Release Notes

Changes in CSIT-2101.1

1. VPP PERFORMANCE TESTS

   • MLRsearch improvements: Added support for multiple packet throughput rates in a single search, each rate is associated with a distinct Packet Loss Ratio (PLR) criterion. Previously only Non Drop Rate (NDR) (PLR=0) and single Partial Drop Rate (PDR) (PLR<0.5%) were supported. Implemented number of optimizations improving rate discovery efficiency.

   • Reduction of tests: Removed obsolete VPP use cases and superfluous test combinations from continuous and report test executions, including:

     • All vts tests, obsolete use cases.

     • dot1q tests apart from dot1q-l2bd, superfluous combinations.

     • -100flows, -100kflows in all acl tests.

     • nat44 tests

       • -pps tests, replaced by -tput tests.

       • h1-p1-s1 single session tests, unessential combination.

       • h4096-p63-s258048 tests, unessential scale combination.

     • ipsec tests

       • ethip4ipsectptlispgpe.

       • policy-aes128gcm.

       • policy-aes128cbc-hmac256sha.

       • policy-aes128cbc-hmac512sha.

       • int-aes128cbc-hmac256sha.

       • scale of

         • 400tnlsw.

         • 5000tnlsw.

         • 20000tnlsw.

         • 60000tnlsw.

2. TEST FRAMEWORK

   • Telemetry retouch: Refactored telemetry retrieval from DUTs and SUTs. Included VPP perfmon plugin telemetry with all perfmon bundles available in VPP release.

   • Upgrade to Ubuntu 20.04 LTS: Re-installed base operating system to Ubuntu 20.04.2 LTS. Upgrade included also baseline Docker containers used for spawning topology.

   • TRex upgrade v2.86 to v2.88: Included move to DPDK 21.02 and changed the way egress low latency queues are used in FVL NICs. This broke latency measurements for majority of FVL NICs in CSIT. Latency values look better after upgrading FVL FW on TRex servers, but still somewhat higher than before the TRex upgrade. Tracked by CSIT-1790.

   • CSIT test environment version has been updated to ver. 7, see Environment Versioning.

   • CSIT PAPI support: Due to issues with PAPI performance, VAT is still used in CSIT for all VPP scale tests. See known issues below.

   • General Code Housekeeping: Ongoing code optimizations and bug fixes.

3. PRESENTATION AND ANALYTICS LAYER

   • Graphs improvements: Updated Packet Latency graphs, see Packet Latency.

Known Issues

List of known issues in CSIT-2101.1 for VPP performance tests:

#	JiraID	Issue Description
1	CSIT-1763	Adapt ramp-up phase of nat44 tests for different frame sizes. Currently ramp-up phase rate and duration values are correctly set for tests with 64B frame size.
2	CSIT-1671	All CSIT scale tests can not use PAPI due to much slower performance compared to VAT/CLI (it takes much longer to program VPP). This needs to be addressed on the PAPI side. The usual PAPI library spends too much time parsing arguments, so even with async processing (hundreds of commands in flight over socket), the VPP configuration for large scale tests (millions of messages) takes too long.
	VPP-1763
3	CSIT-1790	Broken TRex latency measurements with TRex v2.88, DPDK 21.02 and FVL FW 6.01. After upgrading TRex to v2.88, we also needed to upgrade FVL firmware on TG machines, to avoid high latency O(5msec) for all VPP and testpmd/l3fwd test cases.
4	CSIT-1789	AVF driver does not perform RSS in a deterministic way. This increases standard deviation of tests with small number of flows (mainly ipsec).
5	CSIT-1780	IPSEC SW async scheduler MRR tests failing with no traffic forwarded.
6	CSIT-1785	NAT44ED tests failing to establish all TCP sessions.

Root Cause Analysis for Performance Changes

List of RCAs in CSIT-2101.1 for VPP performance changes:

#	JiraID	Issue Description
1	VPP-1972	One VPP change has decreased performance of NAT44ed processing, both slow path and fast path.