is hbase 'checkAndPut' latency significantly higher than simple 'put'?

["Dhodapkar, Chinmay" <ch...@qti.qualcomm.com>]

Logically, it seems checkAndPut will take more time. I am interested specifically in load scenarios where we see avg checkandput latency of 15ms (for 17KB per row of data)....will converting the checkAndPut to simple 'PUT ' significantly reduce avg latency.

For 99% of use cases the row key that we write to does not even exist when we do a checkAndPut.

Thanks,
Chinmay

Re: is hbase 'checkAndPut' latency significantly higher than simple 'put'?

[Jean-Daniel Cryans <jd...@apache.org>]

The "check" part is basically doing a Get before the Put, and this is where
additional latency would come from. If the Get reads cold data then it's
gonna be a little slower. You could try adding bloom filters on that
family. Benchmarking with straight Puts is also something you should be
doing in order to figure where the slowness comes from.

J-D

On Thu, Nov 15, 2012 at 11:12 AM, Dhodapkar, Chinmay <
chinmayd@qti.qualcomm.com> wrote:

> Logically, it seems checkAndPut will take more time. I am interested
> specifically in load scenarios where we see avg checkandput latency of 15ms
> (for 17KB per row of data)....will converting the checkAndPut to simple
> 'PUT ' significantly reduce avg latency.
>
> For 99% of use cases the row key that we write to does not even exist when
> we do a checkAndPut.
>
> Thanks,
> Chinmay
>
>