Mailing-List: contact user-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@hbase.apache.org
Received-SPF: softfail (nike.apache.org: transitioning domain of
 michael_segel@hotmail.com does not designate 173.15.87.35 as permitted
 sender)
Content-Type: text/plain; charset=iso-8859-1
Mime-Version: 1.0 (Mac OS X Mail 6.5 \(1508\))
Subject: Re: Can I make use of TableSplit across Regions to make my MR job
 faster?
From: Michael Segel <michael_segel@hotmail.com>
In-Reply-To: 
 <CAPvS-K_B2XxwnLC3wQLMYy0AFpiSbfQM_QHgCeAqRnNup3ph1A@mail.gmail.com>
Date: Mon, 26 Aug 2013 05:48:50 -0500
Content-Transfer-Encoding: quoted-printable
Message-Id: <C69D7C62-DD83-4365-B954-E9118FB70438@hotmail.com>
References: 
 <CAPvS-K_B2XxwnLC3wQLMYy0AFpiSbfQM_QHgCeAqRnNup3ph1A@mail.gmail.com>
To: user@hbase.apache.org

A 'table split' is a region split and as you split regions, balance the =
regions, you should see some parallelism in your M/R jobs.=20

Of course depending on your choice of row keys... YMMV.

HTH

-Mike

On Aug 26, 2013, at 2:16 AM, Pavan Sudheendra <pavan0591@gmail.com> =
wrote:

> Hi all,
>=20
> How to make use of a TableSplit or a Region Split? How is it used in
> TableInputFormatBase#
> getSplits() ?
>=20
>=20
> I have 6 Region Servers across the cluster for the map-reduce task =
which i
> am using, How to leverage this so that the table is split across the
> clusters and the map-reduce application finishes fast.. Right now, it =
is
> very slow.. For aggregating 3 table values, 1 with 100,000 rows and =
other
> two tables i'm only using get operating to get the value by passing =
the
> key.. For this setup, it takes 40-50 mins.. Which is worse.. The first
> table would eventually be around 20-25m rows.. Please lead me in the =
right
> way.. I will paste the code if anybody is interested.
>=20
>=20
> --=20
> Regards-
> Pavan

The opinions expressed here are mine, while they may reflect a cognitive =
thought, that is purely accidental.=20
Use at your own risk.=20
Michael Segel
michael_segel (AT) hotmail.com