On Thu, Aug 30, 2012 at 9:24 AM, Jay T <[email protected]> wrote: > Thanks Stack for pointing us in the right direction. Indeed it was the > tcpNodeDelay setting. We set these to be true. > > ipc.server.tcpnodelay ==> true > hbase.ipc.client.tcpnodelay ==> true > > All reads that previously had the 40ms overhead are now between 2 and 3 ms > like we would expect them to be. > > Are these settings worthy of being put in the Hbase online book ? >
I added a note on nagles to the book at your suggestion. We've toyed w/ making tcpnodelay true as default but as has been reasoned in the issue cited, it depends on cell sizes whether it effective or not. We should give this config. a higher profile than it currently has it seems. St.Ack
