cyb70289 opened a new pull request #9902:
URL: https://github.com/apache/arrow/pull/9902
Original uncompressed payload size is used in throughput calculation,
not the compressed size. So the result can be compared directly with
non-compression benchmark to see if compression helps performance.
**NOTE:** Flight benchmark uses random data for payload, it's hard to be
compressed. Actually the compressed size (zstd, lz4) is larger than
original size. So the benchmark result reflects the worse case how
compression can influence performance.
Example test result:
- No compression
$ arrow-flight-benchmark -num_streams 1 -num_threads 1 -test_put
*Speed: 2941.53 MB/s*
*Latency: 40 us*
- ZSTD with default compression level
$ arrow-flight-benchmark -num_streams 1 -num_threads 1 -test_put
-compression zstd
*Speed: 1205.66 MB/s*
*Latency: 101 us*
- ZSTD with high compression level
$ arrow-flight-benchmark -num_streams 1 -num_threads 1 -test_put
-compression zstd:20
*Speed: 47.6882 MB/s*
*Latency: 2618 us*
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
