Not that kind of Shard

Tokutek steps up with new solution for MongoDB blues

Lucy Carey

With TokuMX version 1.4, company is offering improved sharding that resolves the bottlenecks experienced by Mongo users.


Are MongoDB bottlenecks giving you grief? John Partridge, CEO of database solution provider Tokutek, mighthave a solution for you. Here, speaks to JAXenter about the company’s latest offering: TokuMX 1.4. This release builds on the work the team have done to help deliver a highly scalable database boosting solution with newly enhanced sharding, which Tokutek claims will resolve performance issues in the biggest NoSQL database around.

JAX: How did you identify the MongoDB ‘pain points’ that you looked to address with this release?

Partridge: We read the comments posted to various MongoDB discussion groups on Google Groups, Reddit, Hacker News, etc. We also spoke at length with our existing TokuDB customers who were also using MongoDB or contemplating doing so.

Did you identify any other issues that your team will be focusing on for future releases?

Yes and they all revolve around improving MongoDB performance for challenging Big Data workloads. They relate to scaling out a MongoDB deployment, particularly with an eye towards how MongoDB impacts DevOps at large scale.

What were the biggest challenges faced in developing TokuMX 1.4?

Frankly, the biggest issue was not to get too far ahead of ourselves and risk a drop in robustness and reliability. We wouldn’t want to introduce bugs in our desire to rapidly bring new features and performance to market.

Tokutek implements Fractal Tree indexes to offer high performance on write-intensive workloads with the TokuDB database. How did you arrive at the conclusion that this was the best solution for data management?

The founders [of Tokutek] are academic researchers from MIT, Rutgers, and SUNY Stony Brook with commercial experience at companies like Google, Akamai, Thinking Machines, and Cray. When they started digging into the problems that databases face when handling Big Data workloads they quickly discovered that indexing was the bottleneck.

In hindsight, that’s not a big surprise since the traditional algorithm that Oracle, Microsoft, MySQL, MongoDB, basically everybody, uses was developed in the 1970s.

It’s called the B-tree, and when the founder discovered a better algorithm that can insert up to 200x (two hundred; that’s not a typo) faster than a B-tree they knew it was time to start a company.

Do you see MongoDB’s issues with scaling out as potentially detrimental to the future of the database, especially with so many rivals nipping at its heels?

MongoDB has done a terrific job on ease of use, both at the language level and at the systems management level. Tokutek adds enterprise class performance to that already very attractive value proposition and we don’t see any rival with a technology that can touch the scaling capabilities of TokuMX.

What do you think the NoSQL market will be like in the next few years? As convergence picks up pace within the sector, what do you think the implications will be for Tokutek?

As the NoSQL market matures I think we’ll see some consolidation, especially as the developer community establishes best practices and gains an understanding of which tools are best for which types of applications.

Tokutek stands to benefit under all scenarios because what we do – advanced indexing – is essential for any workload where the application needs to find a specific piece of information more than once. Today, we’re uplifting MongoDB performance by replacing its B-tree indexing with our Fractal Tree indexing.

If tomorrow a new product shows up on the scene that is just as popular, the customers will want the performance and cost savings that only our Fractal Tree indexing can deliver.

Inline Feedbacks
View all comments