.4   80   6  18289
 9:10:00.000 9:15:00.000 203.0.113.3 39586 198.51.100.17  80   6   3734*

  Figure 28: Distributed Interval Imposition for Time Series per Source

   Subsequent steps are as in Section 8.1; the results, to be exported
   using the Template shown in Figure 11, are shown in Figure 29, with
   Aggregated Flows differing from the example in Section 8.1 indicated
   by "*".



Trammell, et al.             Standards Track                   [Page 45]

RFC 7015                    IPFIX Aggregation             September 2013


   start time |end time   |source ip4 |octets
   9:00:00.000 9:05:00.000 192.0.2.2    21087*
   9:00:00.000 9:05:00.000 192.0.2.3    20041
   9:00:00.000 9:05:00.000 192.0.2.4     8350
   9:00:00.000 9:05:00.000 203.0.113.3   5394*
   9:05:00.000 9:10:00.000 192.0.2.2     9609*
   9:05:00.000 9:10:00.000 192.0.2.3     1284
   9:05:00.000 9:10:00.000 203.0.113.3   8601*
   9:10:00.000 9:15:00.000 192.0.2.2     2869
   9:10:00.000 9:15:00.000 192.0.2.3    20614
   9:10:00.000 9:15:00.000 192.0.2.4     3587
   9:10:00.000 9:15:00.000 203.0.113.3   3734*

          Figure 29: Aggregated Flows for Time Series per Source
                         with Counter Distribution

9.  Security Considerations



   This document specifies the operation of an Intermediate Aggregation
   Process with the IPFIX protocol; the Security Considerations for the
   protocol itself in Section 11 of [RFC7011] therefore apply.  In the
   common case that aggregation is performed on a Mediator, the Security
   Considerations for Mediators in Section 9 of [RFC6183] apply as well.

   As mentioned in Section 3, certain aggregation operations may tend to
   have an anonymizing effect on Flow data by obliterating sensitive
   identifiers.  Aggregation may also be combined with anonymization
   within a Mediator, or as part of a chain of Mediators, to further
   leverage this effect.  In any case in which an Intermediate
   Aggregation Process is applied as part of a data anonymization or
   protection scheme, or is used together with anonymization as
   described in [RFC6235], the Security Considerations in Section 9 of
   [RFC6235] apply.

10.  IANA Considerations



   This document specifies the creation of new IPFIX Information
   Elements in the IPFIX Information Element registry [IANA-IPFIX], as
   defined in Section 7 above.  IANA has assigned Information Element
   numbers to these Information Elements, and entered them into the
   registry.

11.  Acknowledgments



   Special thanks to Elisa Boschi for early work on the concepts laid
   out in this document.  Thanks to Lothar Braun, Christian Henke, and
   Rahul Patel for their reviews and valuable feedback, with special




Trammell, et al.             Standards Track                   [Page 46]

RFC 7015                    IPFIX Aggregation             September 2013


   thanks to Paul Aitken for his multiple detailed reviews.  This work
   is materially supported by the European Union Seventh Framework
   Programme under grant agreement 257315 (DEMONS).

12.  References



12.1.  Normative References



   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC5226]  Narten, T. and H. Alvestrand, "Guidelines for Writing an
              IANA Considerations Section in RFCs", BCP 26, RFC 5226,
              May 2008.

   [RFC7011]  Claise, B., Ed., Trammell, B., Ed., and P. Aitken,
              "Specification of the IP Flow Information Export (IPFIX)
              Protocol for the Exchange of Flow Information", STD 77,
              RFC 7011, September 2013.

12.2.  Informative References



   [RFC3917]  Quittek, J., Zseby, T., Claise, B., and S. Zander,
              "Requirements for IP Flow Information Export (IPFIX)", RFC
              3917, October 2004.

   [RFC5470]  Sadasivan, G., Brownlee, N., Claise, B., and J. Quittek,
              "Architecture for IP Flow Information Export", RFC 5470,
              March 2009.

   [RFC5472]  Zseby, T., Boschi, E., Brownlee, N., and B. Claise, "IP
              Flow Information Export (IPFIX) Applicability", RFC 5472,
              March 2009.

   [RFC5476]  Claise, B., Johnson, A., and J. Quittek, "Packet Sampling
              (PSAMP) Protocol Specifications", RFC 5476, March 2009.

   [RFC5655]  Trammell, B., Boschi, E., Mark, L., Zseby, T., and A.
              Wagner, "Specification of the IP Flow Information Export
              (IPFIX) File Format", RFC 5655, October 2009.

   [RFC5982]  Kobayashi, A. and B. Claise, "IP Flow Information Export
              (IPFIX) Mediation: Problem Statement", RFC 5982, August
              2010.

   [RFC6183]  Kobayashi, A., Claise, B., Muenz, G., and K. Ishibashi,
              "IP Flow Information Export (IPFIX) Mediation: Framework",
              RFC 6183, April 2011.



Trammell, et al.             Standards Track                   [Page 47]

RFC 7015                    IPFIX Aggregation             September 2013


   [RFC6235]  Boschi, E. and B. Trammell, "IP Flow Anonymization
              Support", RFC 6235, May 2011.

   [RFC6728]  Muenz, G., Claise, B., and P. Aitken, "Configuration Data
              Model for the IP Flow Information Export (IPFIX) and
              Packet Sampling (PSAMP) Protocols", RFC 6728, October
              2012.

   [RFC7012]  Claise, B., Ed. and B. Trammell, Ed., "Information Model
              for IP Flow Information Export (IPFIX)", RFC 7012,
              September 2013.

   [RFC7013]  Trammell, B. and B. Claise, "Guidelines for Authors and
              Reviewers of IP Flow Information Export (IPFIX)
              Information Elements", BCP 184, RFC 7013, September 2013.

   [RFC7014]  D'Antonio, S., Zseby, T., Henke, C., and L. Peluso, "Flow
              Selection Techniques", RFC 7014, September 2013.

   [IANA-IPFIX]
              IANA, "IP Flow Information Export (IPFIX) Entities",
              <http://www.iana.org/assignments/ipfix>.

   [IPFIX-MED-PROTO]
              Claise, B., Kobayashi, A., and B. Trammell, "Operation of
              the IP Flow Information Export (IPFIX) Protocol on IPFIX
              Mediators", Work in Progress, July 2013.
























Trammell, et al.             Standards Track                   [Page 48]

RFC 7015                    IPFIX Aggregation             September 2013


Authors' Addresses



   Brian Trammell
   Swiss Federal Institute of Technology Zurich
   Gloriastrasse 35
   8092 Zurich
   Switzerland

   Phone: +41 44 632 70 13
   EMail: trammell@tik.ee.ethz.ch


   Arno Wagner
   Consecom AG
   Bleicherweg 64a
   8002 Zurich
   Switzerland

   EMail: arno@wagner.name


   Benoit Claise
   Cisco Systems, Inc.
   De Kleetlaan 6a b1
   1831 Diegem
   Belgium

   Phone: +32 2 704 5622
   EMail: bclaise@cisco.com






















Trammell, et al.             Standards Track                   [Page 49]