chia7712 commented on pull request #9206: URL: https://github.com/apache/kafka/pull/9206#issuecomment-678722242
> Do you have any data on the improvement this achieves? I profiled (1 minute) one of servers (10 cores * 128 GB) without /without this patch. The tuple allocations is changed from 483 MB to 260 MB. > Generally, the tuples allocated here (if not elided by the optimizer) would be pretty tiny compared to allocating the buffers for the record batches. you are totally right. Almost all memory are allocated for keeping record batches. Perhaps we should offer a pluginable memory pool so users who do care for GC pressure can take custom memory pool to deal with large memory allocation/release and they don't need to change kafka source code. At any rate, this patch aims to reduce some memory usage by a small patch :) ---------------------------------------------------------------- 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: us...@infra.apache.org
