[GitHub] [flink-benchmarks] zhijiangW commented on pull request #2: [FLINK-19003][checkpointing] Add micro-benchmark for unaligned checkpoints

[GitBox <gi...@apache.org>]

zhijiangW commented on pull request #2:
URL: https://github.com/apache/flink-benchmarks/pull/2#issuecomment-678116298


   Thanks for the review and discussions, @pnowojski !
   
   > but maybe it would be also better to use FLIP-27 sources here?
   
   Considering the current situation of mostly used legacy source, it is also meaningful to measure the performance in this case. But I agree we can also supplement FLIP-27 source as an follow-up option.
    
   > 1. how fast quickly can we trigger and complete N checkpoints, by setting checkpointing interval to 0ms (or as small number as possible), make sure the number of concurrent checkpoints is set to 1, and just measure how quick we can complete those checkpoints (note, we should also in that case modify @OperationsPerInvocation to N. This with an induced backpressure.
   
   I think my current way is exactly for this purpose, just set `N =5` and the default concurrent cp should be 1 if I remembered correctly. I am not quite sure whether it is really necessary to set even smaller interval since 100ms is already small enough in most practical cases. If we give much smaller values, the backpressure might not be triggered before N checkpoints finished.
   
   > 2. what is the throughput of the job with unaligned checkpoints compared to aligned checkpoints. Set checkpointing interval to some reasonable value and measure how fast can we process N records without any backpressure.
   
   I agree to supplement this case as well, similar with we did in `RemoteBalanceBenchmark`.
   


----------------------------------------------------------------
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:
users@infra.apache.org