RFC 1944
This document is obsolete. Please refer to RFC 2544.

Network Working Group                                         S. Bradner
Request for Comments: 1944                            Harvard University
Category: Informational                                       J. McQuaid
                                                            Bay Networks
                                                                May 1996

       Benchmarking Methodology for Network Interconnect Devices

Status of This Memo

   This memo provides information for the Internet community.  This memo
   does not specify an Internet standard of any kind.  Distribution of
   this memo is unlimited.


   This document discusses and defines a number of tests that may be
   used to describe the performance characteristics of a network
   interconnecting  device.  In addition to defining the tests this
   document also describes specific formats for reporting the results of
   the tests.  Appendix A lists the tests and conditions that we believe
   should be included for specific cases and gives additional
   information about testing practices.  Appendix B is a reference
   listing of maximum frame rates to be used with specific frame sizes
   on various media and Appendix C gives some examples of frame formats
   to be used in testing.

1. Introduction

   Vendors often engage in "specsmanship" in an attempt to give their
   products a better position in the marketplace.  This often involves
   "smoke & mirrors" to confuse the potential users of the products.

   This document defines a specific set of tests that vendors can use to
   measure and report the performance characteristics of network
   devices.  The results of these tests will provide the user comparable
   data from different vendors with which to evaluate these devices.

   A previous document, "Benchmarking Terminology for Network
   Interconnect Devices" (RFC 1242), defined many of the terms that are
   used in this document.  The terminology document should be consulted
   before attempting to make use of this document.

Bradner & McQuaid            Informational                      [Page 1]

RFC 1944                Benchmarking Methodology                May 1996

2. Real world

   In producing this document the authors attempted to keep in mind the
   requirement that apparatus to perform the described tests must
   actually be built.  We do not know of "off the shelf" equipment
   available to implement all of the tests but it is our opinion that
   such equipment can be constructed.

3. Tests to be run

   There are a number of tests described in this document.  Not all of
   the tests apply to all types of devices under test (DUTs). Vendors
   should perform all of the tests that can be supported by a specific
   type of product.  The authors understand that it will take a
   considerable period of time to perform all of the recommended tests
   nder  all of the recommended conditions. We believe that the results
   are worth the effort.  Appendix A lists some of the tests and
   conditions that we believe should be included for specific cases.

4. Evaluating the results

   Performing all of the recommended tests will result in a great deal
   of data. Much of this data will not apply to the evaluation of the
   devices under each circumstance.  For example, the rate at which a
   router forwards IPX frames will be of little use in selecting a
   router for an environment that does not (and will not) support that
   protocol.  Evaluating even that data which is relevant to a
   particular network installation will require experience which may not
   be readily available. Furthermore, selection of the tests to be run
   and evaluation of the test data must be done with an understanding of
   generally accepted testing practices regarding repeatability,
   variance and statistical significance of small numbers of trials.

5. Requirements

   In this document, the words that are used to define the significance
   of each particular requirement are capitalized. These words are:

    * "MUST" This word, or the words "REQUIRED" and "SHALL" mean that
   the item is an absolute requirement of the specification.

    * "SHOULD" This word or the adjective "RECOMMENDED" means that there
   may exist valid reasons in particular circumstances to ignore this
   item, but the full implications should be understood and the case
   carefully weighed before choosing a different course.

    * "MAY" This word or the adjective "OPTIONAL" means that this item
   is truly optional.  One vendor may choose to include the item because

Bradner & McQuaid            Informational                      [Page 2]

RFC 1944                Benchmarking Methodology                May 1996

   a particular marketplace requires it or because it enhances the
   product, for example; another vendor may omit the same item.

   An implementation is not compliant if it fails to satisfy one or more
   of the MUST requirements for the protocols it implements.  An
   implementation that satisfies all the MUST and all the SHOULD
   requirements for its protocols is said to be "unconditionally
   compliant"; one that satisfies all the MUST requirements but not all
   the SHOULD requirements for its protocols is said to be
   "conditionally compliant".

6. Test set up

   The ideal way to implement this series of tests is to use a tester
   with both transmitting and receiving ports.  Connections are made
   from the sending ports of the tester to the receiving ports of the
   DUT and from the sending ports of the DUT back to the tester. (see
   Figure 1)  Since the tester both sends the test traffic and receives
   it back, after the traffic has been forwarded but the DUT, the tester
   can easily determine if all of the transmitted packets were received
   and verify that the correct packets were received.  The same
   functionality can be obtained with separate transmitting and
   receiving devices (see Figure 2) but unless they are remotely
   controlled by some computer in a way that simulates the single
   tester, the labor required to accurately perform some of the tests
   (particularly the throughput test) can be prohibitive.

                            |            |
               +------------|  tester    |<-------------+
               |            |            |              |
               |            +------------+              |
               |                                        |
               |            +------------+              |
               |            |            |              |
               +----------->|    DUT     |--------------+
                            |            |
                              Figure 1

         +--------+         +------------+          +----------+
         |        |         |            |          |          |
         | sender |-------->|    DUT     |--------->| receiver |
         |        |         |            |          |          |
         +--------+         +------------+          +----------+
                              Figure 2

Bradner & McQuaid            Informational                      [Page 3]

RFC 1944                Benchmarking Methodology                May 1996

6.1 Test set up for multiple media types

   Two different setups could be used to test a DUT which is used in
   real-world networks to connect networks of differing media type,
   local Ethernet to a backbone FDDI ring for example.  The tester could
   support both media types in which case the set up shown in Figure 1
   would be used.

   Two identical DUTs are used in the other test set up. (see Figure 3)
   In many cases this set up may more accurately simulate the real
   world.  For example, connecting two LANs together with a WAN link or
   high speed backbone.  This set up would not be as good at simulating
   a system where clients on a Ethernet LAN were interacting with a
   server on an FDDI backbone.

                               |           |
         +---------------------|  tester   |<---------------------+
         |                     |           |                      |
         |                     +-----------+                      |
         |                                                        |
         |        +----------+               +----------+         |
         |        |          |               |          |         |
         +------->|  DUT 1   |-------------->|   DUT 2  |---------+
                  |          |               |          |
                  +----------+               +----------+

                                  Figure 3

7. DUT set up

   Before starting to perform the tests, the DUT to be tested MUST be
   configured following the instructions provided to the user.
   Specifically, it is expected that all of the supported protocols will
   be configured and enabled during this set up (See Appendix A).  It is
   expected that all of the tests will be run without changing the
   configuration or setup of the DUT in any way other than that required
   to do the specific test.  For example, it is not acceptable to change
   the size of frame handling buffers between tests of frame handling
   rates or to disable all but one transport protocol when testing the
   throughput of that protocol.  It is necessary to modify the
   configuration when starting a test to determine the effect of filters
   on throughput, but the only change MUST be to enable the specific
   filter. The DUT set up SHOULD include the normally recommended
   routing update intervals and keep alive frequency.  The specific
   version of the software and the exact DUT configuration, including
   what functions are disabled, used during the tests MUST be included
   as part of the report of the results.

Bradner & McQuaid            Informational                      [Page 4]

RFC 1944                Benchmarking Methodology                May 1996

8. Frame formats

   The formats of the test frames to use for TCP/IP over Ethernet are
   shown in Appendix C: Test Frame Formats.  These exact frame formats
   SHOULD be used in the tests described in this document for this
   protocol/media combination and that these frames will be used as a
   template for testing other protocol/media combinations.  The specific
   formats that are used to define the test frames for a particular test
   series MUST be included in the report of the results.

9. Frame sizes

   All of the described tests SHOULD be performed at a number of frame
   sizes. Specifically, the sizes SHOULD include the maximum and minimum
   legitimate sizes for the protocol under test on the media under test
   and enough sizes in between to be able to get a full characterization
   of the DUT performance.  Except where noted, at least five frame
   sizes SHOULD be tested for each test condition.

   Theoretically the minimum size UDP Echo request frame would consist
   of an IP header (minimum length 20 octets), a UDP header (8 octets)
   and whatever MAC level header is required by the media in use.  The
   theoretical maximum frame size is determined by the size of the
   length field in the IP header.  In almost all cases the actual
   maximum and minimum sizes are determined by the limitations of the

   In theory it would be ideal to distribute the frame sizes in a way
   that would evenly distribute the theoretical frame rates.  These
   recommendations incorporate this theory but specify frame sizes which
   are easy to understand and remember.  In addition, many of the same
   frame sizes are specified on each of the media types to allow for
   easy performance comparisons.

   Note: The inclusion of an unrealistically small frame size on some of
   the media types (i.e. with little or no space for data) is to help
   characterize the per-frame processing overhead of the DUT.

   9.1 Frame sizes to be used on Ethernet

       64, 128, 256, 512, 1024, 1280, 1518

      These sizes include the maximum and minimum frame sizes permitted
      by the Ethernet standard and a selection of sizes between these
      extremes with a finer granularity for the smaller frame sizes and
      higher frame rates.

Bradner & McQuaid            Informational                      [Page 5]

RFC 1944                Benchmarking Methodology                May 1996

   9.2 Frame sizes to be used on 4Mb and 16Mb token ring

       54, 64, 128, 256, 1024, 1518, 2048, 4472

      The frame size recommendations for token ring assume that there is
      no RIF field in the frames of routed protocols.  A RIF field would
      be present in any direct source route bridge performance test.
      The minimum size frame for UDP on token ring is 54 octets.  The
      maximum size of 4472 octets is recommended for 16Mb token ring
      instead of the theoretical size of 17.9Kb because of the size
      limitations imposed by many token ring interfaces.  The reminder
      of the sizes are selected to permit direct comparisons with other
      types of media.  An IP (i.e. not UDP) frame may be used in
      addition if a higher data rate is desired, in which case the
      minimum frame size is 46 octets.

   9.3 Frame sizes to be used on FDDI

       54, 64, 128, 256, 1024, 1518, 2048, 4472

      The minimum size frame for UDP on FDDI is 53 octets, the minimum
      size of 54 is recommended to allow direct comparison to token ring
      performance.  The maximum size of 4472 is recommended instead of
      the theoretical maximum size of 4500 octets to permit the same
      type of comparison. An IP (i.e. not UDP) frame may be used in
      addition if a higher data rate is desired, in which case the
      minimum frame size is 45 octets.

   9.4 Frame sizes in the presence of disparate MTUs

      When the interconnect DUT supports connecting links with disparate
      MTUs, the frame sizes for the link with the *larger* MTU SHOULD be
      used, up to the limit of the protocol being tested. If the
      interconnect DUT does not support the fragmenting of frames in the
      presence of MTU mismatch, the forwarding rate for that frame size
      shall be reported as zero.

      For example, the test of IP forwarding with a bridge or router
      that joins FDDI and Ethernet should use the frame sizes of FDDI
      when going from the FDDI to the Ethernet link. If the bridge does
      not support IP fragmentation, the forwarding rate for those frames
      too large for Ethernet should be reported as zero.

10. Verifying received frames

   The test equipment SHOULD discard any frames received during a test
   run that are not actual forwarded test frames.  For example, keep-
   alive and routing update frames SHOULD NOT be included in the count

Bradner & McQuaid            Informational                      [Page 6]

RFC 1944                Benchmarking Methodology                May 1996

   of received frames.  In any case, the test equipment SHOULD verify
   the length of the received frames and check that they match the
   expected length.

   Preferably, the test equipment SHOULD include sequence numbers in the
   transmitted frames and check for these numbers on the received
   frames.  If this is done, the reported results SHOULD include in
   addition to the number of frames dropped, the number of frames that
   were received out of order, the number of duplicate frames received
   and the number of gaps in the received frame numbering sequence.
   This functionality is required for some of the described tests.

11. Modifiers

   It might be useful to know the DUT performance under a number of
   conditions; some of these conditions are noted below.  The reported
   results SHOULD include as many of these conditions as the test
   equipment is able to generate.  The suite of tests SHOULD be first
   run without any modifying conditions and then repeated under each of
   the conditions separately.  To preserve the ability to compare the
   results of these tests any frames that are required to generate the
   modifying conditions (management queries for example) will be
   included in the same data stream as the normal test frames in place
   of one of the test frames and not be supplied to the DUT on a
   separate network port.

   11.1 Broadcast frames

      In most router designs special processing is required when frames
      addressed to the hardware broadcast address are received.  In
      bridges (or in bridge mode on routers) these broadcast frames must
      be flooded to a number of ports.  The stream of test frames SHOULD
      be augmented with 1% frames addressed to the hardware broadcast
      address.  The frames sent to the broadcast address should be of a
      type that the router will not need to process.  The aim of this
      test is to determine if there is any effect on the forwarding rate
      of the other data in the stream.  The specific frames that should
      be used are included in the test frame format document. The
      broadcast frames SHOULD be evenly distributed throughout the data
      stream, for example, every 100th frame.

      The same test SHOULD be performed on bridge-like DUTs but in this
      case the broadcast packets will be processed and flooded to all

      It is understood that a level of broadcast frames of 1% is much
      higher than many networks experience but, as in drug toxicity
      evaluations, the higher level is required to be able to gage the

Bradner & McQuaid            Informational                      [Page 7]

RFC 1944                Benchmarking Methodology                May 1996

      effect which would otherwise often fall within the normal
      variability of the system performance.  Due to design factors some
      test equipment will not be able to generate a level of alternate
      frames this low.  In these cases the percentage SHOULD be as small
      as the equipment can provide and that the actual level be
      described in the report of the test results.

   11.2 Management frames

      Most data networks now make use of management protocols such as
      SNMP.  In many environments there can be a number of management
      stations sending queries to the same DUT at the same time.

      The stream of test frames SHOULD be augmented with one management
      query as the first frame sent each second during the duration of
      the trial.  The result of the query must fit into one response
      frame. The response frame SHOULD be verified by the test
      equipment. One example of the specific query frame that should be
      used is shown in Appendix C.

   11.3 Routing update frames

      The processing of dynamic routing protocol updates could have a
      significant impact on the ability of a router to forward data
      frames.  The stream of test frames SHOULD be augmented with one
      routing update frame transmitted as the first frame transmitted
      during the trial.  Routing update frames SHOULD be sent at the
      rate specified in Appendix C for the specific routing protocol
      being used in the test. Two routing update frames are defined in
      Appendix C for the TCP/IP over Ethernet example.  The routing
      frames are designed to change the routing to a number of networks
      that are not involved in the forwarding of the test data.  The
      first frame sets the routing table state to "A", the second one
      changes the state to "B".  The frames MUST be alternated during
      the trial.

      The test SHOULD verify that the routing update was processed by
      the DUT.

   11.4 Filters

      Filters are added to routers and bridges to selectively inhibit
      the forwarding of frames that would normally be forwarded.  This
      is usually done to implement security controls on the data that is
      accepted between one area and another. Different products have
      different capabilities to implement filters.

Bradner & McQuaid            Informational                      [Page 8]

RFC 1944                Benchmarking Methodology                May 1996

      The DUT SHOULD be first configured to add one filter condition and
      the tests performed.  This filter SHOULD permit the forwarding of
      the test data stream. In routers this filter SHOULD be of the

       forward input_protocol_address to output_protocol_address

      In bridges the filter SHOULD be of the form:

       forward destination_hardware_address

      The DUT SHOULD be then reconfigured to implement a total of 25
      filters.  The first 24 of these filters SHOULD be of the form:

       block input_protocol_address to output_protocol_address

      The 24 input and output protocol addresses SHOULD not be any that
      are represented in the test data stream.  The last filter SHOULD
      permit the forwarding of the test data stream.  By "first" and
      "last" we mean to ensure that in the second case, 25 conditions
      must be checked before the data frames will match the conditions
      that permit the forwarding of the frame. Of course, if the DUT
      reorders the filters or does not use a linear scan of the filter
      rules the effect of the sequence in which the filters are input is
      properly lost.

      The exact filters configuration command lines used SHOULD be
      included with the report of the results.

      11.4.1 Filter Addresses

         Two sets of filter addresses are required, one for the single
         filter case and one for the 25 filter case.

         The single filter case should permit traffic from IP address to IP address and deny all other

         The 25 filter case should follow the following sequence.

          deny aa.ba.1.1 to aa.ba.100.1
          deny aa.ba.2.2 to aa.ba.101.2
          deny aa.ba.3.3 to aa.ba.103.3
          deny aa.ba.12.12 to aa.ba.112.12
          allow aa.bc.1.2 to aa.bc.65.1
          deny aa.ba.13.13 to aa.ba.113.13
          deny aa.ba.14.14 to aa.ba.114.14

Bradner & McQuaid            Informational                      [Page 9]

RFC 1944                Benchmarking Methodology                May 1996

          deny aa.ba.24.24 to aa.ba.124.24
          deny all else

         All previous filter conditions should be cleared from the
         router before this sequence is entered.  The sequence is
         selected to test to see if the router sorts the filter
         conditions or accepts them in the order that they were entered.
         Both of these procedures will result in a greater impact on
         performance than will some form of hash coding.

12. Protocol addresses

   It is easier to implement these tests using a single logical stream
   of  data, with one source protocol address and one destination
   protocol address, and for some conditions like the filters described
   above, a practical requirement. Networks in the real world are not
   limited to single streams of data. The test suite SHOULD be first run
   with a single protocol (or hardware for bridge tests) source and
   destination address pair.  The tests SHOULD then be repeated with
   using a random destination address.  While testing routers the
   addresses SHOULD be random and uniformly distributed over a range of
   256 networks and random and uniformly distributed over the full MAC
   range for bridges.  The specific address ranges to use for IP are
   shown in Appendix C.

13. Route Set Up

   It is not reasonable that all of the routing information necessary to
   forward the test stream, especially in the multiple address case,
   will be manually set up.  At the start of each trial a routing update
   MUST be sent to the DUT. This routing update MUST include all of the
   network addresses that will be required for the trial.  All of the
   addresses SHOULD resolve to the same "next-hop". Normally this will
   be the address of the receiving side of the test equipment. This
   routing update will have to be repeated at the interval required by
   the routing protocol being used.  An example of the format and
   repetition interval of the update frames is given in Appendix C.

14. Bidirectional traffic

   Normal network activity is not all in a single direction.  To test
   the bidirectional performance of a DUT, the test series SHOULD be run
   with the same data rate being offered from each direction. The sum of
   the data rates should not exceed the theoretical limit for the media.

Bradner & McQuaid            Informational                     [Page 10]

RFC 1944                Benchmarking Methodology                May 1996

15. Single stream path

   The full suite of tests SHOULD be run along with whatever modifier
   conditions that are relevant using a single input and output network
   port on the DUT. If the internal design of the DUT has multiple
   distinct pathways, for example, multiple interface cards each with
   multiple network ports, then all possible types of pathways SHOULD be
   tested separately.

16. Multi-port

   Many current router and bridge products provide many network ports in
   the same module. In performing these tests first half of the ports
   are designated as "input ports" and half are designated as "output
   ports".  These ports SHOULD be evenly distributed across the DUT
   architecture. For example if a DUT has two interface cards each of
   which has four ports, two ports on each interface card are designated
   as input and two are designated as output.  The specified tests are
   run using the same data rate being offered to each of the input
   ports.  The addresses in the input data streams SHOULD be set so that
   a frame will be directed to each of the output ports in sequence so
   that all "output" ports will get an even distribution of packets from
   this input.  The same configuration MAY be used to perform a
   bidirectional multi-stream test.  In this case all of the ports are
   considered both input and output ports and each data stream MUST
   consist of frames addressed to all of the other ports.

   Consider the following 6 port DUT:

                     ---------| in A  out X|--------
                     ---------| in B  out Y|--------
                     ---------| in C  out Z|--------

   The addressing of the data streams for each of the inputs SHOULD be:

    stream sent to input A:
      packet to out X, packet to out Y, packet to out Z
    stream sent to input B:
      packet to out X, packet to out Y, packet to out Z
    stream sent to input C
      packet to out X, packet to out Y, packet to out Z

   Note that these streams each follow the same sequence so that 3
   packets will arrive at output X at the same time, then 3 packets at
   Y, then 3 packets at Z. This procedure ensures that, as in the real
   world, the DUT will have to deal with multiple packets addressed to

Bradner & McQuaid            Informational                     [Page 11]

RFC 1944                Benchmarking Methodology                May 1996

   the same output at the same time.

17. Multiple protocols

   This document does not address the issue of testing the effects of a
   mixed protocol environment other than to suggest that if such tests
   are wanted then frames SHOULD be distributed between all of the test
   protocols.  The distribution MAY approximate the conditions on the
   network in which the DUT would be used.

18. Multiple frame sizes

   This document does not address the issue of testing the effects of a
   mixed frame size environment other than to suggest that if such tests
   are wanted then frames SHOULD be distributed between all of the
   listed sizes for the protocol under test.  The distribution MAY
   approximate the conditions on the network in which the DUT would be
   used. The authors do not have any idea how the results of such a test
   would be interpreted other than to directly compare multiple DUTs in
   some very specific simulated network.

19. Testing performance beyond a single DUT.

   In the performance testing of a single DUT, the paradigm can be
   described as applying some input to a DUT and monitoring the output.
   The results of which can be used to form a basis of characterization
   of that device under those test conditions.

   This model is useful when the test input and output are homogenous
   (e.g., 64-byte IP, 802.3 frames into the DUT; 64 byte IP, 802.3
   frames out), or the method of test can distinguish between dissimilar
   input/output. (E.g., 1518 byte IP, 802.3 frames in; 576 byte,
   fragmented IP, X.25 frames out.)

   By extending the single DUT test model, reasonable benchmarks
   regarding multiple DUTs or heterogeneous environments may be
   collected. In this extension, the single DUT is replaced by a system
   of interconnected network DUTs. This test methodology would support
   the benchmarking of a variety of device/media/service/protocol
   combinations. For example, a configuration for a LAN-to-WAN-to-LAN
   test might be:

   (1) 802.3-> DUT 1 -> X.25 @ 64kbps -> DUT 2 -> 802.3

   Or a mixed LAN configuration might be:

   (2) 802.3 -> DUT 1 -> FDDI -> DUT 2 -> FDDI -> DUT 3 -> 802.3

Bradner & McQuaid            Informational                     [Page 12]

RFC 1944                Benchmarking Methodology                May 1996

   In both examples 1 and 2, end-to-end benchmarks of each system could
   be empirically ascertained. Other behavior may be characterized
   through the use of intermediate devices. In example 2, the
   configuration may be used to give an indication of the FDDI to FDDI
   capability exhibited by DUT 2.

   Because multiple DUTs are treated as a single system, there are
   limitations to this methodology. For instance, this methodology may
   yield an aggregate benchmark for a tested system. That benchmark
   alone, however, may not necessarily reflect asymmetries in behavior
   between the DUTs, latencies introduce by other apparatus (e.g.,
   CSUs/DSUs, switches), etc.

   Further, care must be used when comparing benchmarks of different
   systems by ensuring that the DUTs' features/configuration of the
   tested systems have the appropriate common denominators to allow

20. Maximum frame rate

   The maximum frame rates that should be used when testing LAN
   connections SHOULD be the listed theoretical maximum rate for the
   frame size on the media.

   The maximum frame rate that should be used when testing WAN
   connections SHOULD be greater than the listed theoretical maximum
   rate for the frame size on that speed connection.  The higher rate
   for WAN tests is to compensate for the fact that some vendors employ
   various forms of header compression.

   A list of maximum frame rates for LAN connections is included in
   Appendix B.

21. Bursty traffic

   It is convenient to measure the DUT performance under steady state
   load but this is an unrealistic way to gauge the functioning of a DUT
   since actual network traffic normally consists of bursts of frames.
   Some of the tests described below SHOULD be performed with both
   steady state traffic and with traffic consisting of repeated bursts
   of frames.  The frames within a burst are transmitted with the
   minimum legitimate inter-frame gap.

   The objective of the test is to determine the minimum interval
   between bursts which the DUT can process with no frame loss. During
   each test the number of frames in each burst is held constant and the
   inter-burst interval varied.  Tests SHOULD be run with burst sizes of
   16, 64, 256 and 1024 frames.

Bradner & McQuaid            Informational                     [Page 13]

RFC 1944                Benchmarking Methodology                May 1996

22. Frames per token

   Although it is possible to configure some token ring and FDDI
   interfaces to transmit more than one frame each time that the token
   is received, most of the network devices currently available transmit
   only one frame per token.  These tests SHOULD first be performed
   while transmitting only one frame per token.

   Some current high-performance workstation servers do transmit more
   than one frame per token on FDDI to maximize throughput.  Since this
   may be a common feature in future workstations and servers,
   interconnect devices with FDDI interfaces SHOULD be tested with 1, 4,
   8, and 16 frames per token.  The reported frame rate SHOULD be the
   average rate of frame transmission over the total trial period.

23. Trial description

   A particular test consists of multiple trials.  Each trial returns
   one piece of information, for example the loss rate at a particular
   input frame rate.  Each trial consists of a number of phases:

    a) If the DUT is a router, send the routing update to the "input"
   port and pause two seconds to be sure that the routing has settled.

    b)  Send the "learning frames" to the "output" port and wait 2
   seconds to be sure that the learning has settled.  Bridge learning
   frames are frames with source addresses that are the same as the
   destination addresses used by the test frames.  Learning frames for
   other protocols are used to prime the address resolution tables in
   the DUT.  The formats of the learning frame that should be used are
   shown in the Test Frame Formats document.

    c) Run the test trial.

    d) Wait for two seconds for any residual frames to be received.

    e) Wait for at least five seconds for the DUT to restabilize.

24. Trial duration

   The aim of these tests is to determine the rate continuously
   supportable by the DUT.  The actual duration of the test trials must
   be a compromise between this aim and the duration of the benchmarking
   test suite.  The duration of the test portion of each trial SHOULD be
   at least 60 seconds.  The tests that involve some form of "binary
   search", for example the throughput test, to determine the exact
   result MAY use a shorter trial duration to minimize the length of the
   search procedure, but  the final determination SHOULD be made with

Bradner & McQuaid            Informational                     [Page 14]

RFC 1944                Benchmarking Methodology                May 1996

   full length trials.

25. Address resolution

   The DUT SHOULD be able to respond to address resolution requests sent
   by the DUT wherever the protocol requires such a process.

26. Benchmarking tests:

   Note: The notation "type of data stream" refers to the above
   modifications to a frame stream with a constant inter-frame gap, for
   example, the addition of traffic filters to the configuration of the

   26.1 Throughput

      To determine the DUT throughput as defined in RFC 1242.

      Send a specific number of frames at a specific rate through the
      DUT and then count the frames that are transmitted by the DUT. If
      the count of offered frames is equal to the count of received
      frames, the rate of the offered stream is raised and the test
      rerun.  If fewer frames are received than were transmitted, the
      rate of the offered stream is reduced and the test is rerun.

      The throughput is the fastest rate at which the count of test
      frames transmitted by the DUT is equal to the number of test
      frames sent to it by the test equipment.

      Reporting format:
      The results of the throughput test SHOULD be reported in the form
      of a graph. If it is, the x coordinate SHOULD be the frame size,
      the y coordinate SHOULD be the frame rate.  There SHOULD be at
      least two lines on the graph.  There SHOULD be one line showing
      the theoretical frame rate for the media at the various frame
      sizes.  The second line SHOULD be the plot of the test results.
      Additional lines MAY be used on the graph to report the results
      for each type of data stream tested.  Text accompanying the graph
      SHOULD indicate the protocol, data stream format, and type of
      media used in the tests.

      We assume that if a single value is desired for advertising
      purposes the vendor will select the rate for the minimum frame
      size for the media. If this is done then the figure MUST be
      expressed in frames per second.  The rate MAY also be expressed in
      bits (or bytes) per second if the vendor so desires.  The

Bradner & McQuaid            Informational                     [Page 15]

RFC 1944                Benchmarking Methodology                May 1996

      statement of performance MUST include a/ the measured maximum
      frame rate, b/ the size of the frame used, c/ the theoretical
      limit of the media for that frame size, and d/ the type of
      protocol used in the test.  Even if a single value is used as part
      of the advertising copy, the full table of results SHOULD be
      included in the product data sheet.

   26.2 Latency

      To determine the latency as defined in RFC 1242.

      First determine the throughput for DUT at each of the listed frame
      sizes. Send a stream of frames at a particular frame size through
      the DUT at the determined throughput rate to a specific
      destination.  The stream SHOULD be at least 120 seconds in
      duration.  An identifying tag SHOULD be included in one frame
      after 60 seconds with the type of tag being implementation
      dependent. The time at which this frame is fully transmitted is
      recorded (timestamp A).  The receiver logic in the test equipment
      MUST recognize the tag information in the frame stream and record
      the time at which the tagged frame was received (timestamp B).

      The latency is timestamp B minus timestamp A as per the relevant
      definition frm RFC 1242, namely latency as defined for store and
      forward devices or latency as defined for bit forwarding devices.

      The test MUST be repeated at least 20 times with the reported
      value being the average of the recorded values.

      This test SHOULD be performed with the test frame addressed to the
      same destination as the rest of the data stream and also with each
      of the test frames addressed to a new destination network.

      Reporting format:
      The report MUST state which definition of latency (from RFC 1242)
      was used for this test.  The latency results SHOULD be reported
      in the format of a table with a row for each of the tested frame
      sizes.  There SHOULD be columns for the frame size, the rate at
      which the latency test was run for that frame size, for the media
      types tested, and for the resultant latency values for each
      type of data stream tested.

Bradner & McQuaid            Informational                     [Page 16]

RFC 1944                Benchmarking Methodology                May 1996

   26.3 Frame loss rate

      To determine the frame loss rate, as defined in RFC 1242, of a DUT
      throughout the entire range of input data rates and frame sizes.

      Send a specific number of frames at a specific rate through the
      DUT to be tested and count the frames that are transmitted by the
      DUT.   The frame loss rate at each point is calculated using the
      following equation:

       ( ( input_count - output_count ) * 100 ) / input_count

      The first trial SHOULD be run for the frame rate that corresponds
      to 100% of the maximum rate for the frame size on the input media.
      Repeat the procedure for the rate that corresponds to 90% of the
      maximum rate used and then for 80% of this rate.  This sequence
      SHOULD be continued (at reducing 10% intervals) until there are
      two successive trials in which no frames are lost. The maximum
      granularity of the trials MUST be 10% of the maximum rate, a finer
      granularity is encouraged.

      Reporting format:
      The results of the frame loss rate test SHOULD be plotted as a
      graph.  If this is done then the X axis MUST be the input frame
      rate as a percent of the theoretical rate for the media at the
      specific frame size. The Y axis MUST be the percent loss at the
      particular input rate.  The left end of the X axis and the bottom
      of the Y axis MUST be 0 percent; the right end of the X axis and
      the top of the Y axis MUST be 100 percent.  Multiple lines on the
      graph MAY used to report the frame loss rate for different frame
      sizes, protocols, and types of data streams.

      Note: See section 18 for the maximum frame rates that SHOULD be

   26.4 Back-to-back frames

      To characterize the ability of a DUT to process back-to-back
      frames as defined in RFC 1242.

      Send a burst of frames with minimum inter-frame gaps to the DUT
      and count the number of frames forwarded by the DUT.  If the count
      of transmitted frames is equal to the number of frames forwarded
      the length of the burst is increased and the test is rerun.  If

Bradner & McQuaid            Informational                     [Page 17]

RFC 1944                Benchmarking Methodology                May 1996

      the number of forwarded frames is less than the number
      transmitted, the length of the burst is reduced and the test is

      The back-to-back value is the number of frames in the longest
      burst that the DUT will handle without the loss of any frames.
      The trial length MUST be at least 2 seconds and SHOULD be
      repeated at least 50 times with the average of the recorded values
      being reported.

      Reporting format:
      The back-to-back results SHOULD be reported in the format of a
      table with a row for each of the tested frame sizes.  There SHOULD
      be columns for the frame size and for the resultant average frame
      count for each type of data stream tested.  The standard deviation
      for each measurement MAY also be reported.

   26.5 System recovery

      To characterize the speed at which a DUT recovers from an overload

      First determine the throughput for a DUT at each of the listed
      frame sizes.

      Send a stream of frames at a rate 110% of the recorded throughput
      rate or the maximum rate for the media, whichever is lower, for at
      least 60 seconds.  At Timestamp A reduce the frame rate to 50% of
      the above rate and record the time of the last frame lost
      (Timestamp B). The system recovery time is determined by
      subtracting Timestamp B from Timestamp A.  The test SHOULD be
      repeated a number of times and the average of the recorded values
      being reported.

      Reporting format:
      The system recovery results SHOULD be reported in the format of a
      table with a row for each of the tested frame sizes.  There SHOULD
      be columns for the frame size, the frame rate used as the
      throughput rate for each type of data stream tested, and for the
      measured recovery time for each type of data stream tested.

   26.6 Reset

      To characterize the speed at which a DUT recovers from a device or
      software reset.

Bradner & McQuaid            Informational                     [Page 18]

RFC 1944                Benchmarking Methodology                May 1996

      First determine the throughput for the DUT for the minimum frame
      size on the media used in the testing.

      Send a continuous stream of frames at the determined throughput
      rate for the minimum sized frames. Cause a reset in the DUT.
      Monitor the output until frames begin to be forwarded and record
      the time that the last frame (Timestamp A) of the initial stream
      and the first frame of the new stream (Timestamp B) are received.
      A power interruption reset test is performed as above except that
      the power to the DUT should be interrupted for 10 seconds in place
      of causing a reset.

      This test SHOULD only be run using frames addressed to networks
      directly connected to the DUT so that there is no requirement to
      delay until a routing update is received.

      The reset value is obtained by subtracting Timestamp A from
      Timestamp B.

      Hardware and software resets, as well as a power interruption
      SHOULD be tested.

      Reporting format:
      The reset value SHOULD be reported in a simple set of statements,
      one for each reset type.

27. Security Considerations

   Security issues are not addressed in this document.

Bradner & McQuaid            Informational                     [Page 19]

RFC 1944                Benchmarking Methodology                May 1996

28. Editors' Addresses

   Scott Bradner
   Harvard University
   1350 Mass. Ave, room 813
   Cambridge, MA 02138

   Phone +1 617 495-3864
   Fax +1 617 496-8500
   EMail: sob@harvard.edu

   Jim McQuaid
   Bay Networks
   3 Federal Street
   Billerica, MA 01821

   Phone +1 508 436-3915
   Fax: +1 508 670-8145
   EMail: jmcquaid@baynetworks.com

Bradner & McQuaid            Informational                     [Page 20]

RFC 1944                Benchmarking Methodology                May 1996

Appendix A: Testing Considerations

A.1 Scope Of This Appendix

   This appendix discusses certain issues in the benchmarking
   methodology where experience or judgment may play a role in the tests
   selected to be run or in the approach to constructing the test with a
   particular DUT.  As such, this appendix MUST not be read as an
   amendment to the methodology described in the body of this document
   but as a guide to testing practice.

   1. Typical testing practice has been to enable all protocols to be
      tested and conduct all testing with no further configuration of
      protocols, even though a given set of trials may exercise only one
      protocol at a time. This minimizes the opportunities to "tune" a
      DUT for a single protocol.

   2. The least common denominator of the available filter functions
      should be used to ensure that there is a basis for comparison
      between vendors. Because of product differences, those conducting
      and evaluating tests must make a judgment about this issue.

   3. Architectural considerations may need to be considered.  For
      example, first perform the tests with the stream going between
      ports on the same interface card and the repeat the tests with the
      stream going into a port on one interface card and out of a port
      on a second interface card. There will almost always be a best
      case and worst case configuration for a given DUT architecture.

   4. Testing done using traffic streams consisting of mixed protocols
      has not shown much difference between testing with individual
      protocols.  That is, if protocol A testing and protocol B testing
      give two different performance results, mixed protocol testing
      appears to give a result which is the average of the two.

   5. Wide Area Network (WAN) performance may be tested by setting up
      two identical devices connected by the appropriate short- haul
      versions of the WAN modems.  Performance is then measured between
      a LAN interface on one DUT to a LAN interface on the other DUT.

   The maximum frame rate to be used for LAN-WAN-LAN configurations is a
   judgment that can be based on known characteristics of the overall
   system including compression effects, fragmentation, and gross link
   speeds. Practice suggests that the rate should be at least 110% of
   the slowest link speed. Substantive issues of testing compression
   itself are beyond the scope of this document.

Bradner & McQuaid            Informational                     [Page 21]

RFC 1944                Benchmarking Methodology                May 1996

Appendix B: Maximum frame rates reference

   (Provided by Roger Beeman, Cisco Systems)

     Size       Ethernet    16Mb Token Ring      FDDI
    (bytes)       (pps)           (pps)         (pps)

    64            14880           24691         152439
    128            8445           13793          85616
    256            4528            7326          45620
    512            2349            3780          23585
    768            1586            2547          15903
    1024           1197            1921          11996
    1280            961            1542           9630
    1518            812            1302           8138

   Ethernet size
    Preamble 64 bits
    Frame 8 x N bits
    Gap  96 bits

   16Mb Token Ring size
      SD               8 bits
      AC               8 bits
      FC               8 bits
      DA              48 bits
      SA              48 bits
      RI              48 bits ( 06 30 00 12 00 30 )
        DSAP           8 bits
        SSAP           8 bits
        Control        8 bits
        Vendor        24 bits
        Type          16 bits
      Data 8 x ( N - 18) bits
      FCS             32 bits
      ED               8 bits
      FS               8 bits

   Tokens or idles between packets are not included

   FDDI size
      Preamble        64 bits
      SD               8 bits
      FC               8 bits
      DA              48 bits
      SA              48 bits

Bradner & McQuaid            Informational                     [Page 22]

RFC 1944                Benchmarking Methodology                May 1996

        DSAP           8 bits
        SSAP           8 bits
        Control        8 bits
        Vendor        24 bits
        Type          16 bits
      Data 8 x ( N - 18) bits
      FCS             32 bits
      ED               4 bits
      FS              12 bits

Appendix C: Test Frame Formats

   This appendix defines the frame formats that may be used with these
   tests.  It also includes protocol specific parameters for TCP/IP over
   Ethernet to be used with the tests as an example.

C.1. Introduction

   The general logic used in the selection of the parameters and the
   design of the frame formats is explained for each case within the
   TCP/IP section.  The same logic has been used in the other sections.
   Comments are used in these sections only if there is a protocol
   specific feature to be explained.  Parameters and frame formats for
   additional protocols can be defined by the reader by using the same

C.2. TCP/IP Information

   The following section deals with the TCP/IP protocol suite.

   C.2.1 Frame Type.
      An application level datagram echo request is used for the test
      data frame in the protocols that support such a function.  A
      datagram protocol is used to minimize the chance that a router
      might expect a specific session initialization sequence, as might
      be the case for a reliable stream protocol. A specific defined
      protocol is used because some routers verify the protocol field
      and refuse to forward unknown protocols.

      For TCP/IP a UDP Echo Request is used.

   C.2.2 Protocol Addresses
      Two sets of addresses must be defined: first the addresses
      assigned to the router ports, and second the address that are to
      be used in the frames themselves and in the routing updates.

      The network addresses through are have
      been assigned to the BMWG by the IANA for this purpose.  This

Bradner & McQuaid            Informational                     [Page 23]

RFC 1944                Benchmarking Methodology                May 1996

      assignment was made to minimize the chance of conflict in case a
      testing device were to be accidentally connected to part of the
      Internet.  The specific use of the addresses is detailed below.

      C.2.2.1 Router port protocol addresses
         Half of the ports on a multi-port router are referred to as
         "input" ports and the other half as "output" ports even though
         some of the tests use all ports both as input and output.  A
         contiguous series of IP Class C network addresses from to have been assigned for use on the
         "input" ports.  A second series from to
         have been assigned for use on the "output" ports. In all cases
         the router port is node 1 on the appropriate network.  For
         example, a two port DUT would have an IP address of
         on one port and on the other port.

         Some of the tests described in the methodology memo make use of
         an SNMP management connection to the DUT.  The management
         access address for the DUT is assumed to be the first of the
         "input" ports (

      C.2.2.2 Frame addresses
         Some of the described tests assume adjacent network routing
         (the reboot time test for example).  The IP address used in the
         test frame is that of node 2 on the appropriate Class C
         network. ( for example)

         If the test involves non-adjacent network routing the phantom
         routers are located at node 10 of each of the appropriate Class
         C networks.  A series of Class C network addresses from to has been assigned for use as the
         networks accessible through the phantom routers on the "input"
         side of DUT.  The series of Class C networks from
         to have been assigned to be used as the networks
         visible through the phantom routers on the "output" side of the

   C.2.3 Routing Update Frequency

      The update interval for each routing protocol is may have to be
      determined by the specifications of the individual protocol.  For
      IP RIP, Cisco IGRP and for OSPF a routing update frame or frames
      should precede each stream of test frames by 5 seconds.  This
      frequency is sufficient for trial durations of up to 60 seconds.
      Routing updates must be mixed with the stream of test frames if
      longer trial periods are selected.  The frequency of updates
      should be taken from the following table.

Bradner & McQuaid            Informational                     [Page 24]

RFC 1944                Benchmarking Methodology                May 1996

       IP-RIP  30 sec
       IGRP  90 sec
       OSPF  90 sec

   C.2.4 Frame Formats - detailed discussion

      C.2.4.1 Learning Frame
         In most protocols a procedure is used to determine the mapping
         between the protocol node address and the MAC address.  The
         Address Resolution Protocol (ARP) is used to perform this
         function in TCP/IP.  No such procedure is required in XNS or
         IPX because the MAC address is used as the protocol node

         In the ideal case the tester would be able to respond to ARP
         requests from the DUT.  In cases where this is not possible an
         ARP request should be sent to the router's "output" port.  This
         request should be seen as coming from the immediate destination
         of the test frame stream. (i.e. the phantom router (Figure 2)
         or the end node if adjacent network routing is being used.) It
         is assumed that the router will cache the MAC address of the
         requesting device.  The ARP request should be sent 5 seconds
         before the test frame stream starts in each trial.  Trial
         lengths of longer than 50 seconds may require that the router
         be configured for an extended ARP timeout.

                      +--------+            +------------+
                      |        |            |  phantom   |------ P LAN
            IN A------|   DUT  |------------|            |------ P LAN
                      |        |   OUT A    |  router    |------ P LAN
                      +--------+            +------------+

                                 Figure 2

           In the case where full routing is being used

      C.2.4.2 Routing Update Frame
         If the test does not involve adjacent net routing the tester
         must supply proper routing information using a routing update.
         A single routing update is used before each trial on each
         "destination" port (see section C.24).  This update includes
         the network addresses that are reachable through a phantom
         router on the network attached to the port.  For a full mesh
         test, one destination network address is present in the routing
         update for each of the "input" ports.  The test stream on each

Bradner & McQuaid            Informational                     [Page 25]

RFC 1944                Benchmarking Methodology                May 1996

         "input" port consists of a repeating sequence of frames, one to
         each of the "output" ports.

      C.2.4.3 Management Query Frame
         The management overhead test uses SNMP to query a set of
         variables that should be present in all DUTs that support SNMP.
         The variables for a single interface only are read by an NMS
         at the appropriate intervals.  The list of variables to
         retrieve follow:


      C.2.4.4 Test Frames
         The test frame is an UDP Echo Request with enough data to fill
         out the required frame size.  The data should not be all bits
         off or all bits on since these patters can cause a "bit
         stuffing" process to be used to maintain clock synchronization
         on WAN links.  This process will result in a longer frame than
         was intended.

      C.2.4.5 Frame Formats - TCP/IP on Ethernet
         Each of the frames below are described for the 1st pair of DUT
         ports, i.e. "input" port #1 and "output" port #1.  Addresses
         must be changed if the frame is to be used for other ports.

      C.2.6.1 Learning Frame

          ARP Request on Ethernet

          -- DATAGRAM HEADER
          offset data (hex)            description
          00     FF FF FF FF FF FF     dest MAC address send to
         broadcast address
          06     xx xx xx xx xx xx     set to source MAC address
          12     08 06                 ARP type
          14     00 01                 hardware type Ethernet = 1
          16     08 00                 protocol type IP = 800
          18     06                    hardware address length 48 bits
         on Ethernet
          19     04                    protocol address length 4 octets
         for IP
          20     00 01                 opcode request = 1
          22     xx xx xx xx xx xx     source MAC address
          28     xx xx xx xx           source IP address

Bradner & McQuaid            Informational                     [Page 26]

RFC 1944                Benchmarking Methodology                May 1996

          32     FF FF FF FF FF FF     requesting DUT's MAC address
          38     xx xx xx xx           DUT's IP address

      C.2.6.2 Routing Update Frame

          -- DATAGRAM HEADER
          offset data (hex)            description
          00     FF FF FF FF FF FF     dest MAC address is broadcast
          06     xx xx xx xx xx xx     source hardware address
          12     08 00                 type

          -- IP HEADER
          14     45                    IP version - 4, header length (4
         byte units) - 5
          15     00                    service field
          16     00 EE                 total length
          18     00 00                 ID
          20     40 00                 flags (3 bits) 4 (do not
                                       fragment offset-0
          22     0A                    TTL
          23     11                    protocol - 17 (UDP)
          24     C4 8D                 header checksum
          26     xx xx xx xx           source IP address
          30     xx xx xx              destination IP address
          33     FF                    host part = FF for broadcast

          -- UDP HEADER
          34     02 08                 source port 208 = RIP
          36     02 08                 destination port 208 = RIP
          38     00 DA                 UDP message length
          40     00 00                 UDP checksum

          -- RIP packet
          42     02                  command = response
          43     01                  version = 1
          44     00 00               0

          -- net 1
          46     00 02               family = IP
          48     00 00               0
          50     xx xx xx            net 1 IP address
          53     00                  net not node
          54     00 00 00 00         0
          58     00 00 00 00         0
          62     00 00 00 07         metric 7

          -- net 2

Bradner & McQuaid            Informational                     [Page 27]

RFC 1944                Benchmarking Methodology                May 1996

          66     00 02               family = IP
          68     00 00               0
          70     xx xx xx            net 2 IP address
          73     00                  net not node
          74     00 00 00 00         0
          78     00 00 00 00         0
          82     00 00 00 07         metric 7

          -- net 3
          86     00 02               family = IP
          88     00 00               0
          90     xx xx xx            net 3 IP address
          93     00                  net not node
          94     00 00 00 00         0
          98     00 00 00 00         0
          102    00 00 00 07         metric 7

          -- net 4
          106    00 02               family = IP
          108    00 00               0
          110    xx xx xx            net 4 IP address
          113    00                  net not node
          114    00 00 00 00         0
          118    00 00 00 00         0
          122    00 00 00 07         metric 7

          -- net 5
          126    00 02               family = IP
          128    00 00               0
          130    00                  net 5 IP address
          133    00                  net not node
          134    00 00 00 00         0
          138    00 00 00 00         0
          142    00 00 00 07         metric 7

          -- net 6
          146    00 02               family = IP
          148    00 00               0
          150    xx xx xx            net 6 IP address
          153    00                  net not node
          154    00 00 00 00         0
          158    00 00 00 00         0
          162    00 00 00 07         metric 7

      C.2.4.6 Management Query Frame

         To be defined.

Bradner & McQuaid            Informational                     [Page 28]

RFC 1944                Benchmarking Methodology                May 1996

      C.2.6.4 Test Frames
              UDP echo request on Ethernet

          -- DATAGRAM HEADER
          offset data (hex)            description
          00     xx xx xx xx xx xx     set to dest MAC address
          06     xx xx xx xx xx xx     set to source MAC address
          12     08 00                 type

          -- IP HEADER
          14     45                    IP version - 4 header length 5 4
         byte units
          15     00                    TOS
          16     00 2E                 total length*
          18     00 00                 ID
          20     00 00                 flags (3 bits) - 0 fragment
          22     0A                    TTL
          23     11                    protocol - 17 (UDP)
          24     C4 8D                 header checksum*
          26     xx xx xx xx           set to source IP address**
          30     xx xx xx xx           set to destination IP address**

          -- UDP HEADER
          34     C0 20                 source port
          36     00 07                 destination port 07 = Echo
          38     00 1A                 UDP message length*
          40     00 00                 UDP checksum

          -- UDP DATA
          42     00 01 02 03 04 05 06 07    some data***
          50     08 09 0A 0B 0C 0D 0E 0F

         * - change for different length frames

         ** - change for different logical streams

         *** - fill remainder of frame with incrementing octets,
         repeated if required by frame length

Bradner & McQuaid            Informational                     [Page 29]

RFC 1944                Benchmarking Methodology                May 1996

   Values to be used in Total Length and UDP message length fields:

          frame size   total length  UDP message length
             64            00 2E          00 1A
             128           00 6E          00 5A
             256           00 EE          00 9A
             512           01 EE          01 9A
             768           02 EE          02 9A
             1024          03 EE          03 9A
             1280          04 EE          04 9A
             1518          05 DC          05 C8

Bradner & McQuaid            Informational                     [Page 30]