Dashboard Insight’s Maroushka Kanywani spoke with Jesse Fountain, DATAllegro’s VP, Pre-Sales Services and Joe Longtin, Product Marketing Manager, also at DATAllegro and they discussed corporate strategy and product development at DATAllegro.
Maroushka Kanywani: Could you please tell me about DATAllegro’s history?
Jesse Fountain: DATAllegro was founded in 2003 and we came out with a new idea on how to approach this data warehouse appliance thing. What we did was we took off-the-shelf hardware and we designed a very, very fast, massively parallel, processing system to break data up into very small pieces and then run queries concurrently against each of those small pieces. So then instead of queries taking hours, days and weeks to run, they would take seconds or minutes to run.
To offer some context, the world has been sort of suffering - data warehouses have become more and more mainstream and everyone wants to have everything in a data warehouse. Data warehouses have grown by leaps and bounds; we like to say they have grown geometrically at least - if not exponentially. Now companies are faced with a problem: they’ve got massive amounts of data and a large number of people who want to do trending, analysis and so forth and they just can’t do it and they’re really having a struggle managing that.
If these companies were to take these very large amounts of data and run them on conventional platforms like Oracle or DB2 or similar ones, while it can be done, there will be a couple of problems that the companies will run into.
The first problem is that it’s going to be extremely expensive for the companies to support that kind of environment on conventional platforms which we call SMPs, single multi-processing. What happens here is that everything is shared; shared CPUs, storage across multiple instances of Oracle or DB2. This works out to be very inefficient for very large databases. It also becomes prohibitively expensive because one is required to buy more and more iron in order to support basic functionality like running a query against the data warehouse.
MK: So essentially as the data grows, so does the expense of running and maintaining it.
JF: It grows almost exponentially – especially in conventional systems. So for the last 27 years the incumbent, called Teradata, over the decades has provided a largely separate and distinct proprietary platform that runs very large databases at extremely fast speeds. Teradata has enjoyed that luxury for the last couple of decades but there are more players in that space now – DATAllegro being one of them.
MK: You mention the issue of a proprietary platform and I understand that this is one of the distinguishing marks between DATAllegro and Netezza - DATAllegro does not use a proprietary platform. Could you tell me more about this, please?
JF: You’ve actually corrected me without even realizing it. Netezza is purely a proprietary environment and this has worked very well for them. They have found however, that it is difficult to get into certain companies which are resistant to proprietary architectures. In other words, for these companies, whatever vendor they deal with, they want them coming in with standard operating systems, standard storage, standard processors and so forth and that is not what Netezza is about.
MK: I’m not familiar with non-proprietary platforms; could you please tell me more about them?
JF: Let me give you an analogy that might make it easier. Let’s say you buy a car in the U.S. and you want to get maintenance done on that car. It would be really nice if you could walk into a garage which has standard parts, standard sizes and standard fittings because you would get much cheaper rates. On the other hand, if you buy a Lamborghini and go to the corner garage, it’s unlikely they’ll be able to fix that car since it does not use standard parts, making it very expensive and necessitating a visit to a specialist.
The same holds true with our appliances. If I brought something in that was entirely proprietary, my existing data center staff would not know how to maintain it since it is not standard storage, power, or behavior. A lot of big companies don’t like proprietary technology and while historically proprietary technology may be a great market disrupter it’s been proven that it does not go on indefinitely. It’s not scalable, interchangeable or re-usable.
DATAllegro uses Infiniband or GigE and these are standard communication protocols. Teradata, by comparison, uses BYNET, which is very proprietary. It is an intelligent network technology and could be considered the secret element to Teradata’s technology and it is what Teradata sells the most of.
MK: In terms of players in the same space, you’ve mentioned Teradata and Netezza. Who else would you include in the same space?
JF: Our third and only real main competitor in this space at this juncture in time is a company called Greenplum. Greenplum has been around longer than DATAllegro has and they’ve tried to find a home and it looks like they have. That home is Sun. Greenplum has partnered very closely with Sun and of all the companies out there, their architecture is most closely aligned, in terms of architecture with what DATAllegro does. They are an open architecture, they can run on just about any platform out there although they don’t sell themselves that way – they sell themselves exclusively on Sun.