hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jonathan Gray" <jl...@streamy.com>
Subject RE: Unofficial HBase Blog
Date Tue, 02 Sep 2008 01:53:24 GMT
Leitao,

That type of functionality can and is being used with HBase, but not at all
as a fully integrated part of HBase.  Joins and sorts are not part of
BigTable or HBase "out of the box" because that type of functionality is not
what the column-oriented data model is about.

There was a discussion on this list last week that went into detail about
the differences of HBase and relational databases.  Reading those posts
might give some clarity as to why this is.

Of course this functionality is what many want/need, so there are some
solutions today (Lucene indexing, etc) with more likely down the road as the
number of contributors and users grow.

The solution for many is to build an logic layer on top of HBase that can
perform these joins/sorts/etc, often combined with some type of cache to
compensate for HBase's relatively poor random-read performance.

Jonathan Gray


-----Original Message-----
From: Guo Leitao [mailto:leitao.guo@gmail.com] 
Sent: Monday, September 01, 2008 6:37 PM
To: hbase-user@hadoop.apache.org
Subject: Re: Unofficial HBase Blog

Hi J-D,

Sorry to hear that there is no plan for HBase to develop JDBC/ODBC API.
Actually, we are now trying to implement some SQL function on Bigtable
(HBase or Hypertable implementation, not deceide which one is suitable for
us), such as JOIN/GROUP/SORT/COUNT.

Is there any plan for Hbase to implement such functions?

Many thanks!

Leitao


2008/9/1 Jean-Daniel Cryans <jdcryans@apache.org>

> Leitao,
>
> A JDBC plugin is not currently planned and I'm not sure that it would be
in
> the best interest for HBase users. The shift in paradigm requires moving
> OLAP (or any complicated query) to MapReduce which is offline and where
> JDBC
> won't be of any use. It may also mislead new users into thinking that
HBase
> is relational.
>
> But, if anyone in the community did code a JDBC plugin, I'm sure we could
> put it in a "contrib" folder in HBase distribution.
>
> I think that a nice alternative for you would be to abstract database
> interactions behind some database access objects (DAO) if it's not already
> your situation. This way your developers unfamiliar with HBase would be
> shielded from the new API.
>
> Regards,
>
> J-D
>
> On Mon, Sep 1, 2008 at 4:07 AM, Guo Leitao <leitao.guo@gmail.com> wrote:
>
> > Hi J-D,
> >
> > Thanks for your blog!
> >
> > We have an application running on Sybase, and now, we want to migrate
> this
> > application onto HBase. I'd like to know whether there is a plan for
> HBase
> > to develop a stardard(subset) ODBC/JDBC API?
> >
> > Cheers!
> >
> > Leitao Guo from ChinaMobile
> >
> > 2008/8/30 Jean-Daniel Cryans <jdcryans@apache.org>
> >
> > > Hi community,
> > >
> > > In a move to better inform the new and current users of HBase, I
> decided
> > to
> > > start a blog. For the moment, this will be personal project where I
> will
> > > share my thoughts, tricks and other useful stuff. You can find it
here:
> > >
> > > http://jdcryans.blogspot.com/
> > >
> > > Happy hbasing!
> > >
> > > J-D
> > >
> >
>


Mime
View raw message