Data warehousing update 3: HP NeoView

HP now has eight customers (excluding itself) for its NeoView data warehousing/appliance product. Two of these are in Europe (including the recently announced Rabobank), one in India, two in the Far East and the remainder in North America (including, most notably, Wal-Mart). Most of its sales have been to Teradata sites although at least one was previously using Oracle as its data warehouse and, again, most of these will be running alongside the existing installations though HP is claiming that it will be replacing Teradata completely at one of its customers. At present the company is targeting heavily committed existing HP users and/or sites where the company can leverage its new recruit’s ex-Teradata contacts.

Good, so the product is progressing nicely even though it is still early days. Now let me tell you where I don’t think the company has got things quite right.

It starts with the market positioning. HP is primarily targeting the traditional enterprise data warehouse (EDW) market as epitomised by Teradata. Fine, it seems to be enjoying some success in that market. However, it also describes the product as being competitive with the likes of Netezza, DATAllegro and Greenplum. What it actually says is that NeoView is delivered as an appliance, and it can be used in that way, but you can also open up the hood and twiddle the dials for indexing and materialised views and such stuff if you want to (which you would for an EDW). In other words, the company is trying to have it both ways and I don’t think they can.

The problem is that I don’t believe that NeoView is competitive with Netezza and its ilk in either performance or price terms. The latter is easier to deal with: NeoView is priced to be competitive with Teradata so of course it won’t compete with a pure play appliance vendor (albeit that prices are elastic).

In so far as performance is concerned, I have reservations about the NeoView architecture: it uses a switched fabric to connect its disks to its processing nodes. Now, this is fabulous for resilience but I don’t see how it can compete with direct attached disk for pure performance. Experience appears to bear this out: in the only customer proof of concept that I know about in which both HP and Netezza were involved, it was Netezza that came out on top (and for interest, Teradata came third: at least it was ahead of the two merchant databases included in the bake-off).

So, my view is that HP should forget about competing with Greenplum et al (at least while NeoView retains the current architecture) and concentrate on Teradata. Yes, it might win a few blow-in appliance deals but talking about that market as if it can compete there is simply muddying the waters: it should stick to EDWs and taking on Teradata (and Oracle and IBM).

To conclude, my report card would read “has made some good initial progress and poses a significant threat to the incumbents in the EDW market; but should pay more attention to details: could do better”.