create an index in riak

John Muellerleile johnm at basho.com
Thu Nov 19 15:48:00 EST 2009


Yes - for optimization of queries where some or all of the data involved is
serendipitously node local.  Riak Search does not do away with any of the
fundamentals Riak provides, namely replicas, in this case; there are
instances where some or all operations that make up a query can be executed
on one node (as opposed to a distributed model) because the involved data
sets have replicas that are coincident.  This possibility allows a system
like the one described in the paper to become useful.  I referenced it
because I am developing optimizations that capitalize on node coincident
data, and it reflects the spirit of part of this endeavor, if not the
implementation.

On Wed, Nov 18, 2009 at 5:14 PM, Jeff Hammerbacher <hammer at cloudera.com>wrote:

>
> G. Candea, N. Polyzotis, R. Vingralek.  A Scalable, Predictable Join
>> Operator for Highly Concurrent Data Warehouses.
>>
>>
> Really? This paper had nothing at all to do with distributed systems, and
> when pressed in person, the authors indicated that they did not intend the
> research to extended beyond the single node case.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.basho.com/pipermail/riak-users_lists.basho.com/attachments/20091119/a5d980be/attachment-0002.html>


More information about the riak-users mailing list