Benchmarks and Methods for Multihomed
EVPNAT&T Labs200 Laurel Avenue SouthMiddletown,NJ07748USA+1 732 420 1571+1 732 368 1192acm@research.att.comAT&T Labs200 Laurel Avenue SouthMiddletown,NJ07748USAuttaro@att.comFundamental Benchmarking Methodologies for Network Interconnect
Devices of interest to the IETF are defined in RFC 2544. Key benchmarks
applicable to restoration and multi-homed sites are in RFC 6894. This
memo applies these methods to Multihomed nodes implemented on Ethernet
Virtual Private Networks (EVPN).The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
"OPTIONAL" in this document are to be interpreted as described in BCP
14 when, and only when,
they appear in all capitals, as shown here.The IETF's fundamental Benchmarking Methodologies are defined in, supported by the terms and definitions in , and actually obsoletes an
earlier specification, .This memo recognizes the importance of Ethernet Virtual Private
Network (EVPN) Multihoming connectivity scenarios, where a CE device is
connected to 2 or more PEs using an instance of an Ethernet Segment.In an all-active or Active-Active scenario, CE-PE traffic is
load-balanced across two or more PEs.Mass-withdrawal of routes may take place when an autodiscovery route
is used on a per Ethernet Segment basis, and there is a link failure on
one of the Ethernet Segment links (or when configuration changes take
place).Although EVPN depends on address-learning in the control-plane, the
Ethernet Segment Instance is permitted to use "the method best suited to
the CE: data-plane learning, IEEE 802.1x, the Link Layer Discovery
Protocol (LLDP), IEEE 802.1aq, Address Resolution Protocol (ARP),
management plane, or other protocols" [RFC7432].This memo seeks to benchmark these important cases (and others).The scope of this memo is to define a method to unambiguously perform
tests, measure the benchmark(s), and report the results for Capacity of
EVPN Multihoming connectivity scenarios, and other key restoration
activities (such as address withdrawl) covering link failure in the
Active-Active scenario.The goal is to provide more efficient test procedures where possible,
and to expand reporting with additional interpretation of the results.
The tests described in this memo address some key multihoming scenarios
implemented on a Device Under Test (DUT) or System Under Test (SUT).The Multihoming scenarios described in this memo emphsize features
with practical value to the industry that have seen deployment.
Therefore, these scenarios derserve further attention that follows from
benchmarking activities and further study.For simple Capacity/Throughput Benchmarks, the Test Setup MUST be
consistent with Figure 1 of , or Figure 2 when
the tester's sender and receiver are different devices.In this case, the System Under Test (SUT) is comprised of a single CE
device and two or more PE devices. The tester SHALL be connected to all
CE and PE, and be capable of simulateneously sending and receiving
frames on all ports in use. The tester SHALL be capable of generating
multiple flows (according to a 5-tuple definition, or any sub-set of the
5-tuple). The tester SHALL be able to control the IP capacity of sets of
individual flows, and the presence of sets of flows on specific
interface ports.Other mandatory testing aspects described in
MUST be included, unless explicitly modified in the next section.The ingress and egress link speeds and link layer protocols MUST be
specified and used to compute the maximum theoretical frame rate when
respecting the minimum inter-frame gap.A second test case is where a BGP backbone implements MPLS-LDP to
provide connectivity between multiple PE - ESI - CE locations.All Link speeds MUST be reported, along with complete device
configurations in the SUT and Test Device(s).Additional Test Setups and configurations will be provided in this
section, after review.One capacity benchmark pertains to the number of ESI that a network
with multiple PE - ESI - CE locations can support.Objective: To characterize the ability of a DUT/SUT to process frames
between CE and one or more PEs in a multihomed connectivity scenario.
Figure 1 gives the test setup.The Procedure follows."For every address, learning frames MUST be sent to the DUT/SUT to
allow the DUT/SUT to update its address tables properly."
[RFC2889]Each trial in the test requires confiuring a number of flows (from
100 to 100k) and a fixed frame size (64 octets to 128, 256, 512, 1024,
1280 and 1518 bytes, as per ).The Procedure SHALL follow section 5.1 of .The test MUST be repeated N times for each frame size in the subset
list, and each Throughput value made available for further processing
(below).For each Frame size, calculate the following summary statistics for
Throughput values over the N tests:Average (Benchmark)MinimumMaximumStandard DeviationComparison will determine how the load was balanced among
PEs.Objective: To characterize the ability of a DUT/SUT to process frames
between CE and one or more PE in a multihomed connectivity scenario when
a mass withdrawal takes place. Figure 2 gives the test setup.The Procedure follows."For every address, learning frames MUST be sent to the DUT/SUT to
allow the DUT/SUT update its address tables properly." Each trial in the test requires Confiuring a number of flows (from
100 to 100k) and a fixed frame size (64 octets to 128, 256, 512, 1024,
1280 and 1518 bytes, as per ).The Offered Load SHALL be transmitted at the Throughput level
corrsponding to previously determined for the selected Frame size and
number of Flows in use.The Procedure SHALL follow section 5.1 of
(except there is no need to search for the Throughput level).When traffic has been sent for 5 seconds one of the CE-PE links on
the ESI SHALL be disabled, and the time of this action SHALL be
recorded for further calculations. For example, if the CE1 link to PE1
is disabled, this should trigger a Mass withdrawal of EVI-1 addresses,
and the subsequent re-routing of traffic to PE2.Frame losses are expected to be recorded during the restoration
time. Time for restoration may be estimated as described in section
3.5 of.The test MUST be repeated N times for each frame size in the subset
list, and each restoration time value made available for further
processing (below).For each Frame size and number of flows, calculate the following
summary statistics for Loss (or Time to return to Throughput level
after restoration) values over the N tests:Average (Benchmark)MinimumMaximumStandard DeviationThe results SHOULD be reported in the format of a table with a row
for each of the tested frame sizes and Number of Flows. There SHOULD be
columns for the frame size with number of flows, and for the resultant
average frame count (or time) for each type of data stream tested.The number of tests Averaged for the Benchmark, N, MUST be
reported.The Minimum, Maximum, and Standard Deviation across all complete
tests SHOULD also be reported.The Corrected DUT Restoration Time SHOULD also be reported, as
applicable.Frame Size, octets + # FlowsAve Benchmark, fps, frames or timeMin,Max,StdDevCalculated Time, Sec64,1002600025500,27000,200.00004Static and configuration parameters:Number of test repetitions, NMinimum Step Size (during searches), in frames.Benchmarking activities as described in this memo are limited to
technology characterization using controlled stimuli in a laboratory
environment, with dedicated address space and the other constraints
.The benchmarking network topology will be an independent test setup
and MUST NOT be connected to devices that may forward the test traffic
into a production network, or misroute traffic to the test management
network. See .Further, benchmarking is performed on a "black-box" basis, relying
solely on measurements observable external to the DUT/SUT.Special capabilities SHOULD NOT exist in the DUT/SUT specifically for
benchmarking purposes. Any implications for network security arising
from the DUT/SUT SHOULD be identical in the lab and in production
networks.This memo makes no requests of IANA.Thanks to Aman Shaikh for sharing his comments on the draft directly
with the authors.Dataplane Performance, Capacity, and Benchmarking in
OPNFVIntel Corp.AT&T LabsSpirent CommunicationsBack2Back Testing Time Series (from CI)Evolution of Repeatability in Benchmarking: Fraser Plugfest
(Summary for IETF BMWG)AT&T LabsSpirent CommunicationsETSI GS NFV-TST 009 V3.1.1 (2018-10), "Network Functions
Virtualisation (NFV) Release 3; Testing; Specification of Networking
Benchmarks and Measurement Methods for NFVI"ETSI Network Function Virtualization
ISG