hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Yu <yuzhih...@gmail.com>
Subject Re: Measuring read latency with multiple threads
Date Thu, 07 Jan 2016 14:05:21 GMT
Were the reads performed by the 100 threads going to very few regions or many regions across
the cluster ?

Which hbase release did you test against ?

Thanks

> On Jan 7, 2016, at 4:21 AM, Sreeram <sreeram.v@gmail.com> wrote:
> 
> Hi,
> 
> I am trying to measure HBase read latency with concurrent users. I am
> doing this by starting multiple threads in a Java program and measuring the
> latency at a thread level.
> 
> I was expecting a minor increase in per-thread latency - but I am observing
> that it is much more.
> 
> For example, in non-threaded mode I got 10 ms for a read operation. On
> running with 100 threads, the whole work load complete in roughly 1 sec.
> And every thread takes close to 1 second for end to end execution (all
> threads seem to start and end at the same time).
> 
> I will appreciate any input that will help me understand this.
> 
> Thank you.
> 
> Sreeram

Mime
View raw message