RFC 9320




Internet Engineering Task Force (IETF)                           N. Finn
Request for Comments: 9320                   Huawei Technologies Co. Ltd
Category: Informational                                  J.-Y. Le Boudec
ISSN: 2070-1721                                          E. Mohammadpour
                                                                    EPFL
                                                                J. Zhang
                                             Huawei Technologies Co. Ltd
                                                                B. Varga
                                                                Ericsson
                                                           November 2022


           Deterministic Networking (DetNet) Bounded Latency

Abstract



   This document presents a timing model for sources, destinations, and
   Deterministic Networking (DetNet) transit nodes.  Using the model, it
   provides a methodology to compute end-to-end latency and backlog
   bounds for various queuing methods.  The methodology can be used by
   the management and control planes and by resource reservation
   algorithms to provide bounded latency and zero congestion loss for
   the DetNet service.

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 candidates for any level of Internet
   Standard; see Section 2 of RFC 7841.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   https://www.rfc-editor.org/info/rfc9320.

Copyright Notice



   Copyright (c) 2022 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
   (https://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 Revised BSD License text as described in Section 4.e of the
   Trust Legal Provisions and are provided without warranty as described
   in the Revised BSD License.

Table of Contents



   1.  Introduction
   2.  Terminology and Definitions
   3.  DetNet Bounded Latency Model
     3.1.  Flow Admission
       3.1.1.  Static Latency Calculation
       3.1.2.  Dynamic Latency Calculation
     3.2.  Relay Node Model
   4.  Computing End-to-End Delay Bounds
     4.1.  Non-queuing Delay Bound
     4.2.  Queuing Delay Bound
       4.2.1.  Per-Flow Queuing Mechanisms
       4.2.2.  Aggregate Queuing Mechanisms
     4.3.  Ingress Considerations
     4.4.  Interspersed DetNet-Unaware Transit Nodes
   5.  Achieving Zero Congestion Loss
   6.  Queuing Techniques
     6.1.  Queuing Data Model
     6.2.  Frame Preemption
     6.3.  Time-Aware Shaper
     6.4.  Credit-Based Shaper with Asynchronous Traffic Shaping
       6.4.1.  Delay Bound Calculation
       6.4.2.  Flow Admission
     6.5.  Guaranteed Service
     6.6.  Cyclic Queuing and Forwarding
   7.  Example Application on DetNet IP Network
   8.  Security Considerations
   9.  IANA considerations
   10. References
     10.1.  Normative References
     10.2.  Informative References
   Acknowledgments
   Contributors

   Authors' Addresses



1.  Introduction



   The ability for IETF Deterministic Networking (DetNet) or IEEE 802.1
   Time-Sensitive Networking [IEEE8021TSN] to provide the DetNet
   services of bounded latency and zero congestion loss depends upon

   A.  configuring and allocating network resources for the exclusive
       use of DetNet flows;



   B.  identifying, in the data plane, the resources to be utilized by
       any given packet; and



   C.  the detailed behavior of those resources, especially transmission
       queue selection, so that latency bounds can be reliably assured.



   As explained in [RFC8655], DetNet flows are notably characterized by

   1.  a maximum bandwidth, guaranteed either by the transmitter or by
       strict input metering, and

   2.  a requirement for a guaranteed worst-case end-to-end latency.



   That latency guarantee, in turn, provides the opportunity for the
   network to supply enough buffer space to guarantee zero congestion
   loss.  In this document, it is assumed that the paths of DetNet flows
   are fixed.  Before the transmission of a DetNet flow, it is possible
   to calculate end-to-end latency bounds and the amount of buffer space
   required at each hop to ensure zero congestion loss; this can be used
   by the applications identified in [RFC8578].

   This document presents a timing model for sources, destinations, and
   the DetNet transit nodes; using this model, it provides a methodology
   to compute end-to-end latency and backlog bounds for various queuing
   mechanisms that can be used by the management and control planes to
   provide DetNet qualities of service.  The methodology used in this
   document accounts for the possibility of packet reordering within a
   DetNet node.  The bounds on the amount of packet reordering is out of
   the scope of this document and can be found in
   [PacketReorderingBounds].  Moreover, this document references
   specific queuing mechanisms, mentioned in [RFC8655], as proofs of
   concept that can be used to control packet transmission at each
   output port and achieve the DetNet quality of service (QoS).

   Using the model presented in this document, it is possible for an
   implementer, user, or standards development organization to select a
   set of queuing mechanisms for each device in a DetNet network and to
   select a resource reservation algorithm for that network so that
   those elements can work together to provide the DetNet service.
   Section 7 provides an example application of the timing model
   introduced in this document on a DetNet IP network with a combination
   of different queuing mechanisms.

   This document does not specify any resource reservation protocol or
   control plane function.  It does not describe all of the requirements
   for that protocol or control plane function.  It does describe
   requirements for such resource reservation methods and for queuing
   mechanisms that, if met, will enable them to work together.

2.  Terminology and Definitions

   This document uses the terms defined in [RFC8655].  Moreover, the
   following terms are used in this document:

   T-SPEC
      TrafficSpecification, as defined in Section 5.5 of [RFC9016].

   arrival curve
      An arrival curve function alpha(t) is an upper bound on the number
      of bits seen at an observation point within any time interval t.

   CQF
      Cyclic Queuing and Forwarding.

   CBS
      Credit-Based Shaper.

   TSN
      Time-Sensitive Networking.

   PREOF
      A collective name for Packet Replication, Elimination, and
      Ordering Functions.

   POF
      A Packet Ordering Function is a function that reorders packets
      within a DetNet flow that are received out of order.  This
      function can be implemented by a DetNet edge node, a DetNet relay
      node, or an end system.

3.  DetNet Bounded Latency Model



3.1.  Flow Admission



   This document assumes that the following paradigm is used to admit
   DetNet flows:

   1.  Perform any configuration required by the DetNet transit nodes in
       the network for aggregates of DetNet flows.  This configuration
       is done beforehand and not tied to any particular DetNet flow.

   2.  Characterize the new DetNet flow, particularly in terms of
       required bandwidth.

   3.  Establish the path that the DetNet flow will take through the
       network from the source to the destination(s).  This can be a
       point-to-point or a point-to-multipoint path.

   4.  Compute the worst-case end-to-end latency for the DetNet flow
       using one of the methods below (Sections 3.1.1 and 3.1.2).  In
       the process, determine whether sufficient resources are available
       for the DetNet flow to guarantee the required latency and to
       provide zero congestion loss.

   5.  Assuming that the resources are available, commit those resources
       to the DetNet flow.  This may require adjusting the parameters
       that control the filtering and/or queuing mechanisms at each hop
       along the DetNet flow's path.

   This paradigm can be implemented using peer-to-peer protocols or
   using a central controller.  In some situations, a lack of resources
   can require backtracking and recursing through the above list.

   Issues, such as service preemption of a DetNet flow in favor of
   another, when resources are scarce, are not considered here.  Also
   not addressed is the question of how to choose the path to be taken
   by a DetNet flow.

3.1.1.  Static Latency Calculation



   The static problem:
           Given a network and a set of DetNet flows, compute an end-to-
           end latency bound (if computable) for each DetNet flow and
           compute the resources, particularly buffer space, required in
           each DetNet transit node to achieve zero congestion loss.

   In this calculation, all of the DetNet flows are known before the
   calculation commences.  This problem is of interest to relatively
   static networks or static parts of larger networks.  It provides
   bounds on latency and buffer size.  The calculations can be extended
   to provide global optimizations, such as altering the path of one
   DetNet flow in order to make resources available to another DetNet
   flow with tighter constraints.

   This calculation may be more difficult to perform than the dynamic
   calculation (Section 3.1.2) because the DetNet flows passing through
   one port on a DetNet transit node affect each other's latency.  The
   effects can even be circular, from node A to B to C and back to A.
   On the other hand, the static calculation can often accommodate
   queuing methods, such as transmission selection by strict priority,
   that are unsuitable for the dynamic calculation.

3.1.2.  Dynamic Latency Calculation



   The dynamic problem:
           Given a network whose maximum capacity for DetNet flows is
           bounded by a set of static configuration parameters applied
           to the DetNet transit nodes and given just one DetNet flow,
           compute the worst-case end-to-end latency that can be
           experienced by that flow, no matter what other DetNet flows
           (within the network's configured parameters) might be created
           or deleted in the future.  Also, compute the resources,
           particularly buffer space, required in each DetNet transit
           node to achieve zero congestion loss.

   This calculation is dynamic, in the sense that DetNet flows can be
   added or deleted at any time, with a minimum of computation effort
   and without affecting the guarantees already given to other DetNet
   flows.

   Dynamic latency calculation can be done based on the static one
   described in Section 3.1.1; when a new DetNet flow is created or
   deleted, the entire calculation for all DetNet flows is repeated.  If
   an already-established DetNet flow would be pushed beyond its latency
   requirements by the new DetNet flow request, then the new DetNet flow
   request can be refused or some other suitable action can be taken.

   The choice of queuing methods is critical to the applicability of the
   dynamic calculation.  Some queuing methods (e.g., CQF, Section 6.6)
   make it easy to configure bounds on the network's capacity and to
   make independent calculations for each DetNet flow.  Some other
   queuing methods (e.g., strict priority with the credit-based shaper
   defined in Section 8.6.8.2 of [IEEE8021Q]) can be used for dynamic
   DetNet flow creation but yield poorer latency and buffer space
   guarantees than when that same queuing method is used for static
   DetNet flow creation (Section 3.1.1).

3.2.  Relay Node Model



   A model for the operation of a DetNet transit node is required in
   order to define the latency and buffer calculations.  In Figure 1, we
   see a breakdown of the per-hop latency experienced by a packet
   passing through a DetNet transit node in terms that are suitable for
   computing both hop-by-hop latency and per-hop buffer requirements.

         DetNet transit node A            DetNet transit node B
      +-------------------------+       +------------------------+
      |              Queuing    |       |              Queuing   |
      |   Regulator subsystem   |       |   Regulator subsystem  |
      |   +-+-+-+-+ +-+-+-+-+   |       |   +-+-+-+-+ +-+-+-+-+  |
   -->+   | | | | | | | | | +   +------>+   | | | | | | | | | +  +--->
      |   +-+-+-+-+ +-+-+-+-+   |       |   +-+-+-+-+ +-+-+-+-+  |
      |                         |       |                        |
      +-------------------------+       +------------------------+
      |<->|<------>|<------->|<->|<---->|<->|<------>|<------>|<->|<--
   2,3  4      5        6      1    2,3   4      5        6     1   2,3
             1: Output delay             4: Processing delay
             2: Link delay               5: Regulation delay
             3: Frame preemption delay   6: Queuing subsystem delay

                  Figure 1: Timing Model for DetNet or TSN

   In Figure 1, we see two DetNet transit nodes that are connected via a
   link.  In this model, the only queues that we deal with explicitly
   are attached to the output port; other queues are modeled as
   variations in the other delay times (e.g., an input queue could be
   modeled as either a variation in the link delay (2) or the processing
   delay (4)).  There are six delays that a packet can experience from
   hop to hop.

   1.  Output delay

       This is the time taken from the selection of a packet for output
       from a queue to the transmission of the first bit of the packet
       on the physical link.  If the queue is directly attached to the
       physical port, output delay can be a constant.  However, in many
       implementations, a multiplexed connection separates the queuing
       mechanism from a multi-port Network Interface Card (NIC).  This
       causes variations in the output delay that are hard for the
       forwarding node to predict or control.

   2.  Link delay

       This is the time taken from the transmission of the first bit of
       the packet to the reception of the last bit, assuming that the
       transmission is not suspended by a frame preemption event.  This
       delay has two components: the first-bit-out to first-bit-in delay
       and the first-bit-in to last-bit-in delay that varies with packet
       size.  The former is typically constant.  However, a virtual
       "link" could exhibit a variable link delay.

   3.  Frame preemption delay

       If the packet is interrupted in order to transmit another packet
       or packets (e.g., frame preemption, as in [IEEE8023], clause 99),
       an arbitrary delay can result.

   4.  Processing delay



       This delay covers the time from the reception of the last bit of
       the packet to the time the packet is enqueued in the regulator
       (queuing subsystem if there is no regulator), as shown in
       Figure 1.  This delay can be variable and depends on the details
       of the operation of the forwarding node.

   5.  Regulator queuing delay



       A regulator, also known as shaper in [RFC2475], delays some or
       all of the packets in a traffic stream in order to bring the
       stream into compliance with an arrival curve; an arrival curve
       'alpha(t)' is an upper bound on the number of bits observed
       within any interval t.  The regulator delay is the time spent
       from the insertion of the last bit of a packet into a regulation
       queue until the time the packet is declared eligible according to
       its regulation constraints.  We assume that this time can be
       calculated based on the details of regulation policy.  If there
       is no regulation, this time is zero.

   6.  Queuing subsystem delay



       This is the time spent for a packet from being declared eligible
       until being selected for output on the next link.  We assume that
       this time is calculable based on the details of the queuing
       mechanism.  If there is no regulation, this time is from the
       insertion of the packet into a queue until it is selected for
       output on the next link.

   Not shown in Figure 1 are the other output queues that we presume are
   also attached to that same output port as the queue shown, and
   against which this shown queue competes for transmission
   opportunities.

   In this analysis, the measurement is from the point at which a packet
   is selected for output in a node to the point at which it is selected
   for output in the next downstream node (i.e., the definition of a
   "hop").  In general, any queue selection method that is suitable for
   use in a DetNet network includes a detailed specification as to
   exactly when packets are selected for transmission.  Any variations
   in any of the delay times 1-4 result in a need for additional buffers
   in the queue.  If all delays 1-4 are constant, then any variation in
   the time at which packets are inserted into a queue depends entirely
   on the timing of packet selection in the previous node.  If delays
   1-4 are not constant, then additional buffers are required in the
   queue to absorb these variations.  Thus:

   *  Variations in the output delay (1) require buffers to absorb that
      variation in the next hop, so the output delay variations of the
      previous hop (on each input port) must be known in order to
      calculate the buffer space required on this hop.

   *  Variations in the processing delay (4) require additional output
      buffers in the queues of that same DetNet transit node.  Depending
      on the details of the queuing subsystem delay (6) calculations,
      these variations need not be visible outside the DetNet transit
      node.

4.  Computing End-to-End Delay Bounds

4.1.  Non-queuing Delay Bound



   End-to-end latency bounds can be computed using the delay model in
   Section 3.2.  Here, it is important to be aware that, for several
   queuing mechanisms, the end-to-end latency bound is less than the sum
   of the per-hop latency bounds.  An end-to-end latency bound for one
   DetNet flow can be computed as

      end_to_end_delay_bound = non_queuing_delay_bound +
      queuing_delay_bound

   The two terms in the above formula are computed as follows.

   First, at the h-th hop along the path of this DetNet flow, obtain an
   upper-bound per-hop_non_queuing_delay_bound[h] on the sum of the
   bounds over delays 1, 2, 3, and 4 of Figure 1.  These upper bounds
   are expected to depend on the specific technology of the DetNet
   transit node at the h-th hop but not on the T-SPEC of this DetNet
   flow [RFC9016].  Then, set non_queuing_delay_bound = the sum of per-
   hop_non_queuing_delay_bound[h] over all hops h.

   Second, compute queuing_delay_bound as an upper bound to the sum of
   the queuing delays along the path.  The value of queuing_delay_bound
   depends on the information on the arrival curve of this DetNet flow
   and possibly of other flows in the network, as well as the specifics
   of the queuing mechanisms deployed along the path of this DetNet
   flow.  Note that arrival curve of the DetNet flow at the source is
   immediately specified by the T-SPEC of this flow.  The computation of
   queuing_delay_bound is described in Section 4.2 as a separate
   section.

4.2.  Queuing Delay Bound



   For several queuing mechanisms, queuing_delay_bound is less than the
   sum of upper bounds on the queuing delays (5 and 6) at every hop.
   This occurs with (1) per-flow queuing and (2) aggregate queuing with
   regulators, as explained in Sections 4.2.1, 4.2.2, and 6.  For other
   queuing mechanisms, the only available value of queuing_delay_bound
   is the sum of the per-hop queuing delay bounds.

   The computation of per-hop queuing delay bounds must account for the
   fact that the arrival curve of a DetNet flow is no longer satisfied
   at the ingress of a hop, since burstiness increases as one flow
   traverses one DetNet transit node.  If a regulator is placed at a
   hop, an arrival curve of a DetNet flow at the entrance of the queuing
   subsystem of this hop is the one configured at the regulator (also
   called shaping curve in [NetCalBook]); otherwise, an arrival curve of
   the flow can be derived using the delay jitter of the flow from the
   last regulation point (the last regulator in the path of the flow if
   there is any, otherwise the source of the flow) to the ingress of the
   hop; more formally, assume a DetNet flow has an arrival curve at the
   last regulation point equal to 'alpha(t)' and the delay jitter from
   the last regulation point to the ingress of the hop is 'V'.  Then,
   the arrival curve at the ingress of the hop is 'alpha(t+V)'.

   For example, consider a DetNet flow with T-SPEC "Interval: tau,
   MaxPacketsPerInterval: K, MaxPayloadSize: L" at the source.  Then, a
   leaky-bucket arrival curve for such flow at the source is alpha(t)=r
   * t+ b, t>0; alpha(0)=0, where r is the rate and b is the bucket
   size, computed as

      r = K * (L+L') / tau,

      b = K * (L+L').

   where L' is the size of any added networking technology-specific
   encapsulation (e.g., MPLS label(s), UDP, or IP headers).  Now, if the
   flow has a delay jitter of 'V' from the last regulation point to the
   ingress of a hop, an arrival curve at this point is r * t + b + r *
   V, implying that the burstiness is increased by r*V.  More detailed
   information on arrival curves is available in [NetCalBook].

4.2.1.  Per-Flow Queuing Mechanisms



   With such mechanisms, each flow uses a separate queue inside every
   node.  The service for each queue is abstracted with a guaranteed
   rate and a latency.  For every DetNet flow, a per-node latency bound,
   as well as an end-to-end latency bound, can be computed from the
   traffic specification of this DetNet flow at its source and from the
   values of rates and latencies at all nodes along its path.  An
   instance of per-flow queuing is Guaranteed Service [RFC2212], for
   which the details of latency bound calculation are presented in
   Section 6.5.

4.2.2.  Aggregate Queuing Mechanisms



   With such mechanisms, multiple flows are aggregated into macro-flows
   and there is one FIFO queue per macro-flow.  A practical example is
   the credit-based shaper defined in Section 8.6.8.2 of [IEEE8021Q],
   where a macro-flow is called a "class".  One key issue in this
   context is how to deal with the burstiness cascade; individual flows
   that share a resource dedicated to a macro-flow may see their
   burstiness increase, which may in turn cause increased burstiness to
   other flows downstream of this resource.  Computing delay upper
   bounds for such cases is difficult and, in some conditions,
   impossible [CharnyDelay] [BennettDelay].  Also, when bounds are
   obtained, they depend on the complete configuration and must be
   recomputed when one flow is added (i.e., the dynamic calculation in
   Section 3.1.2).

   A solution to deal with this issue for the DetNet flows is to reshape
   them at every hop.  This can be done with per-flow regulators (e.g.,
   leaky-bucket shapers), but this requires per-flow queuing and defeats
   the purpose of aggregate queuing.  An alternative is the interleaved
   regulator, which reshapes individual DetNet flows without per-flow
   queuing [SpechtUBS] [IEEE8021Qcr].  With an interleaved regulator,
   the packet at the head of the queue is regulated based on its (flow)
   regulation constraints; it is released at the earliest time at which
   this is possible without violating the constraint.  One key feature
   of a per-flow or interleaved regulator is that it does not increase
   worst-case latency bounds [LeBoudecTheory].  Specifically, when an
   interleaved regulator is appended to a FIFO subsystem, it does not
   increase the worst-case delay of the latter.  In Figure 1, when the
   order of packets from the output of a queuing subsystem at node A to
   the entrance of a regulator at node B is preserved, then the
   regulator does not increase the worst-case latency bounds.  This is
   made possible if all the systems are FIFO or a DetNet Packet Ordering
   Function (POF) is implemented just before the regulator.  This
   property does not hold if packet reordering occurs from the output of
   a queuing subsystem to the entrance of the next downstream
   interleaved regulator, e.g., at a non-FIFO switching fabric.

   Figure 2 shows an example of a network with 5 nodes, an aggregate
   queuing mechanism, and interleaved regulators, as in Figure 1.  An
   end-to-end delay bound for DetNet flow f, traversing nodes 1 to 5, is
   calculated as follows:

      end_to_end_latency_bound_of_flow_f = C12 + C23 + C34 + S4

   In the above formula, Cij is a bound on the delay of the queuing
   subsystem in node i and interleaved regulator of node j, and S4 is a
   bound on the delay of the queuing subsystem in node 4 for DetNet flow
   f.  In fact, using the delay definitions in Section 3.2, Cij is a
   bound on a sum of delays 1, 2, 3, and 6 of node i and delays 4 and 5
   of node j.  Similarly, S4 is a bound on sum of delays 1, 2, 3, and 6
   of node 4.  A practical example of the queuing model and delay
   calculation is presented Section 6.4.

                               f
                     ----------------------------->
                   +---+   +---+   +---+   +---+   +---+
                   | 1 |---| 2 |---| 3 |---| 4 |---| 5 |
                   +---+   +---+   +---+   +---+   +---+
                      \__C12_/\__C23_/\__C34_/\_S4_/

               Figure 2: End-to-End Delay Computation Example

   If packet reordering does not occur, the end-to-end latency bound
   calculation provided here gives a tighter latency upper bound than
   would be obtained by adding the latency bounds of each node in the
   path of a DetNet flow [TSNwithATS].

4.3.  Ingress Considerations



   A sender can be a DetNet node that uses exactly the same queuing
   methods as its adjacent DetNet transit node so that the latency and
   buffer bounds calculations at the first hop are indistinguishable
   from those at a later hop within the DetNet domain.  On the other
   hand, the sender may be DetNet unaware; in which case, some
   conditioning of the DetNet flow may be necessary at the ingress
   DetNet transit node.  The ingress conditioning typically consists of
   the regulators described in Section 3.2.

4.4.  Interspersed DetNet-Unaware Transit Nodes



   It is sometimes desirable to build a network that has both DetNet-
   aware transit nodes and DetNet-unaware transit nodes and for a DetNet
   flow to traverse an island of DetNet-unaware transit nodes while
   still allowing the network to offer delay and congestion loss
   guarantees.  This is possible under certain conditions.

   In general, when passing through a DetNet-unaware island, the island
   may cause delay variation in excess of what would be caused by DetNet
   nodes.  That is, the DetNet flow might be "lumpier" after traversing
   the DetNet-unaware island.  DetNet guarantees for delay and buffer
   requirements can still be calculated and met if and only if the
   following are true:

   1.  The latency variation across the DetNet-unaware island must be
       bounded and calculable.

   2.  An ingress conditioning function (Section 4.3) is required at the
       reentry to the DetNet-aware domain.  This will, at least, require
       some extra buffering to accommodate the additional delay
       variation and thus further increases the latency bound.

   The ingress conditioning is exactly the same problem as that of a
   sender at the edge of the DetNet domain.  The requirement for bounds
   on the latency variation across the DetNet-unaware island is
   typically the most difficult to achieve.  Without such a bound, it is
   obvious that DetNet cannot deliver its guarantees, so a DetNet-
   unaware island that cannot offer bounded latency variation cannot be
   used to carry a DetNet flow.

5.  Achieving Zero Congestion Loss

   When the input rate to an output queue exceeds the output rate for a
   sufficient length of time, the queue must overflow.  This is
   congestion loss, and this is what DetNet seeks to avoid.

   To avoid congestion losses, an upper bound on the backlog present in
   the regulator and queuing subsystem of Figure 1 must be computed
   during resource reservation.  This bound depends on the set of flows
   that use these queues, the details of the specific queuing mechanism,
   and an upper bound on the processing delay (4).  The queue must
   contain the packet in transmission, plus all other packets that are
   waiting to be selected for output.  A conservative backlog bound that
   applies to all systems can be derived as follows.

   The backlog bound is counted in data units (bytes or words of
   multiple bytes) that are relevant for buffer allocation.  For every
   flow or an aggregate of flows, we need one buffer space for the
   packet in transmission, plus space for the packets that are waiting
   to be selected for output.

   Let

   *  total_in_rate be the sum of the line rates of all input ports that
      send traffic to this output port.  The value of total_in_rate is
      in data units (e.g., bytes) per second.

   *  nb_input_ports be the number of input ports that send traffic to
      this output port.

   *  max_packet_length be the maximum packet size for packets that may
      be sent to this output port.  This is counted in data units.

   *  max_delay456 be an upper bound, in seconds, on the sum of the
      processing delay (4) and the queuing delays (5 and 6) for any
      packet at this output port.

   Then, a bound on the backlog of traffic in the queue at this output
   port is

      backlog_bound = (nb_input_ports * max_packet_length) +
      (total_in_rate * max_delay456)

   The above bound is over the backlog caused by the traffic entering
   the queue from the input ports of a DetNet node.  If the DetNet node
   also generates packets (e.g., creation of new packets or replication
   of arriving packets), the bound must accordingly incorporate the
   introduced backlog.

6.  Queuing Techniques

   In this section, we present a general queuing data model, as well as
   some examples of queuing mechanisms.  For simplicity of latency bound
   computation, we assume a leaky-bucket arrival curve for each DetNet
   flow at the source.  Also, at each DetNet transit node, the service
   for each queue is abstracted with a minimum guaranteed rate and a
   latency [NetCalBook].

6.1.  Queuing Data Model



   Sophisticated queuing mechanisms are available in Layer 3 (L3) (e.g.,
   see [RFC7806] for an overview).  In general, we assume that "Layer 3"
   queues, shapers, meters, etc., are precisely the "regulators" shown
   in Figure 1.  The "queuing subsystems" in this figure are FIFO.  They
   are not the province solely of bridges; they are an essential part of
   any DetNet transit node.  As illustrated by numerous implementation
   examples, some of the "Layer 3" mechanisms described in documents,
   such as [RFC7806], are often integrated in an implementation, with
   the "Layer 2" mechanisms also implemented in the same node.  An
   integrated model is needed in order to successfully predict the
   interactions among the different queuing mechanisms needed in a
   network carrying both DetNet flows and non-DetNet flows.

   Figure 3 shows the general model for the flow of packets through the
   queues of a DetNet transit node.  The DetNet packets are mapped to a
   number of regulators.  Here, we assume that the Packet Replication,
   Elimination, and Ordering Functions (PREOF) are performed before the
   DetNet packets enter the regulators.  All packets are assigned to a
   set of queues.  Packets compete for the selection to be passed to
   queues in the queuing subsystem.  Packets again are selected for
   output from the queuing subsystem.

                                    |
   +--------------------------------V----------------------------------+
   |                          Queue assignment                         |
   +--+------+----------+---------+-----------+-----+-------+-------+--+
      |      |          |         |           |     |       |       |
   +--V-+ +--V-+     +--V--+   +--V--+     +--V--+  |       |       |
   |Flow| |Flow|     |Flow |   |Flow |     |Flow |  |       |       |
   |  0 | |  1 | ... |  i  |   | i+1 | ... |  n  |  |       |       |
   | reg| | reg|     | reg |   | reg |     | reg |  |       |       |
   +--+-+ +--+-+     +--+--+   +--+--+     +--+--+  |       |       |
      |      |          |         |           |     |       |       |
   +--V------V----------V--+   +--V-----------V--+  |       |       |
   |  Trans.  selection    |   | Trans. select.  |  |       |       |
   +----------+------------+   +-----+-----------+  |       |       |
              |                      |              |       |       |
           +--V--+                +--V--+        +--V--+ +--V--+ +--V--+
           | out |                | out |        | out | | out | | out |
           |queue|                |queue|        |queue| |queue| |queue|
           |  1  |                |  2  |        |  3  | |  4  | |  5  |
           +--+--+                +--+--+        +--+--+ +--+--+ +--+--+
              |                      |              |       |       |
   +----------V----------------------V--------------V-------V-------V--+
   |                      Transmission selection                       |
   +---------------------------------+---------------------------------+
                                     |
                                     V

               Figure 3: IEEE 802.1Q Queuing Model: Data Flow

   Some relevant mechanisms are hidden in this figure and are performed
   in the queue boxes:

   *  discarding packets because a queue is full

   *  discarding packets marked "yellow" by a metering function in
      preference to discarding "green" packets [RFC2697]

   Ideally, neither of these actions are performed on DetNet packets.
   Full queues for DetNet packets occur only when a DetNet flow is
   misbehaving, and the DetNet QoS does not include "yellow" service for
   packets in excess of a committed rate.

   The queue assignment function can be quite complex, even in a bridge
   [IEEE8021Q], because of the introduction of per-stream filtering and
   policing ([IEEE8021Q], clause 8.6.5.1).  In addition to the Layer 2
   priority expressed in the 802.1Q VLAN tag, a DetNet transit node can
   utilize the information from the non-exhaustive list below to assign
   a packet to a particular queue:

   *  input port

   *  selector based on a rotating schedule that starts at regular,
      time-synchronized intervals and has nanosecond precision

   *  MAC addresses, VLAN ID, IP addresses, Layer 4 port numbers, and
      Differentiated Services Code Point (DSCP) [RFC8939] [RFC8964]

   *  the queue assignment function can contain metering and policing
      functions

   *  MPLS and/or pseudowire labels [RFC6658]

   The "Transmission selection" function decides which queue is to
   transfer its oldest packet to the output port when a transmission
   opportunity arises.

6.2.  Frame Preemption



   In [IEEE8021Q] and [IEEE8023], the transmission of a frame can be
   interrupted by one or more "express" frames; then, the interrupted
   frame can continue transmission.  The frame preemption is modeled as
   consisting of two MAC/PHY stacks: one for packets that can be
   interrupted and one for packets that can interrupt the interruptible
   packets.  Only one layer of frame preemption is supported -- a
   transmitter cannot have more than one interrupted frame in progress.
   DetNet flows typically pass through the interrupting MAC.  For those
   DetNet flows with T-SPEC, latency bounds can be calculated by the
   methods provided in the following sections that account for the
   effect of frame preemption, according to the specific queuing
   mechanism that is used in DetNet nodes.  Best-effort queues pass
   through the interruptible MAC and can thus be preempted.

6.3.  Time-Aware Shaper



   In [IEEE8021Q], the notion of time-scheduling queue gates is
   described in Section 8.6.8.4.  On each node, the transmission
   selection for packets is controlled by time-synchronized gates; each
   output queue is associated with a gate.  The gates can be either open
   or closed.  The states of the gates are determined by the gate
   control list (GCL).  The GCL specifies the opening and closing times
   of the gates.  The design of the GCL must satisfy the requirement of
   latency upper bounds of all DetNet flows; therefore, those DetNet
   flows that traverse a network that uses this kind of shaper must have
   bounded latency if the traffic and nodes are conformant.

   Note that scheduled traffic service relies on a synchronized network
   and coordinated GCL configuration.  Synthesis of the GCL on multiple
   nodes in a network is a scheduling problem considering all DetNet
   flows traversing the network, which is a nondeterministic polynomial-
   time hard (NP-hard) problem [Sch8021Qbv].  Also, at the time of
   writing, scheduled traffic service supports no more than eight
   traffic queues, typically using up to seven priority queues and at
   least one best effort.

6.4.  Credit-Based Shaper with Asynchronous Traffic Shaping



   In this queuing model, it is assumed that the DetNet nodes are FIFO.
   We consider the four traffic classes (Definition 3.268 of
   [IEEE8021Q]): control-data traffic (CDT), class A, class B, and best
   effort (BE) in decreasing order of priority.  Flows of classes A and
   B are DetNet flows that are less critical than CDT (such as studio
   audio and video traffic, as in IEEE 802.1BA Audio-Video-Bridging).
   This model is a subset of Time-Sensitive Networking, as described
   next.

   Based on the timing model described in Figure 1, contention occurs
   only at the output port of a DetNet transit node; therefore, the
   focus of the rest of this subsection is on the regulator and queuing
   subsystem in the output port of a DetNet transit node.  The input
   flows are identified using the information in (Section 5.1 of
   [RFC8939]).  Then, they are aggregated into eight macro-flows based
   on their service requirements; we refer to each macro-flow as a
   class.  The output port performs aggregate scheduling with eight
   queues (queuing subsystems): one for CDT, one for class A flows, one
   for class B flows, and five for BE traffic denoted as BE0-BE4.  The
   queuing policy for each queuing subsystem is FIFO.  In addition, each
   node output port also performs per-flow regulation for class A and B
   flows using an interleaved regulator (IR).  This regulation is called
   asynchronous traffic shaping [IEEE8021Qcr].  Thus, at each output
   port of a node, there is one interleaved regulator per input port and
   per class; the interleaved regulator is mapped to the regulator
   depicted in Figure 1.  The detailed picture of scheduling and
   regulation architecture at a node output port is given by Figure 4.
   The packets received at a node input port for a given class are
   enqueued in the respective interleaved regulator at the output port.
   Then, the packets from all the flows, including CDT and BE flows, are
   enqueued in a queuing subsystem; there is no regulator for CDT and BE
   flows.

         +--+   +--+ +--+   +--+
         |  |   |  | |  |   |  |
         |IR|   |IR| |IR|   |IR|
         |  |   |  | |  |   |  |
         +-++XXX++-+ +-++XXX++-+
           |     |     |     |
           |     |     |     |
   +---+ +-v-XXX-v-+ +-v-XXX-v-+ +-----+ +-----+ +-----+ +-----+ +-----+
   |   | |         | |         | |Class| |Class| |Class| |Class| |Class|
   |CDT| | Class A | | Class B | | BE4 | | BE3 | | BE2 | | BE1 | | BE0 |
   |   | |         | |         | |     | |     | |     | |     | |     |
   +-+-+ +----+----+ +----+----+ +--+--+ +--+--+ +--+--+ +--+--+ +--+--+
     |        |           |         |       |       |       |       |
     |      +-v-+       +-v-+       |       |       |       |       |
     |      |CBS|       |CBS|       |       |       |       |       |
     |      +-+-+       +-+-+       |       |       |       |       |
     |        |           |         |       |       |       |       |
   +-v--------v-----------v---------v-------V-------v-------v-------v--+
   |                     Strict Priority selection                     |
   +--------------------------------+----------------------------------+
                                    |
                                    V

   Figure 4: The Architecture of an Output Port inside a Relay Node with
        Interleaved Regulators (IRs) and a Credit-Based Shaper (CBS)

   Each of the queuing subsystems for classes A and B contains a credit-
   based shaper (CBS).  The CBS serves a packet from a class according
   to the available credit for that class.  As described in
   Section 8.6.8.2 and Annex L.1 of [IEEE8021Q], the credit for each
   class A or B increases based on the idle slope (as guaranteed rate)
   and decreases based on the sendslope (typically equal to the
   difference between the guaranteed and the output link rates), both of
   which are parameters of the CBS.  The CDT and BE0-BE4 flows are
   served by separate queuing subsystems.  Then, packets from all flows
   are served by a transmission selection subsystem that serves packets
   from each class based on its priority.  All subsystems are non-
   preemptive.  Guarantees for class A and B traffic can be provided
   only if CDT is bounded.  It is assumed that the CDT has a leaky-
   bucket arrival curve with two parameters: r_h as rate and b_h as
   bucket size.  That is, the amount of bits entering a node within a
   time interval t is bounded by r_h * t + b_h.

   Additionally, it is assumed that the class A and B flows are also
   regulated at their source according to a leaky-bucket arrival curve.
   At the source, the traffic satisfies its regulation constraint, i.e.,
   the delay due to interleaved regulator at the source is ignored.

   At each DetNet transit node implementing an interleaved regulator,
   packets of multiple flows are processed in one FIFO queue.  The
   packet at the head of the queue is regulated based on its leaky-
   bucket parameters.  It is released at the earliest time at which this
   is possible without violating the constraint.

   The regulation parameters for a flow (leaky-bucket rate and bucket
   size) are the same at its source and at all DetNet transit nodes
   along its path in the case where all clocks are perfect.  However, in
   reality, there is clock non-ideality throughout the DetNet domain,
   even with clock synchronization.  This phenomenon causes inaccuracy
   in the rates configured at the regulators that may lead to network
   instability.  To avoid instability, the rates are set as the source
   rates with some positive margin when configuring regulators.
   [ThomasTime] describes and provides solutions to this issue.

6.4.1.  Delay Bound Calculation



   A delay bound of the queuing subsystem ((4) in Figure 1) of a given
   DetNet node for a flow of class A or B can be computed if the
   following condition holds:

      The sum of leaky-bucket rates of all flows of this class at this
      transit node <= R, where R is given below for every class

   If the condition holds, the delay bounds for a flow of class X (A or
   B) is d_X and calculated as:

      d_X = T_X + (b_t_X-L_min_X)/R_X - L_min_X/c

   where L_min_X is the minimum packet lengths of class X (A or B); c is
   the output link transmission rate; and b_t_X is the sum of the b term
   (bucket size) for all the flows of the class X.  Parameters R_X and
   T_X are calculated as follows for class A and B, separately.

   If the flow is of class A:

      R_A = I_A * (c-r_h)/ c

      T_A = (L_nA + b_h + r_h * L_n/c)/(c-r_h)

   where I_A is the idle slope for class A; L_nA is the maximum packet
   length of class B and BE packets; L_n is the maximum packet length of
   classes A, B, and BE; and r_h is the rate and b_h is the bucket size
   of CDT leaky-bucket arrival curve.

   If the flow is of class B:

      R_B = I_B * (c-r_h)/ c

      T_B = (L_BE + L_A + L_nA * I_A/(c_h-I_A) + b_h + r_h * L_n/
      c)/(c-r_h)

   where I_B is the idle slope for class B; L_A is the maximum packet
   length of class A; and L_BE is the maximum packet length of class BE.

   Then, as discussed in Section 4.2.2, an interleaved regulator does
   not increase the delay bound of the upstream queuing subsystem;
   therefore, an end-to-end delay bound for a DetNet flow of class X (A
   or B) is the sum of d_X_i for all node i in the path of the flow,
   where d_X_i is the delay bound of queuing subsystem in node i, which
   is computed as above.  According to the notation in Section 4.2.2,
   the delay bound of the queuing subsystem in a node i and interleaved
   regulator in node j, i.e., Cij, is:

      Cij = d_X_i

   More information of delay analysis in such a DetNet transit node is
   described in [TSNwithATS].

6.4.2.  Flow Admission



   The delay bound calculation requires some information about each
   node.  For each node, it is required to know the idle slope of the
   CBS for each class A and B (I_A and I_B), as well as the transmission
   rate of the output link (c).  Besides, it is necessary to have the
   information on each class, i.e., maximum packet length of classes A,
   B, and BE.  Moreover, the leaky-bucket parameters of CDT (r_h, b_h)
   must be known.  To admit a flow or flows of classes A and B, their
   delay requirements must be guaranteed not to be violated.  As
   described in Section 3.1, the two problems (static and dynamic) are
   addressed separately.  In either of the problems, the rate and delay
   must be guaranteed.  Thus,

   The static admission control:
           The leaky-bucket parameters of all class A or B flows are
           known; therefore, for each flow f of either class A or B, a
           delay bound can be calculated.  The computed delay bound for
           every flow of class A or B must not be more than its delay
           requirement.  Moreover, the sum of the rate of each flow
           (r_f) must not be more than the rate allocated to each class
           (R).  If these two conditions hold, the configuration is
           declared admissible.

   The dynamic admission control:
           For dynamic admission control, we allocate a static value
           for rate (R) and a maximum bucket size (b_t) to every node
           and each class A or B.  In addition, for every node and
           each class A or B, two counters are maintained:
              R_acc is equal to the sum of the leaky-bucket rates of all
              flows of this class already admitted at this node; at all
              times, we must have:

              R_acc <= R, (Eq. 1)

              b_acc is equal to the sum of the bucket sizes of all flows
              of this class already admitted at this node; at all times,
              we must have:

              b_acc <= b_t.  (Eq. 2)

           A new class A or B flow is admitted at this node if Eqs. (1)
           and (2) continue to be satisfied after adding its leaky-
           bucket rate and bucket size to R_acc and b_acc.  A class A or
           B flow is admitted in the network if it is admitted at all
           nodes along its path.  When this happens, all variables R_acc
           and b_acc along its path must be incremented to reflect the
           addition of the flow.  Similarly, when a class A or B flow
           leaves the network, all variables R_acc and b_acc along its
           path must be decremented to reflect the removal of the flow.

   The choice of the static values of R and b_t at all nodes and classes
   must be done in a prior configuration phase: R controls the bandwidth
   allocated to this class at this node, and b_t affects the delay bound
   and the buffer requirement.  The value of R must be set such that

      R <= I_X*(c-r_h)/c

   where I_X is the idleslope of credit-based shaper for class X={A,B},
   c is the transmission rate of the output link, and r_h is the leaky-
   bucket rate of the CDT class.

6.5.  Guaranteed Service



   The Guaranteed Service is defined in [RFC2212].  The flow, at the
   source, has a leaky-bucket arrival curve with two parameters: r as
   rate and b as bucket size, i.e., the amount of bits entering a node
   within a time interval t is bounded by r * t + b.

   If a resource reservation on a path is applied, a node provides a
   guaranteed rate R and maximum service latency of T.  This can be
   interpreted in a way that the bits might have to wait up to T before
   being served with a rate greater or equal to R.  The delay bound of
   the flow traversing the node is T + b / R.

   Consider a Guaranteed Service [RFC2212] path including a sequence of
   nodes, where the i-th node provides a guaranteed rate R_i and maximum
   service latency of T_i.  Then, the end-to-end delay bound for a flow
   on this can be calculated as sum(T_i) + b / min(R_i).

   The provided delay bound is based on a simple case of Guaranteed
   Service, where only a guaranteed rate and maximum service latency and
   a leaky-bucket arrival curve are available.  If more information
   about the flow is known, e.g., the peak rate, the delay bound is more
   complicated; the details are available in [RFC2212] and Section 1.4.1
   of [NetCalBook].

6.6.  Cyclic Queuing and Forwarding



   Annex T of [IEEE8021Q] describes Cyclic Queuing and Forwarding (CQF),
   which provides bounded latency and zero congestion loss using the
   time-scheduled gates of Section 8.6.8.4 of [IEEE8021Q].  For a given
   class of DetNet flows, a set of two or more buffers is provided at
   the output queue layer of Figure 3.  A cycle time T_c is configured
   for each class of DetNet flows c, and all of the buffer sets in a
   class of DetNet flows swap buffers simultaneously throughout the
   DetNet domain at that cycle rate, all in phase.  In such a mechanism,
   the regulator, as mentioned in Figure 1, is not required.

   In the case of two-buffer CQF, each class of DetNet flows c has two
   buffers, namely buffer1 and buffer2.  In a cycle (i) when buffer1
   accumulates received packets from the node's reception ports, buffer2
   transmits the already stored packets from the previous cycle (i-1).
   In the next cycle (i+1), buffer2 stores the received packets and
   buffer1 transmits the packets received in cycle (i).  The duration of
   each cycle is T_c.

   The cycle time T_c must be carefully chosen; it needs to be large
   enough to accommodate all the DetNet traffic, plus at least one
   maximum packet (or fragment) size from lower priority queues, which
   might be received within a cycle.  Also, the value of T_c includes a
   time interval, called dead time (DT), which is the sum of delays 1,
   2, 3, and 4 defined in Figure 1.  The value of DT guarantees that the
   last packet of one cycle in a node is fully delivered to a buffer of
   the next node in the same cycle.  A two-buffer CQF is recommended if
   DT is small compared to T_c.  For a large DT, CQF with more buffers
   can be used, and a cycle identification label can be added to the
   packets.

   The per-hop latency is determined by the cycle time T_c: a packet
   transmitted from a node at a cycle (i) is transmitted from the next
   node at cycle (i+1).  Then, if the packet traverses h hops, the
   maximum latency experienced by the packet is from the beginning of
   cycle (i) to the end of cycle (i+h); also, the minimum latency is
   from the end of cycle (i), before the DT, to the beginning of cycle
   (i+h).  Then, the maximum latency is:

      (h+1) T_c

   and the minimum latency is:

      (h-1) T_c + DT.

   Ingress conditioning (Section 4.3) may be required if the source of a
   DetNet flow does not itself employ CQF.  Since there are no per-flow
   parameters in the CQF technique, per-hop configuration is not
   required in the CQF forwarding nodes.

7.  Example Application on DetNet IP Network



   This section provides an example application of the timing model
   presented in this document to control the admission of a DetNet flow
   on a DetNet-enabled IP network.  Consider Figure 5, taken from
   Section 3 of [RFC8939], which shows a simple IP network:

   *  End system 1 implements Guaranteed Service [RFC2212], as in
      Section 6.5, between itself and relay node 1.

   *  Sub-network 1 is a TSN network.  The nodes in sub-network 1
      implement credit-based shapers with asynchronous traffic shaping,
      as in Section 6.4.

   *  Sub-network 2 is a TSN network.  The nodes in sub-network 2
      implement Cyclic Queuing and Forwarding with two buffers, as in
      Section 6.6.

   *  The relay nodes 1 and 2 implement credit-based shapers with
      asynchronous traffic shaping, as in Section 6.4.  They also
      perform the aggregation and mapping of IP DetNet flows to TSN
      streams (Section 4.4 of [RFC9023]).

    DetNet IP       Relay                        Relay       DetNet IP
    End System      Node 1                       Node 2      End System
        1                                                        2
   +----------+                                             +----------+
   |   Appl.  |<------------ End-to-End Service ----------->|   Appl.  |
   +----------+  ............                 ...........   +----------+
   | Service  |<-: Service  :-- DetNet flow --: Service  :->| Service  |
   +----------+  +----------+                 +----------+  +----------+
   |Forwarding|  |Forwarding|                 |Forwarding|  |Forwarding|
   +--------.-+  +-.------.-+                 +-.---.----+  +-------.--+
            : Link :       \      ,-----.      /     \   ,-----.   /
            +......+        +----[  Sub- ]----+       +-[  Sub- ]-+
                                 [Network]              [Network]
                                  `--1--'                `--2--'

            |<--------------------- DetNet IP --------------------->|

   |<--- d1 --->|<--------------- d2_p --------------->|<-- d3_p -->|

     Figure 5: A Simple DetNet-Enabled IP Network, Taken from RFC 8939

   Consider a fully centralized control plane for the network of
   Figure 5, as described in Section 3.2 of [DETNET-CONTROL-PLANE].
   Suppose end system 1 wants to create a DetNet flow with a traffic
   specification destined to end system 2 with end-to-end delay bound
   requirement D.  Therefore, the control plane receives a flow
   establishment request and calculates a number of valid paths through
   the network (Section 3.2 of [DETNET-CONTROL-PLANE]).  To select a
   proper path, the control plane needs to compute an end-to-end delay
   bound at every node of each selected path p.

   The end-to-end delay bound is d1 + d2_p + d3_p, where d1 is the delay
   bound from end system 1 to the entrance of relay node 1, d2_p is the
   delay bound for path p from relay node 1 to the entrance of the first
   node in sub-network 2, and d3_p is the delay bound of path p from the
   first node in sub-network 2 to end system 2.  The computation of d1
   is explained in Section 6.5.  Since the relay node 1, sub-network 1,
   and relay node 2 implement aggregate queuing, we use the results in
   Sections 4.2.2 and 6.4 to compute d2_p for the path p.  Finally, d3_p
   is computed using the delay bound computation of Section 6.6.  Any
   path p, such that d1 + d2_p + d3_p <= D, satisfies the delay bound
   requirement of the flow.  If there is no such path, the control plane
   may compute a new set of valid paths and redo the delay bound
   computation or reject the DetNet flow.

   As soon as the control plane selects a path that satisfies the delay
   bound constraint, it allocates and reserves the resources in the path
   for the DetNet flow (Section 4.2 of [DETNET-CONTROL-PLANE]).

8.  Security Considerations



   Detailed security considerations for DetNet are cataloged in
   [RFC9055], and more general security considerations are described in
   [RFC8655].

   Security aspects that are unique to DetNet are those whose aim is to
   provide the specific QoS aspects of DetNet, specifically bounded end-
   to-end delivery latency and zero congestion loss.  Achieving such
   loss rates and bounded latency may not be possible in the face of a
   highly capable adversary, such as the one envisioned by the Internet
   Threat Model of BCP 72 [RFC3552], which can arbitrarily drop or delay
   any or all traffic.  In order to present meaningful security
   considerations, we consider a somewhat weaker attacker who does not
   control the physical links of the DetNet domain but may have the
   ability to control or change the behavior of some resources within
   the boundary of the DetNet domain.

   Latency bound calculations use parameters that reflect physical
   quantities.  If an attacker finds a way to change the physical
   quantities, unknown to the control and management planes, the latency
   calculations fail and may result in latency violation and/or
   congestion losses.  An example of such attacks is to make some
   traffic sources under the control of the attacker send more traffic
   than their assumed T-SPECs.  This type of attack is typically avoided
   by ingress conditioning at the edge of a DetNet domain.  However, it
   must be insured that such ingress conditioning is done per flow and
   that the buffers are segregated such that if one flow exceeds its
   T-SPEC, it does not cause buffer overflow for other flows.

   Some queuing mechanisms require time synchronization and operate
   correctly only if the time synchronization works correctly.  In the
   case of CQF, the correct alignments of cycles can fail if an attack
   against time synchronization fools a node into having an incorrect
   offset.  Some of these attacks can be prevented by cryptographic
   authentication as in Annex K of [IEEE1588] for the Precision Time
   Protocol (PTP).  However, the attacks that change the physical
   latency of the links used by the time synchronization protocol are
   still possible even if the time synchronization protocol is protected
   by authentication and cryptography [DelayAttack].  Such attacks can
   be detected only by their effects on latency bound violations and
   congestion losses, which do not occur in normal DetNet operation.

9.  IANA considerations



   This document has no IANA actions.

10.  References



10.1.  Normative References



   [IEEE8021Q]
              IEEE, "IEEE Standard for Local and Metropolitan Area
              Networks--Bridges and Bridged Networks", IEEE Std 802.1Q-
              2018, DOI 10.1109/IEEESTD.2018.8403927, July 2018,
              <https://ieeexplore.ieee.org/document/8403927>.

   [RFC2212]  Shenker, S., Partridge, C., and R. Guerin, "Specification
              of Guaranteed Quality of Service", RFC 2212,
              DOI 10.17487/RFC2212, September 1997,
              <https://www.rfc-editor.org/info/rfc2212>.

   [RFC2475]  Blake, S., Black, D., Carlson, M., Davies, E., Wang, Z.,
              and W. Weiss, "An Architecture for Differentiated
              Services", RFC 2475, DOI 10.17487/RFC2475, December 1998,
              <https://www.rfc-editor.org/info/rfc2475>.

   [RFC6658]  Bryant, S., Ed., Martini, L., Swallow, G., and A. Malis,
              "Packet Pseudowire Encapsulation over an MPLS PSN",
              RFC 6658, DOI 10.17487/RFC6658, July 2012,
              <https://www.rfc-editor.org/info/rfc6658>.

   [RFC7806]  Baker, F. and R. Pan, "On Queuing, Marking, and Dropping",
              RFC 7806, DOI 10.17487/RFC7806, April 2016,
              <https://www.rfc-editor.org/info/rfc7806>.

   [RFC8655]  Finn, N., Thubert, P., Varga, B., and J. Farkas,
              "Deterministic Networking Architecture", RFC 8655,
              DOI 10.17487/RFC8655, October 2019,
              <https://www.rfc-editor.org/info/rfc8655>.

   [RFC8939]  Varga, B., Ed., Farkas, J., Berger, L., Fedyk, D., and S.
              Bryant, "Deterministic Networking (DetNet) Data Plane:
              IP", RFC 8939, DOI 10.17487/RFC8939, November 2020,
              <https://www.rfc-editor.org/info/rfc8939>.

   [RFC8964]  Varga, B., Ed., Farkas, J., Berger, L., Malis, A., Bryant,
              S., and J. Korhonen, "Deterministic Networking (DetNet)
              Data Plane: MPLS", RFC 8964, DOI 10.17487/RFC8964, January
              2021, <https://www.rfc-editor.org/info/rfc8964>.

   [RFC9016]  Varga, B., Farkas, J., Cummings, R., Jiang, Y., and D.
              Fedyk, "Flow and Service Information Model for
              Deterministic Networking (DetNet)", RFC 9016,
              DOI 10.17487/RFC9016, March 2021,
              <https://www.rfc-editor.org/info/rfc9016>.

10.2.  Informative References



   [BennettDelay]
              Bennett, J. C. R., Benson, K., Charny, A., Courtney, W.
              F., and J.-Y. Le Boudec, "Delay jitter bounds and packet
              scale rate guarantee for expedited forwarding",
              DOI 10.1109/TNET.2002.801404, August 2002,
              <https://dl.acm.org/citation.cfm?id=581870>.

   [CharnyDelay]
              Charny, A. and J.-Y. Le Boudec, "Delay Bounds in a Network
              with Aggregate Scheduling", DOI 10.1007/3-540-39939-9_1,
              September 2002, <https://link.springer.com/
              chapter/10.1007/3-540-39939-9_1>.

   [DelayAttack]
              Barreto, S., Suresh, A., and J. L. Boudec, "Cyber-attack
              on packet-based time synchronization protocols: The
              undetectable Delay Box", DOI 10.1109/I2MTC.2016.7520408,
              May 2016, <https://ieeexplore.ieee.org/document/7520408>.

   [DETNET-CONTROL-PLANE]
              Malis, A., Geng, A., Ed., Chen, M., Qin, F., and B. Varga,
              "Deterministic Networking (DetNet) Controller Plane
              Framework", Work in Progress, Internet-Draft, draft-ietf-
              detnet-controller-plane-framework-02, 28 June 2022,
              <https://datatracker.ietf.org/doc/html/draft-ietf-detnet-
              controller-plane-framework-02>.

   [IEEE1588] IEEE, "IEEE Standard for a Precision Clock Synchronization
              Protocol for Networked Measurement and Control Systems",
              IEEE Std 1588-2008, DOI 10.1109/IEEESTD.2008.4579760, July
              2008, <https://ieeexplore.ieee.org/document/4579760>.

   [IEEE8021Qcr]
              IEEE 802.1, "802.1Qcr-2020 - IEEE Standard for Local and
              Metropolitan Area Networks--Bridges and Bridged Networks
              Amendment 34:Asynchronous Traffic Shaping", November 2020,
              <https://ieeexplore.ieee.org/document/9253013>.

   [IEEE8021TSN]
              IEEE 802.1, "802.1 Time-Sensitive Networking (TSN) Task
              Group", <https://1.ieee802.org/tsn/>.

   [IEEE8023] IEEE, "IEEE Standard for Ethernet", IEEE Std 802.3-2018,
              DOI 10.1109/IEEESTD.2018.8457469, August 2018,
              <http://ieeexplore.ieee.org/document/8457469>.

   [LeBoudecTheory]
              Le Boudec, J.-Y., "A Theory of Traffic Regulators for
              Deterministic Networks With Application to Interleaved
              Regulators", DOI 10.1109/TNET.2018.2875191, November 2018,
              <https://ieeexplore.ieee.org/document/8519761>.

   [NetCalBook]
              Le Boudec, J.-Y. and P. Thiran, "Network Calculus: A
              Theory of Deterministic Queuing Systems for the Internet",
              Springer Science & Business Media, vol. 2050, 2001,
              <https://leboudec.github.io/netcal/latex/netCalBook.pdf>.

   [PacketReorderingBounds]
              Mohammadpour, E. and J.-Y. Le Boudec, "On Packet
              Reordering in Time-Sensitive Networks",
              DOI 10.1109/TNET.2021.3129590, December 2021,
              <https://ieeexplore.ieee.org/document/9640523>.

   [RFC2697]  Heinanen, J. and R. Guerin, "A Single Rate Three Color
              Marker", RFC 2697, DOI 10.17487/RFC2697, September 1999,
              <https://www.rfc-editor.org/info/rfc2697>.

   [RFC3552]  Rescorla, E. and B. Korver, "Guidelines for Writing RFC
              Text on Security Considerations", BCP 72, RFC 3552,
              DOI 10.17487/RFC3552, July 2003,
              <https://www.rfc-editor.org/info/rfc3552>.

   [RFC8578]  Grossman, E., Ed., "Deterministic Networking Use Cases",
              RFC 8578, DOI 10.17487/RFC8578, May 2019,
              <https://www.rfc-editor.org/info/rfc8578>.

   [RFC9023]  Varga, B., Ed., Farkas, J., Malis, A., and S. Bryant,
              "Deterministic Networking (DetNet) Data Plane: IP over
              IEEE 802.1 Time-Sensitive Networking (TSN)", RFC 9023,
              DOI 10.17487/RFC9023, June 2021,
              <https://www.rfc-editor.org/info/rfc9023>.

   [RFC9055]  Grossman, E., Ed., Mizrahi, T., and A. Hacker,
              "Deterministic Networking (DetNet) Security
              Considerations", RFC 9055, DOI 10.17487/RFC9055, June
              2021, <https://www.rfc-editor.org/info/rfc9055>.

   [Sch8021Qbv]
              Craciunas, S., Oliver, R., Chmelik, M., and W. Steiner,
              "Scheduling Real-Time Communication in IEEE 802.1Qbv Time
              Sensitive Networks", DOI 10.1145/2997465.2997470, October
              2016, <https://dl.acm.org/doi/10.1145/2997465.2997470>.

   [SpechtUBS]
              Specht, J. and S. Samii, "Urgency-Based Scheduler for
              Time-Sensitive Switched Ethernet Networks",
              DOI 10.1109/ECRTS.2016.27, July 2016,
              <https://ieeexplore.ieee.org/abstract/document/7557870>.

   [ThomasTime]
              Thomas, L. and J.-Y. Le Boudec, "On Time Synchronization
              Issues in Time-Sensitive Networks with Regulators and
              Nonideal Clocks", DOI 10.1145/3393691.339420, June 2020,
              <https://dl.acm.org/doi/10.1145/3393691.3394206>.

   [TSNwithATS]
              Mohammadpour, E., Stai, E., Mohiuddin, M., and J.-Y. Le
              Boudec, "Latency and Backlog Bounds in Time-Sensitive
              Networking with Credit Based Shapers and Asynchronous
              Traffic Shaping", DOI 10.1109/ITC30.2018.10053, September
              2018, <https://ieeexplore.ieee.org/document/8493026>.

Acknowledgments

   We would like to thank Lou Berger, Tony Przygienda, John Scudder,
   Watson Ladd, Yoshifumi Nishida, Ralf Weber, Robert Sparks, Gyan
   Mishra, Martin Duke, Éric Vyncke, Lars Eggert, Roman Danyliw, and
   Paul Wouters for their useful feedback on this document.

Contributors

   RFC 7322 limits the number of authors listed on the front page to a
   maximum of 5.  The editor wishes to thank and acknowledge the
   following author for contributing text to this document:

   Janos Farkas
   Ericsson
   Email: janos.farkas@ericsson.com


Authors' Addresses



   Norman Finn
   Huawei Technologies Co. Ltd
   3101 Rio Way
   Spring Valley, California 91977
   United States of America
   Phone: +1 925 980 6430
   Email: nfinn@nfinnconsulting.com


   Jean-Yves Le Boudec
   EPFL
   IC Station 14
   CH-1015 Lausanne
   Switzerland
   Email: jean-yves.leboudec@epfl.ch


   Ehsan Mohammadpour
   EPFL
   IC Station 14
   CH-1015 Lausanne
   Switzerland
   Email: ehsan.mohammadpour@epfl.ch


   Jiayi Zhang
   Huawei Technologies Co. Ltd
   Q27, No.156 Beiqing Road
   Beijing
   100095
   China
   Email: zhangjiayi11@huawei.com


   Balázs Varga
   Ericsson
   Budapest
   Konyves Kálmán krt. 11/B
   1097
   Hungary
   Email: balazs.a.varga@ericsson.com