The most significant difference is that Calpont’s system does not have a database, operating system or disks. Instead Calpont holds all the data in memory (SDRAM). In other words the database is implemented as an SQL on a chip.
Architecturally we’re very different, said Patti Shaffner, vice president of marketing at Rockwell, Texas-based Calpont, in a recent interview with Computer Business Review. We’re actually closest to a true plug-and-play appliance given our small form factor, modular and scalable architecture, and our low administration.
A quick glance at the background of Calpont’s executives explains how the company is attacking the database problem from another angle. They are applying their networking backgrounds to data warehousing as an efficient way to route and process data.
Ms Shaffner points out that While [other appliance vendors] have attacked the database taken from macro-level that rely heavily on processors and disks, we’ve taken it at the micro-level, and do a lot of processing across the chip itself.
Because each piece of data has in effect its own IP address our system is good at efficient storage, routing and access.
Calpont admits that the small form factor of its system limits it to sub-terabyte and departmental applications – the capacity ranges from 20 gigabytes up to 128 gigabytes. We have no intention of doing the tens of terabytes that both Netezza and Datallegro target, Ms Shaffner said.
But Calpont’s system can also be implemented as an extensible network meaning there is no serious technical barrier to scaling-up to larger terabyte-class data warehousing as well. Right now we’re focusing on data mart configurations of up to one terabyte, said Ms Shaffner.
Calpont says its system is much cheaper that other appliances since there are no database license fees. It also has a low administration overhead since data is not logically mapped to a physical store in the relational sense – rather Calpont uses trees and multidimensional directed graphs.
This means the system is self-indexing and tuning since the need to devise and maintain disk partitions, striping and other typical database performance enhancing features are effectively removed.
There’s perhaps a bigger market for smaller implementations of data appliances and for these applications maintenance is an issue, notes Ms Shaffner.
Having recently bagged its first venture capital funding, Calpont is still very much in stealth mode at the moment. Ms Shaffner says the company is currently heads-down in product development – specifically optimizing its SQL processor – and expects to detail its go to market plans later this summer.
The data appliance trail is now well trodden path, blazed by vendors like NCR Teradata and WhiteCross Systems back in the 1980s. More recently companies appliance vendors like Netezza and Datallegro have jumped in with nimbler and cheaper appliances that promise to break traditional price-performance barriers associated with high-end data warehousing.
It is too early to gauge Calpont’s chances of success or say if it is going to be a disruptive technology. But the prospect of a data warehousing system that strips away up-front costs and maintenance overheads while improving performance is likely to pique interest among mid-market companies or departments that are willing to try something new. If beta customers like what they see, and Calpont can prove its technology in a few referenceable sites, it could well shake up the data warehousing world.
