This is an old revision of the document!
The test scratchpad (for tests that are WIP/being defined/looked at by the community) can be found here: https://etherpad.opnfv.org/p/vSwitchTestsDrafts
Test | Priority | JIRA |
---|---|---|
Scalability Tests adding More VMs in succession and building a performance profile as we add more VMs. | 1 | |
OverlayNetworking Tests: VXLAN performance testing, encap, decap, encap and decap | 1 | |
Matchaction performance testing? The cost of the different actions supported by a vSwitch | 1 | |
Classifying L2, L3 and L4 traffic Profile Tests, Flows, new flows/sec, L2/L3/L4 etc. entries in"fast" path tables, … | ||
Stream/bulk Data transfer "unidirectional stream" performance request& response/transaction rate tests | ||
Performance testing with Mirroring enabled on the switch | ||
TCP Maxconnections per second, Max # of active sessions, Max transactions per second | ||
IPv6 Support and Testing | ||
Best ofN and Worst of N Tests | ||
Measuring how many instances of a vSwitch would be required to support 1 million simulataneuos flows | ||
Resilience tests | ||
Performance characterization runs vs.number of "x". Tests with Restricted Resources (Host with X Cores, vSwitch with Y Cores? | ||
Tests to evaluate vSwitch contribution to 3GPP performance objectives (3GPP TS 23.203 Policy and Charging Control Architecture) | ||
VM2VM Tests, with scaling number of VMs (in path) | ||
PHY-VNF-vSwitch-VNF-PHY | ||
Programming rate of Forwarding_table /flow_table | ||
SOAK tests (using RFC 2889 Max Forwarding Rate, not RFC 2544, to collect intermediate results) | ||
Tests suggested by OVS developers (there were a few) | ||
Box-full tests, where all CPU cores are doing useful work: vNet4 vSwitch with Nx VNFs, vNet6 vSwitch with Nx VNF chained topologies | ||
LatencyAndLatencyVariation with dedicated latency measurement streams | ||
Topology | Priority | JIRA |
---|---|---|
vNet vComp2 for VM2VM virt interface testing |
We should consider recommending a minimum number of iterations to repeat the test and either average out the result or calculate the variance across trials