I'm testing Hbase read performance recently. the Random-Read-Latency in my cluster is beyond 200 ms.So Slowly.
I dumped 100G data(about 1G keys with 100byte value each) in the cluster,and set java head size to 10G,tuning the block.cache.size to 0.35 but it does not work. How could I decrease Random Read Latency？ I wonder what takes so much time，so I turn on RPC-level logging. The log outputs: 2011-07-26 15:35:20,354 DEBUG org.apache.hadoop.ipc.HBaseServer: got #64449 2011-07-26 15:35:20,354 DEBUG org.apache.hadoop.ipc.HBaseServer: IPC Server handler 83 on 50061: has #64449 from 192.168.1.108:59377 2011-07-26 15:35:20,354 DEBUG org.apache.hadoop.ipc.HBaseServer: Served: get queueTime= 0 procesingTime= 0 2011-07-26 15:35:20,354 DEBUG org.apache.hadoop.ipc.HBaseServer: IPC Server Responder: responding to #64449 from 192.168.1.108:59377 2011-07-26 15:35:20,354 DEBUG org.apache.hadoop.ipc.HBaseServer: IPC Server Responder: responding to #64449 from 192.168.1.108:59377 Wrote 11 bytes. what's the log's meaning? did it say the Region Server take 0 ms to process Get Op?
