[ https://issues.apache.org/jira/browse/HBASE-15403?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15181985#comment-15181985 ]
Anoop Sam John commented on HBASE-15403: ---------------------------------------- --rows dont mean total rows which will be written.. I see, this is the #times one client thread loop runs. The write tests allows for multi? If so this will be #multi requests. So then that will come out like #threads * --rows * puts per multi... For scan --rows also means same way. If we test with ScanRange1000 and --rows = 1000 and 10 threads. total 1000* 1000 * 10 rows read overall.. Yes there may be duplicate rks coming out across diff threads. Need to check more on code. > Performance Evaluation tool isn't working as expected > ----------------------------------------------------- > > Key: HBASE-15403 > URL: https://issues.apache.org/jira/browse/HBASE-15403 > Project: HBase > Issue Type: Bug > Components: Performance > Affects Versions: 1.2.0 > Reporter: Appy > Priority: Critical > > hbase pe --nomapred --rows=100 --table='t4' randomWrite 10 > # count on t4 gives 620 rows > hbase pe --nomapred --rows=200 --table='t5' randomWrite 10 > # count on t5 gives 1257 rows > hbase pe --nomapred --table='t6' --rows=200 randomWrite 1 > # count on t6 gives 126 rows > I was working with 1.2.0, but it's likely that it'll also be affecting master. -- This message was sent by Atlassian JIRA (v6.3.4#6332)