[roda] Ignore POST params unless required - #1414
Conversation
|
/benchmark -f roda |
|
👋 Benchmark request received. A collaborator will review and approve the run. |
Benchmark ResultsFramework:
Full log |
|
/benchmark -f roda --save |
|
👋 Benchmark request received. A collaborator will review and approve the run. |
|
|
request.params merges the GET and POST params. If we only need to read the GET params we don't need to parse the body for POST params.
cf93941 to
c0af9c4
Compare
|
/benchmark -f roda --save |
|
👋 Benchmark request received. A collaborator will review and approve the run. |
Benchmark ResultsFramework:
Full log |
|
this is one of the frameworks that holds a very odd baseline behavior the "warm up" round overshoots, this is likely due to a defect in the baseline test we have which has been causing swings in results since it was created. Basically baseline is a little bit different from every other test, it has 3 different requests which are distributed across the available connections but these are persistent long lived connections that live for the benchmark duration, of course killed at the end of each run. However.. some frameworks like this one, have likely a very different performance for a CL and a TE:chunked request types, both are exercised here and this seems to cause a "short circuit" where the connections that use CL happen to be "favoured" by the runtime causing a spikes and swings in the results. This issue will be fixed after baseline test rework where we will not use long lived connections anymore for this test and force an homogeneous request type distribution, rotated every x requests |
|
Hmm interesting! |
request.params merges the GET and POST params.
If we only need to read the GET params we don't need to parse the body for POST params.
used, and profiles the other framework does not run show
n/arather than a delta.