Release Notes¶
Changes in CSIT-1908.2¶
VPP PERFORMANCE TESTS
Intel Xeon 2n-skx, 3n-skx testbeds: VPP performance test data is provided using a different CSIT test environment compared to CSIT-1908.1 and CSIT-1908. The changes were applied during the CSIT-2001 development cycle.
CSIT test environment is now versioned, with ver. 1 associated with CSIT rls1908 git branch as of 2019-08-21, and ver. 2 associated with CSIT rls2001 git branch as of 2020-03-27.
To identify performance changes due to VPP code changes from v19.08.1 to v19.08.2, both have been tested in CSIT environment ver. 2. See Current vs. Previous Release and Known Issues.
Intel Xeon 2n-clx testbeds: VPP performance test data is now included in this report. See Known Issues.
Service density 2n-skx tests: Added higher NF density tests with 802.1q (vlan) and VXLAN encapsulation from Traffic Generator.
GBP tests: Added GBP (Group Based Policy) routing test cases with 802.1q (vlan) external traffic.
AVF IPv4 scale tests: Increased coverage of AVF IPv4 base and scale test cases (Fortville NICs only).
2n-skx tests: Increased coverage of selected (COP, iACL, Policer) test cases.
IPsec scale tests: Added IPsec interface mode scale tests with 1, 40, 400, 1000, 5000, 10000, 20000, 40000, 60000 IPsec tunnels. Removed DPDK backend dependency. Major IPsec test code refactoring.
Hoststack TCP/IP tests: Major refactor of Hoststack TCP performance tests using WRK generator talking to the VPP HTTP static server plugin measuring connections per second and requests per second.
Changed methodology of dot1q tests in 2-Node testbeds: dot1q encapsulation is now used on both links of SUT. Previously dot1q was used only on a single link with the other link carrying untagged Ethernet frames. This change results in slightly lower throughput in CSIT-1908 for all dot1q tests in all 2-Node testbeds.
KVM VM vhost-user tests: completed move to Kernel-VM for all tests. In addition to running DPDK Testpmd as VM workload, new tests created with VPP as VM workload. VPP in VM is the same version as the DUT VPP (acting as vSwitch) and its configuration depends on the test type. For all L2 Ethernet Switching tests it’s vpp-l2xc (L2 cross-connect), for all IPv4 Routing tests it’s vpp-ip4 (VPP IPv4 routing).
TEST FRAMEWORK
CSIT PAPI Support: Finished conversion of CSIT VAT L1 keywords to PAPI L1 KWs in CSIT using VPP Python bindings (VPP PAPI). Redesign of key components of PAPI Socket Executor and PAPI history. 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 RF keywords optimizations, removal of redundant RF keywords and aligning of suite/test setup/teardowns.
PRESENTATION AND ANALYTICS LAYER
Graphs Layout Improvements: Improved performance graphs layout for better readibility and maintenance: test grouping, axis labels, descriptions, other informative decoration.
Known Issues¶
List of known issues in CSIT-1908.2 for VPP performance tests:
# |
JiraID |
Issue Description |
---|---|---|
1 |
Sporadic (1 in 200) NDR discovery test failures on x520. DPDK reporting rx-errors, indicating L1 issue. Suspected issue with HW combination of X710-X520 in LF testbeds. Not observed outside of LF testbeds. |
|
2 |
9000B packets not supported by NICs VIC1227 and VIC1387. |
|
3 |
Memif tests are sporadically failing on initialization of memif connection. |
|
4 |
9000B ip4 memif errors - ip4-input: ip4 length > l2 length. IP4 jumbo frames (9000B) are dropped in case of tests with memif. |
|
5 |
9000B ip4 nat44: VPP crash + coredump. VPP crashes very often in case that NAT44 is configured and it has to process IP4 jumbo frames (9000B). |
|
6 |
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. |
|
7 |
IPv4 AVF 9000B packet tests are failing on 3n-skx while passing on 2n-skx. |