hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrew Purtell <apurt...@yahoo.com>
Subject Re: [hbase] table HRegionServer affinity
Date Tue, 08 Jan 2008 06:01:11 GMT
> stack <stack@duboce.net> wrote:
> Andrew Purtell wrote:
> > 1) The ability to fix a given key range to a region.
> > This would both assign a range to a given region, and
> > also disable splitting over that range. Aside from API
> > changes, ideally there would be a HBase shell command
> > to support this.
> Do you really want to be able to set the start and end
> key on a single region or is it more that you want the
> japanese info to be kept in regions that are hosted at
> the japanese datacenter, the european regions in the
> european datacenter, etc?

Initially that is the case indeed. While the algorithms
distributed over the clusters would be location agnostic,
the particulars of the workload would not be, so the
(data-driven) result would be region specific. 


> Hbase has no notion of constraining data mobility
> currently.  To add it, I'd imagine you'd want to enhance
> the master region assignment algorithm so it takes into
> consideration geographic location.
> It has been a (secondary) intent that hbase servers can
> be sub-classed so alternative or enhanced implementations
> can be plugged in.  This facility has been little
> exercised.  Let us know if there are any changes you need
> made to get you going.

Thanks. This is encouraging. We'll give it a go.

Going forward what would be the protocol for bug reports or
discussions regarding perhaps minor enhancements? Would it
be appropriate for me to submit an issue under
contrib/hbase on JIRA? 

Thanks again,

Andrew Purtell
Advanced Threats Research
Trend Micro, Inc, Pasadena, CA USA
(personal mail)

Looking for last minute shopping deals?  
Find them fast with Yahoo! Search.  http://tools.search.yahoo.com/newsearch/category.php?category=shopping

View raw message