ParACCEL is the latest company to announce that it is joining the data warehouse appliance mêlée. That said, and before I discuss ParACCEL, another company that is rumoured to be about to enter the market is Dataupia — more on that when it actually makes any announcements.
ParACCEL isn't actually an appliance vendor per se, though it will deliver a pre-installed system if that is what you want. In fact, the software has been designed to run on any 64-bit blade cluster. On its web site the company describes its architecture as based on memory mirroring though it is not exactly clear what that means. However, I can say that the use of in-memory processing is a substantial part of what the company will be bringing to market. Beyond that I cannot go, as I am limited by a non-disclosure agreement. I can tell you, however, that the product is based on PostgreSQL but that the company, like Netezza, is not contributing its development back to the PostgreSQL community, which effectively makes it proprietary.
This should not particularly be a surprise: the CTO of ParACCEL was previously the VP of Architecture for Netezza. The CMO was previously VP Marketing for SAND.
ParACCEL is targeting the Microsoft market and it sees itself specifically as a vendor that is complementary to existing or future enterprise data warehouses rather than replacing these. While cognisant of the fact that appliances, or virtual appliances as in this case, can be used in a variety of environments (data marts, aggregation engines and so on) the company sees its initial target market as a query accelerator sitting alongside SQL Server 2005. Initially, it is targeting something in the 500Gb to 3Tb range, though the product will scale (a lot) further.
This is not a bad strategy: Microsoft is clearly targeting the low-end warehouse market where it will surely gain traction. However, its performance, especially for analytic queries, drops off rapidly at a fairly low level, so there should be ample opportunity for ParACCEL to capitalise on this fact. Moreover, none of the other appliance or virtual appliance (or, as I have previously alluded to them: semi-appliance) vendors are targeting this market, so ParACCEL should have this space to itself, at least for the time being.
One further point of interest is that ParACCEL, which is not generally available yet, is not adopting the "let's go to market and then find customers" model, which was what DATAllegro did, for example; rather, ParACCEL is working on the basis of beta customers with whom the product can be proved, prior to coming to market.
This program is already up and running, which you might think unusual for a company that you have only just heard about. However, ParACCEL is what used to be XPrime or, rather, ParACCEL has bought XPrime's intellectual property, as well as having taken on board some of its personnel. Thus, this is not a brand new company but one that is, effectively, a couple of years along.
ParACCEL is entering a space that is getting increasingly crowded (and there is more to come) but it has a marketing model that is about making noise before it formally comes to market (unlike relatively secretive companies such as Vertica and Dataupia) and it is targeting a market sector that nobody else is addressing—that's a good starting point—we'll have to wait and see what the company makes of it.
Copyright © 2006, IT-Analysis.com