hbase optimizations

[vipul jhawar <vi...@gmail.com>]

Kylin team

Are there any specific points or notes on the internet that you would refer
for the hbase optimizations or tweaks recommended for data north of 20 TB
on a cube.

Have checked variety of literature available out there but wanted to check
with you guys as with your production experience, there must be specific
insights that you may want to share or recommend as we are facing GC,
compaction issues which lead sporadic issues/outages with the application
we are rolling out on top of kylin.

Thanks

Re: hbase optimizations

[Li Yang <li...@apache.org>]

For Kylin, HBase is best tuned in a read intensive way.  Kylin uses HBase
mostly in a readonly fashion.  Cube is built and BULK LOAD into HTable.
Once created, the HTable never update.  There is writes about cube
metadata, but the volume is tiny compare to cube read.

So the read tuning section of HBase book is the place to get started.

http://hbase.apache.org/book.html#perf.reading



On Fri, Aug 14, 2015 at 10:42 AM, vipul jhawar <vi...@gmail.com>
wrote:

> Hi Team
>
> Are there any pointers/practices that you would like to share.
>
> Thanks
>
> On Thu, Aug 13, 2015 at 6:33 PM, vipul jhawar <vi...@gmail.com>
> wrote:
>
> > Kylin team
> >
> > Are there any specific points or notes on the internet that you would
> > refer for the hbase optimizations or tweaks recommended for data north of
> > 20 TB on a cube.
> >
> > Have checked variety of literature available out there but wanted to
> check
> > with you guys as with your production experience, there must be specific
> > insights that you may want to share or recommend as we are facing GC,
> > compaction issues which lead sporadic issues/outages with the application
> > we are rolling out on top of kylin.
> >
> > Thanks
> >
>

Re: hbase optimizations

[vipul jhawar <vi...@gmail.com>]

Hi Team

Are there any pointers/practices that you would like to share.

Thanks

On Thu, Aug 13, 2015 at 6:33 PM, vipul jhawar <vi...@gmail.com>
wrote:

> Kylin team
>
> Are there any specific points or notes on the internet that you would
> refer for the hbase optimizations or tweaks recommended for data north of
> 20 TB on a cube.
>
> Have checked variety of literature available out there but wanted to check
> with you guys as with your production experience, there must be specific
> insights that you may want to share or recommend as we are facing GC,
> compaction issues which lead sporadic issues/outages with the application
> we are rolling out on top of kylin.
>
> Thanks
>