Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
HBase >> mail # dev >> what's the roadmap of secondary index of hbase?


Copy link to this message
-
RE: what's the roadmap of secondary index of hbase?
Right.  Rather than another table, you use another column family in the primary table a la Megastore and Lily.

> -----Original Message-----
> From: Ted Yu [mailto:[EMAIL PROTECTED]]
> Sent: Tuesday, March 01, 2011 9:42 AM
> To: [EMAIL PROTECTED]; [EMAIL PROTECTED]
> Subject: Re: what's the roadmap of secondary index of hbase?
>
> >> Some extra support in the master to colocate secondary index regions
> >> with
> primary table regions is an interesting idea.
> Google Megastore can embed index within the primary table.
> If we can mark index rows, this approach would be feasible.
>
> On Tue, Mar 1, 2011 at 9:14 AM, Andrew Purtell <[EMAIL PROTECTED]>
> wrote:
>
> > > [Lars]
> > > What is the reason not to use JDs replication WLA tracking (using
> > > ZK) and add another "scope" like attribute to the coldefs to define
> > > the indexing?
> >
> > This is similar to what I proposed in the original description on
> > HBASE-3257 (
> > http://mail-archives.apache.org/mod_mbox/hbase-
> issues/201011.mbox/%3C2
> > 5844899.223551290369437343.JavaMail.jira@thor%3E
> > )
> >
> > Regarding Dhruba's comment, using a table as a log+workqueue of
> > secondary index updates is pretty much what Lily does:
> > http://www.lilyproject.org/lily/about/playground/hbaserowlog.html
> >
> > Some extra support in the master to colocate secondary index regions
> > with primary table regions is an interesting idea. The logic should be
> > equally pluggable as that for (secondary) index key generators.
> > Perhaps a balance/migration calculation with a pluggable term that can
> > contribute some affinity given a list of regions already on the RS.
> >
> >    - Andy
> >
> >
> >
> >
> >