Re: Read TPS in 0.8.6

2011-12-23 Thread Radim Kolar
cassandra read performance depends on your disk cache (free memory at node not used by cassandra) and disk IOPS peformance. In ideal case (no need to merge sstables) cassandra needs 2 IOPS per data read if cassandra key/row caches are not used. Standard hard drive has about 150 IOPS. If you ha

Read TPS in 0.8.6

2011-12-23 Thread Jeesoo Shin
Hi, I'm doing stress test with tools/stress (java version) I used 3 EC2 Xlarge with 4 raid instance storage for cluster. I get write TPS min 4000 & up to 1 but I only get 50 for read TPS. is this right? what am I doing wrong? these are option. java -jar stress.jar --nodes=ip1,ip2,ip3 --consist