Hi Vladimir,

As far as I know, Redis' SortedSet is not shared. Is this correct? Is it
even possible to support similar functionality for a partitioned dataset?

-Val

On Fri, Sep 4, 2020 at 8:58 AM Denis Magda <dma...@apache.org> wrote:

> Hi Vladimir,
>
> Usually, LSM-storage engines serve the range searches the best (Cassandra
> is one of the examples). The SortedSet of Redis is one of the typical
> components you can find in LSM-engines.
>
> Presently, Ignite neither supports an LSM store nor a SortedSet data
> structure. However, the range searches with Ignite still have O(log(n))
> complexity, it just takes more steps for the B-tree to collect all the
> records of a range.
>
> -
> Denis
>
>
> On Fri, Sep 4, 2020 at 3:26 AM Vladimir Pligin <vova199...@yandex.ru>
> wrote:
>
> > Hi Igniters,
> >
> > It seems that it's not possible to implement effective leader-board with
> > the
> > current Ignite indexes.
> > Leader-board stores a score and an id of a player of some game. Score is
> > indexed. One of the possible requests to that data structure is to get
> some
> > range of scores based on their rank (it's effectively a number of a
> row). I
> > suppose it's required for pagination. For example Redis has an index
> > (https://redis.io/topics/data-types#sorted-sets) that can be scanned
> > (https://redis.io/commands/zrange) to a particular line with O(log(n))
> > complexity. As far as I know it's node-local. Correct me if I'm wrong but
> > in
> > Ignite we scan an index until we find a row corresponding to a
> limit/offset
> > clause. It looks like a linear complexity. I suppose it could be possible
> > to
> > implement it for REPLICATED caches and local queries. But it's really
> > difficult for me to estimate the efforts. By the way it would be useful
> for
> > analytical tools, most of them paginate. So what do you think is it
> > possible
> > to make that happen in Ignite and do we need it at all? Ignite 3.0 maybe?
> >
> >
> >
> > --
> > Sent from: http://apache-ignite-developers.2346864.n4.nabble.com/
> >
>

Reply via email to