cwang9208 opened a new issue, #34478:
URL: https://github.com/apache/arrow/issues/34478

   ### Describe the usage question you have. Please include as many useful 
details as  possible.
   
   
   I'm benchmarking the performance of flight RPC using the built-in C++ 
benchmark server/client. Both the client and server run on the same machine.
   
   Below is the result for records_per_batch = 1000000. 
   ```
   $ ./arrow-flight-benchmark -records_per_stream 1000000000 -records_per_batch 
1000000 -num_perf_runs 2 -num_streams 10 -num_threads 10
   Testing method: DoGet
   Using spawned TCP server
   Server running with pid 3207456
   Server host: localhost
   Server port: 31337
   Number of perf runs: 2
   Number of concurrent gets/puts: 10
   Batch size: 32000000
   Batches read: 20000
   Bytes read: 640000000000
   Nanos: 55198332155
   Speed: 11057.4 MB/s
   Throughput: 362.33 batches/s
   Latency mean: 26931 us
   Latency quantile=0.5: 26270 us
   Latency quantile=0.95: 41160 us
   Latency quantile=0.99: 51025 us
   Latency max: 159917 us
   ```
   And the result for records_per_batch = 5000000. 
   ```
   $ ./arrow-flight-benchmark -records_per_stream 1000000000 -records_per_batch 
5000000 -num_perf_runs 2 -num_streams 10 -num_threads 10
   Testing method: DoGet
   Using spawned TCP server
   Server running with pid 3206446
   Server host: localhost
   Server port: 31337
   Number of perf runs: 2
   Number of concurrent gets/puts: 10
   Batch size: 160000000
   Batches read: 4000
   Bytes read: 640000000000
   Nanos: 79217518482
   Speed: 7704.75 MB/s
   Throughput: 50.4939 batches/s
   Latency mean: 184366 us
   Latency quantile=0.5: 167930 us
   Latency quantile=0.95: 309036 us
   Latency quantile=0.99: 487428 us
   Latency max: 744113 us
   ```
   I'm wondering why the Speed drops when the records_per_batch grows bigger?
   
   Can anyone help explain the reason behind this?
   
   ### Component(s)
   
   Benchmarking, C++, FlightRPC


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to