RFC 6645

Internet Engineering Task Force (IETF)                          J. Novak
Request for Comments: 6645                           Cisco Systems, Inc.
Category: Informational                                        July 2012
ISSN: 2070-1721

                   IP Flow Information Accounting and
                    Export Benchmarking Methodology


   This document provides a methodology and framework for quantifying
   the performance impact of the monitoring of IP flows on a network
   device and the export of this information to a Collector.  It
   identifies the rate at which the IP flows are created, expired, and
   successfully exported as a new performance metric in combination with
   traditional throughput.  The metric is only applicable to the devices
   compliant with RFC 5470, "Architecture for IP Flow Information
   Export".  The methodology quantifies the impact of the IP flow
   monitoring process on the network equipment.

Status of This Memo

   This document is not an Internet Standards Track specification; it is
   published for informational purposes.

   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).  Not all documents
   approved by the IESG are a candidate for any level of Internet
   Standard; see 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

Novak                         Informational                     [Page 1]

RFC 6645              Flow Monitoring Benchmarking             July 2012

Copyright Notice

   Copyright (c) 2012 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.

Novak                         Informational                     [Page 2]

RFC 6645              Flow Monitoring Benchmarking             July 2012

Table of Contents

   1. Introduction ....................................................4
   2. Terminology .....................................................5
      2.1. Existing Terminology .......................................5
      2.2. New Terminology ............................................6
   3. Flow Monitoring Performance Benchmark ...........................8
      3.1. Definition .................................................8
      3.2. Device Applicability .......................................8
      3.3. Measurement Concept ........................................8
      3.4. The Measurement Procedure Overview .........................9
   4. Measurement Setup ..............................................11
      4.1. Measurement Topology ......................................11
      4.2. Baseline DUT Setup ........................................13
      4.3. Flow Monitoring Configuration .............................13
      4.4. Collector .................................................19
      4.5. Sampling ..................................................19
      4.6. Frame Formats .............................................19
      4.7. Frame Sizes ...............................................20
      4.8. Flow Export Data Packet Sizes .............................20
      4.9. Illustrative Test Setup Examples ..........................20
   5. Flow Monitoring Throughput Measurement Methodology .............22
      5.1. Flow Monitoring Configuration .............................23
      5.2. Traffic Configuration .....................................24
      5.3. Cache Population ..........................................25
      5.4. Measurement Time Interval .................................25
      5.5. Flow Export Rate Measurement ..............................26
      5.6. The Measurement Procedure .................................27
   6. RFC 2544 Measurements ..........................................28
      6.1. Flow Monitoring Configuration..............................28
      6.2. Measurements with the Flow Monitoring Throughput Setup ....29
      6.3. Measurements with Fixed Flow Export Rate...................29
   7. Flow Monitoring Accuracy .......................................30
   8. Evaluating Flow Monitoring Applicability .......................31
   9. Acknowledgements ...............................................32
   10. Security Considerations .......................................32
   11. References ....................................................33
      11.1. Normative References .....................................33
      11.2. Informative References ...................................33
   Appendix A. Recommended Report Format .............................35
   Appendix B. Miscellaneous Tests ...................................36
       B.1. DUT Under Traffic Load ...................................36
       B.2. In-Band Flow Export ......................................36
       B.3. Variable Packet Rate .....................................37
       B.4. Bursty Traffic ...........................................37
       B.5. Various Flow Monitoring Configurations ...................38
       B.6. Tests with Bidirectional Traffic .........................38
       B.7. Instantaneous Flow Export Rate ...........................39

Novak                         Informational                     [Page 3]

RFC 6645              Flow Monitoring Benchmarking             July 2012

1.  Introduction

   Monitoring IP flows (Flow monitoring) is defined in the "Architecture
   for IP Flow Information Export" [RFC5470] and related IPFIX documents
   specified in Section 1.2 of [RFC5470].  It analyzes the traffic using
   predefined fields from the packet header as keys and stores the
   traffic and other internal information in the DUT (Device Under Test)
   memory.  This cached flow information is then formatted into records
   (see Section 2.1 for term definitions) and exported from the DUT to
   an external data collector for analysis.  More details on the
   measurement architecture are provided in Section 3.3.

   Flow monitoring on network devices is widely deployed and has
   numerous uses in both service-provider and enterprise segments as
   detailed in the "Requirements for IP Flow Information Export (IPFIX)"
   [RFC3917].  This document provides a methodology for measuring Flow
   monitoring performance so that network operators have a framework to
   measure the impact on the network and network equipment.

   This document's goal is to provide a series of methodology
   specifications for the measurement of Flow monitoring performance in
   a way that is comparable amongst various implementations, platforms,
   and vendor devices.

   Flow monitoring is, in most cases, run on network devices that also
   forward packets.  Therefore, this document also provides the
   methodology for [RFC2544] measurements in the presence of Flow
   monitoring.  It is applicable to IPv6 and MPLS traffic with their
   specifics defined in [RFC5180] and [RFC5695], respectively.

   This document specifies a methodology to measure the maximum IP Flow
   Export Rate that a network device can sustain without impacting the
   Forwarding Plane, without losing any IP flow information and without
   compromising IP flow accuracy (see Section 7 for details).

   [RFC2544], [RFC5180], and [RFC5695] specify benchmarking of network
   devices forwarding IPv4, IPv6, and MPLS [RFC3031] traffic,
   respectively.  The methodology specified in this document stays the
   same for any traffic type.  The only restriction may be the DUT's
   lack of support for Flow monitoring of a particular traffic type.

   A variety of different DUT architectures exist that are capable of
   Flow monitoring and export.  As such, this document does not attempt
   to list the various white-box variables (e.g., CPU load, memory
   utilization, hardware resources utilization, etc.) that could be
   gathered as they always help in comparison evaluations.  A more
   complete understanding of the stress points of a particular device

Novak                         Informational                     [Page 4]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   can be attained using this internal information, and the tester MAY
   choose to gather this information during the measurement iterations.

2.  Terminology

   The terminology used in this document is based on that defined in
   [RFC5470], [RFC2285], and [RFC1242], as summarized in Section 2.1.
   The only new terms needed for this methodology are defined in Section

   Additionally, the key words "MUST", "MUST NOT", "REQUIRED", "SHALL",
   "OPTIONAL" in this document are to be interpreted as described  in
   RFC 2119 [RFC2119].

2.1.  Existing Terminology

    Device Under Test (DUT)   [RFC2285, Section 3.1.1]

    Flow                      [RFC5101, Section 2]

    Flow Key                  [RFC5101, Section 2]

    Flow Record               [RFC5101, Section 2]

    Template Record           [RFC5101, Section 2]

    Observation Point         [RFC5470, Section 2]

    Metering Process          [RFC5470, Section 2]

    Exporting Process         [RFC5470, Section 2]

    Exporter                  [RFC5470, Section 2]

    Collector                 [RFC5470, Section 2]

    Control Information       [RFC5470, Section 2]

    Data Stream               [RFC5470, Section 2]

    Flow Expiration           [RFC5470, Section 5.1.1]

    Flow Export               [RFC5470, Section 5.1.2]

    Throughput                [RFC1242, Section 3.17]

Novak                         Informational                     [Page 5]

RFC 6645              Flow Monitoring Benchmarking             July 2012

2.2.  New Terminology

2.2.1.  Cache

      Memory area held and dedicated by the DUT to store Flow
      information prior to the Flow Expiration.

2.2.2.  Cache Size

      The size of the Cache in terms of how many entries the Cache can

      This term is typically represented as a configurable option in the
      particular Flow monitoring implementation.  Its highest value will
      depend on the memory available in the network device.

   Measurement units:
      Number of Cache entries

2.2.3.  Active Timeout

      For long-running Flows, the time interval after which the Metering
      Process expires a Cache entry to ensure Flow data is regularly

      This term is typically presented as a configurable option in the
      particular Flow monitoring implementation.  See Section 5.1.1 of
      [RFC5470] for a more detailed discussion.

      Flows are considered long running when they last longer than
      several multiples of the Active Timeout.  If the Active Timeout is
      zero, then Flows are considered long running if they contain many
      more packets (tens of packets) than usually observed in a single

   Measurement units:

Novak                         Informational                     [Page 6]

RFC 6645              Flow Monitoring Benchmarking             July 2012

2.2.4.  Idle Timeout

      The time interval used by the Metering Process to expire an entry
      from the Cache when no more packets belonging to that specific
      Cache entry have been observed during the interval.

      Idle Timeout is typically represented as a configurable option in
      the particular Flow monitoring implementation.  See Section 5.1.1
      of [RFC5470] for more detailed discussion.  Note that some
      documents in the industry refer to "Idle Timeout" as "inactive

   Measurement units:

2.2.5.  Flow Export Rate

      The number of Cache entries that expire from the Cache (as defined
      by the Flow Expiration term) and are exported to the Collector
      within a measurement time interval.  There SHOULD NOT be any
      export filtering, so that all the expired Cache entries are
      exported.  If there is export filtering and it can't be disabled,
      this MUST be indicated in the measurement report.

      The measured Flow Export Rate MUST include both the Data Stream
      and the Control Information, as defined in Section 2 of [RFC5470].

      The Flow Export Rate is measured using Flow Export data observed
      at the Collector by counting the exported Flow Records during the
      measurement time interval (see Section 5.4).  The value obtained
      is an average of the instantaneous export rates observed during
      the measurement time interval.  The smallest possible measurement
      interval (if attempting to measure a nearly instantaneous export
      rate rather than average export rate on the DUT) is limited by the
      export capabilities of the particular Flow monitoring
      implementation (when physical-layer issues between the DUT and the
      Collector are excluded).

   Measurement units:
      Number of Flow Records per second

Novak                         Informational                     [Page 7]

RFC 6645              Flow Monitoring Benchmarking             July 2012

3.  Flow Monitoring Performance Benchmark

3.1.  Definition

   Flow Monitoring Throughput

      The maximum Flow Export Rate the DUT can sustain without losing a
      single Cache entry.  Additionally, for packet forwarding devices,
      the maximum Flow Export Rate the DUT can sustain without dropping
      packets in the Forwarding Plane (see Figure 1).

   Measurement units:
      Number of Flow Records per second

      The losses of Cache entries, or forwarded packets per this
      definition are assumed to happen due to the lack of DUT resources
      to process any additional traffic information or lack of resources
      to process Flow Export data.  The physical-layer issues, like
      insufficient bandwidth from the DUT to the Collector or lack of
      Collector resources, MUST be excluded as detailed in Section 4.

3.2.  Device Applicability

   The Flow monitoring performance metric is applicable to network
   devices that deploy the architecture described in [RFC5470].  These
   devices can be network packet forwarding devices or appliances that
   analyze traffic but do not forward traffic (e.g., probes, sniffers,

   This document does not intend to measure Collector performance, it
   only requires sufficient Collector resources (as specified in Section
   4.4) in order to measure the DUT characteristics.

3.3.  Measurement Concept

   Figure 1 presents the functional block diagram of the DUT.  The
   traffic in the figure represents test traffic sent to the DUT and
   forwarded by the DUT, if possible.  When testing devices that do not
   act as network packet forwarding devices (such as probes, sniffers,
   and replicators), the Forwarding Plane is simply an Observation Point
   as defined in Section 2 of [RFC5470].  The Throughput of such devices
   will always be zero, and the only applicable performance metric is
   the Flow Monitoring Throughput.  Netflow is specified by [RFC3954].

Novak                         Informational                     [Page 8]

RFC 6645              Flow Monitoring Benchmarking             July 2012

              +------------------------- +
              | IPFIX | NetFlow | Others |
              +------------------------- +
              |            ^             |
              |       Flow Export        |
              |            ^             |
              |     +-------------+      |
              |     | Monitoring  |      |
              |     |   Plane     |      |
              |     +-------------+      |
              |            ^             |
              |     traffic information  |
              |            ^             |
              |     +-------------+      |
              |     |             |      |
   traffic ---|---->| Forwarding  |------|---->
              |     |    Plane    |      |
              |     +-------------+      |
              |                          |
              |           DUT            |
              +------------------------- +

   Figure 1.  The Functional Block Diagram of the DUT

   Flow monitoring is represented in Figure 1 by the Monitoring Plane;
   it is enabled as specified in Section 4.3.  It uses the traffic
   information provided by the Forwarding Plane and configured Flow Keys
   to create Cache entries representing the traffic forwarded (or
   observed) by the DUT in the DUT Cache.  The Cache entries are expired
   from the Cache depending on the Cache configuration (e.g., the Active
   and Idle Timeouts, the Cache Size), number of Cache entries, and the
   traffic pattern.  The Cache entries are used by the Exporting Process
   to format the Flow Records, which are then exported from the DUT to
   the Collector (see Figure 2 in Section 4).

   The Forwarding Plane and Monitoring Plane represent two separate
   functional blocks, each with its own performance capability.  The
   Forwarding Plane handles user data packets and is fully characterized
   by the metrics defined by [RFC1242].

   The Monitoring Plane handles Flows that reflect the analyzed traffic.
   The metric for Monitoring Plane performance is the Flow Export Rate,
   and the benchmark is the Flow Monitoring Throughput.

3.4.  The Measurement Procedure Overview

   The measurement procedure is fully specified in Sections 4, 5, and 6.
   This section provides an overview of principles for the measurements.

Novak                         Informational                     [Page 9]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   The basic measurement procedure of the performance characteristics of
   a DUT with Flow monitoring enabled is a conventional Throughput
   measurement using a search algorithm to determine the maximum packet
   rate at which none of the offered packets and corresponding Flow
   Records are dropped by the DUT as described in [RFC1242] and Section
   26.1 of [RFC2544].

   The DUT with Flow monitoring enabled contains two functional blocks
   that need to be measured using characteristics applicable to one or
   both blocks (see Figure 1).  See Sections 3.4.1 and 3.4.2 for further

   On one hand, the Monitoring Plane and Forwarding Plane (see Figure 1)
   need to be looked at as two independent blocks, and the performance
   of each measured independently.  On the other hand, when measuring
   the performance of one, the status and performance of the other MUST
   be known and benchmarked when both are present.

3.4.1.  Monitoring Plane Performance Measurement

   The Flow Monitoring Throughput MUST be (and can only be) measured
   with one packet per Flow as specified in Section 5.  This traffic
   type represents the most demanding traffic from the Flow monitoring
   point of view and will exercise the Monitoring Plane (see Figure 1)
   of the DUT most.  In this scenario, every packet seen by the DUT
   creates a new Cache entry and forces the DUT to fill the Cache
   instead of just updating the packet and byte counters of an already
   existing Cache entry.

   The exit criteria for the Flow Monitoring Throughput measurement are
   one of the following (e.g., if any of the conditions are reached):

   a. The Flow Export Rate at which the DUT starts to lose Flow
      Information or the Flow Information gets corrupted.

   b. The Flow Export Rate at which the Forwarding Plane starts to drop
      or corrupt packets (if the Forwarding Plane is present).

   A corrupted packet here means packet header corruption (resulting in
   the cyclic redundancy check failure on the transmission level and
   consequent packet drop) or packet payload corruption, which leads to
   lost application-level data.

3.4.2.  Forwarding Plane Performance Measurement

   The Forwarding Plane (see Figure 1) performance metrics are fully
   specified by [RFC1242] and MUST be measured accordingly.  A detailed
   traffic analysis (see below) with relation to Flow monitoring MUST be

Novak                         Informational                    [Page 10]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   performed prior of any [RFC2544] measurements.  Most importantly, the
   Flow Export Rate caused by the test traffic during an [RFC2544]
   measurement MUST be known and reported.

   The required test traffic analysis mainly involves the following:

   a. Which packet header parameters are incremented or changed during
      traffic generation.

   b. Which Flow Keys the Flow monitoring configuration uses to generate
      Flow Records.

   The performance metrics described in RFC 1242 can be measured in one
   of the three modes:

   a. As a baseline of forwarding performance without Flow monitoring.

   b. At a certain level of Flow monitoring activity specified by a Flow
      Export Rate lower than the Flow Monitoring Throughput.

   c. At the maximum level of Flow monitoring performance, e.g., using
      traffic conditions representing a measurement of Flow Monitoring

   The above mentioned measurement mode in point a.  represents an
   ordinary Throughput measurement specified in RFC 2544.  The details
   of how to set up the measurements in points b. and c. are given in
   Section 6.

4.  Measurement Setup

   This section concentrates on the setup of all components necessary to
   perform Flow monitoring performance measurement.  The recommended
   reporting format can be found in Appendix A.

4.1.  Measurement Topology

   The measurement topology described in this section is applicable only
   to the measurements with packet forwarding network devices.  The
   possible architectures and implementation of the traffic monitoring
   appliances (see Section 3.2) are too various to be covered in this
   document.  Instead of the Forwarding Plane, these appliances
   generally have some kind of feed (e.g., an optical splitter, an
   interface sniffing traffic on a shared media, or an internal channel
   on the DUT providing a copy of the traffic) providing the information
   about the traffic necessary for Flow monitoring analysis.  The
   measurement topology then needs to be adjusted to the appliance
   architecture and MUST be part of the measurement report.

Novak                         Informational                    [Page 11]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   The measurement setup is identical to that used by [RFC2544], with
   the addition of a Collector to analyze the Flow Export (see Figure

   In the measurement topology with unidirectional traffic, the traffic
   is transmitted from the sender to the receiver through the DUT.  The
   received traffic is analyzed to check that it is identical to the
   generated traffic.

   The ideal way to implement the measurement is by using a single
   device to provide the sender and receiver capabilities with one
   sending port and one receiving port.  This allows for an easy check
   as to whether all the traffic sent by the sender was re-transmitted
   by the DUT and received at the receiver.

                       |           |
                       | Collector |
                       |           |
                       |Flow Record|
                       | analysis  |
                       |           |
                             | Flow Export
                             | Export Interface
   +--------+         +-------------+          +----------+
   |        |         |             |          | traffic  |
   | traffic|      (*)|             |          | receiver |
   | sender |-------->|     DUT     |--------->|          |
   |        |         |             |          | traffic  |
   |        |         |             |          | analysis |
   +--------+         +-------------+          +----------+

   Figure 2.  Measurement Topology with Unidirectional Traffic

   The DUT's export interface (connecting the Collector) MUST NOT be
   used for forwarding test traffic but only for the Flow Export data
   containing the Flow Records.  In all measurements, the export
   interface MUST have enough bandwidth to transmit Flow Export data
   without congestion.  In other words, the export interface MUST NOT be
   a bottleneck during the measurement.

   The traffic receiver MUST have sufficient resources to measure all
   test traffic transferred successfully by the DUT.  This may be
   checked through measurements with and without the DUT.

Novak                         Informational                    [Page 12]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   Note that more complex topologies might be required.  For example, if
   the effects of enabling Flow monitoring on several interfaces is of
   concern, or the maximum speed of media transmission is less than the
   DUT Throughput, the topology can be expanded with several input and
   output ports.  However, the topology MUST be clearly written in the
   measurement report.

4.2.  Baseline DUT Setup

   The baseline DUT setup and the way the setup is reported in the
   measurement results is fully specified in Section 7 of [RFC2544].

   The baseline DUT configuration might include other features, like
   packet filters or quality of service on the input and/or output
   interfaces, if there is the need to study Flow monitoring in the
   presence of those features.  The Flow monitoring measurement
   procedures do not change in this case.  Consideration needs to be
   made when evaluating measurement results to take into account the
   possible change of packet rates offered to the DUT and Flow
   monitoring after application of the features to the configuration.
   Any such feature configuration MUST be part of the measurement

   The DUT export interface (see Figure 2) SHOULD be configured with
   sufficient output buffers to avoid dropping the Flow Export data due
   to a simple lack of resources in the interface hardware.  The applied
   configuration MUST be part of the measurement report.

   The test designer has the freedom to run tests in multiple
   configurations.  It is therefore possible to run both non-production
   and real deployment configurations in the laboratory, according to
   the needs of the tester.  All configurations MUST be part of the
   measurement report.

4.3.  Flow Monitoring Configuration

   This section covers all of the aspects of the Flow monitoring
   configuration necessary on the DUT in order to perform the Flow
   monitoring performance measurement.  The necessary configuration has
   a number of components (see [RFC5470]), namely Observation Points,
   Metering Process, and Exporting Process as detailed below.

   The DUT MUST support the Flow monitoring architecture as specified by
   [RFC5470].  The DUT SHOULD support IPFIX [RFC5101] to allow a
   meaningful results comparison due to the standardized export

Novak                         Informational                    [Page 13]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   The DUT configuration, any existing Cache, and Cache entries MUST be
   erased before the application of any new configuration for the
   currently executed measurement.

4.3.1.  Observation Points

   The Observation Points specify the interfaces and direction in which
   the Flow monitoring traffic analysis is to be performed.

   The (*) in Figure 2 designates the Observation Points in the default
   configuration.  Other DUT Observation Points might be configured
   depending on the specific measurement needs as follows:

   a. ingress port/ports only
   b. egress port/ports only
   c. both ingress and egress

   This test topology corresponds to unidirectional traffic only with
   traffic analysis performed on the input and/or output interface.
   Testing with bidirectional traffic is discussed in Appendix B.

   Generally, the placement of Observation Points depends upon the
   position of the DUT in the deployed network and the purpose of Flow
   monitoring.  See [RFC3917] for detailed discussion.  The measurement
   procedures are otherwise the same for all these possible

   In the case of both ingress and egress Flow monitoring being enabled
   on one DUT, the resulting analysis should consider that each Flow
   will be represented in the DUT Cache by two Flow Records (one for
   each direction).  Therefore, the Flow Export will also contain those
   two Flow Records.

   If more than one Observation Point for one direction is defined on
   the DUT, the traffic passing through each of the Observation Points
   MUST be configured in such a way that it creates Flows and Flow
   Records that do not overlap.  Each packet (or set of packets if
   measuring more than one packet per Flow - see Section 6.3.1) sent to
   the DUT on different ports still creates one unique Flow Record.

   The specific Observation Points and associated monitoring direction
   MUST be included as part of the measurement report.

4.3.2.  Metering Process

   The Metering Process MUST be enabled in order to create the Cache in
   the DUT and configure the Cache related parameters.

Novak                         Informational                    [Page 14]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   The Cache Size available to the DUT MUST be known and taken into
   account when designing the measurement as specified in Section 5.
   Typically, the Cache Size will be present in the "show" commands of
   the Flow monitoring process, in either the actual configuration or
   the product documentation from the DUT vendor.  The Cache Size MUST
   have a fixed value for the entire duration of the measurement.  This
   method is not applicable to benchmarking any Flow monitoring
   applications that dynamically change their Cache Size.

   The configuration of the Metering Process MUST be included as part of
   the measurement report.  For example, when a Flow monitoring
   implementation uses timeouts to expire entries from the Cache, the
   Cache's Idle and Active Timeouts MUST be known and taken into account
   when designing the measurement as specified in Section 5.  If the
   Flow monitoring implementation allows only timeouts equal to zero
   (e.g., immediate timeout or non-existent Cache), then the measurement
   conditions in Section 5 are fulfilled inherently without any
   additional configuration.  The DUT simply exports information about
   every packet immediately, subject to the Flow Export Rate definition
   in Section 2.2.5.

   If the Flow monitoring implementation allows configuration of
   multiple Metering Processes on a single DUT, the exact configuration
   of each process MUST be included in the measurement report.  Only
   measurements with the same number of Metering Processes can be

   The Cache Size and the Idle and Active Timeouts MUST be included in
   the measurement report.

4.3.3.  Exporting Process

   The Exporting Process MUST be configured in order to export the Flow
   Record data to the Collector.

   The Exporting Process MUST be configured in such a way that all Flow
   Records from all configured Observation Points are exported towards
   the Collector, after the expiration policy, which is composed of the
   Idle and Active Timeouts and Cache Size.

   The Exporting Process SHOULD be configured with IPFIX [RFC5101] as
   the protocol used to format the Flow Export data.  If the Flow
   monitoring implementation does not support IPFIX, proprietary
   protocols MAY be used.  Only measurements with the same export
   protocol SHOULD be compared since the protocols may differ in their
   export efficiency.  The export efficiency might also be influenced by
   the Template Record used and the ordering of the individual export
   fields within the template.

Novak                         Informational                    [Page 15]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   The Template Records used by the tested implementations SHOULD be
   analyzed and documented as part of the measurement report.  Ideally,
   only tests with same Template Records should be compared.

   Various Flow monitoring implementations might use different default
   values regarding the export of Control Information [RFC5470];
   therefore, the Flow Export corresponding to Control Information
   SHOULD be analyzed and reported as a separate item on the measurement
   report.  The export of Control Information SHOULD always be
   configured consistently across all testing and configured to the
   minimal possible value.  Ideally, just one set of Control Information
   should be exported during each measurement.  Note that Control
   Information includes options and Template Records [RFC5470].

   Section 10 of [RFC5101] and Section 8.1 of [RFC5470] discuss the
   possibility of deploying various transport-layer protocols to deliver
   Flow Export data from the DUT to the Collector.  The selected
   protocol MUST be included in the measurement report.  Only benchmarks
   with the same transport-layer protocol SHOULD be compared.  If the
   Flow monitoring implementation allows the use of multiple transport-
   layer protocols, each of the protocols SHOULD be measured in a
   separate measurement run and the results reported independently in
   the measurement report.

   If a reliable transport protocol is used for the transmission of the
   Flow Export data from the DUT, the configuration of the Transport
   session MUST allow for non-blocking data transmission.  An example of
   parameters to look at would be the TCP window size and maximum
   segment size (MSS).  The most substantial transport-layer parameters
   should be included in the measurement report.

4.3.4.  Flow Records

   A Flow Record contains information about a specific Flow observed at
   an Observation Point.  A Flow Record contains measured properties of
   the Flow (e.g., the total number of bytes for all the Flow packets)
   and usually characteristic properties of the Flow (e.g., source IP

   The Flow Record definition is implementation specific.  A Flow
   monitoring implementation might allow for only a fixed Flow Record
   definition, based on the most common IP parameters in the IPv4 or
   IPv6 headers -- for example, source and destination IP addresses, IP
   protocol numbers, or transport-level port numbers.  Another
   implementation might allow the user to define their own arbitrary
   Flow Record to monitor the traffic.  The only requirement for the
   measurements defined in this document is the need for a large

Novak                         Informational                    [Page 16]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   number of Cache entries in the Cache.  The Flow Keys needed to
   achieve that will typically be source and destination IP addresses
   and transport-level port numbers.

   The recommended full IPv4, IPv6, or MPLS Flow Record is shown below.
   The IP address indicates either IPv4 or IPv6, depending on the
   traffic type being tested.  The Flow Record configuration is Flow
   monitoring implementation-specific; therefore, the examples below
   cannot provide an exact specification of individual entries in each
   Flow Record.  The best set of key fields to use is left to the test
   designer using the capabilities of the specific Flow monitoring

      Flow Keys:
         Source IP address
         Destination IP address
         MPLS label (for MPLS traffic type only)
         Transport-layer source port
         Transport-layer destination port
         IP protocol number (IPv6 next header)
         IP type of service (IPv6 traffic class)

      Other fields:
         Packet counter
         Byte counter

      Table 1: Recommended Configuration

   If the Flow monitoring allows for user-defined Flow Records, the
   minimal Flow Record configurations allowing large numbers of Cache
   entries are, for example:

      Flow Keys:
         Source IP address
         Destination IP address

      Other fields:
         Packet counter

      Flow Keys:
         Transport-layer source port
         Transport-layer destination port

      Other fields:
         Packet counter

      Table 2: User-Defined Configuration

Novak                         Informational                    [Page 17]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   The Flow Record configuration MUST be clearly noted in the
   measurement report.  The Flow Monitoring Throughput measurements on
   different DUTs, or different Flow monitoring implementations, MUST be
   only compared for exactly the same Flow Record configuration.

4.3.5.  Flow Monitoring with Multiple Configurations

   The Flow monitoring architecture as specified in [RFC5470] allows for
   more complicated configurations with multiple Metering and Exporting
   Processes on a single DUT.  Depending on the particular Flow
   monitoring implementation, it might affect the measured DUT
   performance.  Therefore, the measurement report should contain
   information about how many Metering and Exporting Processes were
   configured on the DUT for the selected Observation Points.

   The examples of such possible configurations are:

   a. Several Observation Points with a single Metering Process and a
      single Exporting Process.

   b. Several Observation Points, each with one Metering Process but all
      using just one instance of Exporting Process.

   c. Several Observation Points with per-Observation-Point Metering
      Process and Exporting Process.

4.3.6.  MPLS Measurement Specifics

   The Flow Record configuration for measurements with MPLS encapsulated
   traffic SHOULD contain the MPLS label.  For this document's purposes,
   "MPLS Label" is the entire 4 byte MPLS header.  Typically, the label
   of the interest will be at the top of the label stack, but this
   depends on the details of the MPLS test setup.

   The tester SHOULD ensure that the data received by the Collector
   contains the expected MPLS labels.

   The MPLS forwarding performance document [RFC5695] specifies a number
   of possible MPLS label operations to test.  The Observation Points
   MUST be placed on all the DUT test interfaces where the particular
   MPLS label operation takes place.  The performance measurements
   SHOULD be performed with only one MPLS label operation at the time.

   The DUT MUST be configured in such a way that all the traffic is
   subject to the measured MPLS label operation.

Novak                         Informational                    [Page 18]

RFC 6645              Flow Monitoring Benchmarking             July 2012

4.4.  Collector

   The Collector is needed in order to capture the Flow Export data,
   which allows the Flow Monitoring Throughput to be measured.

   The Collector can be used exclusively as a capture device, providing
   just hexadecimal format of the Flow Export data.  In such a case, it
   does not need to have any additional Flow Export decoding
   capabilities and all the decoding is done offline.

   However, if the Collector is also used to decode the Flow Export
   data, it SHOULD support IPFIX [RFC5101] for meaningful results
   analysis.  If proprietary Flow Export is deployed, the Collector MUST
   support it; otherwise, the Flow Export data analysis is not possible.

   The Collector MUST be capable of capturing the export packets sent
   from the DUT at the full rate without losing any of them.  When using
   reliable transport protocols (see also Section 4.3.3) to transmit
   Flow Export data, the Collector MUST have sufficient resources to
   guarantee non-blocking data transmission on the transport-layer

   During the analysis, the Flow Export data needs to be decoded and the
   received Flow Records counted.

   The capture buffer MUST be cleared at the beginning of each

4.5.  Sampling

   Packet sampling and flow sampling is out of the scope of this
   document.  This document applies to situations without packet, flow,
   or export sampling.

4.6.  Frame Formats

   Flow monitoring itself is not dependent in any way on the media used
   on the input and output ports.  Any media can be used as supported by
   the DUT and the test equipment.  This applies both to data forwarding
   interfaces and to the export interface (see Figure 2).

   At the time of this writing, the most common transmission media and
   corresponding frame formats (e.g., Ethernet, Packet over SONET) for
   IPv4, IPv6, and MPLS traffic are specified within [RFC2544],
   [RFC5180], and [RFC5695].

   The presented frame formats MUST be recorded in the measurement

Novak                         Informational                    [Page 19]

RFC 6645              Flow Monitoring Benchmarking             July 2012

4.7.  Frame Sizes

   Frame sizes of the traffic to be analyzed by the DUT are specified in
   Section 9 of [RFC2544] for Ethernet type interfaces (64, 128, 256,
   1024, 1280, 1518 bytes) and in Section 5 of [RFC5180] for Packet over
   SONET interfaces (47, 64, 128, 256, 1024, 1280, 1518, 2048, 4096

   When measuring with large frame sizes, care needs to be taken to
   avoid any packet fragmentation on the DUT interfaces that could
   negatively affect measured performance values.

   The presented frame sizes MUST be recorded in the measurement report.

4.8.  Flow Export Data Packet Sizes

   The Flow monitoring performance will be affected by the packet size
   that the particular implementation uses to transmit Flow Export data
   to the Collector.  The used packet size MUST be part of the
   measurement report and only measurements with same packet sizes
   SHOULD be compared.

   The DUT export interface (see Figure 2) maximum transmission unit
   (MTU) SHOULD be configured to the largest available value for the
   media.  The Flow Export MTU MUST be recorded in the measurement

4.9.  Illustrative Test Setup Examples

   The examples below represent a hypothetical test setup to clarify the
   use of Flow monitoring parameters and configuration, together with
   traffic parameters to test Flow monitoring.  The actual benchmarking
   specifications are in Sections 5 and 6.

4.9.1.  Example 1 - Idle Timeout Flow Expiration

   The traffic generator sends 1000 packets per second in 10000 defined
   streams, each stream identified by a unique destination IP address.
   Therefore, each stream has a packet rate of 0.1 packets per second.

   The packets are sent in a round-robin fashion (stream 1 to 10000)
   while incrementing the destination IP address for each sent packet.
   After a packet for stream 10000 is sent, the next packet destination
   IP address corresponds to stream 1's address again.

   The configured Cache Size is 20000 Flow Records.  The configured
   Active Timeout is 100 seconds, and the Idle Timeout is 5 seconds.

Novak                         Informational                    [Page 20]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   Flow monitoring on the DUT uses the destination IP address as the
   Flow Key.

   A packet with the destination IP address equal to A is sent every 10
   seconds, so the Cache entry is refreshed in the Cache every 10
   seconds.  However, the Idle Timeout is 5 seconds, so the Cache
   entries will expire from the Cache due to the Idle Timeout, and when
   a new packet is sent with the same IP address A, it will create a new
   entry in the Cache.  This behavior depends upon the design and
   efficiency of the Cache ager, and incidences of multi-packet flows
   observed during this test should be noted.

   The measured Flow Export Rate in this case will be 1000 Flow Records
   per second since every single sent packet will always create a new
   Cache entry and 1000 packets per second are sent.

   The expected number of Cache entries in the Cache during the whole
   measurement is around 5000.  It corresponds to the Idle Timeout being
   5 seconds; during those five seconds, 5000 entries are created.  This
   expectation might change in real measurement setups with large Cache
   Sizes and a high packet rate where the DUT's actual export rate might
   be limited and lower than the Flow Expiration activity caused by the
   traffic offered to the DUT.  This behavior is entirely

4.9.2.  Example 2 - Active Timeout Flow Expiration

   The traffic generator sends 1000 packets per second in 100 defined
   streams, each stream identified by a unique destination IP address.
   Each stream has a packet rate of 10 packets per second.  The packets
   are sent in a round-robin fashion (stream 1 to 100) while
   incrementing the destination IP address for each sent packet.  After
   a packet for stream 100 is sent, the next packet destination IP
   address corresponds to stream 1's address again.

   The configured Cache Size is 1000 Flow Records.  The configured
   Active Timeout is 100 seconds.  The Idle Timeout is 10 seconds.

   Flow monitoring on the DUT uses the destination IP address as the
   Flow Key.

   After the first 100 packets are sent, 100 Cache entries will have
   been created in the Flow monitoring Cache.  The subsequent packets
   will be counted against the already created Cache entries since the
   destination IP address (Flow Key) has already been seen by the DUT
   (provided the Cache entries did not expire yet as described below).

Novak                         Informational                    [Page 21]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   A packet with the destination IP address equal to A is sent every 0.1
   second, so the Cache entry is refreshed in the Cache every 0.1
   second, while the Idle Timeout is 10 seconds.  In this case, the
   Cache entries will not expire until the Active Timeout expires, e.g.,
   they will expire every 100 seconds and then the Cache entries will be
   created again.

   If the test measurement time is 50 seconds from the start of the
   traffic generator, then the measured Flow Export Rate is 0 since
   during this period nothing expired from the Cache.

   If the test measurement time is 100 seconds from the start of the
   traffic generator, then the measured Flow Export Rate is 1 Flow
   Record per second.

   If the test measurement time is 290 seconds from the start of the
   traffic generator, then the measured Flow Export Rate is 2/3 of a
   Flow Record per second since the Cache expired the same number of
   Flows twice (100) during the 290-seconds period.

5.  Flow Monitoring Throughput Measurement Methodology


      To measure the Flow monitoring performance in a manner that is
      comparable between different Flow monitoring implementations.

   Metric definition:

      Flow Monitoring Throughput - see Section 3.


      Different Flow monitoring implementations might choose to handle
      Flow Export from a partially empty Cache differently than in the
      case of the Cache being fully occupied.  Similarly, software- and
      hardware-based DUTs can handle the same situation as stated above
      differently.  The purpose of the benchmark measurement in this
      section is to define one measurement procedure covering all the
      possible behaviors.

      The only criteria is to measure as defined here until Flow Record
      or packet losses are seen.  The decision whether to dive deeper
      into the conditions under which the packet losses happen is left
      to the tester.

Novak                         Informational                    [Page 22]

RFC 6645              Flow Monitoring Benchmarking             July 2012

5.1.  Flow Monitoring Configuration

   Cache Size
      Cache Size configuration is dictated by the expected position of
      the DUT in the network and by the chosen Flow Keys of the Flow
      Record.  The number of unique sets of Flow Keys that the traffic
      generator (sender) provides should be multiple times larger than
      the Cache Size.  This ensures that the existing Cache entries are
      never updated by a packet from the sender before the particular
      Flow Expiration and Flow Export.  This condition is simple to
      fulfill with linearly incremented Flow Keys (for example, IP
      addresses or transport-layer ports) where the range of values must
      be larger than the Cache Size.  When randomized traffic generation
      is in use, the generator must ensure that the same Flow Keys are
      not repeated within a range of randomly generated values.

      The Cache Size MUST be known in order to define the measurement
      circumstances properly.  Typically, the Cache Size will be found
      using the "show" commands of the Flow monitoring implementation in
      the actual configuration or in the product documentation from the

   Idle Timeout
      Idle Timeout is set (if configurable) to the minimum possible
      value on the DUT.  This ensures that the Cache entries are expired
      as soon as possible and exported out of the DUT Cache.  It MUST be
      known in order to define the measurement circumstances completely
      and equally across implementations.

   Active Timeout
      Active Timeout is set (if configurable) to a value equal to or
      higher than the Idle Timeout.  It MUST be known in order to define
      the measurement circumstances completely and equally across

   Flow Keys Definition:
      The test needs large numbers of unique Cache entries to be created
      by incrementing values of one or several Flow Keys.  The number of
      unique combinations of Flow Keys values SHOULD be several times
      larger than the DUT Cache Size.  This makes sure that any incoming
      packet will never refresh any already existing Cache entry.

   The availability of Cache Size, Idle Timeout, and Active Timeout as
   configuration parameters is implementation-specific.  If the Flow
   monitoring implementation does not support these parameters, the test
   possibilities, as specified by this document, are restricted.  Some

Novak                         Informational                    [Page 23]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   testing might be viable if the implementation follows the guidance
   provided in the [IPFIX-CONFIG] document and is considered on a case-
   by-case basis.

5.2.  Traffic Configuration

   Traffic Generation
      The traffic generator needs to increment the Flow Keys values with
      each sent packet.  This way, each packet represents one Cache
      entry in the DUT Cache.

      A particular Flow monitoring implementation might choose to deploy
      a hashing mechanism to match incoming data packets to a certain
      Flow.  In such a case, the combination of how the traffic is
      constructed and the hashing might influence the DUT Flow
      monitoring performance.  For example, if IP addresses are used as
      Flow Keys, this means there could be a performance difference for
      linearly incremented addresses (in ascending or descending order)
      as opposed to IP addresses randomized in a certain range.  If
      randomized IP address sequences are used, then the traffic
      generator needs to be able to reproduce the randomization (e.g.,
      the same set of IP addresses sent in the same order in different
      test runs) in order to compare various DUTs and Flow monitoring

      If the test traffic rate is below the maximum media rate for the
      particular packet size, the traffic generator MUST send the
      packets in equidistant time intervals.  Traffic generators that do
      not fulfill this condition MUST NOT and cannot be used for the
      Flow Monitoring Throughput measurement.  An example of this
      behavior is if the test traffic rate is one half of the media
      rate.  The traffic generator achieves this rate by sending packets
      each half of each second at the full media rate and sending
      nothing for the second half of each second.  In such conditions,
      it would be impossible to distinguish if the DUT failed to handle
      the Flows due to the shortage of input buffers during the burst or
      due to the limits in the Flow monitoring performance.

   Measurement Duration
      The measurement duration (e.g., how long the test traffic is sent
      to the DUT) MUST be at least two-times longer than the Idle
      Timeout; otherwise, no Flow Export would be seen.  The measurement
      duration SHOULD guarantee that the number of Cache entries created
      during the measurement exceeds the available Cache Size.

Novak                         Informational                    [Page 24]

RFC 6645              Flow Monitoring Benchmarking             July 2012

5.3.  Cache Population

   The product of the Idle Timeout and the packet rate offered to the
   DUT (Cache population) during one measurement determines the total
   number of Cache entries in the DUT Cache during the measurement
   (while taking into account some margin for dynamic behavior during
   high DUT loads when processing the Flows).

   The Flow monitoring implementation might behave differently depending
   on the relation of the Cache population to the available Cache Size
   during the measurement.  This behavior is fully implementation-
   specific and will also be influenced if the DUT architecture is
   software based or hardware based.

   The Cache population (if it is lower or higher than the available
   Cache Size) during a particular benchmark measurement SHOULD be
   noted, and mainly only measurements with the same Cache population
   SHOULD be compared.

5.4.  Measurement Time Interval

   The measurement time interval is the time value that is used to
   calculate the measured Flow Export Rate from the captured Flow Export
   data.  It is obtained as specified below.

   RFC 2544 specifies, with the precision of the packet beginning and
   ending, the time intervals to be used to measure the DUT time
   characteristics.  In the case of a Flow Monitoring Throughput
   measurement, the start and stop time needs to be clearly defined, but
   the granularity of this definition can be limited to just marking the
   start and stop time with the start and stop of the traffic generator.
   This assumes that the traffic generator and DUT are collocated and
   the variance in transmission delay from the generator to the DUT is
   negligible as compared to the total time of traffic generation.

   The measurement start time:
      the time when the traffic generator is started

   The measurement stop time: the time when the traffic generator is

   The measurement time interval is then calculated as the difference
   (stop time) - (start time) - (Idle Timeout).

   This supposes that the Cache Size is large enough that the time
   needed to fill it with Cache entries is longer than the Idle Timeout.
   Otherwise, the time needed to fill the Cache needs to be used to
   calculate the measurement time interval in place of the Idle Timeout.

Novak                         Informational                    [Page 25]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   Instead of measuring the absolute values of the stop and start times,
   it is possible to set up the traffic generator to send traffic for a
   certain predefined time interval, which is then used in the above
   definition instead of the difference (stop time) - (start time).

   The Collector MUST stop collecting the Flow Export data at the
   measurement stop time.

   The Idle Timeout (or the time needed to fill the Cache) causes delay
   of the Flow Export data behind the test traffic that is analyzed by
   the DUT.  For example, if the traffic starts at time point X, Flow
   Export will start only at the time point X + Idle Timeout (or X +
   time to fill the Cache).  Since Flow Export capture needs to stop
   with the traffic (because that's when the DUT stops processing the
   Flows at the given rate), the time interval during which the DUT kept
   exporting data is shorter by the Idle Timeout than the time interval
   when the test traffic was sent from the traffic generator to the DUT.

5.5.  Flow Export Rate Measurement

   The Flow Export Rate needs to be measured in two consequent steps.
   The purpose of the first step (point a. below) is to gain the actual
   value for the rate; the second step (point b. below) needs to be done
   in order to verify that no Flow Record are dropped during the

   a. In the first step, the captured Flow Export data MUST be analyzed
      only for the capturing interval (measurement time interval) as
      specified in Section 5.4.  During this period, the DUT is forced
      to process Cache entries at the rate the packets are sent.  When
      traffic generation finishes, the behavior when emptying the Cache
      is completely implementation-specific; therefore, the Flow Export
      data from this period cannot be used for benchmarking.

   b. In the second step, all the Flow Export data from the DUT MUST be
      captured in order to determine the Flow Record losses.  It needs
      to be taken into account that especially when large Cache Sizes
      (in order of magnitude of hundreds of thousands of entries and
      higher) are in use, the Flow Export can take many multiples of
      Idle Timeout to empty the Cache after the measurement.  This
      behavior is completely implementation-specific.

   If the Collector has the capability to redirect the Flow Export data
   after the measurement time interval into a different capture buffer
   (or time stamp the received Flow Export data after that), this can be
   done in one step.  Otherwise, each Flow Monitoring Throughput
   measurement at a certain packet rate needs to be executed twice --
   once to capture the Flow Export data just for the measurement time

Novak                         Informational                    [Page 26]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   interval (to determine the actual Flow Export Rate) and a second time
   to capture all Flow Export data in order to determine Flow Record
   losses at that packet rate.

   At the end of the measurement time interval, the DUT might still be
   processing Cache entries that belong to the Flows expired from the
   Cache before the end of the interval.  These Flow Records might
   appear in an export packet sent only after the end of the measurement
   interval.  This imprecision can be mitigated by use of large amounts
   of Flow Records during the measurement (so that the few Flow Records
   in one export packet can be ignored) or by use of timestamps exported
   with the Flow Records.

5.6.  The Measurement Procedure

   The measurement procedure is the same as the Throughput measurement
   in Section 26.1 of [RFC2544] for the traffic sending side.  The DUT
   output analysis is done on the traffic generator receiving side for
   the test traffic, the same way as for RFC 2544 measurements.

   An additional analysis is performed using data captured by the
   Collector.  The purpose of this analysis is to establish the value of
   the Flow Export Rate during the current measurement step and to
   verify that no Flow Records were dropped during the measurement.  The
   procedure for measuring the Flow Export Rate is described in Section

   The Flow Export performance can be significantly affected by the way
   the Flow monitoring implementation formats the Flow Records into the
   Flow Export packets.  The ordering and frequency in which Control
   Information is exported and the number of Flow Records in one Flow
   Export packet are of interest.  In the worst case scenario, there is
   just one Flow Record in every Flow Export packet.

   Flow Export data should be sanity checked during the benchmark
   measurement for:

   a. the number of Flow Records per packet, by simply calculating the
      ratio of exported Flow Records to the number of Flow Export
      packets captured during the measurement (which should be available
      as a counter on the Collector capture buffer).

   b. the number of Flow Records corresponding to the export of Control
      Information per Flow Export packet (calculated as the ratio of the
      total number of such Flow Records in the Flow Export data and the
      number of Flow Export packets).

Novak                         Informational                    [Page 27]

RFC 6645              Flow Monitoring Benchmarking             July 2012

6.  RFC 2544 Measurements

   RFC 2544 measurements can be performed under two Flow monitoring
   setups (see also Section 3.4.2).  This section details both and
   specifies ways to construct the test traffic so that RFC 2544
   measurements can be performed in a controlled environment from the
   Flow monitoring point of view.  A controlled Flow monitoring
   environment means that the tester always knows what Flow monitoring
   activity (Flow Export Rate) the traffic offered to the DUT causes.

   This section is applicable mainly for the Throughput (RFC 2544,
   Section 26.1) and latency (RFC 2544, Section 26.2 ) measurements.  It
   could also be used to measure frame loss rate (RFC 2544, Section
   26.3) and back-to-back frames (RFC 2544, Section 26.4).  Flow Export
   requires DUT resources to be generated and transmitted; therefore,
   the Throughput in most cases will be much lower when Flow monitoring
   is enabled on the DUT than when it is not.


      Provide RFC 2544 network device characteristics in the presence of
      Flow monitoring on the DUT.  RFC 2544 studies numerous
      characteristics of network devices.  The DUT forwarding and time
      characteristics without Flow monitoring present on the DUT can
      vary significantly when Flow monitoring is deployed on the network

   Metric definition:

      Metric as specified in [RFC2544].

   The measured Throughput MUST NOT include the packet rate
   corresponding to the Flow Export data, because it is not user traffic
   forwarded by the DUT.  It is generated by the DUT as a result of
   enabling Flow monitoring and does not contribute to the test traffic
   that the DUT can handle.  Flow Export requires DUT resources to be
   generated and transmitted; therefore, the Throughput in most cases
   will be much lower when Flow monitoring is enabled on the DUT than
   when it is not.

6.1.  Flow Monitoring Configuration

   Flow monitoring configuration (as detailed in Section 4.3) needs to
   be applied the same way as discussed in Section 5 with the exception
   of the Active Timeout configuration.

   The Active Timeout SHOULD be configured to exceed several times the
   measurement time interval (see Section 5.4).  This ensures that if

Novak                         Informational                    [Page 28]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   measurements with two traffic components are performed (see Section
   6.3.2), there is no Flow monitoring activity related to the second
   traffic component.

   The Flow monitoring configuration does not change in any other way
   for the measurement performed in this section.  What changes and
   makes the difference is the traffic configurations as specified in
   the sections below.

6.2.  Measurements with the Flow Monitoring Throughput Setup

   To perform a measurement with Flow Monitoring Throughput setup, the
   major requirement is that the traffic and Flow monitoring be
   configured in such a way that each sent packet creates one entry in
   the DUT Cache.  This restricts the possible setups only to the
   measurement with two traffic components as specified in Section

6.3.  Measurements with a Fixed Flow Export Rate

   This section covers the measurements where the RFC 2544 metrics need
   to be measured with Flow monitoring enabled, but at a certain Flow
   Export Rate that is lower than the Flow Monitoring Throughput.

   The tester here has both options as specified in Sections 6.3.1 and

6.3.1.  Measurements with a Single Traffic Component

   Section 12 of [RFC2544] discusses the use of protocol source and
   destination addresses for defined measurements.  To perform all the
   RFC 2544 type measurements with Flow monitoring enabled, the defined
   Flow Keys SHOULD contain an IP source and destination address.  The
   RFC 2544 type measurements with Flow monitoring enabled then can be
   executed under these additional conditions:

   a. the test traffic is not limited to a single, unique pair of source
      and destination addresses.

   b. the traffic generator defines test traffic as follows: it allows
      for a parameter to send N (where N is an integer number starting
      at 1 and is incremented in small steps) packets with source IP
      address A and destination IP address B before changing both IP
      addresses to the next value.

   This test traffic definition allows execution of the Flow monitoring
   measurements with a fixed Flow Export Rate while measuring the DUT
   RFC 2544 characteristics.  This setup is the better option since it

Novak                         Informational                    [Page 29]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   best simulates the live network traffic scenario with Flows
   containing more than just one packet.

   The initial packet rate at N equal to 1 defines the Flow Export Rate
   for the whole measurement procedure.  Subsequent increases of N will
   not change the Flow Export Rate as the time and Cache characteristics
   of the test traffic stay the same.  This setup is suitable for
   measurements with Flow Export Rates below the Flow Monitoring

6.3.2 Measurements with Two Traffic Components

   The test traffic setup described in Section 6.3.1 might be difficult
   to achieve with commercial traffic generators or if the granularity
   of the traffic rates as defined by the initial packet rate at N equal
   to 1 are unsuitable for the required measurement.  An alternative
   mechanism is to define two traffic components in the test traffic:
   one to populate Flow monitoring Cache and the second to execute the
   RFC 2544 measurements.

   a. Flow monitoring test traffic component -- the exact traffic
      definition as specified in Section 5.2.

   b. RFC 2544 Test Traffic Component -- test traffic as specified by
      RFC 2544 MUST create just one entry in the DUT Cache.  In the
      particular setup discussed here, this would mean a traffic stream
      with just one pair of unique source and destination IP addresses
      (but could be avoided if Flow Keys were, for example, UDP/TCP
      source and destination ports and Flow Keys did not contain the

   The Flow monitoring traffic component will exercise the DUT in terms
   of Flow activity, while the second traffic component will measure the
   RFC 2544 characteristics.

   The measured Throughput is the sum of the packet rates of both
   traffic components.  The definition of other RFC 1242 metrics remains

7.  Flow Monitoring Accuracy

   The pure Flow Monitoring Throughput measurement described in Section
   5 provides the capability to verify the Flow monitoring accuracy in
   terms of the exported Flow Record data.  Since every Cache entry
   created in the Cache is populated by just one packet, the full set of
   captured data on the Collector can be parsed (e.g., providing the
   values of all Flow Keys and other Flow Record fields, not only the
   overall Flow Record count in the exported data), and each set of

Novak                         Informational                    [Page 30]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   parameters from each Flow Record can be checked against the
   parameters as configured on the traffic generator and set in packets
   sent to the DUT.  The exported Flow Record is considered accurate if:

   a. all the Flow Record fields are present in each exported Flow

   b. all the Flow Record fields' values match the value ranges set by
      the traffic generator (for example, an IP address falls within the
      range of the IP address increments on the traffic generator).

   c. all the possible Flow Record field values as defined at the
      traffic generator have been found in the captured export data on
      the Collector.  This check needs to be offset against detected
      packet losses at the DUT during the measurement.

   For a DUT with packet forwarding, the Flow monitoring accuracy also
   involves data checks on the received traffic, as already discussed in
   Section 4.

8.  Evaluating Flow Monitoring Applicability

   The measurement results, as discussed in this document and obtained
   for certain DUTs, allow for a preliminary analysis of a Flow
   monitoring deployment based on the traffic analysis data from the
   providers' network.  An example of such traffic analysis in the
   Internet is provided by [CAIDA]; the way it can be used is discussed
   below.  The data needed to estimate if a certain network device can
   manage the particular amount of live traffic with Flow monitoring
   enabled is:

   Average packet size:            350 bytes
   Number of packets per IP flow:  20

   Expected data rate on the network device: 1 Gbit/s

   The average number of Flows created per second in the network device
   is needed and is determined as follows:

                       Expected packet rate
   Flows per second =  --------------------
                       Packet per flow

   When using the above example values, the network device is required
   to process 18000 Flows per second.  By executing the benchmarking as
   specified in this document, a platform capable of this processing can
   be determined for the deployment in that particular part of the user

Novak                         Informational                    [Page 31]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   Keep in mind that the above is a very rough and averaged Flow
   activity estimate, which cannot account for traffic anomalies; for
   example, a large number of DNS request packets that are typically
   small packets coming from many different sources and represent mostly
   just one packet per Flow.

9.  Acknowledgements

   This work was performed thanks to the patience and support of Cisco
   Systems NetFlow development team, namely Paul Aitken, Paul Atkins,
   and Andrew Johnson.  Thanks to Benoit Claise for numerous detailed
   reviews and presentations of the document, and to Aamer Akhter for
   initiating this work.  A special acknowledgment to the entire BMWG
   working group, especially to the chair, Al Morton, for the support
   and work on this document and Paul Aitken for a very detailed
   technical review.

10.  Security Considerations

   Documents of this type do not directly affect the security of the
   Internet or corporate networks as long as benchmarking is not
   performed on devices or systems connected to operating networks.

   Benchmarking activities, as described in this memo, are limited to
   technology characterization using controlled stimuli in a laboratory
   environment, with dedicated address space and the constraints
   specified in sections above.

   The benchmarking network topology will be an independent test setup
   and MUST NOT be connected to devices that may forward the test
   traffic into a production network, or misroute traffic to the test
   management network.

   Further, benchmarking is performed on a "black-box" basis, relying
   solely on measurements observable external to the DUT.

   Special capabilities SHOULD NOT exist in the DUT specifically for
   benchmarking purposes.  Any implications for network security arising
   from the DUT SHOULD be identical in the lab and in production

Novak                         Informational                    [Page 32]

RFC 6645              Flow Monitoring Benchmarking             July 2012

11.  References

11.1.  Normative References

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

   [RFC2544]   Bradner, S. and J. McQuaid, "Benchmarking Methodology for
               Network Interconnect Devices", RFC 2544, March 1999.

11.2.  Informative References

   [RFC1242]   Bradner, S., "Benchmarking Terminology for Network
               Interconnection Devices", RFC 1242, July 1991.

   [RFC2285]   Mandeville, R., "Benchmarking Terminology for LAN
               Switching Devices", RFC 2285, February 1998.

   [RFC3031]   Rosen, E., Viswanathan, A., and R. Callon, "Multiprotocol
               Label Switching Architecture", RFC 3031, January 2001.

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

   [RFC3954]   Claise, B., Ed., "Cisco Systems NetFlow Services Export
               Version 9", RFC 3954, October 2004.

   [RFC5101]   Claise, B., Ed., "Specification of the IP Flow
               Information Export (IPFIX) Protocol for the Exchange of
               IP Traffic Flow Information", RFC 5101, January 2008.

   [RFC5180]   Popoviciu, C., Hamza, A., Van de Velde, G., and D.
               Dugatkin, "IPv6 Benchmarking Methodology for Network
               Interconnect Devices", RFC 5180, May 2008.

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

   [RFC5695]   Akhter, A., Asati, R., and C. Pignataro, "MPLS Forwarding
               Benchmarking Methodology for IP Flows", RFC 5695,
               November 2009.

   [CAIDA]     Claffy, K., "The nature of the beast: recent traffic
               measurements from an Internet backbone",

Novak                         Informational                    [Page 33]

RFC 6645              Flow Monitoring Benchmarking             July 2012

               Muenz, G., Muenchen, TU, Claise, B., and P. Aitken,
               "Configuration Data Model for IPFIX and PSAMP", Work in
               Progress, July 2011.

   [PSAMP-MIB] Dietz, T., Claise, B., and J. Quittek, "Definitions of
               Managed Objects for Packet Sampling", Work in Progress,
               October 2011.

   [IPFIX-MIB] Dietz, T., Kobayashi, A., Claise, B., and G. Muenz,
               "Definitions of Managed Objects for IP Flow Information
               Export", Work in Progress, March 2012.

Novak                         Informational                    [Page 34]

RFC 6645              Flow Monitoring Benchmarking             July 2012

Appendix A.  (Informative) Recommended Report Format

Parameter                           Units
----------------------------------- ------------------------------------
Test Case                           test case name (Sections 5 and 6)
Test Topology                       Figure 2, other
Traffic Type                        IPv4, IPv6, MPLS, other

Test Results
  Flow Monitoring Throughput        Flow Records per second or Not
  Flow Export Rate                  Flow Records per second or Not
  Control Information Export Rate   Flow Records per second
  Throughput                        packets per second
  (Other RFC 1242 Metrics)          (as appropriate)

General Parameters
  DUT Interface Type                Ethernet, POS, ATM, other
  DUT Interface Bandwidth           MegaBits per second

Traffic Specifications
  Number of Traffic Components      (see Sections 6.3.1 and 6.3.2)
  For each traffic component:
  Packet Size                       bytes
  Traffic Packet Rate               packets per second
  Traffic Bit Rate                  MegaBits per second
  Number of Packets Sent            number of entries
  Incremented Packet Header Fields  list of fields
  Number of Unique Header Values    number of entries
  Number of Packets per Flow        number of entries
  Traffic Generation                linearly incremented or

Flow monitoring Specifications
  Direction                         ingress, egress, both
  Observation Points                DUT interface names
  Cache Size                        number of entries
  Active Timeout                    seconds
  Idle Timeout                      seconds
  Flow Keys                         list of fields
  Flow Record Fields                total number of fields
  Number of Flows Created           number of entries
  Flow Export Transport Protocol    UDP, TCP, SCTP, other
  Flow Export Protocol              IPFIX, NetFlow, other
  Flow Export data packet size      bytes
  Flow Export MTU                   bytes

Novak                         Informational                    [Page 35]

RFC 6645              Flow Monitoring Benchmarking             July 2012

Parameter                           Units (continued)
----------------------------------- ------------------------------------
MPLS Specifications                 (for traffic type MPLS only)
  Tested Label Operation            imposition, swap, disposition

The format of the report as documented in this appendix is informative,
but the entries in the contents of it are required as specified in the
corresponding sections of this document.

Many of the configuration parameters required by the measurement report
can be retrieved from the [IPFIX-MIB] and [PSAMP-MIB] MIB modules, and
from the [IPFIX-CONFIG] YANG module or other general MIBs.  Therefore,
querying those modules from the DUT would be beneficial: first of all,
to help in populating the required entries of the measurement report,
and also to document all the other configuration parameters from the

Appendix B.  (Informative) Miscellaneous Tests

   This section lists tests that could be useful to asses a proper Flow
   monitoring operation under various operational or stress conditions.
   These tests are not deemed suitable for any benchmarking for various

B.1.  DUT Under Traffic Load

   The Flow Monitoring Throughput should be measured under different
   levels of static traffic load through the DUT.  This can be achieved
   only by using two traffic components as discussed in Section 6.3.2.
   One traffic component exercises the Flow Monitoring Plane.  The
   second traffic component loads only the Forwarding Plane without
   affecting Flow monitoring (i.e., it creates just a certain amount of
   permanent Cache entries).

   The variance in Flow Monitoring Throughput as a function of the
   traffic load should be noted for comparison purposes between two DUTs
   of similar architecture and capability.

B.2.  In-Band Flow Export

   The test topology in Section 4.1 mandates the use of a separate Flow
   Export interface to avoid the Flow Export data generated by the DUT
   to mix with the test traffic from the traffic generator.  This is
   necessary in order to create clear and reproducible test conditions
   for the benchmark measurement.

   The real network deployment of Flow monitoring might not allow for
   such a luxury -- for example, on a very geographically large network.

Novak                         Informational                    [Page 36]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   In such a case, the Flow Export will use an ordinary traffic
   forwarding interface, e.g., in-band Flow Export.

   The Flow monitoring operation should be verified with in-band Flow
   Export configuration while following these test steps:

   a. Perform the benchmark test as specified in Section 5.  One of the
      results will be how much bandwidth Flow Export used on the
      dedicated Flow Export interface.
   b. Change Flow Export configuration to use the test interface.
   c. Repeat the benchmark test while the receiver filters out the Flow
      Export data from analysis.

   The expected result is that the Throughput achieved in step a. is
   same as the Throughput achieved in step c. provided that the
   bandwidth of the output DUT interface is not the bottleneck (in other
   words, it must have enough capacity to forward both test and Flow
   Export traffic).

B.3.  Variable Packet Size

   The Flow monitoring measurements specified in this document would be
   interesting to repeat with variable packet sizes within one
   particular test (e.g., test traffic containing mixed packet sizes).
   The packet forwarding tests specified mainly in [RFC2544] do not
   recommend performing such tests.  Flow monitoring is not dependent on
   packet sizes, so such a test could be performed during the Flow
   Monitoring Throughput measurement, and verification of its value does
   not depend on the offered traffic packet sizes.  The tests must be
   carefully designed in order to avoid measurement errors due to the
   physical bandwidth limitations and changes of the base forwarding
   performance with packet size.

B.4.  Bursty Traffic

   RFC 2544, Section 21 discusses and defines the use of bursty traffic.
   It can be used for Flow monitoring testing to gauge some short-term
   overload DUT capabilities in terms of Flow monitoring.  The test
   benchmark here would not be the Flow Export Rate the DUT can sustain,
   but the absolute number of Flow Records the DUT can process without
   dropping any single Flow Record.  The traffic setup to be used for
   this test is as follows:

   a. each sent packet creates a new Cache entry.
   b. the packet rate is set to the maximum transmission speed of the
      DUT interface used for the test.

Novak                         Informational                    [Page 37]

RFC 6645              Flow Monitoring Benchmarking             July 2012

B.5.  Various Flow Monitoring Configurations

   This section translates the terminology used in the IPFIX documents
   ([RFC5470], [RFC5101], and others) into the terminology used in this
   document.  Section B.5.2 proposes another measurement that is
   impossible to verify in a black box test manner.

B.5.1.  Throughput without the Metering Process

   If the Metering Process is not defined on the DUT it means no Flow
   monitoring Cache exists and no Flow analysis occurs.  The performance
   measurement of the DUT in such a case is just pure [RFC2544]

B.5.2.  Throughput with the Metering Process

   If only the Metering Process is enabled, Flow analysis on the DUT is
   enabled and operational but no Flow Export happens.  The performance
   measurement of a DUT in such a configuration represents a useful test
   of the DUT's capabilities (this corresponds to the case when the
   network operator uses Flow monitoring, for example, for manual
   detection of denial-of-service attacks, and does not wish to use Flow

   The performance testing on this DUT can be performed as discussed in
   this document, but it is not possible to verify the operation and
   results without interrogating the DUT.

B.5.3.  Throughput with the Metering and Exporting Processes

   This test represents the performance testing as discussed in Section

B.6.  Tests With Bidirectional Traffic

   Bidirectional traffic is not part of the normative benchmarking tests
   based on discussion with and recommendation of the Benchmarking
   working group.  The experienced participants stated that this kind of
   traffic did not provide reproducible results.

   The test topology in Figure 2 can be expanded to verify Flow
   monitoring functionality with bidirectional traffic using the
   interfaces in full duplex mode, e.g., sending and receiving
   simultaneously on each of them.

   The same rules should be applied for Flow creation in the DUT Cache
   (as per Sections 4.1 and 4.3.1) -- traffic passing through each
   Observation Point should always create a new Cache entry in the

Novak                         Informational                    [Page 38]

RFC 6645              Flow Monitoring Benchmarking             July 2012

   Cache, e.g., the same traffic should not be just looped back on the
   receiving interfaces to create the bidirectional traffic flow.

B.7.  Instantaneous Flow Export Rate

   Additional useful information when analyzing the Flow Export data is
   the time distribution of the instantaneous Flow Export Rate.  It can
   be derived during the measurements in two ways:

   a. The Collector might provide the capability to decode Flow Export
      during capturing and at the same time count the Flow Records and
      provide the instantaneous (or simply, an average over shorter time
      interval than specified in Section 5.4) Flow Export Rate.
   b. The Flow Export protocol (like IPFIX [RFC5101]) can provide time
      stamps in the Flow Export packets that would allow time-based
      analysis and calculate the Flow Export Rate as an average over
      much shorter time interval than specified in Section 5.4.

   The accuracy and shortest time average will always be limited by the
   precision of the time stamps (1 second for IPFIX) or by the
   capabilities of the DUT and the Collector.

Author's Address

   Jan Novak (editor)
   Cisco Systems
   United Kingdom
   EMail: janovak@cisco.com

Novak                         Informational                    [Page 39]