The data warehousing space is getting renewed attention with the recent launch of Oracle's OLTP device, the Exadata Database Machine Version 2 - following the launch of the Exadata 1 data warehousing appliance last autumn. Enterprises using data warehousing infrastructures now show a growing desire to have the software and hardware closely integrated on the same device.
Database appliances combine server hardware with pre-configured software including a database and operating system. The ability to fine-tune all components for the sole task of data processing promises considerable improvements in transaction speeds.
According to Marcus Collins, senior analyst with the Burton Group, these appliances are attracting quite some interest now as companies are looking for ways to improve performance and cut running and deployment costs. He says though that it's important to differentiate between machines for OLTP online transaction processing (OLTP) and those for data warehouse processing.
Alan Cornwell, VP of EMEA operations with Netezza, agrees. "When Oracle launched Exadata 1, it said that it was a machine aimed at OLTP and at data warehouses. But these are very different things. In OLTP, a very small transaction is whizzed through very quickly. It's typically a small transaction, something like credit card approval. But with data warehouse analytics, you could be handling many terabytes of data."
The appliance market has grown steadily in recent years. Oracle is now trying to muscle into a market that already attracted plenty of interest. Cornwell puts the appeal of database appliances down to the sheer complexity of data warehouse transactions. "There are plenty of horror stories in data warehouses, about enterprises taking 10 to 12 hours to run processes. We speed that up considerably," he says. Collins adds that it's not just about throughput. Pre-configured appliances also make more sense financially, as they offer considerable cost savings. "Although the throughput is an added bonus," he says.
Cornwell cites mobile operator Orange as an example of a company where producing reports took several weeks. "Executives would get reports that were out of date by the time they got them. Now they can get those reports on a daily basis. We often a see a report that used to take eight or nine hours to be delivered, now taking 30 to 40 minutes. The company can now run follow-up queries to that data, something that they couldn't have dreamed of previously."
According to Burton's Collins, the success of companies like Netezza and Teradata has prompted a flurry of interest from the established hardware providers. "We're not just talking about Oracle here. Dell has done something similar with EMC. We see vendors like IBM entering this market. Cisco, which announced its blade servers at the beginning of the year, will be another competitor."
One of the factors that have driven the need to introduce appliances into the data warehousing space has traditionally been the lack of alterations and enhancements that could be used to improve performance. Data could be compressed to save on storage, but nothing was able to radically improve throughput times.
In the last couple of years however, this has changed. A number of new companies have emerged to address this market by improving the database lookup process. Companies such as Aster and Greenplum have come from nowhere to make their presence felt. These companies use a database improvement technology called MapReduce developed by Google engineers to cope with the vast volume of traffic that data warehouses have to handle.
One such software company is Kognitio, best known for its data warehouse as a service (DaaS) offering. Kognitio's vice president of marketing, Sean Jackson, says there is nothing inherently wrong with appliances, but that it is important to provide choice. "In some cases, an appliance is right, in some cases, data warehouse as a service is best. DaaS is ideal for companies that want to dip their toe into analytics to see if it can be useful to them, without spending a six-figure sum on an appliance."
But despite the advances, speed is still the name of the game. Oracle claims a ten-fold increase over a traditional data warehouse. Netezza's Cornwell advertises that in some cases he has seen speeds increase by a factor 100, and in some cases by a factor of up to 1,000.
So, what should a company looking to install such a device pay attention to? According to Burton's Collins, the time for such devices has come, no caveat necessary. "There's no real downside to them - there are concerns about vendor lock-in, but those disappear with the combination of different vendors supplying the technology."