hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From stack <st...@duboce.net>
Subject Re: Question about how queries are distributed
Date Mon, 04 Aug 2008 18:03:14 GMT
See below:

Leon Mergen wrote:
> ...
> However, as I was suspecting, the Thrift functionality does appear to be
> lacking compared to the native Java API -- since HBase is not a project
> where you want to force your users into a specific development environment,
> I do feel this should be fixed (especially since you cannot efficiently
> write locality aware map/reduce jobs with the Thrift API, unless you will be
> running a Thrift server on every single slave node in your cluster).
Agreed. HBASE-697 is about updating our current IDL so it exposes new 
features added in 0.2.0.

> I am not aware of the possibilities, especially since the HBase client
> connects to the region servers "under the hood", where in all the
> Thrift-like API's you have to manually setup communications. However, if no
> one picked up on this by then, I will be willing to look into it in a few
> months from now.
That'd be great.  Thanks.

Andrew Purtell wrote:
>> Something to think about is integration of Thrift with the master
>> and regionserver themselves as a first class API. I think the
>> Thrift (and also the REST) APIs as clients/front ends are proof-
>> of-concepts more than anything else.
Yes.  Or just make it so all master/regionservers/client traffic is atop 
thrift -- or at least provide the option.  This might not be that hard 
to do with the recent changes in hadoop serialization.


View raw message