In page 3 of http://www.slideshare.net/schubertzhang/hbase-0200-performance-evaluation, for 0.20.0 random reads: 1106, sequential reads: 5433. It is about 1 to 5.
In the wiki, http://wiki.apache.org/hadoop/Hbase/PerformanceEvaluation, for 0.19.0RC1 random reads: 540, sequential reads: 464. It is about 1 to 1. I was doing similar performance evaluation against hbase 0.20.3 hadoop 0.20.1. I did not get as good sequential read as that and would like to know the interesting fact what has changed from 0.19.0RC1 to 0.20.x to make that improvement. Thanks, -gabe On Thu, Feb 4, 2010 at 10:58 PM, Stack <[email protected]> wrote: > On Wed, Feb 3, 2010 at 3:58 PM, Gabriel Ki <[email protected]> wrote: > > Hi, > > > > I was reading > > > http://www.slideshare.net/schubertzhang/hbase-0200-performance-evaluation. > > Could someone explain what has changed to improve the random reads to > > sequential reads ratio from 1:1 to 1:5? I don't seem to reproduce such > good > > sequential reads. > > > Which slide is that Gabriel? Its saying that we do 5 sequential reads > per random read? What are you seeing? Why not scan instead of do > sequential reads? > > St.Ack >
