Yes, Virginia, There is Going to be a TokuMX Storage Engine for MongoDB 2.8

UPDATE: Since the publication of this blog post, Tokutek subsequently published preliminary benchmark results for this new TokuMX sibling, now called TokuMXse. In addition, there have been a number of release candidates.  Finally, MongoDB ultimately decided to call the new release of MongoDB v3.0 (previously it had been called v2.8).

Learn more about the first TokuMXse performance benchmark here; and learn how to download v1.0.0.RC.4 here.

First things first… Congratulations to Keith, Michael, and the rest of the WiredTiger team.  We look forward to seeing you in the MongoDB market in 2015 . That said, while this news is certainly germane, it’s not the sole motivation behind today’s post.

The reason for this post is to formally and officially disclose what we have been telling our customers and prospects (when asked) for weeks: Yes. Tokutek will indeed offer a storage engine API edition of TokuMX for MongoDB.

Background: As most folks in the MongoDB community know, MongoDB recently announced their intention to make a storage engine API available in its forthcoming release of MongoDB, version 2.8. That release was originally slated for this month, but according to recent statements by Eliot Horowitz, the release will most likely be available sometime in early 2015.  In addition to announcing the API, Eliot and team reached out to storage engine vendors, including Tokutek and WiredTiger, encouraging them to build products based on that API.  Tokutek fully embraced this move by MongoDB, and as anyone who follows us in JIRA knows, we started work on a version of TokuMX based on the new APIs soon thereafter.

What it all means: If you believe, as we do, that competition is a good thing — one that results in more / better choices for users — this can only mean good things are going to continue happening in the world of high-performance big data applications built with MongoDB. Soon the MongoDB community will have several storage engine choices and among them will be the new, to-be-named, TokuMX sibling.

Implications for TokuMX: Unfortunately, the new MongoDB storage engine APIs will not enable storage engine vendors to, for example, offer ACID compliant transaction support; a better election protocol; improved replication and sharding; clustered secondary indexes; and so on. These TokuMX improvements to MongoDB require a deeper level of software engineering.  Consequently, Tokutek will offer two products for MongoDB: TokuMX, the product we offer today with the aforementioned improvements and a new, storage-engine-only, sibling.

Watch this space in the coming weeks for more news about our new TokuMX product, and comparisons between it, the full version of TokuMX, WiredTiger, and more. It’s going to be an interesting new year.

More about TokuMX: In June of 2013 Tokutek released a fork of MongoDB built with the same proven, modern database indexing engine it first introduced for MySQL in 2008. Since its introduction, TokuMX has won over numerous MongoDB users with its high-performance ingestion and compression abilities. The MaxCDN and Performance Horizon Group TokuMX success stories are two recently published examples of companies meeting their rapidly accelerating MongoDB data ingestion and storage needs with TokuMX

Your comments are welcome here or you can reach me on Twitter via @dcrosenlund.

Share this post

Comments (8)

  • ddorian benchmarks of tokumx vs mongodb+tokukv

    2.create a sharding layer like leif’s slides (all nodes the same) with regex sharding (redislabs)

    3.introduce sql support by merging tokudb and keep the mongo-api so webscale-believers can transition from mongodb

    4.say no to requests for buyout

    bonus: create tokupg! (no storage engine, but maybe the cstore_fwd may help(probably not))
    rip mongodb

    December 16, 2014 at 4:20 pm
  • sven

    The new version 3.0 of MongoDB allows compression as well. How does this compare with tokuMX-SE in terms of file size and speed?

    February 23, 2015 at 7:33 pm

Comments are closed.

Use Percona's Technical Forum to ask any follow-up questions on this blog topic.