1 CHARACTERIZE VSWITCH PERFORMANCE FOR TELCO NFV USE CASES LEVEL TEST DESIGN
2 ==========================================================================
4 .. contents:: Table of Contents
9 The objective of the OPNFV project titled
10 **“Characterize vSwitch Performance for Telco NFV Use Cases”**, is to
11 evaluate a virtual switch to identify its suitability for a Telco
12 Network Function Virtualization (NFV) environment. The intention of this
13 Level Test Design (LTD) document is to specify the set of tests to carry
14 out in order to objectively measure the current characteristics of a
15 virtual switch in the Network Function Virtualization Infrastructure
16 (NFVI) as well as the test pass criteria. The detailed test cases will
17 be defined in `Section 2 <#DetailsOfTheLevelTestDesign>`__, preceded by
18 the `Document identifier <#DocId>`__ and the `Scope <#Scope>`__.
20 This document is currently in draft form.
22 1.1. Document identifier
23 ------------------------
25 The document id will be used to uniquely
26 identify versions of the LTD. The format for the document id will be:
27 OPNFV\_vswitchperf\_LTD\_ver\_NUM\_MONTH\_YEAR\_STATUS, where by the
28 status is one of: draft, reviewed, corrected or final. The document id
29 for this version of the LTD is:
30 OPNFV\_vswitchperf\_LTD\_ver\_1.6\_Jan\_15\_DRAFT.
35 The main purpose of this project is to specify a suite of
36 performance tests in order to objectively measure the current packet
37 transfer characteristics of a virtual switch in the NFVI. The intent of
38 the project is to facilitate testing of any virtual switch. Thus, a
39 generic suite of tests shall be developed, with no hard dependencies to
40 a single implementation. In addition, the test case suite shall be
41 architecture independent.
43 The test cases developed in this project shall not form part of a
44 separate test framework, all of these tests may be inserted into the
45 Continuous Integration Test Framework and/or the Platform Functionality
46 Test Framework - if a vSwitch becomes a standard component of an OPNFV
52 * `RFC 1242 Benchmarking Terminology for Network Interconnection
53 Devices <http://www.ietf.org/rfc/rfc1242.txt>`__
54 * `RFC 2544 Benchmarking Methodology for Network Interconnect
55 Devices <http://www.ietf.org/rfc/rfc2544.txt>`__
56 * `RFC 2285 Benchmarking Terminology for LAN Switching
57 Devices <http://www.ietf.org/rfc/rfc2285.txt>`__
58 * `RFC 2889 Benchmarking Methodology for LAN Switching
59 Devices <http://www.ietf.org/rfc/rfc2889.txt>`__
60 * `RFC 3918 Methodology for IP Multicast
61 Benchmarking <http://www.ietf.org/rfc/rfc3918.txt>`__
62 * `RFC 4737 Packet Reordering
63 Metrics <http://www.ietf.org/rfc/rfc4737.txt>`__
64 * `RFC 5481 Packet Delay Variation Applicability
65 Statement <http://www.ietf.org/rfc/rfc5481.txt>`__
66 * `RFC 6201 Device Reset
67 Characterization <http://tools.ietf.org/html/rfc6201>`__
69 2. Details of the Level Test Design
70 ===================================
72 This section describes the features to be tested (`cf. 2.1
73 <#FeaturesToBeTested>`__), the test approach (`cf. 2.2 <#Approach>`__);
74 it also identifies the sets of test cases or scenarios (`cf. 2.3
75 <#TestIdentification>`__) along with the pass/fail criteria (`cf. 2.4
76 <#PassFail>`__) and the test deliverables (`cf. 2.5 <#TestDeliverables>`__).
78 2.1. Features to be tested
79 --------------------------
81 Characterizing virtual switches (i.e. Device Under Test (DUT) in this document)
82 includes measuring the following performance metrics:
84 - **Throughput** as defined by `RFC1242
85 <https://www.rfc-editor.org/rfc/rfc1242.txt>`__: The maximum rate at which
86 **none** of the offered frames are dropped by the DUT. The maximum frame
87 rate and bit rate that can be transmitted by the DUT without any error
88 should be recorded. Note there is an equivalent bit rate and a specific
89 layer at which the payloads contribute to the bits. Errors and
90 improperly formed frames or packets are dropped.
91 - **Packet delay** introduced by the DUT and its cumulative effect on
92 E2E networks. Frame delay can be measured equivalently.
93 - **Packet delay variation**: measured from the perspective of the
94 VNF/application. Packet delay variation is sometimes called "jitter".
95 However, we will avoid the term "jitter" as the term holds different
96 meaning to different groups of people. In this document we will
97 simply use the term packet delay variation. The preferred form for this
98 metric is the PDV form of delay variation defined in `RFC5481
99 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__.
100 - **Packet loss** (within a configured waiting time at the receiver): All
101 packets sent to the DUT should be accounted for.
102 - **Burst behaviour**: measures the ability of the DUT to buffer packets.
103 - **Packet re-ordering**: measures the ability of the device under test to
104 maintain sending order throughout transfer to the destination.
105 - **Packet correctness**: packets or Frames must be well-formed, in that
106 they include all required fields, conform to length requirements, pass
107 integrity checks, etc.
108 - **Availability and capacity** of the DUT i.e. when the DUT is fully “up”
111 - Includes power consumption of the CPU (in various power states) and
113 - Includes CPU utilization.
114 - Includes the number of NIC interfaces supported.
115 - Includes headroom of VM workload processing cores (i.e. available
122 In order to determine the packet transfer characteristics of a virtual
123 switch, the tests will be broken down into the following categories:
125 2.2.1 Test Categories
126 ----------------------
127 - **Throughput Tests** to measure the maximum forwarding rate (in
128 frames per second or fps) and bit rate (in Mbps) for a constant load
129 (as defined by `RFC1242 <https://www.rfc-editor.org/rfc/rfc1242.txt>`__)
130 without traffic loss.
131 - **Packet and Frame Delay Tests** to measure average, min and max
132 packet and frame delay for constant loads.
133 - **Stream Performance Tests** (TCP, UDP) to measure bulk data transfer
134 performance, i.e. how fast systems can send and receive data through
136 - **Request/Response Performance** Tests (TCP, UDP) the measure the
137 transaction rate through the switch.
138 - **Packet Delay Tests** to understand latency distribution for
139 different packet sizes and over an extended test run to uncover
141 - **Scalability Tests** to understand how the virtual switch performs
142 as the number of flows, active ports, complexity of the forwarding
143 logic's configuration... it has to deal with increases.
144 - **Control Path and Datapath Coupling** Tests, to understand how
145 closely coupled the datapath and the control path are as well as the
146 effect of this coupling on the performance of the DUT.
147 - **CPU and Memory Consumption Tests** to understand the virtual
148 switch’s footprint on the system, this includes:
153 * Time To Establish Flows Tests.
155 - **Noisy Neighbour Tests**, to understand the effects of resource
156 sharing on the performance of a virtual switch.
158 **Note:** some of the tests above can be conducted simultaneously where
159 the combined results would be insightful, for example Packet/Frame Delay
162 2.2.2 Deployment Scenarios
163 --------------------------
164 The following represents possible deployments which can help to
165 determine the performance of both the virtual switch and the datapath
168 Physical port → vSwitch → physical port
169 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
170 .. code-block:: console
173 +--------------------------------------------------+ |
174 | +--------------------+ | |
177 | +--------------+ +--------------+ | |
178 | | phy port | vSwitch | phy port | | |
179 +---+--------------+------------+--------------+---+ _|
183 +--------------------------------------------------+
185 | traffic generator |
187 +--------------------------------------------------+
190 Physical port → vSwitch → VNF → vSwitch → physical port
191 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
192 .. code-block:: console
195 +---------------------------------------------------+ |
197 | +-------------------------------------------+ | |
198 | | Application | | |
199 | +-------------------------------------------+ | |
203 | +---------------+ +---------------+ | |
204 | | logical port 0| | logical port 1| | |
205 +---+---------------+-----------+---------------+---+ _|
209 +---+---------------+----------+---------------+---+ |
210 | | logical port 0| | logical port 1| | |
211 | +---------------+ +---------------+ | |
215 | +--------------+ +--------------+ | |
216 | | phy port | vSwitch | phy port | | |
217 +---+--------------+------------+--------------+---+ _|
221 +--------------------------------------------------+
223 | traffic generator |
225 +--------------------------------------------------+
228 Physical port → vSwitch → VNF → vSwitch → VNF → vSwitch → physical port
229 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
231 .. code-block:: console
234 +----------------------+ +----------------------+ |
235 | Guest 1 | | Guest 2 | |
236 | +---------------+ | | +---------------+ | |
237 | | Application | | | | Application | | |
238 | +---------------+ | | +---------------+ | |
240 | | v | | | v | | Guests
241 | +---------------+ | | +---------------+ | |
242 | | logical ports | | | | logical ports | | |
243 | | 0 1 | | | | 0 1 | | |
244 +---+---------------+--+ +---+---------------+--+ _|
248 +---+---------------+---------+---------------+--+ |
249 | | 0 1 | | 3 4 | | |
250 | | logical ports | | logical ports | | |
251 | +---------------+ +---------------+ | |
253 | | L-----------------+ v | |
254 | +--------------+ +--------------+ | |
255 | | phy ports | vSwitch | phy ports | | |
256 +---+--------------+----------+--------------+---+ _|
260 +--------------------------------------------------+
262 | traffic generator |
264 +--------------------------------------------------+
267 Physical port → vSwitch → VNF
268 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
270 .. code-block:: console
273 +---------------------------------------------------+ |
275 | +-------------------------------------------+ | |
276 | | Application | | |
277 | +-------------------------------------------+ | |
281 | +---------------+ | |
282 | | logical port 0| | |
283 +---+---------------+-------------------------------+ _|
287 +---+---------------+------------------------------+ |
288 | | logical port 0| | |
289 | +---------------+ | |
293 | +--------------+ | |
294 | | phy port | vSwitch | |
295 +---+--------------+------------ -------------- ---+ _|
299 +--------------------------------------------------+
301 | traffic generator |
303 +--------------------------------------------------+
305 VNF → vSwitch → physical port
306 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
308 .. code-block:: console
311 +---------------------------------------------------+ |
313 | +-------------------------------------------+ | |
314 | | Application | | |
315 | +-------------------------------------------+ | |
319 | +---------------+ | |
320 | | logical port | | |
321 +-------------------------------+---------------+---+ _|
325 +------------------------------+---------------+---+ |
326 | | logical port | | |
327 | +---------------+ | |
331 | +--------------+ | |
332 | vSwitch | phy port | | |
333 +-------------------------------+--------------+---+ _|
337 +--------------------------------------------------+
339 | traffic generator |
341 +--------------------------------------------------+
343 VNF → vSwitch → VNF → vSwitch
344 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
346 .. code-block:: console
349 +-------------------------+ +-------------------------+ |
350 | Guest 1 | | Guest 2 | |
351 | +-----------------+ | | +-----------------+ | |
352 | | Application | | | | Application | | |
353 | +-----------------+ | | +-----------------+ | |
357 | +---------------+ | | +---------------+ | |
358 | | logical port 0| | | | logical port 0| | |
359 +-----+---------------+---+ +---+---------------+-----+ _|
363 +----+---------------+------------+---------------+-----+ |
364 | | port 0 | | port 1 | | |
365 | +---------------+ +---------------+ | |
368 | +--------------------+ | |
371 +-------------------------------------------------------+ _|
373 HOST 1(Physical port → virtual switch → VNF → virtual switch →
374 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
375 Physical port) → HOST 2(Physical port → virtual switch → VNF →
376 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
377 virtual switch → Physical port)
378 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
380 .. code-block:: console
383 +--------------------------------------------+ +------------------------------------------+
384 | +---------------------------------+ | | +--------------------------------+ |
385 | | Application | | | | Application | |
386 | +----------------------------+----+ | | +-------------------------+------+ |
389 | +------------------+ +------------------+ | | +------------------+ +----------------+ |
390 | | Logical port 0 | | Logical port 1 | | | | Logical port 0 | | Logical port 1 | |
391 +-+------------------+--+------------------+-+ +-+------------------+--+----------------+-+
395 +-+------------------+--+------------------+-+ +-+------------------+--+----------------+-+
396 | | Logical port 0 | | Logical port 1 | | | | Logical port 0 | | Logical port 1 | |
397 | +------------------+ +------------------+ | | +------------------+ +----------------+ |
400 | | vswitch v | | | vswitch v |
401 | +--------+---------+ +------------------+ | | +------------------+ +----------------+ |
402 | | phy port | | phy port | | | | phy port | | phy port | |
403 +-+------------------+--+------------------+-+ +-+------------------+--+----------------+-+
406 | ------------------------ v
407 +-------------------------------------------------------------------------------------------+
409 | traffic generator |
411 +-------------------------------------------------------------------------------------------+
413 **Note:** For tests where the traffic generator and/or measurement
414 receiver are implemented on VM and connected to the virtual switch
415 through vNIC, the issues of shared resources and interactions between
416 the measurement devices and the device under test must be considered.
418 2.2.3 General Methodology:
419 --------------------------
420 To establish the baseline performance of the virtual switch, tests would
421 initially be run with a simple workload in the VNF (the recommended
422 simple workload VNF would be `DPDK <http://www.dpdk.org/>`__'s testpmd
423 application forwarding packets in a VM or vloop\_vnf a simple kernel
424 module that forwards traffic between two network interfaces inside the
425 virtualized environment while bypassing the networking stack).
426 Subsequently, the tests would also be executed with a real Telco
427 workload running in the VNF, which would exercise the virtual switch in
428 the context of higher level Telco NFV use cases, and prove that its
429 underlying characteristics and behaviour can be measured and validated.
430 Suitable real Telco workload VNFs are yet to be identified.
432 2.2.3.1 Default Test Parameters
433 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
435 The following list identifies the default parameters for suite of
438 - Reference application: Simple forwarding or Open Source VNF.
439 - Frame size (bytes): 64, 128, 256, 512, 1024, 1280, 1518, 2K, 4k OR
440 Packet size based on use-case (e.g. RTP 64B, 256B).
441 - Reordering check: Tests should confirm that packets within a flow are
443 - Duplex: Unidirectional / Bidirectional. Default: Full duplex with
444 traffic transmitting in both directions, as network traffic generally
445 does not flow in a single direction. By default the data rate of
446 transmitted traffic should be the same in both directions, please
447 note that asymmetric traffic (e.g. downlink-heavy) tests will be
448 mentioned explicitly for the relevant test cases.
449 - Number of Flows: Default for non scalability tests is a single flow.
450 For scalability tests the goal is to test with maximum supported
451 flows but where possible will test up to 10 Million flows. Start with
452 a single flow and scale up. By default flows should be added
453 sequentially, tests that add flows simultaneously will explicitly
454 call out their flow addition behaviour. Packets are generated across
455 the flows uniformly with no burstiness.
456 - Traffic Types: UDP, SCTP, RTP, GTP and UDP traffic.
457 - Deployment scenarios are:
458 - Physical → virtual switch → physical.
459 - Physical → virtual switch → VNF → virtual switch → physical.
460 - Physical → virtual switch → VNF → virtual switch → VNF → virtual
462 - Physical → virtual switch → VNF.
463 - VNF → virtual switch → Physical.
464 - VNF → virtual switch → VNF.
466 Tests MUST have these parameters unless otherwise stated. **Test cases
467 with non default parameters will be stated explicitly**.
469 **Note**: For throughput tests unless stated otherwise, test
470 configurations should ensure that traffic traverses the installed flows
471 through the switch, i.e. flows are installed and have an appropriate
472 time out that doesn't expire before packet transmission starts.
474 2.2.3.2 Flow Classification
475 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
477 Virtual switches classify packets into flows by processing and matching
478 particular header fields in the packet/frame and/or the input port where
479 the packets/frames arrived. The vSwitch then carries out an action on
480 the group of packets that match the classification parameters. Thus a
481 flow is considered to be a sequence of packets that have a shared set of
482 header field values or have arrived on the same port and have the same
483 action applied to them. Performance results can vary based on the
484 parameters the vSwitch uses to match for a flow. The recommended flow
485 classification parameters for L3 vSwitch performance tests are: the
486 input port, the source IP address, the destination IP address and the
487 Ethernet protocol type field. It is essential to increase the flow
488 time-out time on a vSwitch before conducting any performance tests that
489 do not measure the flow set-up time. Normally the first packet of a
490 particular flow will install the flow in the vSwitch which adds an
491 additional latency, subsequent packets of the same flow are not subject
492 to this latency if the flow is already installed on the vSwitch.
494 2.2.3.3 Test Priority
495 ~~~~~~~~~~~~~~~~~~~~~
497 Tests will be assigned a priority in order to determine which tests
498 should be implemented immediately and which tests implementations
501 Priority can be of following types: - Urgent: Must be implemented
502 immediately. - High: Must be implemented in the next release. - Medium:
503 May be implemented after the release. - Low: May or may not be
509 The SUT should be configured to its "default" state. The
510 SUT's configuration or set-up must not change between tests in any way
511 other than what is required to do the test. All supported protocols must
512 be configured and enabled for each test set up.
514 2.2.3.4.1 Port Configuration
515 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
517 The DUT should be configured with n ports where
518 n is a multiple of 2. Half of the ports on the DUT should be used as
519 ingress ports and the other half of the ports on the DUT should be used
520 as egress ports. Where a DUT has more than 2 ports, the ingress data
521 streams should be set-up so that they transmit packets to the egress
522 ports in sequence so that there is an even distribution of traffic
523 across ports. For example, if a DUT has 4 ports 0(ingress), 1(ingress),
524 2(egress) and 3(egress), the traffic stream directed at port 0 should
525 output a packet to port 2 followed by a packet to port 3. The traffic
526 stream directed at port 1 should also output a packet to port 2 followed
527 by a packet to port 3.
529 2.2.3.4.2 Frame Formats
530 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
532 Frame formats Layer 2 (data link layer) protocols
533 ++++++++++++++++++++++++++++++++++++++++++++++++++
536 .. code-block:: console
538 +---------------------+--------------------+-----------+
539 | Ethernet Header | Payload | Check Sum |
540 +---------------------+--------------------+-----------+
541 |_____________________|____________________|___________|
542 14 Bytes 46 - 1500 Bytes 4 Bytes
544 Layer 3 (network layer) protocols
545 ++++++++++++++++++++++++++++++++++
549 .. code-block:: console
551 +---------------------+--------------------+--------------------+-----------+
552 | Ethernet Header | IP Header | Payload | Check Sum |
553 +---------------------+--------------------+--------------------+-----------+
554 |_____________________|____________________|____________________|___________|
555 14 Bytes 20 bytes 26 - 1480 Bytes 4 Bytes
559 .. code-block:: console
561 +---------------------+--------------------+--------------------+-----------+
562 | Ethernet Header | IP Header | Payload | Check Sum |
563 +---------------------+--------------------+--------------------+-----------+
564 |_____________________|____________________|____________________|___________|
565 14 Bytes 40 bytes 26 - 1460 Bytes 4 Bytes
567 Layer 4 (transport layer) protocols
568 ++++++++++++++++++++++++++++++++++++
573 .. code-block:: console
575 +---------------------+--------------------+-----------------+--------------------+-----------+
576 | Ethernet Header | IP Header | Layer 4 Header | Payload | Check Sum |
577 +---------------------+--------------------+-----------------+--------------------+-----------+
578 |_____________________|____________________|_________________|____________________|___________|
579 14 Bytes 40 bytes 20 Bytes 6 - 1460 Bytes 4 Bytes
581 Layer 5 (application layer) protocols
582 +++++++++++++++++++++++++++++++++++++
586 .. code-block:: console
588 +---------------------+--------------------+-----------------+--------------------+-----------+
589 | Ethernet Header | IP Header | Layer 4 Header | Payload | Check Sum |
590 +---------------------+--------------------+-----------------+--------------------+-----------+
591 |_____________________|____________________|_________________|____________________|___________|
592 14 Bytes 20 bytes 20 Bytes Min 6 Bytes 4 Bytes
595 2.2.3.4.3 Packet Throughput
596 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
597 There is a difference between an Ethernet frame,
598 an IP packet, and a UDP datagram. In the seven-layer OSI model of
599 computer networking, packet refers to a data unit at layer 3 (network
600 layer). The correct term for a data unit at layer 2 (data link layer) is
601 a frame, and at layer 4 (transport layer) is a segment or datagram.
603 Important concepts related to 10GbE performance are frame rate and
604 throughput. The MAC bit rate of 10GbE, defined in the IEEE standard 802
605 .3ae, is 10 billion bits per second. Frame rate is based on the bit rate
606 and frame format definitions. Throughput, defined in IETF RFC 1242, is
607 the highest rate at which the system under test can forward the offered
610 The frame rate for 10GbE is determined by a formula that divides the 10
611 billion bits per second by the preamble + frame length + inter-frame
614 The maximum frame rate is calculated using the minimum values of the
615 following parameters, as described in the IEEE 802 .3ae standard:
617 - Preamble: 8 bytes \* 8 = 64 bits
618 - Frame Length: 64 bytes (minimum) \* 8 = 512 bits
619 - Inter-frame Gap: 12 bytes (minimum) \* 8 = 96 bits
621 Therefore, Maximum Frame Rate (64B Frames)
622 = MAC Transmit Bit Rate / (Preamble + Frame Length + Inter-frame Gap)
623 = 10,000,000,000 / (64 + 512 + 96)
624 = 10,000,000,000 / 672
625 = 14,880,952.38 frame per second (fps)
627 2.2.3.4.4 System isolation and validation
628 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
630 A key consideration when conducting any sort of benchmark is trying to
631 ensure the consistency and repeatability of test results between runs.
632 When benchmarking the performance of a virtual switch there are many
633 factors that can affect the consistency of results. This section
634 describes these factors and the measures that can be taken to limit
635 their effects. In addition, this section will outline some system tests
636 to validate the platform and the VNF before conducting any vSwitch
641 When conducting a benchmarking test on any SUT, it is essential to limit
642 (and if reasonable, eliminate) any noise that may interfere with the
643 accuracy of the metrics collected by the test. This noise may be
644 introduced by other hardware or software (OS, other applications), and
645 can result in significantly varying performance metrics being collected
646 between consecutive runs of the same test. In the case of characterizing
647 the performance of a virtual switch, there are a number of configuration
648 parameters that can help increase the repeatability and stability of
649 test results, including:
651 - OS/GRUB configuration:
653 - maxcpus = n where n >= 0; limits the kernel to using 'n'
654 processors. Only use exactly what you need.
655 - isolcpus: Isolate CPUs from the general scheduler. Isolate all
656 CPUs bar one which will be used by the OS.
657 - use taskset to affinitize the forwarding application and the VNFs
658 onto isolated cores. VNFs and the vSwitch should be allocated
659 their own cores, i.e. must not share the same cores. vCPUs for the
660 VNF should be affinitized to individual cores also.
661 - Limit the amount of background applications that are running and
662 set OS to boot to runlevel 3. Make sure to kill any unnecessary
663 system processes/daemons.
664 - Only enable hardware that you need to use for your test – to
665 ensure there are no other interrupts on the system.
666 - Configure NIC interrupts to only use the cores that are not
667 allocated to any other process (VNF/vSwitch).
669 - NUMA configuration: Any unused sockets in a multi-socket system
671 - CPU pinning: The vSwitch and the VNF should each be affinitized to
672 separate logical cores using a combination of maxcpus, isolcpus and
674 - BIOS configuration: BIOS should be configured for performance where
675 an explicit option exists, sleep states should be disabled, any
676 virtualization optimization technologies should be enabled, and
677 hyperthreading should also be enabled.
681 System validation is broken down into two sub-categories: Platform
682 validation and VNF validation. The validation test itself involves
683 verifying the forwarding capability and stability for the sub-system
684 under test. The rationale behind system validation is two fold. Firstly
685 to give a tester confidence in the stability of the platform or VNF that
686 is being tested; and secondly to provide base performance comparison
687 points to understand the overhead introduced by the virtual switch.
689 * Benchmark platform forwarding capability: This is an OPTIONAL test
690 used to verify the platform and measure the base performance (maximum
691 forwarding rate in fps and latency) that can be achieved by the
692 platform without a vSwitch or a VNF. The following diagram outlines
693 the set-up for benchmarking Platform forwarding capability:
695 .. code-block:: console
698 +--------------------------------------------------+ |
699 | +------------------------------------------+ | |
701 | | l2fw or DPDK L2FWD app | | Host
703 | +------------------------------------------+ | |
705 +---+------------------------------------------+---+ __|
709 +--------------------------------------------------+
711 | traffic generator |
713 +--------------------------------------------------+
715 * Benchmark VNF forwarding capability: This test is used to verify
716 the VNF and measure the base performance (maximum forwarding rate in
717 fps and latency) that can be achieved by the VNF without a vSwitch.
718 The performance metrics collected by this test will serve as a key
719 comparison point for NIC passthrough technologies and vSwitches. VNF
720 in this context refers to the hypervisor and the VM. The following
721 diagram outlines the set-up for benchmarking VNF forwarding
724 .. code-block:: console
727 +--------------------------------------------------+ |
728 | +------------------------------------------+ | |
732 | +------------------------------------------+ | |
733 | | Passthrough/SR-IOV | | Host
734 | +------------------------------------------+ | |
736 +---+------------------------------------------+---+ __|
740 +--------------------------------------------------+
742 | traffic generator |
744 +--------------------------------------------------+
747 Methodology to benchmark Platform/VNF forwarding capability
748 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
750 The recommended methodology for the platform/VNF validation and
751 benchmark is: - Run `RFC2889 <https://www.rfc-editor.org/rfc/rfc2289.txt>`__
752 Maximum Forwarding Rate test, this test will produce maximum
753 forwarding rate and latency results that will serve as the
754 expected values. These expected values can be used in
755 subsequent steps or compared with in subsequent validation tests. -
756 Transmit bidirectional traffic at line rate/max forwarding rate
757 (whichever is higher) for at least 72 hours, measure throughput (fps)
758 and latency. - Note: Traffic should be bidirectional. - Establish a
759 baseline forwarding rate for what the platform can achieve. - Additional
760 validation: After the test has completed for 72 hours run bidirectional
761 traffic at the maximum forwarding rate once more to see if the system is
762 still functional and measure throughput (fps) and latency. Compare the
763 measure the new obtained values with the expected values.
765 **NOTE 1**: How the Platform is configured for its forwarding capability
766 test (BIOS settings, GRUB configuration, runlevel...) is how the
767 platform should be configured for every test after this
769 **NOTE 2**: How the VNF is configured for its forwarding capability test
770 (# of vCPUs, vNICs, Memory, affinitization…) is how it should be
771 configured for every test that uses a VNF after this.
773 2.2.4 RFCs for testing switch performance
774 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
776 The starting point for defining the suite of tests for benchmarking the
777 performance of a virtual switch is to take existing RFCs and standards
778 that were designed to test their physical counterparts and adapting them
779 for testing virtual switches. The rationale behind this is to establish
780 a fair comparison between the performance of virtual and physical
781 switches. This section outlines the RFCs that are used by this
784 RFC 1242 Benchmarking Terminology for Network Interconnection
785 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
786 Devices RFC 1242 defines the terminology that is used in describing
787 performance benchmarking tests and their results. Definitions and
788 discussions covered include: Back-to-back, bridge, bridge/router,
789 constant load, data link frame size, frame loss rate, inter frame gap,
790 latency, and many more.
792 RFC 2544 Benchmarking Methodology for Network Interconnect Devices
793 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
794 RFC 2544 outlines a benchmarking methodology for network Interconnect
795 Devices. The methodology results in performance metrics such as latency,
796 frame loss percentage, and maximum data throughput.
798 In this document network “throughput” (measured in millions of frames
799 per second) is based on RFC 2544, unless otherwise noted. Frame size
800 refers to Ethernet frames ranging from smallest frames of 64 bytes to
801 largest frames of 4K bytes.
805 1. Throughput test defines the maximum number of frames per second
806 that can be transmitted without any error.
808 2. Latency test measures the time required for a frame to travel from
809 the originating device through the network to the destination device.
810 Please note that RFC2544 Latency measurement will be superseded with
811 a measurement of average latency over all successfully transferred
814 3. Frame loss test measures the network’s
815 response in overload conditions - a critical indicator of the
816 network’s ability to support real-time applications in which a
817 large amount of frame loss will rapidly degrade service quality.
819 4. Burst test assesses the buffering capability of a switch. It
820 measures the maximum number of frames received at full line rate
821 before a frame is lost. In carrier Ethernet networks, this
822 measurement validates the excess information rate (EIR) as defined in
825 5. System recovery to characterize speed of recovery from an overload
828 6. Reset to characterize speed of recovery from device or software
829 reset. This type of test has been updated by `RFC6201 <https://www.rfc-editor.org/rfc/rfc6201.txt>`__ as such,
830 the methodology defined by this specification will be that of RFC 6201.
832 Although not included in the defined RFC 2544 standard, another crucial
833 measurement in Ethernet networking is packet delay variation. The
834 definition set out by this specification comes from
835 `RFC5481 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__.
837 RFC 2285 Benchmarking Terminology for LAN Switching Devices
838 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
839 RFC 2285 defines the terminology that is used to describe the
840 terminology for benchmarking a LAN switching device. It extends RFC
841 1242 and defines: DUTs, SUTs, Traffic orientation and distribution,
842 bursts, loads, forwarding rates, etc.
844 RFC 2889 Benchmarking Methodology for LAN Switching
845 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
846 RFC 2889 outlines a benchmarking methodology for LAN switching, it
847 extends RFC 2544. The outlined methodology gathers performance
848 metrics for forwarding, congestion control, latency, address handling
849 and finally filtering.
851 RFC 3918 Methodology for IP Multicast Benchmarking
852 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
853 RFC 3918 outlines a methodology for IP Multicast benchmarking.
855 RFC 4737 Packet Reordering Metrics
856 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
857 RFC 4737 describes metrics for identifying and counting re-ordered
858 packets within a stream, and metrics to measure the extent each
859 packet has been re-ordered.
861 RFC 5481 Packet Delay Variation Applicability Statement
862 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
863 RFC 5481 defined two common, but different forms of delay variation
864 metrics, and compares the metrics over a range of networking
865 circumstances and tasks. The most suitable form for vSwitch
866 benchmarking is the "PDV" form.
868 RFC 6201 Device Reset Characterization
869 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
870 RFC 6201 extends the methodology for characterizing the speed of
871 recovery of the DUT from device or software reset described in RFC
874 2.2.5 Details of the Test Report
875 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
877 There are a number of parameters related to the system, DUT and tests
878 that can affect the repeatability of a test results and should be
879 recorded. In order to minimise the variation in the results of a test,
880 it is recommended that the test report includes the following information:
882 - Hardware details including:
886 - Memory information (see below)
887 - Number of enabled cores.
888 - Number of cores used for the test.
889 - Number of physical NICs, as well as their details (manufacturer,
890 versions, type and the PCI slot they are plugged into).
891 - NIC interrupt configuration.
892 - BIOS version, release date and any configurations that were
895 - Software details including:
897 - OS version (for host and VNF)
898 - Kernel version (for host and VNF)
899 - GRUB boot parameters (for host and VNF).
900 - Hypervisor details (Type and version).
901 - Selected vSwitch, version number or commit id used.
902 - vSwitch launch command line if it has been parameterised.
903 - Memory allocation to the vSwitch – which NUMA node it is using,
904 and how many memory channels.
905 - Where the vswitch is built from source: compiler details including
906 versions and the flags that were used to compile the vSwitch.
907 - DPDK or any other SW dependency version number or commit id used.
908 - Memory allocation to a VM - if it's from Hugpages/elsewhere.
909 - VM storage type: snapshot/independent persistent/independent
912 - Number of Virtual NICs (vNICs), versions, type and driver.
913 - Number of virtual CPUs and their core affinity on the host.
914 - Number vNIC interrupt configuration.
915 - Thread affinitization for the applications (including the vSwitch
917 - Details of Resource isolation, such as CPUs designated for
918 Host/Kernel (isolcpu) and CPUs designated for specific processes
937 - Traffic Information:
939 - Traffic type - UDP, TCP, IMIX / Other.
942 - Deployment Scenario.
944 **Note**: Tests that require additional parameters to be recorded will
945 explicitly specify this.
947 2.3. Test identification
948 ------------------------
949 2.3.1 Throughput tests
950 ~~~~~~~~~~~~~~~~~~~~~~
951 The following tests aim to determine the maximum forwarding rate that
952 can be achieved with a virtual switch. The list is not exhaustive but
953 should indicate the type of tests that should be required. It is
954 expected that more will be added.
956 Test ID: LTD.Throughput.RFC2544.PacketLossRatio
957 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
958 **Title**: RFC 2544 X% packet loss ratio Throughput and Latency Test
960 **Prerequisite Test**: N/A
966 This test determines the DUT's maximum forwarding rate with X% traffic
967 loss for a constant load (fixed length frames at a fixed interval time).
968 The default loss percentages to be tested are: - X = 0% - X = 10^-7%
970 Note: Other values can be tested if required by the user.
972 The selected frame sizes are those previously defined under `Default
973 Test Parameters <#DefaultParams>`__. The test can also be used to
974 determine the average latency of the traffic.
976 Under the `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__
977 test methodology, the test duration will
978 include a number of trials; each trial should run for a minimum period
979 of 60 seconds. A binary search methodology must be applied for each
980 trial to obtain the final result.
982 **Expected Result**: At the end of each trial, the presence or absence
983 of loss determines the modification of offered load for the next trial,
984 converging on a maximum rate, or
985 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__ Throughput with X% loss.
986 The Throughput load is re-used in related
987 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__ tests and other
990 **Metrics Collected**:
992 The following are the metrics collected for this test:
994 - The maximum forwarding rate in Frames Per Second (FPS) and Mbps of
995 the DUT for each frame size with X% packet loss.
996 - The average latency of the traffic flow when passing through the DUT
997 (if testing for latency, note that this average is different from the
998 test specified in Section 26.3 of
999 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__).
1000 - CPU and memory utilization may also be collected as part of this
1001 test, to determine the vSwitch's performance footprint on the system.
1003 Test ID: LTD.Throughput.RFC2544.PacketLossRatioFrameModification
1004 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1005 **Title**: RFC 2544 X% packet loss Throughput and Latency Test with
1008 **Prerequisite Test**: N/A
1014 This test determines the DUT's maximum forwarding rate with X% traffic
1015 loss for a constant load (fixed length frames at a fixed interval time).
1016 The default loss percentages to be tested are: - X = 0% - X = 10^-7%
1018 Note: Other values can be tested if required by the user.
1020 The selected frame sizes are those previously defined under `Default
1021 Test Parameters <#DefaultParams>`__. The test can also be used to
1022 determine the average latency of the traffic.
1024 Under the `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__
1025 test methodology, the test duration will
1026 include a number of trials; each trial should run for a minimum period
1027 of 60 seconds. A binary search methodology must be applied for each
1028 trial to obtain the final result.
1030 During this test, the DUT must perform the following operations on the
1033 - Perform packet parsing on the DUT's ingress port.
1034 - Perform any relevant address look-ups on the DUT's ingress ports.
1035 - Modify the packet header before forwarding the packet to the DUT's
1036 egress port. Packet modifications include:
1038 - Modifying the Ethernet source or destination MAC address.
1039 - Modifying/adding a VLAN tag. (**Recommended**).
1040 - Modifying/adding a MPLS tag.
1041 - Modifying the source or destination ip address.
1042 - Modifying the TOS/DSCP field.
1043 - Modifying the source or destination ports for UDP/TCP/SCTP.
1044 - Modifying the TTL.
1046 **Expected Result**: The Packet parsing/modifications require some
1047 additional degree of processing resource, therefore the
1048 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__
1049 Throughput is expected to be somewhat lower than the Throughput level
1050 measured without additional steps. The reduction is expected to be
1051 greatest on tests with the smallest packet sizes (greatest header
1054 **Metrics Collected**:
1056 The following are the metrics collected for this test:
1058 - The maximum forwarding rate in Frames Per Second (FPS) and Mbps of
1059 the DUT for each frame size with X% packet loss and packet
1060 modification operations being performed by the DUT.
1061 - The average latency of the traffic flow when passing through the DUT
1062 (if testing for latency, note that this average is different from the
1063 test specified in Section 26.3 of
1064 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__).
1065 - The `RFC5481 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__
1066 PDV form of delay variation on the traffic flow,
1067 using the 99th percentile.
1068 - CPU and memory utilization may also be collected as part of this
1069 test, to determine the vSwitch's performance footprint on the system.
1071 Test ID: LTD.Throughput.RFC2544.Profile
1072 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1073 **Title**: RFC 2544 Throughput and Latency Profile
1075 **Prerequisite Test**: N/A
1081 This test reveals how throughput and latency degrades as the offered
1082 rate varies in the region of the DUT's maximum forwarding rate as
1083 determined by LTD.Throughput.RFC2544.PacketLossRatio (0% Packet Loss).
1084 For example it can be used to determine if the degradation of throughput
1085 and latency as the offered rate increases is slow and graceful or sudden
1088 The selected frame sizes are those previously defined under `Default
1089 Test Parameters <#DefaultParams>`__.
1091 The offered traffic rate is described as a percentage delta with respect
1092 to the DUT's maximum forwarding rate as determined by
1093 LTD.Throughput.RFC2544.PacketLoss Ratio (0% Packet Loss case). A delta
1094 of 0% is equivalent to an offered traffic rate equal to the maximum
1095 forwarding rate; A delta of +50% indicates an offered rate half-way
1096 between the maximum forwarding rate and line-rate, whereas a delta of
1097 -50% indicates an offered rate of half the maximum rate. Therefore the
1098 range of the delta figure is natuarlly bounded at -100% (zero offered
1099 traffic) and +100% (traffic offered at line rate).
1101 The following deltas to the maximum forwarding rate should be applied:
1103 - -50%, -10%, 0%, +10% & +50%
1105 **Expected Result**: For each packet size a profile should be produced
1106 of how throughput and latency vary with offered rate.
1108 **Metrics Collected**:
1110 The following are the metrics collected for this test:
1112 - The forwarding rate in Frames Per Second (FPS) and Mbps of the DUT
1113 for each delta to the maximum forwarding rate and for each frame
1115 - The average latency for each delta to the maximum forwarding rate and
1116 for each frame size.
1117 - CPU and memory utilization may also be collected as part of this
1118 test, to determine the vSwitch's performance footprint on the system.
1119 - Any failures experienced (for example if the vSwitch crashes, stops
1120 processing packets, restarts or becomes unresponsive to commands)
1121 when the offered load is above Maximum Throughput MUST be recorded
1122 and reported with the results.
1124 Test ID: LTD.Throughput.RFC2544.SystemRecoveryTime
1125 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1126 **Title**: RFC 2544 System Recovery Time Test
1128 **Prerequisite Test** LTD.Throughput.RFC2544.PacketLossRatio
1134 The aim of this test is to determine the length of time it takes the DUT
1135 to recover from an overload condition for a constant load (fixed length
1136 frames at a fixed interval time). The selected frame sizes are those
1137 previously defined under `Default Test Parameters <#DefaultParams>`__,
1138 traffic should be sent to the DUT under normal conditions. During the
1139 duration of the test and while the traffic flows are passing though the
1140 DUT, at least one situation leading to an overload condition for the DUT
1141 should occur. The time from the end of the overload condition to when
1142 the DUT returns to normal operations should be measured to determine
1143 recovery time. Prior to overloading the DUT, one should record the
1144 average latency for 10,000 packets forwarded through the DUT.
1146 The overload condition SHOULD be to transmit traffic at a very high
1147 frame rate to the DUT (150% of the maximum 0% packet loss rate as
1148 determined by LTD.Throughput.RFC2544.PacketLossRatio or line-rate
1149 whichever is lower), for at least 60 seconds, then reduce the frame rate
1150 to 75% of the maximum 0% packet loss rate. A number of time-stamps
1151 should be recorded: - Record the time-stamp at which the frame rate was
1152 reduced and record a second time-stamp at the time of the last frame
1153 lost. The recovery time is the difference between the two timestamps. -
1154 Record the average latency for 10,000 frames after the last frame loss
1155 and continue to record average latency measurements for every 10,000
1156 frames, when latency returns to within 10% of pre-overload levels record
1159 **Expected Result**:
1161 **Metrics collected**
1163 The following are the metrics collected for this test:
1165 - The length of time it takes the DUT to recover from an overload
1167 - The length of time it takes the DUT to recover the average latency to
1168 pre-overload conditions.
1170 **Deployment scenario**:
1172 - Physical → virtual switch → physical.
1174 Test ID: LTD.Throughput.RFC2544.BackToBackFrames
1175 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1176 **Title**: RFC2544 Back To Back Frames Test
1178 **Prerequisite Test**: N
1184 The aim of this test is to characterize the ability of the DUT to
1185 process back-to-back frames. For each frame size previously defined
1186 under `Default Test Parameters <#DefaultParams>`__, a burst of traffic
1187 is sent to the DUT with the minimum inter-frame gap between each frame.
1188 If the number of received frames equals the number of frames that were
1189 transmitted, the burst size should be increased and traffic is sent to
1190 the DUT again. The value measured is the back-to-back value, that is the
1191 maximum burst size the DUT can handle without any frame loss.
1193 **Expected Result**:
1195 Tests of back-to-back frames with physical devices have produced
1196 unstable results in some cases. All tests should be repeated in multiple
1197 test sessions and results stability should be examined.
1199 **Metrics collected**
1201 The following are the metrics collected for this test:
1203 - The back-to-back value, which is the the number of frames in the
1204 longest burst that the DUT will handle without the loss of any
1206 - CPU and memory utilization may also be collected as part of this
1207 test, to determine the vSwitch's performance footprint on the system.
1209 **Deployment scenario**:
1211 - Physical → virtual switch → physical.
1213 Test ID: LTD.Throughput.RFC2889.Soak
1214 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1215 **Title**: RFC 2889 X% packet loss Throughput Soak Test
1217 **Prerequisite Test** LTD.Throughput.RFC2544.PacketLossRatio
1223 The aim of this test is to understand the Throughput stability over an
1224 extended test duration in order to uncover any outliers. To allow for an
1225 extended test duration, the test should ideally run for 24 hours or, if
1226 this is not possible, for at least 6 hours. For this test, each frame
1227 size must be sent at the highest Throughput with X% packet loss, as
1228 determined in the prerequisite test. The default loss percentages to be
1229 tested are: - X = 0% - X = 10^-7%
1231 Note: Other values can be tested if required by the user.
1233 **Expected Result**:
1235 **Metrics Collected**:
1237 The following are the metrics collected for this test:
1239 - Throughput stability of the DUT.
1241 - This means reporting the number of packets lost per time interval
1242 and reporting any time intervals with packet loss. The
1243 `RFC2889 <https://www.rfc-editor.org/rfc/rfc2289.txt>`__
1244 Forwarding Rate shall be measured in each interval.
1245 An interval of 60s is suggested.
1247 - CPU and memory utilization may also be collected as part of this
1248 test, to determine the vSwitch's performance footprint on the system.
1249 - The `RFC5481 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__
1250 PDV form of delay variation on the traffic flow,
1251 using the 99th percentile.
1253 Test ID: LTD.Throughput.RFC2889.SoakFrameModification
1254 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1255 **Title**: RFC 2889 Throughput Soak Test with Frame Modification
1257 **Prerequisite Test**: LTD.Throughput.RFC2544.PacketLossRatioFrameModification (0% Packet Loss)
1263 The aim of this test is to understand the throughput stability over an
1264 extended test duration in order to uncover any outliers. To allow for an
1265 extended test duration, the test should ideally run for 24 hours or, if
1266 this is not possible, for at least 6 hour. For this test, each frame
1267 size must be sent at the highest Throughput with 0% packet loss, as
1268 determined in the prerequisite test.
1270 During this test, the DUT must perform the following operations on the
1273 - Perform packet parsing on the DUT's ingress port.
1274 - Perform any relevant address look-ups on the DUT's ingress ports.
1275 - Modify the packet header before forwarding the packet to the DUT's
1276 egress port. Packet modifications include:
1278 - Modifying the Ethernet source or destination MAC address.
1279 - Modifying/adding a VLAN tag (**Recommended**).
1280 - Modifying/adding a MPLS tag.
1281 - Modifying the source or destination ip address.
1282 - Modifying the TOS/DSCP field.
1283 - Modifying the source or destination ports for UDP/TCP/SCTP.
1284 - Modifying the TTL.
1286 **Expected Result**:
1288 **Metrics Collected**:
1290 The following are the metrics collected for this test:
1292 - Throughput stability of the DUT.
1294 - This means reporting the number of packets lost per time interval
1295 and reporting any time intervals with packet loss. The
1296 `RFC2889 <https://www.rfc-editor.org/rfc/rfc2289.txt>`__
1297 Forwarding Rate shall be measured in each interval.
1298 An interval of 60s is suggested.
1300 - CPU and memory utilization may also be collected as part of this
1301 test, to determine the vSwitch's performance footprint on the system.
1302 - The `RFC5481 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__ PDV form of delay variation on the traffic flow,
1303 using the 99th percentile.
1305 Test ID: LTD.Throughput.RFC6201.ResetTime
1306 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1307 **Title**: RFC 6201 Reset Time Test
1309 **Prerequisite Test**: N/A
1315 The aim of this test is to determine the length of time it takes the DUT
1316 to recover from a reset.
1318 Two reset methods are defined - planned and unplanned. A planned reset
1319 requires stopping and restarting the virtual switch by the usual
1320 'graceful' method defined by it's documentation. An unplanned reset
1321 requires simulating a fatal internal fault in the virtual switch - for
1322 example by using kill -SIGKILL on a Linux environment.
1324 Both reset methods SHOULD be exercised.
1326 For each frame size previously defined under `Default Test
1327 Parameters <#DefaultParams>`__, traffic should be sent to the DUT under
1328 normal conditions. During the duration of the test and while the traffic
1329 flows are passing through the DUT, the DUT should be reset and the Reset
1330 time measured. The Reset time is the total time that a device is
1331 determined to be out of operation and includes the time to perform the
1332 reset and the time to recover from it (cf. `RFC6201 <https://www.rfc-editor.org/rfc/rfc6201.txt>`__).
1334 `RFC6201 <https://www.rfc-editor.org/rfc/rfc6201.txt>`__ defines two methods to measure the Reset time:
1335 - Frame-Loss Method: which requires the monitoring of the number of
1336 lost frames and calculates the Reset time based on the number of
1337 frames lost and the offered rate according to the following
1340 .. code-block:: console
1342 Frames_lost (packets)
1343 Reset_time = -------------------------------------
1344 Offered_rate (packets per second)
1346 - Timestamp Method: which measures the time from which the last frame
1347 is forwarded from the DUT to the time the first frame is forwarded
1348 after the reset. This involves time-stamping all transmitted frames
1349 and recording the timestamp of the last frame that was received prior
1350 to the reset and also measuring the timestamp of the first frame that
1351 is received after the reset. The Reset time is the difference between
1352 these two timestamps.
1354 According to `RFC6201 <https://www.rfc-editor.org/rfc/rfc6201.txt>`__ the choice of method depends on the test
1355 tool's capability; the Frame-Loss method SHOULD be used if the test tool
1356 supports: - Counting the number of lost frames per stream. -
1357 Transmitting test frame despite the physical link status.
1359 whereas the Timestamp method SHOULD be used if the test tool supports: -
1360 Timestamping each frame. - Monitoring received frame's timestamp. -
1361 Transmitting frames only if the physical link status is up.
1363 **Expected Result**:
1365 **Metrics collected**
1367 The following are the metrics collected for this test: - Average Reset
1368 Time over the number of trials performed.
1370 Results of this test should include the following information: - The
1371 reset method used. - Throughput in Fps and Mbps. - Average Frame Loss
1372 over the number of trials performed. - Average Reset Time in
1373 milliseconds over the number of trials performed. - Number of trials
1374 performed. - Protocol: IPv4, IPv6, MPLS, etc. - Frame Size in Octets -
1375 Port Media: Ethernet, Gigabit Ethernet (GbE), etc. - Port Speed: 10
1376 Gbps, 40 Gbps etc. - Interface Encapsulation: Ethernet, Ethernet VLAN,
1379 **Deployment scenario**:
1381 - Physical → virtual switch → physical.
1383 Test ID: LTD.Throughput.RFC2889.MaxForwardingRate
1384 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1385 **Title**: RFC2889 Forwarding Rate Test
1387 **Prerequisite Test**: LTD.Throughput.RFC2544.PacketLossRatio
1393 This test measures the DUT's Max Forwarding Rate when the Offered Load
1394 is varied between the throughput and the Maximum Offered Load for fixed
1395 length frames at a fixed time interval. The selected frame sizes are
1396 those previously defined under `Default Test
1397 Parameters <#DefaultParams>`__. The throughput is the maximum offered
1398 load with 0% frame loss (measured by the prerequisite test), and the
1399 Maximum Offered Load (as defined by
1400 `RFC2285 <https://www.rfc-editor.org/rfc/rfc2285.txt>`__) is *"the highest
1401 number of frames per second that an external source can transmit to a
1402 DUT/SUT for forwarding to a specified output interface or interfaces"*.
1404 Traffic should be sent to the DUT at a particular rate (TX rate)
1405 starting with TX rate equal to the throughput rate. The rate of
1406 successfully received frames at the destination counted (in FPS). If the
1407 RX rate is equal to the TX rate, the TX rate should be increased by a
1408 fixed step size and the RX rate measured again until the Max Forwarding
1411 The trial duration for each iteration should last for the period of time
1412 needed for the system to reach steady state for the frame size being
1413 tested. Under `RFC2889 <https://www.rfc-editor.org/rfc/rfc2289.txt>`__
1414 (Sec. 5.6.3.1) test methodology, the test
1415 duration should run for a minimum period of 30 seconds, regardless
1416 whether the system reaches steady state before the minimum duration
1419 **Expected Result**: According to
1420 `RFC2889 <https://www.rfc-editor.org/rfc/rfc2289.txt>`__ The Max Forwarding Rate
1421 is the highest forwarding rate of a DUT taken from an iterative set of
1422 forwarding rate measurements. The iterative set of forwarding rate
1423 measurements are made by setting the intended load transmitted from an
1424 external source and measuring the offered load (i.e what the DUT is
1425 capable of forwarding). If the Throughput == the Maximum Offered Load,
1426 it follows that Max Forwarding Rate is equal to the Maximum Offered
1429 **Metrics Collected**:
1431 The following are the metrics collected for this test:
1433 - The Max Forwarding Rate for the DUT for each packet size.
1434 - CPU and memory utilization may also be collected as part of this
1435 test, to determine the vSwitch's performance footprint on the system.
1437 Test ID: LTD.Throughput.RFC2889.ForwardPressure
1438 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1439 **Title**: RFC2889 Forward Pressure Test
1441 **Prerequisite Test**: LTD.Throughput.RFC2889.MaxForwardingRate
1447 The aim of this test is to determine if the DUT transmits frames with an
1448 inter-frame gap that is less than 12 bytes. This test overloads the DUT
1449 and measures the output for forward pressure. Traffic should be
1450 transmitted to the DUT with an inter-frame gap of 11 bytes, this will
1451 overload the DUT by 1 byte per frame. The forwarding rate of the DUT
1454 **Expected Result**: The forwarding rate should not exceed the maximum
1455 forwarding rate of the DUT collected by
1456 LTD.Throughput.RFC2889.MaxForwardingRate.
1458 **Metrics collected**
1460 The following are the metrics collected for this test:
1462 - Forwarding rate of the DUT in FPS or Mbps.
1463 - CPU and memory utilization may also be collected as part of this
1464 test, to determine the vSwitch's performance footprint on the system.
1466 **Deployment scenario**:
1468 - Physical → virtual switch → physical.
1470 Test ID: LTD.Throughput.RFC2889.AddressCachingCapacity
1471 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1472 **Title**: RFC2889 Address Caching Capacity Test
1474 **Prerequisite Test**: N/A
1480 Please note this test is only applicable to switches that are capable of
1481 MAC learning. The aim of this test is to determine the address caching
1482 capacity of the DUT for a constant load (fixed length frames at a fixed
1483 interval time). The selected frame sizes are those previously defined
1484 under `Default Test Parameters <#DefaultParams>`__.
1486 In order to run this test the aging time, that is the maximum time the
1487 DUT will keep a learned address in its flow table, and a set of initial
1488 addresses, whose value should be >= 1 and <= the max number supported by
1489 the implementation must be known. Please note that if the aging time is
1490 configurable it must be longer than the time necessary to produce frames
1491 from the external source at the specified rate. If the aging time is
1492 fixed the frame rate must be brought down to a value that the external
1493 source can produce in a time that is less than the aging time.
1495 Learning Frames should be sent from an external source to the DUT to
1496 install a number of flows. The Learning Frames must have a fixed
1497 destination address and must vary the source address of the frames. The
1498 DUT should install flows in its flow table based on the varying source
1499 addresses. Frames should then be transmitted from an external source at
1500 a suitable frame rate to see if the DUT has properly learned all of the
1501 addresses. If there is no frame loss and no flooding, the number of
1502 addresses sent to the DUT should be increased and the test is repeated
1503 until the max number of cached addresses supported by the DUT
1506 **Expected Result**:
1508 **Metrics collected**:
1510 The following are the metrics collected for this test:
1512 - Number of cached addresses supported by the DUT.
1513 - CPU and memory utilization may also be collected as part of this
1514 test, to determine the vSwitch's performance footprint on the system.
1516 **Deployment scenario**:
1518 - Physical → virtual switch → physical.
1520 Test ID: LTD.Throughput.RFC2889.AddressLearningRate
1521 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1522 **Title**: RFC2889 Address Learning Rate Test
1524 **Prerequisite Test**: LTD.Memory.RFC2889.AddressCachingCapacity
1530 Please note this test is only applicable to switches that are capable of
1531 MAC learning. The aim of this test is to determine the rate of address
1532 learning of the DUT for a constant load (fixed length frames at a fixed
1533 interval time). The selected frame sizes are those previously defined
1534 under `Default Test Parameters <#DefaultParams>`__, traffic should be
1535 sent with each IPv4/IPv6 address incremented by one. The rate at which
1536 the DUT learns a new address should be measured. The maximum caching
1537 capacity from LTD.Memory.RFC2889.AddressCachingCapacity should be taken
1538 into consideration as the maximum number of addresses for which the
1539 learning rate can be obtained.
1541 **Expected Result**: It may be worthwhile to report the behaviour when
1542 operating beyond address capacity - some DUTS may be more friendly to
1543 new addresses than others.
1545 **Metrics collected**:
1547 The following are the metrics collected for this test:
1549 - The address learning rate of the DUT.
1551 **Deployment scenario**:
1553 - Physical → virtual switch → physical.
1555 Test ID: LTD.Throughput.RFC2889.ErrorFramesFiltering
1556 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1557 **Title**: RFC2889 Error Frames Filtering Test
1559 **Prerequisite Test**: N/A
1565 The aim of this test is to determine whether the DUT will propagate any
1566 erroneous frames it receives or whether it is capable of filtering out
1567 the erroneous frames. Traffic should be sent with erroneous frames
1568 included within the flow at random intervals. Illegal frames that must
1569 be tested include: - Oversize Frames. - Undersize Frames. - CRC Errored
1570 Frames. - Dribble Bit Errored Frames - Alignment Errored Frames
1572 The traffic flow exiting the DUT should be recorded and checked to
1573 determine if the erroneous frames where passed through the DUT.
1575 **Expected Result**: Broken frames are not passed!
1577 **Metrics collected**
1579 No Metrics are collected in this test, instead it determines:
1581 - Whether the DUT will propagate erroneous frames.
1582 - Or whether the DUT will correctly filter out any erroneous frames
1583 from traffic flow with out removing correct frames.
1585 **Deployment scenario**:
1587 - Physical → virtual switch → physical.
1589 Test ID: LTD.Throughput.RFC2889.BroadcastFrameForwarding
1590 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1591 **Title**: RFC2889 Broadcast Frame Forwarding Test
1593 **Prerequisite Test**: N
1599 The aim of this test is to determine the maximum forwarding rate of the
1600 DUT when forwarding broadcast traffic. For each frame previously defined
1601 under `Default Test Parameters <#DefaultParams>`__, the traffic should
1602 be set up as broadcast traffic. The traffic throughput of the DUT should
1605 The test should be conducted with at least 4 physical ports on the DUT.
1606 The number of ports used MUST be recorded.
1608 As broadcast involves forwarding a single incoming packet to several
1609 destinations, the latency of a single packet is defined as the average
1610 of the latencies for each of the broadcast destinations.
1612 The incoming packet is transmitted on each of the other physical ports,
1613 it is not transmitted on the port on which it was received. The test MAY
1614 be conducted using different broadcasting ports to uncover any
1615 performance differences.
1617 **Expected Result**:
1619 **Metrics collected**:
1621 The following are the metrics collected for this test:
1623 - The forwarding rate of the DUT when forwarding broadcast traffic.
1624 - The minimum, average & maximum packets latencies observed.
1626 **Deployment scenario**:
1628 - Physical → virtual switch 3x physical.
1630 2.3.2 Packet Latency tests
1631 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
1632 These tests will measure the store and forward latency as well as the packet
1633 delay variation for various packet types through the virtual switch. The
1634 following list is not exhaustive but should indicate the type of tests
1635 that should be required. It is expected that more will be added.
1637 Test ID: LTD.PacketLatency.InitialPacketProcessingLatency
1638 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1639 **Title**: Initial Packet Processing Latency
1641 **Prerequisite Test**: N/A
1647 In some virtual switch architectures, the first packets of a flow will
1648 take the system longer to process than subsequent packets in the flow.
1649 This test determines the latency for these packets. The test will
1650 measure the latency of the packets as they are processed by the
1651 flow-setup-path of the DUT. There are two methods for this test, a
1652 recommended method and a nalternative method that can be used if it is
1653 possible to disable the fastpath of the virtual switch.
1655 Recommended method: This test will send 64,000 packets to the DUT, each
1656 belonging to a different flow. Average packet latency will be determined
1657 over the 64,000 packets.
1659 Alternative method: This test will send a single packet to the DUT after
1660 a fixed interval of time. The time interval will be equivalent to the
1661 amount of time it takes for a flow to time out in the virtual switch
1662 plus 10%. Average packet latency will be determined over 1,000,000
1665 This test is intended only for non-learning switches; For learning
1666 switches use RFC2889.
1668 For this test, only unidirectional traffic is required.
1670 **Expected Result**: The average latency for the initial packet of all
1671 flows should be greater than the latency of subsequent traffic.
1673 **Metrics Collected**:
1675 The following are the metrics collected for this test:
1677 - Average latency of the initial packets of all flows that are
1678 processed by the DUT.
1680 **Deployment scenario**:
1682 - Physical → Virtual Switch → Physical.
1684 Test ID: LTD.PacketDelayVariation.RFC3393.Soak
1685 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1686 **Title**: Packet Delay Variation Soak Test
1688 **Prerequisite Tests**: LTD.Throughput.RFC2544.PacketLossRatio (0% Packet Loss)
1694 The aim of this test is to understand the distribution of packet delay
1695 variation for different frame sizes over an extended test duration and
1696 to determine if there are any outliers. To allow for an extended test
1697 duration, the test should ideally run for 24 hours or, if this is not
1698 possible, for at least 6 hour. For this test, each frame size must be
1699 sent at the highest possible throughput with 0% packet loss, as
1700 determined in the prerequisite test.
1702 **Expected Result**:
1704 **Metrics Collected**:
1706 The following are the metrics collected for this test:
1708 - The packet delay variation value for traffic passing through the DUT.
1709 - The `RFC5481 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__
1710 PDV form of delay variation on the traffic flow,
1711 using the 99th percentile, for each 60s interval during the test.
1712 - CPU and memory utilization may also be collected as part of this
1713 test, to determine the vSwitch's performance footprint on the system.
1715 2.3.3 Scalability tests
1716 ~~~~~~~~~~~~~~~~~~~~~~~~
1717 The general aim of these tests is to understand the impact of large flow
1718 table size and flow lookups on throughput. The following list is not
1719 exhaustive but should indicate the type of tests that should be required.
1720 It is expected that more will be added.
1722 Test ID: LTD.Scalability.RFC2544.0PacketLoss
1723 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1724 **Title**: RFC 2544 0% loss Scalability throughput test
1726 **Prerequisite Test**:
1732 The aim of this test is to measure how throughput changes as the number
1733 of flows in the DUT increases. The test will measure the throughput
1734 through the fastpath, as such the flows need to be installed on the DUT
1735 before passing traffic.
1737 For each frame size previously defined under `Default Test
1738 Parameters <#DefaultParams>`__ and for each of the following number of
1748 - Max supported number of flows.
1750 The maximum 0% packet loss throughput should be determined in a manner
1751 identical to LTD.Throughput.RFC2544.PacketLossRatio.
1753 **Expected Result**:
1755 **Metrics Collected**:
1757 The following are the metrics collected for this test:
1759 - The maximum number of frames per second that can be forwarded at the
1760 specified number of flows and the specified frame size, with zero
1763 Test ID: LTD.MemoryBandwidth.RFC2544.0PacketLoss.Scalability
1764 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1765 **Title**: RFC 2544 0% loss Memory Bandwidth Scalability test
1767 **Prerequisite Tests**:
1773 The aim of this test is to understand how the DUT's performance is
1774 affected by cache sharing and memory bandwidth between processes.
1776 During the test all cores not used by the vSwitch should be running a
1777 memory intensive application. This application should read and write
1778 random data to random addresses in unused physical memory. The random
1779 nature of the data and addresses is intended to consume cache, exercise
1780 main memory access (as opposed to cache) and exercise all memory buses
1781 equally. Furthermore: - the ratio of reads to writes should be recorded.
1782 A ratio of 1:1 SHOULD be used. - the reads and writes MUST be of
1783 cache-line size and be cache-line aligned. - in NUMA architectures
1784 memory access SHOULD be local to the core's node. Whether only local
1785 memory or a mix of local and remote memory is used MUST be recorded. -
1786 the memory bandwidth (reads plus writes) used per-core MUST be recorded;
1787 the test MUST be run with a per-core memory bandwidth equal to half the
1788 maximum system memory bandwidth divided by the number of cores. The test
1789 MAY be run with other values for the per-core memory bandwidth. - the
1790 test MAY also be run with the memory intensive application running on
1793 Under these conditions the DUT's 0% packet loss throughput is determined
1794 as per LTD.Throughput.RFC2544.PacketLossRatio.
1796 **Expected Result**:
1798 **Metrics Collected**:
1800 The following are the metrics collected for this test:
1802 - The DUT's 0% packet loss throughput in the presence of cache sharing and memory bandwidth between processes.
1804 2.3.5 Coupling between control path and datapath Tests
1805 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1806 The following tests aim to determine how tightly coupled the datapath
1807 and the control path are within a virtual switch. The following list
1808 is not exhaustive but should indicate the type of tests that should be
1809 required. It is expected that more will be added.
1811 Test ID: LTD.CPDPCouplingFlowAddition
1812 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1813 **Title**: Control Path and Datapath Coupling
1815 **Prerequisite Test**:
1821 The aim of this test is to understand how exercising the DUT's control
1822 path affects datapath performance.
1824 Initially a certain number of flow table entries are installed in the
1825 vSwitch. Then over the duration of an RFC2544 throughput test
1826 flow-entries are added and removed at the rates specified below. No
1827 traffic is 'hitting' these flow-entries, they are simply added and
1830 The test MUST be repeated with the following initial number of
1831 flow-entries installed: - < 10 - 1000 - 100,000 - 10,000,000 (or the
1832 maximum supported number of flow-entries)
1834 The test MUST be repeated with the following rates of flow-entry
1835 addition and deletion per second: - 0 - 1 (i.e. 1 addition plus 1
1836 deletion) - 100 - 10,000
1838 **Expected Result**:
1840 **Metrics Collected**:
1842 The following are the metrics collected for this test:
1844 - The maximum forwarding rate in Frames Per Second (FPS) and Mbps of
1846 - The average latency of the traffic flow when passing through the DUT
1847 (if testing for latency, note that this average is different from the
1848 test specified in Section 26.3 of
1849 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__).
1850 - CPU and memory utilization may also be collected as part of this
1851 test, to determine the vSwitch's performance footprint on the system.
1853 **Deployment scenario**:
1855 - Physical → virtual switch → physical.
1857 2.3.4 CPU and memory consumption
1858 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1859 The following tests will profile a virtual switch's CPU and memory
1860 utilization under various loads and circumstances. The following
1861 list is not exhaustive but should indicate the type of tests that
1862 should be required. It is expected that more will be added.
1864 Test ID: LTD.CPU.RFC2544.0PacketLoss
1865 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1866 **Title**: RFC 2544 0% Loss Compute Test
1868 **Prerequisite Test**:
1874 The aim of this test is to understand the overall performance of the
1875 system when a CPU intensive application is run on the same DUT as the
1876 Virtual Switch. For each frame size, an
1877 LTD.Throughput.RFC2544.PacketLossRatio (0% Packet Loss) test should be
1878 performed. Throughout the entire test a CPU intensive application should
1879 be run on all cores on the system not in use by the Virtual Switch. For
1880 NUMA system only cores on the same NUMA node are loaded.
1882 It is recommended that stress-ng be used for loading the non-Virtual
1883 Switch cores but any stress tool MAY be used.
1885 **Expected Result**:
1887 **Metrics Collected**:
1889 The following are the metrics collected for this test:
1891 - CPU utilization of the cores running the Virtual Switch.
1892 - The number of identity of the cores allocated to the Virtual Switch.
1893 - The configuration of the stress tool (for example the command line
1894 parameters used to start it.)
1896 2.3.9 Summary List of Tests
1897 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
1900 - Test ID: LTD.Throughput.RFC2544.PacketLossRatio
1901 - Test ID: LTD.Throughput.RFC2544.PacketLossRatioFrameModification
1902 - Test ID: LTD.Throughput.RFC2544.Profile
1903 - Test ID: LTD.Throughput.RFC2544.SystemRecoveryTime
1904 - Test ID: LTD.Throughput.RFC2544.BackToBackFrames
1905 - Test ID: LTD.Throughput.RFC2889.Soak
1906 - Test ID: LTD.Throughput.RFC2889.SoakFrameModification
1907 - Test ID: LTD.Throughput.RFC6201.ResetTime
1908 - Test ID: LTD.Throughput.RFC2889.MaxForwardingRate
1909 - Test ID: LTD.Throughput.RFC2889.ForwardPressure
1910 - Test ID: LTD.Throughput.RFC2889.AddressCachingCapacity
1911 - Test ID: LTD.Throughput.RFC2889.AddressLearningRate
1912 - Test ID: LTD.Throughput.RFC2889.ErrorFramesFiltering
1913 - Test ID: LTD.Throughput.RFC2889.BroadcastFrameForwarding
1915 2. Packet Latency tests
1917 - Test ID: LTD.PacketLatency.InitialPacketProcessingLatency
1918 - Test ID: LTD.PacketDelayVariation.RFC3393.Soak
1920 3. Scalability tests
1922 - Test ID: LTD.Scalability.RFC2544.0PacketLoss
1923 - Test ID: LTD.MemoryBandwidth.RFC2544.0PacketLoss.Scalability
1925 4. Coupling between control path and datapath Tests
1927 - Test ID: LTD.CPDPCouplingFlowAddition
1929 5. CPU and memory consumption
1931 - Test ID: LTD.CPU.RFC2544.0PacketLoss