Jackie-Jiang commented on issue #4362: Optimize MutableOffHeapByteArrayStore by directly calling the PinotDataBuffer API URL: https://github.com/apache/incubator-pinot/pull/4362#issuecomment-506502047 @mcvsubbu As you can see in the benchmark number, for read/write of 128 bytes, there are about 7x performance penalty, and for 1k bytes, 20x if reading/writing byte-by-byte. And these numbers already took GC into account. The whole point of using off-heap is to reduce the long-lived large on-heap garbage but allow short-lived small garbages which can be easily collected (think about how to read from off-heap buffer). The extra byte buffer in the previous implementation is not necessary as well because you can directly call PinotDataBuffer.getByte()/putByte(). This PR focuses on the performance gain from batch read/write, so as the benchmark.
---------------------------------------------------------------- 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] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
