I have another facet to the performance regressions in .18. Let me know if this should be a separate issue.
The performance of large result sets has significantly worsened. One of the ones I'm looking at is about 100% slower, going from 400ms to 800ms.
I believe this is because the new connection code uses active mode sockets and by default is receiving something like a single packet of data per message (byte_size says about 1400). For each tiny message, we append it into the buffer and try to parse a frame from it.
I experimented a bit and while I don't know if this is the right way to solve this, if I set the buffer option on the transport options to some large number, it fixes the performance degredation.
The performance of large result sets has significantly worsened. One of the ones I'm looking at is about 100% slower, going from 400ms to 800ms.
I believe this is because the new connection code uses active mode sockets and by default is receiving something like a single packet of data per message (byte_size says about 1400). For each tiny message, we append it into the buffer and try to parse a frame from it.
I experimented a bit and while I don't know if this is the right way to solve this, if I set the
bufferoption on the transport options to some large number, it fixes the performance degredation.For instance
Thoughts?
Originally posted by @kennethito in https://github.com/whatyouhide/xandra/issues/357#issuecomment-2053441538