> It already has the ability to take a key-value BDB and add range
> queries and alternate indexes to it.

> On Jun 28, 2:17 pm, erdody <erd...@gmail.com> wrote:

> > Hello,

> > We have decided to use Voldemort and there is one problem we need to
> > address.
> > We have a few fields (we use JsonSerializer and the value is a map of
> > several fields) we need to query by, on rare occasions, and we would
> > need range queries on some of these fields. These queries would be
> > made as maintenance tasks so performance is not critical, but we need
> > to be able to perform them somehow.
> > One option I was considering is to take advantage of the Secondary
> > index support in BDB and create an extension of BdbStorageEngine that
> > creates a BDB secondary database for each secondary index we need. The
> > SecondaryKeyCreator would need to deserialize the main value, extract
> > the corresponding value and create the secondary key. This creates a
> > performance overhead, but based on our preliminar tests, it's not very
> > significant.
> > The main advantage of this approach, as opposed to maintaining a
> > secondary index by hand (double StoreClient.put) is that the secondary
> > insert is transactional, handled by BDB.

> > Then, to make the range query, I was thinking of using something
> > similar to the AdminClient.fetchKeys, but taking advantage of the BDB
> > range queries support. I would need to make such query for each node
> > (could be done in parallel), of course, and merge the results.

> > A) Is our idea too crazy?
> > B) Other suggestions?
> > C) Are range queries or secondary indexes support in the Voldemort
> > roadmap?

> > Thanks in advance,

> > Diego Erdody