You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@arrow.apache.org by "Yibo Cai (Jira)" <ji...@apache.org> on 2021/04/28 05:08:00 UTC

[jira] [Created] (ARROW-12581) [C++][FlightRPC] Benchmark compression with real data

Yibo Cai created ARROW-12581:
--------------------------------

             Summary: [C++][FlightRPC] Benchmark compression with real data
                 Key: ARROW-12581
                 URL: https://issues.apache.org/jira/browse/ARROW-12581
             Project: Apache Arrow
          Issue Type: Improvement
          Components: C++, FlightRPC
            Reporter: Yibo Cai


Flight benchmark program is using random test data as payload. Random data is hard to compress. Per my test, the compressed payload (zstd default level) is actually a bit larger than the original uncompressed payload. So compression is a pure loss in current benchmark.

It's better to use real world test set for a more reasonable benchmark about how compression influences FlightRPC performance. Perhaps the benchmark (both client/server) could accept a path to an IPC file as an option.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)