Benchmarking Methodology
charter-ietf-bmwg-06

The Benchmarking Methodology Working Group (BMWG) will continue to
produce a series of recommendations concerning the key performance
characteristics of internetworking technologies, or benchmarks for
network devices, systems, and services. Taking a view of networking
divided into planes, the scope of work includes benchmarks for the
management, control, and forwarding planes.

Each recommendation will describe the class of equipment, system, or
service being addressed; discuss the performance characteristics that
are pertinent to that class; clearly identify a set of metrics that aid
in the description of those characteristics; specify the methodologies
required to collect said metrics; and lastly, present the requirements
for the common, unambiguous reporting of benchmarking results.

The set of relevant benchmarks will be developed with input from the
community of users (e.g, network operators and testing organizations)
and from those affected by the benchmarks when they are published
(networking and test equipment manufacturers). When possible, the
benchmarks and other terminology will be developed jointly with
organizations that are willing to share their expertise. Joint review
requirements for a specific work area will be included in the detailed
description of the task, as listed below.

To better distinguish the BMWG from other measurement initiatives in the
IETF, the scope of the BMWG is limited to the characterization of
implementations of various internetworking technologies
using controlled stimuli in a laboratory environment. Said differently,
the BMWG does not attempt to produce benchmarks for live, operational
networks. Moreover, the benchmarks produced by this WG shall strive to
be vendor independent or otherwise have universal applicability to a
given technology class.

Because the demands of a particular technology may vary from deployment
to deployment, a specific non-goal of the Working Group is to define
acceptance criteria or performance requirements.

An ongoing task is to provide a forum for discussion regarding the
advancement of measurements designed to provide insight on the
capabilities and operation of inter-networking technology
implementations.

The BMWG will communicate with the operations community through
organizations such as NANOG, RIPE, and APRICOT.

In addition to its current work plan, the BMWG is explicitly tasked to
develop benchmarks and methodologies for the following technologies:

• BGP Control-plane Convergence Methodology (Terminology is complete):
  With relevant performance characteristics identified, BMWG will prepare
  a Benchmarking Methodology Document with review from the Routing Area
  (e.g., the IDR working group and/or the RTG-DIR). The Benchmarking
  Methodology will be Last-Called in all the groups that previously
  provided input, including another round of network operator input during
  the last call.

• SIP Networking Devices: Develop new terminology and methods to
  characterize the key performance aspects of network devices using
  SIP, including the signaling plane scale and service rates while
  considering load conditions on both the signaling and media planes. This
  work will be harmonized with related SIP performance metric definitions
  prepared by the PMOL working group.

• Flow Export and Collection: Develop terminology and methods to
  characterize network devices flow monitoring, export, and collection.
  The goal is a methodology to assess the maximum IP flow rate that a
  network device can sustain without losing any IP flow information or
  compromising the accuracy of information exported on the IP flows,
  and to asses the forwarding plane performance (if the forwarding
  function is present) in the presence of Flow Monitoring.

• Data Center Bridging Devices:
  Some key concepts from BMWG's past work are not meaningful when testing
  switches that implement new IEEE specifications in the area of data
  center bridging. For example, throughput as defined in RFC 1242 cannot
  be measured when testing devices that implement three new IEEE
  specifications: priority-based flow control (802.1Qbb); priority groups
  (802.1Qaz); and congestion notification (802.1Qau).
  Since devices that implement these new congestion-management
  specifications should never drop frames, and since the metric of
  throughput distinguishes between non-zero and zero drop rates, no
  throughput measurement is possible using the existing methodology.
  The current emphasis is on the Priority Flow Control aspects of
  Data Center Bridging, and the work will include an investigation
  into whether TRILL RBridges require any specific treatment in the
  methodology. This work will update RFC 2544 and exchange periodic
  Liaisons with IEEE 802.1 DCB Task Group, especially at WG Last Call.

• Content Aware Devices:
  New classes of network devices that operate above the IP layer of the
  network stack require a new methodology to perform adequate
  benchmarking. Existing BMWG RFCs (RFC2647 and RFC3511) provides useful
  measurement and performance statistics, though they may not reflect the
  actual performance of the device when deployed in production networks.
  Operating within the limitations of the charter, namely blackbox
  characterization in laboratory environments, the BMWG will develop a
  methodology that more closely relates the performance of these devices
  to performance in an operational setting. In order to confirm or
  identify key performance characteristics, BMWG will solicit input from
  operations groups such as NANOG, RIP and APRICOT.

• LDP Dataplane Convergence:
  In order to identify key LDP convergence performance characteristics,
  BMWG will solicit input from operations groups such as NANOG, RIP and
  APRICOT. When relevant performance characteristics have been identified,
  BMWG will jointly prepare a Benchmarking Terminology Document with the
  Routing Area (e.g., the MPLS working group and or the RTG-DIR), which
  would define metrics relevant to LDP convergence. The Benchmark
  definition document would be Last-Called in all the working groups that
  produced it, and solicit operator input during the last call. The work
  will then continue in BMWG to define the test methodology, with input
  and review from the aforementioned parties.