Kognitio: clarifying misunderstandings

Kognitio is often misunderstood. The company is not entirely blameless in this regard but it is also the fault of analysts and the market more generally, which likes to pigeonhole vendors and their products.

The first misunderstanding is general. Kognitio is usually categorised as a “data warehousing” vendor. The problem with that is that “data warehousing” is too broad a term. What Kognitio is good at is supporting (very) complex analytics: that’s more of the function of a data mart than a data warehouse. The company describes its product as an analytical platform, which is fair enough.

The second misunderstanding comes about because Kognitio is an “in-memory database” and there’s been a lot of hype around in-memory recently, largely thanks to SAP. The problem arises because people want to put all “in-memory” technologies into a single bucket. And the truth is that that simply isn’t a reasonable proposition. Naturally enough, given all the marketing around SAP HANA, Kognitio needs to position itself against HANA but, in my view at least, they simply aren’t competitive products. In terms of performance and functionality (not to mention that Kognitio has been doing this for 10 times longer) HANA is not yet in the same ballpark as Kognitio let alone able to compete on a level playing field.

Kognitio has not helped itself here by promoting a comparative paper, published by EMA, which compares Kognitio with SAP HANA and Oracle TimesTen. I understand that it was Oracle that suggested the inclusion on TimesTen rather than Exalytics but, frankly, this makes a mockery of the whole paper, since TimesTen is even further from Kognitio’s ballpark than SAP HANA.

A third misunderstanding I need to correct is one that is promulgated on the DBMS2 website run by Curt Monash. He states there that Kognitio isn’t an in-memory database because it stores indexes on disk. This isn’t true. It doesn’t. I am not sure where Curt got this idea from but according to Kognitio it derives from a misunderstanding about the way that Kognitio allows you to remove disk blocks. Anyway, you can run Kognitio entirely without disks if you want to so it patently can’t be the case that indexes have to be on disk.

A potential misunderstanding may be the way that Kognitio runs in conjunction with Hadoop. I say potential only because there is a danger that it again gets lumped into a bucket that assumes that all “data warehousing” vendors offer the same sort of approach to support for Hadoop. In fact, Kognitio’s approach is most similar to the way that DataStax works, whereby you can have a Cassandra implementation and a Hadoop implementation sharing the same cluster. Ditto Kognitio and Hadoop. This is actually a very powerful offering because of the low cost of the cluster hardware, which should mean that you get excellent TCO. In addition, unlike most vendors that merely connect to Hadoop, Kognitio is integrated with it so that you can do things like push down processing into the Hadoop environment, and there’s no need to modify any of your SQL in the process. It is also noteworthy that Kognitio has a partnership with HortonWorks.

Finally, a word or two about the company. The perception has been that the company was a) British and b) too technical for its own good. That’s true but it’s also changing. While the head office is still in the UK, half the company’s sales are now in the United States and with a change in management over the last couple of years Kognitio is now much more market-focused. It still has work to do to focus on solving customer problems rather than just offering great technology but that is happening. Unfortunately, technology-driven companies, like oil tankers, can be slow to change course but then again, once they do change, they can be hard to stop. While Kognitio tends not to be on the radar of the major players as a significant competitor that could change if the company gets its marketing act together. The signs are that that may just be happening.