[ https://issues.apache.org/jira/browse/CASSANDRA-8971?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
C. Scott Andreas updated CASSANDRA-8971: ---------------------------------------- Component/s: CQL > Multiplexed storage protocol > ---------------------------- > > Key: CASSANDRA-8971 > URL: https://issues.apache.org/jira/browse/CASSANDRA-8971 > Project: Cassandra > Issue Type: Improvement > Components: Coordination, Streaming and Messaging > Reporter: Robert Stupp > Priority: Major > > The current storage protocol requires to (de)serialize the whole {{Verb}} > once into a big byte-buffer with the intention not to influence other > concurrent requests. Just want to say that removing the payload size doesn't > eliminate the need to buffer the whole {{Verb}}. > Both streaming-result-set stuff (CASSANDRA-8483) and removing payload size > (CASSANDRA-8100) could benefit from a multiplexed storage protocol - > comparable to the move from thrift to native for the client side. Means: if > we have streams on the storage protocol, we could easily implement both > without influencing other concurrent requests. > With CASSANDRA-8483 working from client via coordinator to replicas, a > request could also become more responsive - i.e. results that are available > early can reach the client early. > Beside improving payload handling a multiplexed storage protocol would also > reduce number of threads and heap pressure required by temporary and possibly > big buffers. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org