1 .. This work is licensed under a Creative Commons Attribution 4.0 International License.
2 .. http://creativecommons.org/licenses/by/4.0
3 .. (c) OPNFV, Intel Corporation, AT&T and others.
11 The objective of the OPNFV project titled
12 **“Characterize vSwitch Performance for Telco NFV Use Cases”**, is to
13 evaluate a virtual switch to identify its suitability for a Telco
14 Network Function Virtualization (NFV) environment. The intention of this
15 Level Test Design (LTD) document is to specify the set of tests to carry
16 out in order to objectively measure the current characteristics of a
17 virtual switch in the Network Function Virtualization Infrastructure
18 (NFVI) as well as the test pass criteria. The detailed test cases will
19 be defined in details-of-LTD_, preceded by the doc-id_ and the scope_.
21 This document is currently in draft form.
29 =========================
31 The document id will be used to uniquely
32 identify versions of the LTD. The format for the document id will be:
33 OPNFV\_vswitchperf\_LTD\_REL\_STATUS, where by the
34 status is one of: draft, reviewed, corrected or final. The document id
35 for this version of the LTD is:
36 OPNFV\_vswitchperf\_LTD\_Brahmaputra\_REVIEWED.
45 The main purpose of this project is to specify a suite of
46 performance tests in order to objectively measure the current packet
47 transfer characteristics of a virtual switch in the NFVI. The intent of
48 the project is to facilitate testing of any virtual switch. Thus, a
49 generic suite of tests shall be developed, with no hard dependencies to
50 a single implementation. In addition, the test case suite shall be
51 architecture independent.
53 The test cases developed in this project shall not form part of a
54 separate test framework, all of these tests may be inserted into the
55 Continuous Integration Test Framework and/or the Platform Functionality
56 Test Framework - if a vSwitch becomes a standard component of an OPNFV
64 * `RFC 1242 Benchmarking Terminology for Network Interconnection
65 Devices <http://www.ietf.org/rfc/rfc1242.txt>`__
66 * `RFC 2544 Benchmarking Methodology for Network Interconnect
67 Devices <http://www.ietf.org/rfc/rfc2544.txt>`__
68 * `RFC 2285 Benchmarking Terminology for LAN Switching
69 Devices <http://www.ietf.org/rfc/rfc2285.txt>`__
70 * `RFC 2889 Benchmarking Methodology for LAN Switching
71 Devices <http://www.ietf.org/rfc/rfc2889.txt>`__
72 * `RFC 3918 Methodology for IP Multicast
73 Benchmarking <http://www.ietf.org/rfc/rfc3918.txt>`__
74 * `RFC 4737 Packet Reordering
75 Metrics <http://www.ietf.org/rfc/rfc4737.txt>`__
76 * `RFC 5481 Packet Delay Variation Applicability
77 Statement <http://www.ietf.org/rfc/rfc5481.txt>`__
78 * `RFC 6201 Device Reset
79 Characterization <http://tools.ietf.org/html/rfc6201>`__
85 ===================================
86 Details of the Level Test Design
87 ===================================
89 This section describes the features to be tested (
90 FeaturesToBeTested_), the test approach (Approach_);
91 it also identifies the sets of test cases or scenarios (
92 TestIdentification_) along with the pass/fail criteria and
93 the test deliverables.
97 .. _FeaturesToBeTested:
100 ==========================
102 Characterizing virtual switches (i.e. Device Under Test (DUT) in this document)
103 includes measuring the following performance metrics:
105 - **Throughput** as defined by `RFC1242
106 <https://www.rfc-editor.org/rfc/rfc1242.txt>`__: The maximum rate at which
107 **none** of the offered frames are dropped by the DUT. The maximum frame
108 rate and bit rate that can be transmitted by the DUT without any error
109 should be recorded. Note there is an equivalent bit rate and a specific
110 layer at which the payloads contribute to the bits. Errors and
111 improperly formed frames or packets are dropped.
112 - **Packet delay** introduced by the DUT and its cumulative effect on
113 E2E networks. Frame delay can be measured equivalently.
114 - **Packet delay variation**: measured from the perspective of the
115 VNF/application. Packet delay variation is sometimes called "jitter".
116 However, we will avoid the term "jitter" as the term holds different
117 meaning to different groups of people. In this document we will
118 simply use the term packet delay variation. The preferred form for this
119 metric is the PDV form of delay variation defined in `RFC5481
120 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__. The most relevant
121 measurement of PDV considers the delay variation of a single user flow,
122 as this will be relevant to the size of end-system buffers to compensate
123 for delay variation. The measurement system's ability to store the
124 delays of individual packets in the flow of interest is a key factor
125 that determines the specific measurement method. At the outset, it is
126 ideal to view the complete PDV distribution. Systems that can capture
127 and store packets and their delays have the freedom to calculate the
128 reference minimum delay and to determine various quantiles of the PDV
129 distribution accurately (in post-measurement processing routines).
130 Systems without storage must apply algorithms to calculate delay and
131 statistical measurements on the fly. For example, a system may store
132 temporary estimates of the mimimum delay and the set of (100) packets
133 with the longest delays during measurement (to calculate a high quantile,
134 and update these sets with new values periodically.
135 In some cases, a limited number of delay histogram bins will be
136 available, and the bin limits will need to be set using results from
137 repeated experiments. See section 8 of `RFC5481
138 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__.
139 - **Packet loss** (within a configured waiting time at the receiver): All
140 packets sent to the DUT should be accounted for.
141 - **Burst behaviour**: measures the ability of the DUT to buffer packets.
142 - **Packet re-ordering**: measures the ability of the device under test to
143 maintain sending order throughout transfer to the destination.
144 - **Packet correctness**: packets or Frames must be well-formed, in that
145 they include all required fields, conform to length requirements, pass
146 integrity checks, etc.
147 - **Availability and capacity** of the DUT i.e. when the DUT is fully “up”
148 and connected, following measurements should be captured for
149 DUT without any network packet load:
151 - Includes average power consumption of the CPUs (in various power states) and
152 system over specified period of time. Time period should not be less
154 - Includes average per core CPU utilization over specified period of time.
155 Time period should not be less than 60 seconds.
156 - Includes the number of NIC interfaces supported.
157 - Includes headroom of VM workload processing cores (i.e. available
167 In order to determine the packet transfer characteristics of a virtual
168 switch, the tests will be broken down into the following categories:
173 ----------------------
174 - **Throughput Tests** to measure the maximum forwarding rate (in
175 frames per second or fps) and bit rate (in Mbps) for a constant load
176 (as defined by `RFC1242 <https://www.rfc-editor.org/rfc/rfc1242.txt>`__)
177 without traffic loss.
178 - **Packet and Frame Delay Tests** to measure average, min and max
179 packet and frame delay for constant loads.
180 - **Stream Performance Tests** (TCP, UDP) to measure bulk data transfer
181 performance, i.e. how fast systems can send and receive data through
183 - **Request/Response Performance** Tests (TCP, UDP) the measure the
184 transaction rate through the virtual switch.
185 - **Packet Delay Tests** to understand latency distribution for
186 different packet sizes and over an extended test run to uncover
188 - **Scalability Tests** to understand how the virtual switch performs
189 as the number of flows, active ports, complexity of the forwarding
190 logic's configuration... it has to deal with increases.
191 - **Control Path and Datapath Coupling** Tests, to understand how
192 closely coupled the datapath and the control path are as well as the
193 effect of this coupling on the performance of the DUT.
194 - **CPU and Memory Consumption Tests** to understand the virtual
195 switch’s footprint on the system, this includes:
197 * CPU core utilization.
198 * CPU cache utilization.
200 * System bus (QPI, PCI, ..) utilization.
201 * Memory lanes utilization.
202 * CPU cycles consumed per packet.
203 * Time To Establish Flows Tests.
205 - **Noisy Neighbour Tests**, to understand the effects of resource
206 sharing on the performance of a virtual switch.
208 **Note:** some of the tests above can be conducted simultaneously where
209 the combined results would be insightful, for example Packet/Frame Delay
215 --------------------------
216 The following represents possible deployment test scenarios which can
217 help to determine the performance of both the virtual switch and the
218 datapaths to physical ports (to NICs) and to logical ports (to VNFs):
222 Physical port → vSwitch → physical port
223 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
224 .. code-block:: console
227 +--------------------------------------------------+ |
228 | +--------------------+ | |
231 | +--------------+ +--------------+ | |
232 | | phy port | vSwitch | phy port | | |
233 +---+--------------+------------+--------------+---+ _|
237 +--------------------------------------------------+
239 | traffic generator |
241 +--------------------------------------------------+
245 Physical port → vSwitch → VNF → vSwitch → physical port
246 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
247 .. code-block:: console
250 +---------------------------------------------------+ |
252 | +-------------------------------------------+ | |
253 | | Application | | |
254 | +-------------------------------------------+ | |
258 | +---------------+ +---------------+ | |
259 | | logical port 0| | logical port 1| | |
260 +---+---------------+-----------+---------------+---+ _|
264 +---+---------------+----------+---------------+---+ |
265 | | logical port 0| | logical port 1| | |
266 | +---------------+ +---------------+ | |
270 | +--------------+ +--------------+ | |
271 | | phy port | vSwitch | phy port | | |
272 +---+--------------+------------+--------------+---+ _|
276 +--------------------------------------------------+
278 | traffic generator |
280 +--------------------------------------------------+
284 Physical port → vSwitch → VNF → vSwitch → VNF → vSwitch → physical port
285 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
287 .. code-block:: console
290 +----------------------+ +----------------------+ |
291 | Guest 1 | | Guest 2 | |
292 | +---------------+ | | +---------------+ | |
293 | | Application | | | | Application | | |
294 | +---------------+ | | +---------------+ | |
296 | | v | | | v | | Guests
297 | +---------------+ | | +---------------+ | |
298 | | logical ports | | | | logical ports | | |
299 | | 0 1 | | | | 0 1 | | |
300 +---+---------------+--+ +---+---------------+--+ _|
304 +---+---------------+---------+---------------+--+ |
305 | | 0 1 | | 3 4 | | |
306 | | logical ports | | logical ports | | |
307 | +---------------+ +---------------+ | |
309 | | L-----------------+ v | |
310 | +--------------+ +--------------+ | |
311 | | phy ports | vSwitch | phy ports | | |
312 +---+--------------+----------+--------------+---+ _|
316 +--------------------------------------------------+
318 | traffic generator |
320 +--------------------------------------------------+
324 Physical port → VNF → vSwitch → VNF → physical port
325 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
327 .. code-block:: console
330 +----------------------+ +----------------------+ |
331 | Guest 1 | | Guest 2 | |
332 |+-------------------+ | | +-------------------+| |
333 || Application | | | | Application || |
334 |+-------------------+ | | +-------------------+| |
335 | ^ | | | ^ | | | Guests
337 |+-------------------+ | | +-------------------+| |
338 || logical ports | | | | logical ports || |
339 || 0 1 | | | | 0 1 || |
340 ++--------------------++ ++--------------------++ _|
342 (PCI passthrough) | | (PCI passthrough)
344 +--------++------------+-+------------++---------+ |
345 | | || 0 | | 1 || | | |
346 | | ||logical port| |logical port|| | | |
347 | | |+------------+ +------------+| | | |
349 | | | L-----------------+ | | | |
351 | | | vSwitch | | | |
352 | | +-----------------------------+ | | |
355 | +--------------+ +--------------+ | |
356 | | phy port/VF | | phy port/VF | | |
357 +-+--------------+--------------+--------------+-+ _|
361 +--------------------------------------------------+
363 | traffic generator |
365 +--------------------------------------------------+
369 Physical port → vSwitch → VNF
370 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
372 .. code-block:: console
375 +---------------------------------------------------+ |
377 | +-------------------------------------------+ | |
378 | | Application | | |
379 | +-------------------------------------------+ | |
383 | +---------------+ | |
384 | | logical port 0| | |
385 +---+---------------+-------------------------------+ _|
389 +---+---------------+------------------------------+ |
390 | | logical port 0| | |
391 | +---------------+ | |
395 | +--------------+ | |
396 | | phy port | vSwitch | |
397 +---+--------------+------------ -------------- ---+ _|
401 +--------------------------------------------------+
403 | traffic generator |
405 +--------------------------------------------------+
409 VNF → vSwitch → physical port
410 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
412 .. code-block:: console
415 +---------------------------------------------------+ |
417 | +-------------------------------------------+ | |
418 | | Application | | |
419 | +-------------------------------------------+ | |
423 | +---------------+ | |
424 | | logical port | | |
425 +-------------------------------+---------------+---+ _|
429 +------------------------------+---------------+---+ |
430 | | logical port | | |
431 | +---------------+ | |
435 | +--------------+ | |
436 | vSwitch | phy port | | |
437 +-------------------------------+--------------+---+ _|
441 +--------------------------------------------------+
443 | traffic generator |
445 +--------------------------------------------------+
449 VNF → vSwitch → VNF → vSwitch
450 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
452 .. code-block:: console
455 +-------------------------+ +-------------------------+ |
456 | Guest 1 | | Guest 2 | |
457 | +-----------------+ | | +-----------------+ | |
458 | | Application | | | | Application | | |
459 | +-----------------+ | | +-----------------+ | |
463 | +---------------+ | | +---------------+ | |
464 | | logical port 0| | | | logical port 0| | |
465 +-----+---------------+---+ +---+---------------+-----+ _|
469 +----+---------------+------------+---------------+-----+ |
470 | | port 0 | | port 1 | | |
471 | +---------------+ +---------------+ | |
474 | +--------------------+ | |
477 +-------------------------------------------------------+ _|
481 HOST 1(Physical port → virtual switch → VNF → virtual switch → Physical port)
482 → HOST 2(Physical port → virtual switch → VNF → virtual switch → Physical port)
484 HOST 1 (PVP) → HOST 2 (PVP)
485 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
487 .. code-block:: console
490 +----------------------+ +----------------------+ |
491 | Guest 1 | | Guest 2 | |
492 | +---------------+ | | +---------------+ | |
493 | | Application | | | | Application | | |
494 | +---------------+ | | +---------------+ | |
496 | | v | | | v | | Guests
497 | +---------------+ | | +---------------+ | |
498 | | logical ports | | | | logical ports | | |
499 | | 0 1 | | | | 0 1 | | |
500 +---+---------------+--+ +---+---------------+--+ _|
504 +---+---------------+--+ +---+---------------+--+ |
505 | | 0 1 | | | | 3 4 | | |
506 | | logical ports | | | | logical ports | | |
507 | +---------------+ | | +---------------+ | |
508 | ^ | | | ^ | | | Hosts
510 | +--------------+ | | +--------------+ | |
511 | | phy ports | | | | phy ports | | |
512 +---+--------------+---+ +---+--------------+---+ _|
514 | +-----------------+ |
516 +--------------------------------------------------+
518 | traffic generator |
520 +--------------------------------------------------+
524 **Note:** For tests where the traffic generator and/or measurement
525 receiver are implemented on VM and connected to the virtual switch
526 through vNIC, the issues of shared resources and interactions between
527 the measurement devices and the device under test must be considered.
529 **Note:** Some RFC 2889 tests require a full-mesh sending and receiving
530 pattern involving more than two ports. This possibility is illustrated in the
531 Physical port → vSwitch → VNF → vSwitch → VNF → vSwitch → physical port
532 diagram above (with 2 sending and 2 receiving ports, though all ports
533 could be used bi-directionally).
535 **Note:** When Deployment Scenarios are used in RFC 2889 address learning
536 or cache capacity testing, an additional port from the vSwitch must be
537 connected to the test device. This port is used to listen for flooded
543 --------------------------
544 To establish the baseline performance of the virtual switch, tests would
545 initially be run with a simple workload in the VNF (the recommended
546 simple workload VNF would be `DPDK <http://www.dpdk.org/>`__'s testpmd
547 application forwarding packets in a VM or vloop\_vnf a simple kernel
548 module that forwards traffic between two network interfaces inside the
549 virtualized environment while bypassing the networking stack).
550 Subsequently, the tests would also be executed with a real Telco
551 workload running in the VNF, which would exercise the virtual switch in
552 the context of higher level Telco NFV use cases, and prove that its
553 underlying characteristics and behaviour can be measured and validated.
554 Suitable real Telco workload VNFs are yet to be identified.
558 Default Test Parameters
559 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
561 The following list identifies the default parameters for suite of
564 - Reference application: Simple forwarding or Open Source VNF.
565 - Frame size (bytes): 64, 128, 256, 512, 1024, 1280, 1518, 2K, 4k OR
566 Packet size based on use-case (e.g. RTP 64B, 256B) OR Mix of packet sizes as
567 maintained by the Functest project <https://wiki.opnfv.org/traffic_profile_management>.
568 - Reordering check: Tests should confirm that packets within a flow are
570 - Duplex: Unidirectional / Bidirectional. Default: Full duplex with
571 traffic transmitting in both directions, as network traffic generally
572 does not flow in a single direction. By default the data rate of
573 transmitted traffic should be the same in both directions, please
574 note that asymmetric traffic (e.g. downlink-heavy) tests will be
575 mentioned explicitly for the relevant test cases.
576 - Number of Flows: Default for non scalability tests is a single flow.
577 For scalability tests the goal is to test with maximum supported
578 flows but where possible will test up to 10 Million flows. Start with
579 a single flow and scale up. By default flows should be added
580 sequentially, tests that add flows simultaneously will explicitly
581 call out their flow addition behaviour. Packets are generated across
582 the flows uniformly with no burstiness. For multi-core tests should
583 consider the number of packet flows based on vSwitch/VNF multi-thread
584 implementation and behavior.
586 - Traffic Types: UDP, SCTP, RTP, GTP and UDP traffic.
587 - Deployment scenarios are:
588 - Physical → virtual switch → physical.
589 - Physical → virtual switch → VNF → virtual switch → physical.
590 - Physical → virtual switch → VNF → virtual switch → VNF → virtual
592 - Physical → VNF → virtual switch → VNF → physical.
593 - Physical → virtual switch → VNF.
594 - VNF → virtual switch → Physical.
595 - VNF → virtual switch → VNF.
597 Tests MUST have these parameters unless otherwise stated. **Test cases
598 with non default parameters will be stated explicitly**.
600 **Note**: For throughput tests unless stated otherwise, test
601 configurations should ensure that traffic traverses the installed flows
602 through the virtual switch, i.e. flows are installed and have an appropriate
603 time out that doesn't expire before packet transmission starts.
608 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
610 Virtual switches classify packets into flows by processing and matching
611 particular header fields in the packet/frame and/or the input port where
612 the packets/frames arrived. The vSwitch then carries out an action on
613 the group of packets that match the classification parameters. Thus a
614 flow is considered to be a sequence of packets that have a shared set of
615 header field values or have arrived on the same port and have the same
616 action applied to them. Performance results can vary based on the
617 parameters the vSwitch uses to match for a flow. The recommended flow
618 classification parameters for L3 vSwitch performance tests are: the
619 input port, the source IP address, the destination IP address and the
620 Ethernet protocol type field. It is essential to increase the flow
621 time-out time on a vSwitch before conducting any performance tests that
622 do not measure the flow set-up time. Normally the first packet of a
623 particular flow will install the flow in the vSwitch which adds an
624 additional latency, subsequent packets of the same flow are not subject
625 to this latency if the flow is already installed on the vSwitch.
630 ~~~~~~~~~~~~~~~~~~~~~
632 Tests will be assigned a priority in order to determine which tests
633 should be implemented immediately and which tests implementations
636 Priority can be of following types: - Urgent: Must be implemented
637 immediately. - High: Must be implemented in the next release. - Medium:
638 May be implemented after the release. - Low: May or may not be
646 The SUT should be configured to its "default" state. The
647 SUT's configuration or set-up must not change between tests in any way
648 other than what is required to do the test. All supported protocols must
649 be configured and enabled for each test set up.
654 ~~~~~~~~~~~~~~~~~~~~~~~~~~
656 The DUT should be configured with n ports where
657 n is a multiple of 2. Half of the ports on the DUT should be used as
658 ingress ports and the other half of the ports on the DUT should be used
659 as egress ports. Where a DUT has more than 2 ports, the ingress data
660 streams should be set-up so that they transmit packets to the egress
661 ports in sequence so that there is an even distribution of traffic
662 across ports. For example, if a DUT has 4 ports 0(ingress), 1(ingress),
663 2(egress) and 3(egress), the traffic stream directed at port 0 should
664 output a packet to port 2 followed by a packet to port 3. The traffic
665 stream directed at port 1 should also output a packet to port 2 followed
666 by a packet to port 3.
671 ~~~~~~~~~~~~~~~~~~~~~
673 **Frame formats Layer 2 (data link layer) protocols**
677 .. code-block:: console
679 +---------------------------+-----------+
680 | Ethernet Header | Payload | Check Sum |
681 +-----------------+---------+-----------+
682 |_________________|_________|___________|
683 14 Bytes 46 - 1500 4 Bytes
687 **Layer 3 (network layer) protocols**
691 .. code-block:: console
693 +-----------------+-----------+---------+-----------+
694 | Ethernet Header | IP Header | Payload | Checksum |
695 +-----------------+-----------+---------+-----------+
696 |_________________|___________|_________|___________|
697 14 Bytes 20 bytes 26 - 1480 4 Bytes
702 .. code-block:: console
704 +-----------------+-----------+---------+-----------+
705 | Ethernet Header | IP Header | Payload | Checksum |
706 +-----------------+-----------+---------+-----------+
707 |_________________|___________|_________|___________|
708 14 Bytes 40 bytes 26 - 1460 4 Bytes
711 **Layer 4 (transport layer) protocols**
717 .. code-block:: console
719 +-----------------+-----------+-----------------+---------+-----------+
720 | Ethernet Header | IP Header | Layer 4 Header | Payload | Checksum |
721 +-----------------+-----------+-----------------+---------+-----------+
722 |_________________|___________|_________________|_________|___________|
723 14 Bytes 40 bytes 20 Bytes 6 - 1460 4 Bytes
727 **Layer 5 (application layer) protocols**
732 .. code-block:: console
734 +-----------------+-----------+-----------------+---------+-----------+
735 | Ethernet Header | IP Header | Layer 4 Header | Payload | Checksum |
736 +-----------------+-----------+-----------------+---------+-----------+
737 |_________________|___________|_________________|_________|___________|
738 14 Bytes 20 bytes 20 Bytes >= 6 Bytes 4 Bytes
743 ~~~~~~~~~~~~~~~~~~~~~~~~~
744 There is a difference between an Ethernet frame,
745 an IP packet, and a UDP datagram. In the seven-layer OSI model of
746 computer networking, packet refers to a data unit at layer 3 (network
747 layer). The correct term for a data unit at layer 2 (data link layer) is
748 a frame, and at layer 4 (transport layer) is a segment or datagram.
750 Important concepts related to 10GbE performance are frame rate and
751 throughput. The MAC bit rate of 10GbE, defined in the IEEE standard 802
752 .3ae, is 10 billion bits per second. Frame rate is based on the bit rate
753 and frame format definitions. Throughput, defined in IETF RFC 1242, is
754 the highest rate at which the system under test can forward the offered
757 The frame rate for 10GbE is determined by a formula that divides the 10
758 billion bits per second by the preamble + frame length + inter-frame
761 The maximum frame rate is calculated using the minimum values of the
762 following parameters, as described in the IEEE 802 .3ae standard:
764 - Preamble: 8 bytes \* 8 = 64 bits
765 - Frame Length: 64 bytes (minimum) \* 8 = 512 bits
766 - Inter-frame Gap: 12 bytes (minimum) \* 8 = 96 bits
768 Therefore, Maximum Frame Rate (64B Frames)
769 = MAC Transmit Bit Rate / (Preamble + Frame Length + Inter-frame Gap)
770 = 10,000,000,000 / (64 + 512 + 96)
771 = 10,000,000,000 / 672
772 = 14,880,952.38 frame per second (fps)
776 System isolation and validation
777 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
779 A key consideration when conducting any sort of benchmark is trying to
780 ensure the consistency and repeatability of test results between runs.
781 When benchmarking the performance of a virtual switch there are many
782 factors that can affect the consistency of results. This section
783 describes these factors and the measures that can be taken to limit
784 their effects. In addition, this section will outline some system tests
785 to validate the platform and the VNF before conducting any vSwitch
788 **System Isolation:**
790 When conducting a benchmarking test on any SUT, it is essential to limit
791 (and if reasonable, eliminate) any noise that may interfere with the
792 accuracy of the metrics collected by the test. This noise may be
793 introduced by other hardware or software (OS, other applications), and
794 can result in significantly varying performance metrics being collected
795 between consecutive runs of the same test. In the case of characterizing
796 the performance of a virtual switch, there are a number of configuration
797 parameters that can help increase the repeatability and stability of
798 test results, including:
800 - OS/GRUB configuration:
802 - maxcpus = n where n >= 0; limits the kernel to using 'n'
803 processors. Only use exactly what you need.
804 - isolcpus: Isolate CPUs from the general scheduler. Isolate all
805 CPUs bar one which will be used by the OS.
806 - use taskset to affinitize the forwarding application and the VNFs
807 onto isolated cores. VNFs and the vSwitch should be allocated
808 their own cores, i.e. must not share the same cores. vCPUs for the
809 VNF should be affinitized to individual cores also.
810 - Limit the amount of background applications that are running and
811 set OS to boot to runlevel 3. Make sure to kill any unnecessary
812 system processes/daemons.
813 - Only enable hardware that you need to use for your test – to
814 ensure there are no other interrupts on the system.
815 - Configure NIC interrupts to only use the cores that are not
816 allocated to any other process (VNF/vSwitch).
818 - NUMA configuration: Any unused sockets in a multi-socket system
820 - CPU pinning: The vSwitch and the VNF should each be affinitized to
821 separate logical cores using a combination of maxcpus, isolcpus and
823 - BIOS configuration: BIOS should be configured for performance where
824 an explicit option exists, sleep states should be disabled, any
825 virtualization optimization technologies should be enabled, and
826 hyperthreading should also be enabled, turbo boost and overclocking
829 **System Validation:**
831 System validation is broken down into two sub-categories: Platform
832 validation and VNF validation. The validation test itself involves
833 verifying the forwarding capability and stability for the sub-system
834 under test. The rationale behind system validation is two fold. Firstly
835 to give a tester confidence in the stability of the platform or VNF that
836 is being tested; and secondly to provide base performance comparison
837 points to understand the overhead introduced by the virtual switch.
839 * Benchmark platform forwarding capability: This is an OPTIONAL test
840 used to verify the platform and measure the base performance (maximum
841 forwarding rate in fps and latency) that can be achieved by the
842 platform without a vSwitch or a VNF. The following diagram outlines
843 the set-up for benchmarking Platform forwarding capability:
845 .. code-block:: console
848 +--------------------------------------------------+ |
849 | +------------------------------------------+ | |
851 | | l2fw or DPDK L2FWD app | | Host
853 | +------------------------------------------+ | |
855 +---+------------------------------------------+---+ __|
859 +--------------------------------------------------+
861 | traffic generator |
863 +--------------------------------------------------+
865 * Benchmark VNF forwarding capability: This test is used to verify
866 the VNF and measure the base performance (maximum forwarding rate in
867 fps and latency) that can be achieved by the VNF without a vSwitch.
868 The performance metrics collected by this test will serve as a key
869 comparison point for NIC passthrough technologies and vSwitches. VNF
870 in this context refers to the hypervisor and the VM. The following
871 diagram outlines the set-up for benchmarking VNF forwarding
874 .. code-block:: console
877 +--------------------------------------------------+ |
878 | +------------------------------------------+ | |
882 | +------------------------------------------+ | |
883 | | Passthrough/SR-IOV | | Host
884 | +------------------------------------------+ | |
886 +---+------------------------------------------+---+ __|
890 +--------------------------------------------------+
892 | traffic generator |
894 +--------------------------------------------------+
897 **Methodology to benchmark Platform/VNF forwarding capability**
900 The recommended methodology for the platform/VNF validation and
901 benchmark is: - Run `RFC2889 <https://www.rfc-editor.org/rfc/rfc2289.txt>`__
902 Maximum Forwarding Rate test, this test will produce maximum
903 forwarding rate and latency results that will serve as the
904 expected values. These expected values can be used in
905 subsequent steps or compared with in subsequent validation tests. -
906 Transmit bidirectional traffic at line rate/max forwarding rate
907 (whichever is higher) for at least 72 hours, measure throughput (fps)
908 and latency. - Note: Traffic should be bidirectional. - Establish a
909 baseline forwarding rate for what the platform can achieve. - Additional
910 validation: After the test has completed for 72 hours run bidirectional
911 traffic at the maximum forwarding rate once more to see if the system is
912 still functional and measure throughput (fps) and latency. Compare the
913 measure the new obtained values with the expected values.
915 **NOTE 1**: How the Platform is configured for its forwarding capability
916 test (BIOS settings, GRUB configuration, runlevel...) is how the
917 platform should be configured for every test after this
919 **NOTE 2**: How the VNF is configured for its forwarding capability test
920 (# of vCPUs, vNICs, Memory, affinitization…) is how it should be
921 configured for every test that uses a VNF after this.
925 RFCs for testing virtual switch performance
926 --------------------------------------------------
928 The starting point for defining the suite of tests for benchmarking the
929 performance of a virtual switch is to take existing RFCs and standards
930 that were designed to test their physical counterparts and adapting them
931 for testing virtual switches. The rationale behind this is to establish
932 a fair comparison between the performance of virtual and physical
933 switches. This section outlines the RFCs that are used by this
938 RFC 1242 Benchmarking Terminology for Network Interconnection
939 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
940 Devices RFC 1242 defines the terminology that is used in describing
941 performance benchmarking tests and their results. Definitions and
942 discussions covered include: Back-to-back, bridge, bridge/router,
943 constant load, data link frame size, frame loss rate, inter frame gap,
944 latency, and many more.
948 RFC 2544 Benchmarking Methodology for Network Interconnect Devices
949 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
950 RFC 2544 outlines a benchmarking methodology for network Interconnect
951 Devices. The methodology results in performance metrics such as latency,
952 frame loss percentage, and maximum data throughput.
954 In this document network “throughput” (measured in millions of frames
955 per second) is based on RFC 2544, unless otherwise noted. Frame size
956 refers to Ethernet frames ranging from smallest frames of 64 bytes to
957 largest frames of 9K bytes.
961 1. Throughput test defines the maximum number of frames per second
962 that can be transmitted without any error, or 0% loss ratio.
963 In some Throughput tests (and those tests with long duration),
964 evaluation of an additional frame loss ratio is suggested. The
965 current ratio (10^-7 %) is based on understanding the typical
966 user-to-user packet loss ratio needed for good application
967 performance and recognizing that a single transfer through a
968 vswitch must contribute a tiny fraction of user-to-user loss.
969 Further, the ratio 10^-7 % also recognizes practical limitations
970 when measuring loss ratio.
972 2. Latency test measures the time required for a frame to travel from
973 the originating device through the network to the destination device.
974 Please note that RFC2544 Latency measurement will be superseded with
975 a measurement of average latency over all successfully transferred
978 3. Frame loss test measures the network’s
979 response in overload conditions - a critical indicator of the
980 network’s ability to support real-time applications in which a
981 large amount of frame loss will rapidly degrade service quality.
983 4. Burst test assesses the buffering capability of a virtual switch. It
984 measures the maximum number of frames received at full line rate
985 before a frame is lost. In carrier Ethernet networks, this
986 measurement validates the excess information rate (EIR) as defined in
989 5. System recovery to characterize speed of recovery from an overload
992 6. Reset to characterize speed of recovery from device or software
993 reset. This type of test has been updated by `RFC6201
994 <https://www.rfc-editor.org/rfc/rfc6201.txt>`__ as such,
995 the methodology defined by this specification will be that of RFC 6201.
997 Although not included in the defined RFC 2544 standard, another crucial
998 measurement in Ethernet networking is packet delay variation. The
999 definition set out by this specification comes from
1000 `RFC5481 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__.
1004 RFC 2285 Benchmarking Terminology for LAN Switching Devices
1005 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1006 RFC 2285 defines the terminology that is used to describe the
1007 terminology for benchmarking a LAN switching device. It extends RFC
1008 1242 and defines: DUTs, SUTs, Traffic orientation and distribution,
1009 bursts, loads, forwarding rates, etc.
1013 RFC 2889 Benchmarking Methodology for LAN Switching
1014 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1015 RFC 2889 outlines a benchmarking methodology for LAN switching, it
1016 extends RFC 2544. The outlined methodology gathers performance
1017 metrics for forwarding, congestion control, latency, address handling
1018 and finally filtering.
1022 RFC 3918 Methodology for IP Multicast Benchmarking
1023 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1024 RFC 3918 outlines a methodology for IP Multicast benchmarking.
1028 RFC 4737 Packet Reordering Metrics
1029 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1030 RFC 4737 describes metrics for identifying and counting re-ordered
1031 packets within a stream, and metrics to measure the extent each
1032 packet has been re-ordered.
1036 RFC 5481 Packet Delay Variation Applicability Statement
1037 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1038 RFC 5481 defined two common, but different forms of delay variation
1039 metrics, and compares the metrics over a range of networking
1040 circumstances and tasks. The most suitable form for vSwitch
1041 benchmarking is the "PDV" form.
1045 RFC 6201 Device Reset Characterization
1046 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1047 RFC 6201 extends the methodology for characterizing the speed of
1048 recovery of the DUT from device or software reset described in RFC
1053 Details of the Test Report
1054 ---------------------------------
1056 There are a number of parameters related to the system, DUT and tests
1057 that can affect the repeatability of a test results and should be
1058 recorded. In order to minimise the variation in the results of a test,
1059 it is recommended that the test report includes the following information:
1061 - Hardware details including:
1064 - Processor details.
1065 - Memory information (see below)
1066 - Number of enabled cores.
1067 - Number of cores used for the test.
1068 - Number of physical NICs, as well as their details (manufacturer,
1069 versions, type and the PCI slot they are plugged into).
1070 - NIC interrupt configuration.
1071 - BIOS version, release date and any configurations that were
1074 - Software details including:
1076 - OS version (for host and VNF)
1077 - Kernel version (for host and VNF)
1078 - GRUB boot parameters (for host and VNF).
1079 - Hypervisor details (Type and version).
1080 - Selected vSwitch, version number or commit id used.
1081 - vSwitch launch command line if it has been parameterised.
1082 - Memory allocation to the vSwitch – which NUMA node it is using,
1083 and how many memory channels.
1084 - Where the vswitch is built from source: compiler details including
1085 versions and the flags that were used to compile the vSwitch.
1086 - DPDK or any other SW dependency version number or commit id used.
1087 - Memory allocation to a VM - if it's from Hugpages/elsewhere.
1088 - VM storage type: snapshot/independent persistent/independent
1091 - Number of Virtual NICs (vNICs), versions, type and driver.
1092 - Number of virtual CPUs and their core affinity on the host.
1093 - Number vNIC interrupt configuration.
1094 - Thread affinitization for the applications (including the vSwitch
1095 itself) on the host.
1096 - Details of Resource isolation, such as CPUs designated for
1097 Host/Kernel (isolcpu) and CPUs designated for specific processes
1116 - Traffic Information:
1118 - Traffic type - UDP, TCP, IMIX / Other.
1121 - Deployment Scenario.
1123 **Note**: Tests that require additional parameters to be recorded will
1124 explicitly specify this.
1126 .. _TestIdentification:
1131 =========================
1136 ----------------------
1137 The following tests aim to determine the maximum forwarding rate that
1138 can be achieved with a virtual switch. The list is not exhaustive but
1139 should indicate the type of tests that should be required. It is
1140 expected that more will be added.
1144 Test ID: LTD.Throughput.RFC2544.PacketLossRatio
1145 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1146 **Title**: RFC 2544 X% packet loss ratio Throughput and Latency Test
1148 **Prerequisite Test**: N/A
1154 This test determines the DUT's maximum forwarding rate with X% traffic
1155 loss for a constant load (fixed length frames at a fixed interval time).
1156 The default loss percentages to be tested are: - X = 0% - X = 10^-7%
1158 Note: Other values can be tested if required by the user.
1160 The selected frame sizes are those previously defined under `Default
1161 Test Parameters <#DefaultParams>`__. The test can also be used to
1162 determine the average latency of the traffic.
1164 Under the `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__
1165 test methodology, the test duration will
1166 include a number of trials; each trial should run for a minimum period
1167 of 60 seconds. A binary search methodology must be applied for each
1168 trial to obtain the final result.
1170 **Expected Result**: At the end of each trial, the presence or absence
1171 of loss determines the modification of offered load for the next trial,
1172 converging on a maximum rate, or
1173 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__ Throughput with X%
1175 The Throughput load is re-used in related
1176 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__ tests and other
1179 **Metrics Collected**:
1181 The following are the metrics collected for this test:
1183 - The maximum forwarding rate in Frames Per Second (FPS) and Mbps of
1184 the DUT for each frame size with X% packet loss.
1185 - The average latency of the traffic flow when passing through the DUT
1186 (if testing for latency, note that this average is different from the
1187 test specified in Section 26.3 of
1188 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__).
1189 - CPU and memory utilization may also be collected as part of this
1190 test, to determine the vSwitch's performance footprint on the system.
1194 Test ID: LTD.Throughput.RFC2544.PacketLossRatioFrameModification
1195 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1196 **Title**: RFC 2544 X% packet loss Throughput and Latency Test with
1199 **Prerequisite Test**: N/A
1205 This test determines the DUT's maximum forwarding rate with X% traffic
1206 loss for a constant load (fixed length frames at a fixed interval time).
1207 The default loss percentages to be tested are: - X = 0% - X = 10^-7%
1209 Note: Other values can be tested if required by the user.
1211 The selected frame sizes are those previously defined under `Default
1212 Test Parameters <#DefaultParams>`__. The test can also be used to
1213 determine the average latency of the traffic.
1215 Under the `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__
1216 test methodology, the test duration will
1217 include a number of trials; each trial should run for a minimum period
1218 of 60 seconds. A binary search methodology must be applied for each
1219 trial to obtain the final result.
1221 During this test, the DUT must perform the following operations on the
1224 - Perform packet parsing on the DUT's ingress port.
1225 - Perform any relevant address look-ups on the DUT's ingress ports.
1226 - Modify the packet header before forwarding the packet to the DUT's
1227 egress port. Packet modifications include:
1229 - Modifying the Ethernet source or destination MAC address.
1230 - Modifying/adding a VLAN tag. (**Recommended**).
1231 - Modifying/adding a MPLS tag.
1232 - Modifying the source or destination ip address.
1233 - Modifying the TOS/DSCP field.
1234 - Modifying the source or destination ports for UDP/TCP/SCTP.
1235 - Modifying the TTL.
1237 **Expected Result**: The Packet parsing/modifications require some
1238 additional degree of processing resource, therefore the
1239 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__
1240 Throughput is expected to be somewhat lower than the Throughput level
1241 measured without additional steps. The reduction is expected to be
1242 greatest on tests with the smallest packet sizes (greatest header
1245 **Metrics Collected**:
1247 The following are the metrics collected for this test:
1249 - The maximum forwarding rate in Frames Per Second (FPS) and Mbps of
1250 the DUT for each frame size with X% packet loss and packet
1251 modification operations being performed by the DUT.
1252 - The average latency of the traffic flow when passing through the DUT
1253 (if testing for latency, note that this average is different from the
1254 test specified in Section 26.3 of
1255 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__).
1256 - The `RFC5481 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__
1257 PDV form of delay variation on the traffic flow,
1258 using the 99th percentile.
1259 - CPU and memory utilization may also be collected as part of this
1260 test, to determine the vSwitch's performance footprint on the system.
1264 Test ID: LTD.Throughput.RFC2544.Profile
1265 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1266 **Title**: RFC 2544 Throughput and Latency Profile
1268 **Prerequisite Test**: N/A
1274 This test reveals how throughput and latency degrades as the offered
1275 rate varies in the region of the DUT's maximum forwarding rate as
1276 determined by LTD.Throughput.RFC2544.PacketLossRatio (0% Packet Loss).
1277 For example it can be used to determine if the degradation of throughput
1278 and latency as the offered rate increases is slow and graceful or sudden
1281 The selected frame sizes are those previously defined under `Default
1282 Test Parameters <#DefaultParams>`__.
1284 The offered traffic rate is described as a percentage delta with respect
1285 to the DUT's RFC 2544 Throughput as determined by
1286 LTD.Throughput.RFC2544.PacketLoss Ratio (0% Packet Loss case). A delta
1287 of 0% is equivalent to an offered traffic rate equal to the RFC 2544
1288 Maximum Throughput; A delta of +50% indicates an offered rate half-way
1289 between the Maximum RFC2544 Throughput and line-rate, whereas a delta of
1290 -50% indicates an offered rate of half the RFC 2544 Maximum Throughput.
1291 Therefore the range of the delta figure is natuarlly bounded at -100%
1292 (zero offered traffic) and +100% (traffic offered at line rate).
1294 The following deltas to the maximum forwarding rate should be applied:
1296 - -50%, -10%, 0%, +10% & +50%
1298 **Expected Result**: For each packet size a profile should be produced
1299 of how throughput and latency vary with offered rate.
1301 **Metrics Collected**:
1303 The following are the metrics collected for this test:
1305 - The forwarding rate in Frames Per Second (FPS) and Mbps of the DUT
1306 for each delta to the maximum forwarding rate and for each frame
1308 - The average latency for each delta to the maximum forwarding rate and
1309 for each frame size.
1310 - CPU and memory utilization may also be collected as part of this
1311 test, to determine the vSwitch's performance footprint on the system.
1312 - Any failures experienced (for example if the vSwitch crashes, stops
1313 processing packets, restarts or becomes unresponsive to commands)
1314 when the offered load is above Maximum Throughput MUST be recorded
1315 and reported with the results.
1319 Test ID: LTD.Throughput.RFC2544.SystemRecoveryTime
1320 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1321 **Title**: RFC 2544 System Recovery Time Test
1323 **Prerequisite Test** LTD.Throughput.RFC2544.PacketLossRatio
1329 The aim of this test is to determine the length of time it takes the DUT
1330 to recover from an overload condition for a constant load (fixed length
1331 frames at a fixed interval time). The selected frame sizes are those
1332 previously defined under `Default Test Parameters <#DefaultParams>`__,
1333 traffic should be sent to the DUT under normal conditions. During the
1334 duration of the test and while the traffic flows are passing though the
1335 DUT, at least one situation leading to an overload condition for the DUT
1336 should occur. The time from the end of the overload condition to when
1337 the DUT returns to normal operations should be measured to determine
1338 recovery time. Prior to overloading the DUT, one should record the
1339 average latency for 10,000 packets forwarded through the DUT.
1341 The overload condition SHOULD be to transmit traffic at a very high
1342 frame rate to the DUT (150% of the maximum 0% packet loss rate as
1343 determined by LTD.Throughput.RFC2544.PacketLossRatio or line-rate
1344 whichever is lower), for at least 60 seconds, then reduce the frame rate
1345 to 75% of the maximum 0% packet loss rate. A number of time-stamps
1346 should be recorded: - Record the time-stamp at which the frame rate was
1347 reduced and record a second time-stamp at the time of the last frame
1348 lost. The recovery time is the difference between the two timestamps. -
1349 Record the average latency for 10,000 frames after the last frame loss
1350 and continue to record average latency measurements for every 10,000
1351 frames, when latency returns to within 10% of pre-overload levels record
1354 **Expected Result**:
1356 **Metrics collected**
1358 The following are the metrics collected for this test:
1360 - The length of time it takes the DUT to recover from an overload
1362 - The length of time it takes the DUT to recover the average latency to
1363 pre-overload conditions.
1365 **Deployment scenario**:
1367 - Physical → virtual switch → physical.
1371 Test ID: LTD.Throughput.RFC2544.BackToBackFrames
1372 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1373 **Title**: RFC2544 Back To Back Frames Test
1375 **Prerequisite Test**: N
1381 The aim of this test is to characterize the ability of the DUT to
1382 process back-to-back frames. For each frame size previously defined
1383 under `Default Test Parameters <#DefaultParams>`__, a burst of traffic
1384 is sent to the DUT with the minimum inter-frame gap between each frame.
1385 If the number of received frames equals the number of frames that were
1386 transmitted, the burst size should be increased and traffic is sent to
1387 the DUT again. The value measured is the back-to-back value, that is the
1388 maximum burst size the DUT can handle without any frame loss. Please note
1389 a trial must run for a minimum of 2 seconds and should be repeated 50
1390 times (at a minimum).
1392 **Expected Result**:
1394 Tests of back-to-back frames with physical devices have produced
1395 unstable results in some cases. All tests should be repeated in multiple
1396 test sessions and results stability should be examined.
1398 **Metrics collected**
1400 The following are the metrics collected for this test:
1402 - The average back-to-back value across the trials, which is
1403 the number of frames in the longest burst that the DUT will
1404 handle without the loss of any frames.
1405 - CPU and memory utilization may also be collected as part of this
1406 test, to determine the vSwitch's performance footprint on the system.
1408 **Deployment scenario**:
1410 - Physical → virtual switch → physical.
1414 Test ID: LTD.Throughput.RFC2889.MaxForwardingRateSoak
1415 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1416 **Title**: RFC 2889 X% packet loss Max Forwarding Rate Soak Test
1418 **Prerequisite Test** LTD.Throughput.RFC2544.PacketLossRatio
1424 The aim of this test is to understand the Max Forwarding Rate stability
1425 over an extended test duration in order to uncover any outliers. To allow
1426 for an extended test duration, the test should ideally run for 24 hours
1427 or, if this is not possible, for at least 6 hours. For this test, each frame
1428 size must be sent at the highest Throughput rate with X% packet loss, as
1429 determined in the prerequisite test. The default loss percentages to be
1430 tested are: - X = 0% - X = 10^-7%
1432 Note: Other values can be tested if required by the user.
1434 **Expected Result**:
1436 **Metrics Collected**:
1438 The following are the metrics collected for this test:
1440 - Max Forwarding Rate stability of the DUT.
1442 - This means reporting the number of packets lost per time interval
1443 and reporting any time intervals with packet loss. The
1444 `RFC2889 <https://www.rfc-editor.org/rfc/rfc2289.txt>`__
1445 Forwarding Rate shall be measured in each interval.
1446 An interval of 60s is suggested.
1448 - CPU and memory utilization may also be collected as part of this
1449 test, to determine the vSwitch's performance footprint on the system.
1450 - The `RFC5481 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__
1451 PDV form of delay variation on the traffic flow,
1452 using the 99th percentile.
1456 Test ID: LTD.Throughput.RFC2889.MaxForwardingRateSoakFrameModification
1457 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1458 **Title**: RFC 2889 Max Forwarding Rate Soak Test with Frame Modification
1460 **Prerequisite Test**:
1461 LTD.Throughput.RFC2544.PacketLossRatioFrameModification (0% Packet Loss)
1467 The aim of this test is to understand the Max Forwarding Rate stability over an
1468 extended test duration in order to uncover any outliers. To allow for an
1469 extended test duration, the test should ideally run for 24 hours or, if
1470 this is not possible, for at least 6 hour. For this test, each frame
1471 size must be sent at the highest Throughput rate with 0% packet loss, as
1472 determined in the prerequisite test.
1474 During this test, the DUT must perform the following operations on the
1477 - Perform packet parsing on the DUT's ingress port.
1478 - Perform any relevant address look-ups on the DUT's ingress ports.
1479 - Modify the packet header before forwarding the packet to the DUT's
1480 egress port. Packet modifications include:
1482 - Modifying the Ethernet source or destination MAC address.
1483 - Modifying/adding a VLAN tag (**Recommended**).
1484 - Modifying/adding a MPLS tag.
1485 - Modifying the source or destination ip address.
1486 - Modifying the TOS/DSCP field.
1487 - Modifying the source or destination ports for UDP/TCP/SCTP.
1488 - Modifying the TTL.
1490 **Expected Result**:
1492 **Metrics Collected**:
1494 The following are the metrics collected for this test:
1496 - Max Forwarding Rate stability of the DUT.
1498 - This means reporting the number of packets lost per time interval
1499 and reporting any time intervals with packet loss. The
1500 `RFC2889 <https://www.rfc-editor.org/rfc/rfc2289.txt>`__
1501 Forwarding Rate shall be measured in each interval.
1502 An interval of 60s is suggested.
1504 - CPU and memory utilization may also be collected as part of this
1505 test, to determine the vSwitch's performance footprint on the system.
1506 - The `RFC5481 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__
1507 PDV form of delay variation on the traffic flow, using the 99th
1512 Test ID: LTD.Throughput.RFC6201.ResetTime
1513 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1514 **Title**: RFC 6201 Reset Time Test
1516 **Prerequisite Test**: N/A
1522 The aim of this test is to determine the length of time it takes the DUT
1523 to recover from a reset.
1525 Two reset methods are defined - planned and unplanned. A planned reset
1526 requires stopping and restarting the virtual switch by the usual
1527 'graceful' method defined by it's documentation. An unplanned reset
1528 requires simulating a fatal internal fault in the virtual switch - for
1529 example by using kill -SIGKILL on a Linux environment.
1531 Both reset methods SHOULD be exercised.
1533 For each frame size previously defined under `Default Test
1534 Parameters <#DefaultParams>`__, traffic should be sent to the DUT under
1535 normal conditions. During the duration of the test and while the traffic
1536 flows are passing through the DUT, the DUT should be reset and the Reset
1537 time measured. The Reset time is the total time that a device is
1538 determined to be out of operation and includes the time to perform the
1539 reset and the time to recover from it (cf. `RFC6201
1540 <https://www.rfc-editor.org/rfc/rfc6201.txt>`__).
1542 `RFC6201 <https://www.rfc-editor.org/rfc/rfc6201.txt>`__ defines two methods
1543 to measure the Reset time:
1545 - Frame-Loss Method: which requires the monitoring of the number of
1546 lost frames and calculates the Reset time based on the number of
1547 frames lost and the offered rate according to the following
1550 .. code-block:: console
1552 Frames_lost (packets)
1553 Reset_time = -------------------------------------
1554 Offered_rate (packets per second)
1556 - Timestamp Method: which measures the time from which the last frame
1557 is forwarded from the DUT to the time the first frame is forwarded
1558 after the reset. This involves time-stamping all transmitted frames
1559 and recording the timestamp of the last frame that was received prior
1560 to the reset and also measuring the timestamp of the first frame that
1561 is received after the reset. The Reset time is the difference between
1562 these two timestamps.
1564 According to `RFC6201 <https://www.rfc-editor.org/rfc/rfc6201.txt>`__ the
1565 choice of method depends on the test tool's capability; the Frame-Loss
1566 method SHOULD be used if the test tool supports:
1568 * Counting the number of lost frames per stream.
1569 * Transmitting test frame despite the physical link status.
1571 whereas the Timestamp method SHOULD be used if the test tool supports:
1572 * Timestamping each frame.
1573 * Monitoring received frame's timestamp.
1574 * Transmitting frames only if the physical link status is up.
1576 **Expected Result**:
1578 **Metrics collected**
1580 The following are the metrics collected for this test:
1582 * Average Reset Time over the number of trials performed.
1584 Results of this test should include the following information:
1586 * The reset method used.
1587 * Throughput in Fps and Mbps.
1588 * Average Frame Loss over the number of trials performed.
1589 * Average Reset Time in milliseconds over the number of trials performed.
1590 * Number of trials performed.
1591 * Protocol: IPv4, IPv6, MPLS, etc.
1592 * Frame Size in Octets
1593 * Port Media: Ethernet, Gigabit Ethernet (GbE), etc.
1594 * Port Speed: 10 Gbps, 40 Gbps etc.
1595 * Interface Encapsulation: Ethernet, Ethernet VLAN, etc.
1597 **Deployment scenario**:
1599 * Physical → virtual switch → physical.
1603 Test ID: LTD.Throughput.RFC2889.MaxForwardingRate
1604 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1605 **Title**: RFC2889 Forwarding Rate Test
1607 **Prerequisite Test**: LTD.Throughput.RFC2544.PacketLossRatio
1613 This test measures the DUT's Max Forwarding Rate when the Offered Load
1614 is varied between the throughput and the Maximum Offered Load for fixed
1615 length frames at a fixed time interval. The selected frame sizes are
1616 those previously defined under `Default Test
1617 Parameters <#DefaultParams>`__. The throughput is the maximum offered
1618 load with 0% frame loss (measured by the prerequisite test), and the
1619 Maximum Offered Load (as defined by
1620 `RFC2285 <https://www.rfc-editor.org/rfc/rfc2285.txt>`__) is *"the highest
1621 number of frames per second that an external source can transmit to a
1622 DUT/SUT for forwarding to a specified output interface or interfaces"*.
1624 Traffic should be sent to the DUT at a particular rate (TX rate)
1625 starting with TX rate equal to the throughput rate. The rate of
1626 successfully received frames at the destination counted (in FPS). If the
1627 RX rate is equal to the TX rate, the TX rate should be increased by a
1628 fixed step size and the RX rate measured again until the Max Forwarding
1631 The trial duration for each iteration should last for the period of time
1632 needed for the system to reach steady state for the frame size being
1633 tested. Under `RFC2889 <https://www.rfc-editor.org/rfc/rfc2289.txt>`__
1634 (Sec. 5.6.3.1) test methodology, the test
1635 duration should run for a minimum period of 30 seconds, regardless
1636 whether the system reaches steady state before the minimum duration
1639 **Expected Result**: According to
1640 `RFC2889 <https://www.rfc-editor.org/rfc/rfc2289.txt>`__ The Max Forwarding
1641 Rate is the highest forwarding rate of a DUT taken from an iterative set of
1642 forwarding rate measurements. The iterative set of forwarding rate measurements
1643 are made by setting the intended load transmitted from an external source and
1644 measuring the offered load (i.e what the DUT is capable of forwarding). If the
1645 Throughput == the Maximum Offered Load, it follows that Max Forwarding Rate is
1646 equal to the Maximum Offered Load.
1648 **Metrics Collected**:
1650 The following are the metrics collected for this test:
1652 - The Max Forwarding Rate for the DUT for each packet size.
1653 - CPU and memory utilization may also be collected as part of this
1654 test, to determine the vSwitch's performance footprint on the system.
1656 **Deployment scenario**:
1658 - Physical → virtual switch → physical. Note: Full mesh tests with
1659 multiple ingress and egress ports are a key aspect of RFC 2889
1660 benchmarks, and scenarios with both 2 and 4 ports should be tested.
1661 In any case, the number of ports used must be reported.
1665 Test ID: LTD.Throughput.RFC2889.ForwardPressure
1666 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1667 **Title**: RFC2889 Forward Pressure Test
1669 **Prerequisite Test**: LTD.Throughput.RFC2889.MaxForwardingRate
1675 The aim of this test is to determine if the DUT transmits frames with an
1676 inter-frame gap that is less than 12 bytes. This test overloads the DUT
1677 and measures the output for forward pressure. Traffic should be
1678 transmitted to the DUT with an inter-frame gap of 11 bytes, this will
1679 overload the DUT by 1 byte per frame. The forwarding rate of the DUT
1682 **Expected Result**: The forwarding rate should not exceed the maximum
1683 forwarding rate of the DUT collected by
1684 LTD.Throughput.RFC2889.MaxForwardingRate.
1686 **Metrics collected**
1688 The following are the metrics collected for this test:
1690 - Forwarding rate of the DUT in FPS or Mbps.
1691 - CPU and memory utilization may also be collected as part of this
1692 test, to determine the vSwitch's performance footprint on the system.
1694 **Deployment scenario**:
1696 - Physical → virtual switch → physical.
1700 Test ID: LTD.Throughput.RFC2889.ErrorFramesFiltering
1701 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1702 **Title**: RFC2889 Error Frames Filtering Test
1704 **Prerequisite Test**: N/A
1710 The aim of this test is to determine whether the DUT will propagate any
1711 erroneous frames it receives or whether it is capable of filtering out
1712 the erroneous frames. Traffic should be sent with erroneous frames
1713 included within the flow at random intervals. Illegal frames that must
1714 be tested include: - Oversize Frames. - Undersize Frames. - CRC Errored
1715 Frames. - Dribble Bit Errored Frames - Alignment Errored Frames
1717 The traffic flow exiting the DUT should be recorded and checked to
1718 determine if the erroneous frames where passed through the DUT.
1720 **Expected Result**: Broken frames are not passed!
1722 **Metrics collected**
1724 No Metrics are collected in this test, instead it determines:
1726 - Whether the DUT will propagate erroneous frames.
1727 - Or whether the DUT will correctly filter out any erroneous frames
1728 from traffic flow with out removing correct frames.
1730 **Deployment scenario**:
1732 - Physical → virtual switch → physical.
1736 Test ID: LTD.Throughput.RFC2889.BroadcastFrameForwarding
1737 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1738 **Title**: RFC2889 Broadcast Frame Forwarding Test
1740 **Prerequisite Test**: N
1746 The aim of this test is to determine the maximum forwarding rate of the
1747 DUT when forwarding broadcast traffic. For each frame previously defined
1748 under `Default Test Parameters <#DefaultParams>`__, the traffic should
1749 be set up as broadcast traffic. The traffic throughput of the DUT should
1752 The test should be conducted with at least 4 physical ports on the DUT.
1753 The number of ports used MUST be recorded.
1755 As broadcast involves forwarding a single incoming packet to several
1756 destinations, the latency of a single packet is defined as the average
1757 of the latencies for each of the broadcast destinations.
1759 The incoming packet is transmitted on each of the other physical ports,
1760 it is not transmitted on the port on which it was received. The test MAY
1761 be conducted using different broadcasting ports to uncover any
1762 performance differences.
1764 **Expected Result**:
1766 **Metrics collected**:
1768 The following are the metrics collected for this test:
1770 - The forwarding rate of the DUT when forwarding broadcast traffic.
1771 - The minimum, average & maximum packets latencies observed.
1773 **Deployment scenario**:
1775 - Physical → virtual switch 3x physical. In the Broadcast rate testing,
1776 four test ports are required. One of the ports is connected to the test
1777 device, so it can send broadcast frames and listen for miss-routed frames.
1781 Test ID: LTD.Throughput.RFC2544.WorstN-BestN
1782 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1783 **Title**: Modified RFC 2544 X% packet loss ratio Throughput and Latency Test
1785 **Prerequisite Test**: N/A
1791 This test determines the DUT's maximum forwarding rate with X% traffic
1792 loss for a constant load (fixed length frames at a fixed interval time).
1793 The default loss percentages to be tested are: X = 0%, X = 10^-7%
1795 Modified RFC 2544 throughput benchmarking methodology aims to quantify
1796 the throughput measurement variations observed during standard RFC 2544
1797 benchmarking measurements of virtual switches and VNFs. The RFC2544
1798 binary search algorithm is modified to use more samples per test trial
1799 to drive the binary search and yield statistically more meaningful
1800 results. This keeps the heart of the RFC2544 methodology, still relying
1801 on the binary search of throughput at specified loss tolerance, while
1802 providing more useful information about the range of results seen in
1803 testing. Instead of using a single traffic trial per iteration step,
1804 each traffic trial is repeated N times and the success/failure of the
1805 iteration step is based on these N traffic trials. Two types of revised
1806 tests are defined - *Worst-of-N* and *Best-of-N*.
1810 *Worst-of-N* indicates the lowest expected maximum throughput for (
1811 packet size, loss tolerance) when repeating the test.
1813 1. Repeat the same test run N times at a set packet rate, record each
1815 2. Take the WORST result (highest packet loss) out of N result samples,
1816 called the Worst-of-N sample.
1817 3. If Worst-of-N sample has loss less than the set loss tolerance, then
1818 the step is successful - increase the test traffic rate.
1819 4. If Worst-of-N sample has loss greater than the set loss tolerance
1820 then the step failed - decrease the test traffic rate.
1825 *Best-of-N* indicates the highest expected maximum throughput for (
1826 packet size, loss tolerance) when repeating the test.
1828 1. Repeat the same traffic run N times at a set packet rate, record
1830 2. Take the BEST result (least packet loss) out of N result samples,
1831 called the Best-of-N sample.
1832 3. If Best-of-N sample has loss less than the set loss tolerance, then
1833 the step is successful - increase the test traffic rate.
1834 4. If Best-of-N sample has loss greater than the set loss tolerance,
1835 then the step failed - decrease the test traffic rate.
1838 Performing both Worst-of-N and Best-of-N benchmark tests yields lower
1839 and upper bounds of expected maximum throughput under the operating
1840 conditions, giving a very good indication to the user of the
1841 deterministic performance range for the tested setup.
1843 **Expected Result**: At the end of each trial series, the presence or
1844 absence of loss determines the modification of offered load for the
1845 next trial series, converging on a maximum rate, or
1846 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__ Throughput
1848 The Throughput load is re-used in related
1849 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__ tests and other
1852 **Metrics Collected**:
1854 The following are the metrics collected for this test:
1856 - The maximum forwarding rate in Frames Per Second (FPS) and Mbps of
1857 the DUT for each frame size with X% packet loss.
1858 - The average latency of the traffic flow when passing through the DUT
1859 (if testing for latency, note that this average is different from the
1860 test specified in Section 26.3 of
1861 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__).
1862 - Following may also be collected as part of this test, to determine
1863 the vSwitch's performance footprint on the system:
1865 - CPU core utilization.
1866 - CPU cache utilization.
1868 - System bus (QPI, PCI, ...) utilization.
1869 - CPU cycles consumed per packet.
1873 Test ID: LTD.Throughput.Overlay.Network.<tech>.RFC2544.PacketLossRatio
1874 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1875 **Title**: <tech> Overlay Network RFC 2544 X% packet loss ratio Throughput and Latency Test
1878 NOTE: Throughout this test, four interchangeable overlay technologies are covered by the
1879 same test description. They are: VXLAN, GRE, NVGRE and GENEVE.
1881 **Prerequisite Test**: N/A
1886 This test evaluates standard switch performance benchmarks for the scenario where an
1887 Overlay Network is deployed for all paths through the vSwitch. Overlay Technologies covered
1888 (replacing <tech> in the test name) include:
1895 Performance will be assessed for each of the following overlay network functions:
1897 - Encapsulation only
1898 - De-encapsulation only
1899 - Both Encapsulation and De-encapsulation
1901 For each native packet, the DUT must perform the following operations:
1903 - Examine the packet and classify its correct overlay net (tunnel) assignment
1904 - Encapsulate the packet
1905 - Switch the packet to the correct port
1907 For each encapsulated packet, the DUT must perform the following operations:
1909 - Examine the packet and classify its correct native network assignment
1910 - De-encapsulate the packet, if required
1911 - Switch the packet to the correct port
1913 The selected frame sizes are those previously defined under `Default
1914 Test Parameters <#DefaultParams>`__.
1916 Thus, each test comprises an overlay technology, a network function,
1917 and a packet size *with* overlay network overhead included
1918 (but see also the discussion at
1919 https://etherpad.opnfv.org/p/vSwitchTestsDrafts ).
1921 The test can also be used to determine the average latency of the traffic.
1923 Under the `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__
1924 test methodology, the test duration will
1925 include a number of trials; each trial should run for a minimum period
1926 of 60 seconds. A binary search methodology must be applied for each
1927 trial to obtain the final result for Throughput.
1929 **Expected Result**: At the end of each trial, the presence or absence
1930 of loss determines the modification of offered load for the next trial,
1931 converging on a maximum rate, or
1932 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__ Throughput with X%
1933 loss (where the value of X is typically equal to zero).
1934 The Throughput load is re-used in related
1935 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__ tests and other
1938 **Metrics Collected**:
1939 The following are the metrics collected for this test:
1941 - The maximum Throughput in Frames Per Second (FPS) and Mbps of
1942 the DUT for each frame size with X% packet loss.
1943 - The average latency of the traffic flow when passing through the DUT
1944 and VNFs (if testing for latency, note that this average is different from the
1945 test specified in Section 26.3 of
1946 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__).
1947 - CPU and memory utilization may also be collected as part of this
1948 test, to determine the vSwitch's performance footprint on the system.
1953 Packet Latency tests
1954 ---------------------------
1955 These tests will measure the store and forward latency as well as the packet
1956 delay variation for various packet types through the virtual switch. The
1957 following list is not exhaustive but should indicate the type of tests
1958 that should be required. It is expected that more will be added.
1962 Test ID: LTD.PacketLatency.InitialPacketProcessingLatency
1963 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1964 **Title**: Initial Packet Processing Latency
1966 **Prerequisite Test**: N/A
1972 In some virtual switch architectures, the first packets of a flow will
1973 take the system longer to process than subsequent packets in the flow.
1974 This test determines the latency for these packets. The test will
1975 measure the latency of the packets as they are processed by the
1976 flow-setup-path of the DUT. There are two methods for this test, a
1977 recommended method and a nalternative method that can be used if it is
1978 possible to disable the fastpath of the virtual switch.
1980 Recommended method: This test will send 64,000 packets to the DUT, each
1981 belonging to a different flow. Average packet latency will be determined
1982 over the 64,000 packets.
1984 Alternative method: This test will send a single packet to the DUT after
1985 a fixed interval of time. The time interval will be equivalent to the
1986 amount of time it takes for a flow to time out in the virtual switch
1987 plus 10%. Average packet latency will be determined over 1,000,000
1990 This test is intended only for non-learning virtual switches; For learning
1991 virtual switches use RFC2889.
1993 For this test, only unidirectional traffic is required.
1995 **Expected Result**: The average latency for the initial packet of all
1996 flows should be greater than the latency of subsequent traffic.
1998 **Metrics Collected**:
2000 The following are the metrics collected for this test:
2002 - Average latency of the initial packets of all flows that are
2003 processed by the DUT.
2005 **Deployment scenario**:
2007 - Physical → Virtual Switch → Physical.
2011 Test ID: LTD.PacketDelayVariation.RFC3393.Soak
2012 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2013 **Title**: Packet Delay Variation Soak Test
2015 **Prerequisite Tests**: LTD.Throughput.RFC2544.PacketLossRatio (0% Packet Loss)
2021 The aim of this test is to understand the distribution of packet delay
2022 variation for different frame sizes over an extended test duration and
2023 to determine if there are any outliers. To allow for an extended test
2024 duration, the test should ideally run for 24 hours or, if this is not
2025 possible, for at least 6 hour. For this test, each frame size must be
2026 sent at the highest possible throughput with 0% packet loss, as
2027 determined in the prerequisite test.
2029 **Expected Result**:
2031 **Metrics Collected**:
2033 The following are the metrics collected for this test:
2035 - The packet delay variation value for traffic passing through the DUT.
2036 - The `RFC5481 <https://www.rfc-editor.org/rfc/rfc5481.txt>`__
2037 PDV form of delay variation on the traffic flow,
2038 using the 99th percentile, for each 60s interval during the test.
2039 - CPU and memory utilization may also be collected as part of this
2040 test, to determine the vSwitch's performance footprint on the system.
2045 ------------------------
2046 The general aim of these tests is to understand the impact of large flow
2047 table size and flow lookups on throughput. The following list is not
2048 exhaustive but should indicate the type of tests that should be required.
2049 It is expected that more will be added.
2053 Test ID: LTD.Scalability.Flows.RFC2544.0PacketLoss
2054 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2055 **Title**: RFC 2544 0% loss Flow Scalability throughput test
2057 **Prerequisite Test**: LTD.Throughput.RFC2544.PacketLossRatio, IF the
2058 delta Throughput between the single-flow RFC2544 test and this test with
2059 a variable number of flows is desired.
2065 The aim of this test is to measure how throughput changes as the number
2066 of flows in the DUT increases. The test will measure the throughput
2067 through the fastpath, as such the flows need to be installed on the DUT
2068 before passing traffic.
2070 For each frame size previously defined under `Default Test
2071 Parameters <#DefaultParams>`__ and for each of the following number of
2081 - Max supported number of flows.
2083 This test will be conducted under two conditions following the
2084 establishment of all flows as required by RFC 2544, regarding the flow
2085 expiration time-out:
2087 1) The time-out never expires during each trial.
2089 2) The time-out expires for all flows periodically. This would require a
2090 short time-out compared with flow re-appearance for a small number of
2091 flows, and may not be possible for all flow conditions.
2093 The maximum 0% packet loss Throughput should be determined in a manner
2094 identical to LTD.Throughput.RFC2544.PacketLossRatio.
2096 **Expected Result**:
2098 **Metrics Collected**:
2100 The following are the metrics collected for this test:
2102 - The maximum number of frames per second that can be forwarded at the
2103 specified number of flows and the specified frame size, with zero
2108 Test ID: LTD.MemoryBandwidth.RFC2544.0PacketLoss.Scalability
2109 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2110 **Title**: RFC 2544 0% loss Memory Bandwidth Scalability test
2112 **Prerequisite Tests**: LTD.Throughput.RFC2544.PacketLossRatio, IF the
2113 delta Throughput between an undisturbed RFC2544 test and this test with
2114 the Throughput affected by cache and memory bandwidth contention is desired.
2120 The aim of this test is to understand how the DUT's performance is
2121 affected by cache sharing and memory bandwidth between processes.
2123 During the test all cores not used by the vSwitch should be running a
2124 memory intensive application. This application should read and write
2125 random data to random addresses in unused physical memory. The random
2126 nature of the data and addresses is intended to consume cache, exercise
2127 main memory access (as opposed to cache) and exercise all memory buses
2128 equally. Furthermore:
2130 - the ratio of reads to writes should be recorded. A ratio of 1:1
2132 - the reads and writes MUST be of cache-line size and be cache-line aligned.
2133 - in NUMA architectures memory access SHOULD be local to the core's node.
2134 Whether only local memory or a mix of local and remote memory is used
2136 - the memory bandwidth (reads plus writes) used per-core MUST be recorded;
2137 the test MUST be run with a per-core memory bandwidth equal to half the
2138 maximum system memory bandwidth divided by the number of cores. The test
2139 MAY be run with other values for the per-core memory bandwidth.
2140 - the test MAY also be run with the memory intensive application running
2143 Under these conditions the DUT's 0% packet loss throughput is determined
2144 as per LTD.Throughput.RFC2544.PacketLossRatio.
2146 **Expected Result**:
2148 **Metrics Collected**:
2150 The following are the metrics collected for this test:
2152 - The DUT's 0% packet loss throughput in the presence of cache sharing and
2153 memory bandwidth between processes.
2157 Test ID: LTD.Scalability.VNF.RFC2544.PacketLossRatio
2158 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2159 **Title**: VNF Scalability RFC 2544 X% packet loss ratio Throughput and
2162 **Prerequisite Test**: N/A
2168 This test determines the DUT's throughput rate with X% traffic loss for
2169 a constant load (fixed length frames at a fixed interval time) when the
2170 number of VNFs on the DUT increases. The default loss percentages
2171 to be tested are: - X = 0% - X = 10^-7% . The minimum number of
2172 VNFs to be tested are 3.
2174 Flow classification should be conducted with L2, L3 and L4 matching
2175 to understand the matching and scaling capability of the vSwitch. The
2176 matching fields which were used as part of the test should be reported
2177 as part of the benchmark report.
2179 The vSwitch is responsible for forwarding frames between the VNFs
2181 The SUT (vSwitch and VNF daisy chain) operation should be validated
2182 before running the test. This may be completed by running a burst or
2183 continuous stream of traffic through the SUT to ensure proper operation
2186 **Note**: The traffic rate used to validate SUT operation should be low
2187 enough not to stress the SUT.
2189 **Note**: Other values can be tested if required by the user.
2191 **Note**: The same VNF should be used in the "daisy chain" formation.
2192 Each addition of a VNF should be conducted in a new test setup (The DUT
2193 is brought down, then the DUT is brought up again). An atlernative approach
2194 would be to continue to add VNFs without bringing down the DUT. The
2195 approach used needs to be documented as part of the test report.
2197 The selected frame sizes are those previously defined under `Default
2198 Test Parameters <#DefaultParams>`__. The test can also be used to
2199 determine the average latency of the traffic.
2201 Under the `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__
2202 test methodology, the test duration will
2203 include a number of trials; each trial should run for a minimum period
2204 of 60 seconds. A binary search methodology must be applied for each
2205 trial to obtain the final result for Throughput.
2207 **Expected Result**: At the end of each trial, the presence or absence
2208 of loss determines the modification of offered load for the next trial,
2209 converging on a maximum rate, or
2210 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__ Throughput with X%
2212 The Throughput load is re-used in related
2213 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__ tests and other
2216 If the test VNFs are rather light-weight in terms of processing, the test
2217 provides a view of multiple passes through the vswitch on logical
2218 interfaces. In other words, the test produces an optimistic count of
2219 daisy-chained VNFs, but the cumulative effect of traffic on the vSwitch is
2220 "real" (assuming that the vSwitch has some dedicated resources, and the
2221 effects on shared resources is understood).
2224 **Metrics Collected**:
2225 The following are the metrics collected for this test:
2227 - The maximum Throughput in Frames Per Second (FPS) and Mbps of
2228 the DUT for each frame size with X% packet loss.
2229 - The average latency of the traffic flow when passing through the DUT
2230 and VNFs (if testing for latency, note that this average is different from the
2231 test specified in Section 26.3 of
2232 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__).
2233 - CPU and memory utilization may also be collected as part of this
2234 test, to determine the vSwitch's performance footprint on the system.
2238 Test ID: LTD.Scalability.VNF.RFC2544.PacketLossProfile
2239 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2240 **Title**: VNF Scalability RFC 2544 Throughput and Latency Profile
2242 **Prerequisite Test**: N/A
2248 This test reveals how throughput and latency degrades as the number
2249 of VNFs increases and offered rate varies in the region of the DUT's
2250 maximum forwarding rate as determined by
2251 LTD.Throughput.RFC2544.PacketLossRatio (0% Packet Loss).
2252 For example it can be used to determine if the degradation of throughput
2253 and latency as the number of VNFs and offered rate increases is slow
2254 and graceful, or sudden and severe. The minimum number of VNFs to
2257 The selected frame sizes are those previously defined under `Default
2258 Test Parameters <#DefaultParams>`__.
2260 The offered traffic rate is described as a percentage delta with respect
2261 to the DUT's RFC 2544 Throughput as determined by
2262 LTD.Throughput.RFC2544.PacketLoss Ratio (0% Packet Loss case). A delta
2263 of 0% is equivalent to an offered traffic rate equal to the RFC 2544
2264 Throughput; A delta of +50% indicates an offered rate half-way
2265 between the Throughput and line-rate, whereas a delta of
2266 -50% indicates an offered rate of half the maximum rate. Therefore the
2267 range of the delta figure is natuarlly bounded at -100% (zero offered
2268 traffic) and +100% (traffic offered at line rate).
2270 The following deltas to the maximum forwarding rate should be applied:
2272 - -50%, -10%, 0%, +10% & +50%
2274 **Note**: Other values can be tested if required by the user.
2276 **Note**: The same VNF should be used in the "daisy chain" formation.
2277 Each addition of a VNF should be conducted in a new test setup (The DUT
2278 is brought down, then the DUT is brought up again). An atlernative approach
2279 would be to continue to add VNFs without bringing down the DUT. The
2280 approach used needs to be documented as part of the test report.
2282 Flow classification should be conducted with L2, L3 and L4 matching
2283 to understand the matching and scaling capability of the vSwitch. The
2284 matching fields which were used as part of the test should be reported
2285 as part of the benchmark report.
2287 The SUT (vSwitch and VNF daisy chain) operation should be validated
2288 before running the test. This may be completed by running a burst or
2289 continuous stream of traffic through the SUT to ensure proper operation
2292 **Note**: the traffic rate used to validate SUT operation should be low
2293 enough not to stress the SUT
2295 **Expected Result**: For each packet size a profile should be produced
2296 of how throughput and latency vary with offered rate.
2298 **Metrics Collected**:
2300 The following are the metrics collected for this test:
2302 - The forwarding rate in Frames Per Second (FPS) and Mbps of the DUT
2303 for each delta to the maximum forwarding rate and for each frame
2305 - The average latency for each delta to the maximum forwarding rate and
2306 for each frame size.
2307 - CPU and memory utilization may also be collected as part of this
2308 test, to determine the vSwitch's performance footprint on the system.
2309 - Any failures experienced (for example if the vSwitch crashes, stops
2310 processing packets, restarts or becomes unresponsive to commands)
2311 when the offered load is above Maximum Throughput MUST be recorded
2312 and reported with the results.
2318 The general aim of these tests is to understand the capacity of the
2319 and speed with which the vswitch can accommodate new flows.
2323 Test ID: LTD.Activation.RFC2889.AddressCachingCapacity
2324 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2325 **Title**: RFC2889 Address Caching Capacity Test
2327 **Prerequisite Test**: N/A
2333 Please note this test is only applicable to virtual switches that are capable of
2334 MAC learning. The aim of this test is to determine the address caching
2335 capacity of the DUT for a constant load (fixed length frames at a fixed
2336 interval time). The selected frame sizes are those previously defined
2337 under `Default Test Parameters <#DefaultParams>`__.
2339 In order to run this test the aging time, that is the maximum time the
2340 DUT will keep a learned address in its flow table, and a set of initial
2341 addresses, whose value should be >= 1 and <= the max number supported by
2342 the implementation must be known. Please note that if the aging time is
2343 configurable it must be longer than the time necessary to produce frames
2344 from the external source at the specified rate. If the aging time is
2345 fixed the frame rate must be brought down to a value that the external
2346 source can produce in a time that is less than the aging time.
2348 Learning Frames should be sent from an external source to the DUT to
2349 install a number of flows. The Learning Frames must have a fixed
2350 destination address and must vary the source address of the frames. The
2351 DUT should install flows in its flow table based on the varying source
2352 addresses. Frames should then be transmitted from an external source at
2353 a suitable frame rate to see if the DUT has properly learned all of the
2354 addresses. If there is no frame loss and no flooding, the number of
2355 addresses sent to the DUT should be increased and the test is repeated
2356 until the max number of cached addresses supported by the DUT
2359 **Expected Result**:
2361 **Metrics collected**:
2363 The following are the metrics collected for this test:
2365 - Number of cached addresses supported by the DUT.
2366 - CPU and memory utilization may also be collected as part of this
2367 test, to determine the vSwitch's performance footprint on the system.
2369 **Deployment scenario**:
2371 - Physical → virtual switch → 2 x physical (one receiving, one listening).
2375 Test ID: LTD.Activation.RFC2889.AddressLearningRate
2376 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2377 **Title**: RFC2889 Address Learning Rate Test
2379 **Prerequisite Test**: LTD.Memory.RFC2889.AddressCachingCapacity
2385 Please note this test is only applicable to virtual switches that are capable of
2386 MAC learning. The aim of this test is to determine the rate of address
2387 learning of the DUT for a constant load (fixed length frames at a fixed
2388 interval time). The selected frame sizes are those previously defined
2389 under `Default Test Parameters <#DefaultParams>`__, traffic should be
2390 sent with each IPv4/IPv6 address incremented by one. The rate at which
2391 the DUT learns a new address should be measured. The maximum caching
2392 capacity from LTD.Memory.RFC2889.AddressCachingCapacity should be taken
2393 into consideration as the maximum number of addresses for which the
2394 learning rate can be obtained.
2396 **Expected Result**: It may be worthwhile to report the behaviour when
2397 operating beyond address capacity - some DUTs may be more friendly to
2398 new addresses than others.
2400 **Metrics collected**:
2402 The following are the metrics collected for this test:
2404 - The address learning rate of the DUT.
2406 **Deployment scenario**:
2408 - Physical → virtual switch → 2 x physical (one receiving, one listening).
2412 Coupling between control path and datapath Tests
2413 -------------------------------------------------------
2414 The following tests aim to determine how tightly coupled the datapath
2415 and the control path are within a virtual switch. The following list
2416 is not exhaustive but should indicate the type of tests that should be
2417 required. It is expected that more will be added.
2421 Test ID: LTD.CPDPCouplingFlowAddition
2422 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2423 **Title**: Control Path and Datapath Coupling
2425 **Prerequisite Test**:
2431 The aim of this test is to understand how exercising the DUT's control
2432 path affects datapath performance.
2434 Initially a certain number of flow table entries are installed in the
2435 vSwitch. Then over the duration of an RFC2544 throughput test
2436 flow-entries are added and removed at the rates specified below. No
2437 traffic is 'hitting' these flow-entries, they are simply added and
2440 The test MUST be repeated with the following initial number of
2441 flow-entries installed: - < 10 - 1000 - 100,000 - 10,000,000 (or the
2442 maximum supported number of flow-entries)
2444 The test MUST be repeated with the following rates of flow-entry
2445 addition and deletion per second: - 0 - 1 (i.e. 1 addition plus 1
2446 deletion) - 100 - 10,000
2448 **Expected Result**:
2450 **Metrics Collected**:
2452 The following are the metrics collected for this test:
2454 - The maximum forwarding rate in Frames Per Second (FPS) and Mbps of
2456 - The average latency of the traffic flow when passing through the DUT
2457 (if testing for latency, note that this average is different from the
2458 test specified in Section 26.3 of
2459 `RFC2544 <https://www.rfc-editor.org/rfc/rfc2544.txt>`__).
2460 - CPU and memory utilization may also be collected as part of this
2461 test, to determine the vSwitch's performance footprint on the system.
2463 **Deployment scenario**:
2465 - Physical → virtual switch → physical.
2469 CPU and memory consumption
2470 ---------------------------------
2471 The following tests will profile a virtual switch's CPU and memory
2472 utilization under various loads and circumstances. The following
2473 list is not exhaustive but should indicate the type of tests that
2474 should be required. It is expected that more will be added.
2478 Test ID: LTD.Stress.RFC2544.0PacketLoss
2479 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2480 **Title**: RFC 2544 0% Loss CPU OR Memory Stress Test
2482 **Prerequisite Test**:
2488 The aim of this test is to understand the overall performance of the
2489 system when a CPU or Memory intensive application is run on the same DUT as
2490 the Virtual Switch. For each frame size, an
2491 LTD.Throughput.RFC2544.PacketLossRatio (0% Packet Loss) test should be
2492 performed. Throughout the entire test a CPU or Memory intensive application
2493 should be run on all cores on the system not in use by the Virtual Switch.
2494 For NUMA system only cores on the same NUMA node are loaded.
2496 It is recommended that stress-ng be used for loading the non-Virtual
2497 Switch cores but any stress tool MAY be used.
2499 **Expected Result**:
2501 **Metrics Collected**:
2503 The following are the metrics collected for this test:
2505 - Memory and CPU utilization of the cores running the Virtual Switch.
2506 - The number of identity of the cores allocated to the Virtual Switch.
2507 - The configuration of the stress tool (for example the command line
2508 parameters used to start it.)
2510 **Note:** Stress in the test ID can be replaced with the name of the
2511 component being stressed, when reporting the results:
2512 LTD.CPU.RFC2544.0PacketLoss or LTD.Memory.RFC2544.0PacketLoss
2516 Summary List of Tests
2517 ----------------------------
2520 - Test ID: LTD.Throughput.RFC2544.PacketLossRatio
2521 - Test ID: LTD.Throughput.RFC2544.PacketLossRatioFrameModification
2522 - Test ID: LTD.Throughput.RFC2544.Profile
2523 - Test ID: LTD.Throughput.RFC2544.SystemRecoveryTime
2524 - Test ID: LTD.Throughput.RFC2544.BackToBackFrames
2525 - Test ID: LTD.Throughput.RFC2889.Soak
2526 - Test ID: LTD.Throughput.RFC2889.SoakFrameModification
2527 - Test ID: LTD.Throughput.RFC6201.ResetTime
2528 - Test ID: LTD.Throughput.RFC2889.MaxForwardingRate
2529 - Test ID: LTD.Throughput.RFC2889.ForwardPressure
2530 - Test ID: LTD.Throughput.RFC2889.ErrorFramesFiltering
2531 - Test ID: LTD.Throughput.RFC2889.BroadcastFrameForwarding
2532 - Test ID: LTD.Throughput.RFC2544.WorstN-BestN
2533 - Test ID: LTD.Throughput.Overlay.Network.<tech>.RFC2544.PacketLossRatio
2535 2. Packet Latency tests
2537 - Test ID: LTD.PacketLatency.InitialPacketProcessingLatency
2538 - Test ID: LTD.PacketDelayVariation.RFC3393.Soak
2540 3. Scalability tests
2542 - Test ID: LTD.Scalability.Flows.RFC2544.0PacketLoss
2543 - Test ID: LTD.MemoryBandwidth.RFC2544.0PacketLoss.Scalability
2544 - LTD.Scalability.VNF.RFC2544.PacketLossProfile
2545 - LTD.Scalability.VNF.RFC2544.PacketLossRatio
2549 - Test ID: LTD.Activation.RFC2889.AddressCachingCapacity
2550 - Test ID: LTD.Activation.RFC2889.AddressLearningRate
2552 5. Coupling between control path and datapath Tests
2554 - Test ID: LTD.CPDPCouplingFlowAddition
2556 6. CPU and memory consumption
2558 - Test ID: LTD.Stress.RFC2544.0PacketLoss
Code Review / vswitchperf.git / blob