Netezza goes spatial

Written By:
Content Copyright © 2008 Bloor. All Rights Reserved.

At last year’s Netezza annual user conference the company announced the Netezza Developer Network, designed to encourage partners to develop and implement functionality within the Netezza appliance at the node level. Or, in Netezza parlance, at the SPU (snippet processing unit) level. The advantage being that by implementing functions at this level you gain all the benefits of Netezza’s massively parallel architecture and therefore get much better performance.

There were a number of these partners at last year’s event and there are on-going developments in cryptography, data mining, image recognition and other areas but one such that is being launched at this year’s conference is an embedded spatial processing capability that has been developed in conjunction with IISI.

At present, the market for data warehouses that have spatial capabilities is dominated by Oracle, which has something approaching 90% of the market, with IBM being the only other serious player. However, by implementing spatial capabilities at the SPU level, Netezza not only becomes a contender in this market but it can offer significant advantages, notably in performance and administrative terms. If we take the latter first there is no need, with Netezza, to define R-trees or to worry about partitioning: true to its appliance roots the whole process is made as simple as possible.

In terms of performance, Netezza is claiming 30 to 40 times improvements compared to Oracle and while I have no way to verify this it seems in line with other performance benefits derived from using on-stream analytics. This potentially has a major impact, not just in its own terms but also more generally. This is because spatial analytics tend to be compute-intensive and the resulting performance overheads often means that you need to have a dedicated system for spatial analytics, or a much bigger main processor, so that the spatial computations don’t impair the performance of other applications. This means that if you want to combine spatial and business intelligence then you have to run the relevant queries on separate systems and then combine the results rather than being able to run this sort of analysis as a single query on a single system, which is what the Netezza/IISI partnership enables.

Given the increasing demand for location intelligence (sample use cases include retail store locations, telecommunications dropped call analysis and insurance distances from hazardous areas) this is potentially a very important development as it could bring spatial capability to a much wider audience at a significantly lower cost.

Note that IISI, in conjunction with Netezza, simply provides the spatial engine; you can then use an appropriate application vendor for the actual analytics. For example, MapInfo (part of Pitney Bowes) is a Netezza partner.

Finally, the spatial package gives Netezza a significant advantage when compared to other appliance or virtual appliance vendors, since none of these has any comparable facility, at least that I know of. So, if location intelligence is a requirement, Netezza is an obvious candidate.