Crunching data on the cheap

Data-warehousing software systems are expensive, but many enterprises have nonetheless been willing to dig deep, betting that analytics will provide new insights into their business and a competitive advantage.

In a report released earlier this year, research firm IDC predicted the data-warehousing platform market will grow from roughly US$7.9 billion in 2009 to about $10.8 billion in 2013.

The good news for IT shops that want to get started in analytics, but don't have the budget right now, is the recent emergence of free software options that pack fairly serious data-crunching firepower.

In October, Greenplum announced a Single Node Edition of its MPP (massively parallel processing) database. MPP architectures split up data workloads into multiple pieces that are managed independently on an array of servers.

The Single Node version can be used in production mode on one x86 server with up to two CPU sockets and unlimited cores. It can also be deployed in a single virtual machine with up to eight virtual cores. There is no storage cap. Single Node Edition can also be tied back into a broader Greenplum implementation.

Also in October, Calpont released InfiniDB Community Edition, an open-source, column-oriented database. The columnar method can in many cases greatly reduce disk I/O demand compared to systems that store data in rows, and also achieve higher levels of compression, said analyst Curt Monash of Monash Research.

InfiniDB Community Edition is also limited to a single server, but has no cap on CPU count, the number of cores, memory, data volume or concurrent users.

Calpont also has a commercial edition of InfiniDB, now in early adopter stage, that allows users to scale out the system to multiple servers.

InfiniDB Community Edition follows the release last year of another open-source columnar data-warehousing platform from Infobright. The latter may have an edge for now over InfiniDB in terms of community support; Infobright recently said the software has been downloaded more than 15,000 times.

Ultimately, though, these free data-warehousing options have their limits and likely usage scenarios, according to Monash.

“If you have a single analyst or small team of analysts doing early exploratory querying against some terabytes of data or less, then these products are likely to do the job,” he said.

Companies may also not have the budget to procure anything else, or can't get funding without conducting an initial proof of concept, Monash added.

“There certainly are workloads for which they are insufficient, and you'll have to pay money for a product that will do the job for you,” he said. “But if you want to get more value out of your data, these free products could be a great place to start.”