Mailing-List: contact user-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@hbase.apache.org
Received-SPF: neutral (nike.apache.org: local policy)
MIME-Version: 1.0
In-Reply-To: 
 <CAMPNBf53YLK-SgrcHOGgO3APGFH_mDmaK4-7wop-0=96Y2b8fg@mail.gmail.com>
References: 
 <CAMPNBf53YLK-SgrcHOGgO3APGFH_mDmaK4-7wop-0=96Y2b8fg@mail.gmail.com>
From: Jean-Marc Spaggiari <jean-marc@spaggiari.org>
Date: Tue, 9 Apr 2013 07:51:14 -0400
Message-ID: 
 <CAPQV63VJ9g5229heD4V1KskvYVOwALU9=_z-8fyCrxiuYqL=zQ@mail.gmail.com>
Subject: Re: Efficient way to use different storage medium
To: user@hbase.apache.org
Content-Type: text/plain; charset=UTF-8

Hi Bing,

If you mount all your drives into HDFS, some blocks are going to be on
SSD and some on regular drives. So from reads are going to be fast,
and some others are going to be slow.

On a single machine, I don't think you can specify which table will be
on which drive since the blocks are going to be spread over the
drives.

You will still see some performances improvement since sometime it's
the SSD which is going to be used.

Also, at some points, your SSD drives might be full before your
regular drives. They will removed from valide WRITE destinations, but
will still be used as READs.

JM

2013/4/9 Bing Jiang <jiangbinglover@gmail.com>:
> hi,
>
> There are some physical machines which each one contains a large ssd(2T)
> and general disk(4T),
> and we want to build our hdfs and hbase environment.
>
> If we use all storage(6T) as each machine provides, I want to know whether
> it is an efficient way to
> make advantage of ssd, or provide different performance for different table
> through hdfs or hbase's configuration as application's requirements.
>
> From my views, different physical storage medium will provide a good chance
> to think over hdfs or hbase's infrastructure, so any views from yours will
> be nice.
>
> Regards
>
>
> --
> Bing Jiang
> weibo: http://weibo.com/jiangbinglover
> BLOG: http://www.binospace.com
> National Research Center for Intelligent Computing Systems
> Institute of Computing technology
> Graduate University of Chinese Academy of Science