Appliance vendor Datallegro Inc might not yet be a household name outside of mainstream data warehousing circles. But the Californian start-up says its ready to make some big waves in the industry, and perhaps ruffle a few feathers of the incumbent data warehousing vendors in the process.
Often thought of as a direct competitor to large data warehousing vendors like IBM Corp, Oracle Corp, IBM Corp and NCR Corp’s Teradata division, Datallegro’s CEO Stuart Frost recently spoke with Computer Business Review to clarify the company’s positioning in the market.
He told Computer Business Review that while the company isn’t targeting the extreme high-end of the data warehousing market per se – i.e. the 100 terabyte-plus range – it is starting to penetrate some key marquee accounts of much larger data warehousing plays like IBM and Teradata.
However he maintains the company isn’t necessarily looking to uproot these expensive implementations.
Large telcos and financial services companies are generally satisfied with their enterprise data warehouses. We’re not playing in that space, he said.
Rather he says the company is targeting a sub-segment of that market, which he calls operational data warehousing.
By this he means a specialized (usually complex and data intensive) set of ad hoc queries which most enterprise data warehouses typically struggle to handle in a cost-effective way due to the sheer weight of summary transactional data tables (billions of rows) and the large-scale aggregations that need to be performed.
Frost believes that companies are better handing off this expensive heavy-duty processing to Datallegro’s system.
We’ll handle the processing grunt for these types of ad hoc analyses with an attractive price-performance point to boot.
Frost says this represents a different category of analytic application than conventional data warehousing.
We’re going after a specific set of complex ad hoc queries against a subset of the data warehouse. So in many ways we’re complementary to enterprise data warehousing infrastructures already in place.
For this reason Datallegro isn’t staking any claims to be a fully-blown master enterprise data warehouse. Architecturally in fact its closer to being a specialized data mart slave.
This positioning seems to suit Datallegro down to a tee. Indeed its appliances seem to be carving out a comfortable niche in the market by simply sitting in front of large enterprise data warehouses, pulling fact [table] data out of them, aggregating it up, and then feeding it back into the central warehouse.
We’re finding that many companies are using [Datallegro] to front-end their enterprise warehouses. Its almost as if they’re turning Teradata into the data mart, which is surreal to us, Frost says.
The price-performance advantages that Datallegro’s system boasts – between 10 to 100 times performance gains at a tenth of the price of less – are certainly impressive. Enough so to raise a few skeptical eyebrows.
Clearly Datallegro has a lot to live up to. So how does it do it? The secret sauce is the number of disks bundled into each of its appliance units. Whereas other appliances typically have one or two disks per unit, a peek under Datallegro’s bonnet reveals six RAID disk arrays per unit comprising of 12 disks which are complemented by Intel processors, an Ingres open source database and some memory thrown in for good measure.
In other words companies can get by with fewer appliance units using Datallegro, which in turn translates to lower hardware costs. Datallegro’s system also implements a rather clever partitioning scheme in the Ingres database (using sequential, rather than random, I/O without tuning or indexes) to manage the various disk arrays and maintain performance.
Of course this means that Datallegro can scale upwards quite easily (and cheaply).
Our sweet spot is still in the five to one hundred terabyte range, Frost admits, but points to larger deployments in its pipeline.
He claims the company is starting to notch up more million-dollar deals based wholly on the price-performance aspects of its appliance solution. We already handle an 86-terabyte system for one of the world’s major retailers.
Its this attractive price-point that’s also starting to prize open new opportunities for companies that were previously priced-out of complex, large-scale data analysis.
We’re pitching that functionality at a price-point that others simply can’t match.
For the functionality we target and the price-point we’re pitching, nothing out there can really touch our C-range appliances. Many of our prospects have told us that IBM, Teradata and even Netezza [which is a rival data warehousing appliance vendor] are too expensive.
We can handle large scale ad hoc analysis and do necessary aggregate grunt-work that’s needed up-front at a fraction of the cost that companies like Teradata quotes up to $15m for.
While Frost acknowledges that rival Netezza Corp has been the most visible data appliance vendor in the market up to now, he notes that Datallegro is starting to steal some of the limelight as well.
We’ve got proof of concepts running and are catching up with Netezza fast. We’ve built a fantastic pipeline and expect to be profitable next quarter.
Datallegro keeps the number of its customers close to its chest. But Frost says to expect some major new customer wins in the next month or two.
However he’s vary of exaggerating expectations given the relative small size of the company. Rival Netezza somewhat shot itself in the foot at the beginning of 2005 after bragging publicly that it would break the $100m barrier by the end of that year. But indications are that it barely surpassed half that figure.
Having a smaller number of larger deals, of up to three million dollars, is quite adequate for a company of our size at the moment. We won’t make the same mistake as Netezza by overstating our revenue goals, says Frost.