hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jean-Daniel Cryans <jdcry...@apache.org>
Subject About data locality (Was: Re: HBase on same boxes as HDFS Data nodes)
Date Thu, 08 Jul 2010 17:58:27 GMT
(changing the subject, let's not hijack threads)

> will the data move over time though...for example if i have lots of access to data in
DataNode A ? without the current work that is in progress..

HBase has no control on that, but data will be moved if those regions
are used. Like the article explains, the first replica goes to the
local node, so through compactions/flushes one replica of each block
will be on the local node.

Also keep in mind that the new datanode may already contain some
replicas of some of the blocks for that region, so it's not just black
and white. This is quite possible on a small cluster, but over 1k
nodes not that much ;)

J-D

On Thu, Jul 8, 2010 at 10:51 AM,  <vramanathan00@aol.com> wrote:
>
>  Thankyou..
> I've some more questions
> I'm spending quite a bit over last few weeks to develop one of our applications using
HBase/Hadoop
> & using 0.20.4
>
> Hbase - Table X
> rows - 1- 100 -> Region A -> RegionServer A     --> DataNode A
> ....
> rows  1500 - 1600 -> Region M -> RegionServer B -> DataNode B
>
> So based on what I have read so far..I'm thinking of Region Server A & Data Node
A pairs on the same host to
> make use of locality..
>
> As per your answer ..If we restart the cluster, because of radom assigment, locality
is gone
> so..Region Server B -..> Region A ---> data blocks will be in Data Node A
> ...if I understand correctly..
> will the data move over time though...for example if i have lots of access to data in
DataNode A ? without the current work that is in progress..
>
> thanks again for your reply
>
> venkatesh
>
>

Mime
View raw message