Make data collection aggregated throughout#362
Conversation
# Conflicts: # cli/benchclient.go # cli/flags.go # go.mod # go.sum # pkg/bench/benchmark.go # pkg/bench/delete.go # pkg/bench/get.go # pkg/bench/list.go # pkg/bench/mixed.go # pkg/bench/multipart.go # pkg/bench/ops.go # pkg/bench/retention.go # pkg/bench/s3zip.go # pkg/bench/select.go # pkg/bench/stat.go # pkg/bench/versioned.go
|
Can you share the https://asciinema.org/ of the new UI? |
|
@harshavardhana If you port it to Windows ;) Let's see if this works: https://github.com/user-attachments/assets/c68b74d1-1748-4e1b-89f4-bbb455c4baff |
4800e8d to
ffda852
Compare
why do you have a more verbose output for default results when the test is over? Can we have the same output as it is today, which mainly talks about the higher-level points? |
|
@harshavardhana Just seemed unnecessary, but I can hook it up. Most infrastructure needed should be there. |
(except on `warp analyze`, where it is always enabled)
|
@harshavardhana Brought back the simple view - thinking of it - it was only on |
This PR will change all data collection to be automatically aggregated as requests happen.
This means individual requests will no longer be recorded unless
-fullis specified.This will allow warp to run for days without consuming significant amount of memory. The only downside is that there are no full-run median request values, but they are instead grouped and averaged.
--autotermfor remote running benchmarks as well.-analyze.vdoes not do anything (all details shown by default)--stdoutto print operations to stdout.Obviously a significant rewrite, so there will likely be uncovered corner cases. But testing checks out fine now.
Pending:
warp cmp. Not strictly needed for release.warp merge. Rarely used.