Thanks to the promise of quick deployments with little to no customization, data warehouse appliances are making
their way into the mainstream data center, according to a recent Forrester Research report.
Companies are increasingly turning to preconfigured data warehouse appliances for specific data management and analytic functions, such as bulk data loading and multidimensional OLAP queries, the report said. Cheaper and less complex than traditional enterprise data warehouses, data warehouse appliances are being touted by vendors as no-fuss, no-muss, plug-and-play alternatives.
"What all these solutions do, at heart, is accelerate the deployment and processing of data warehousing applications, specifically complex queries against structured data sets maintained in star schemas within your analytic databases," said James Kobielus, the report's author and senior analyst with the Cambridge, Mass.-based research firm. "They're all basically OLAP [and data mart] accelerators."
This is good news for information and knowledge management workers, who are increasingly expected to quickly create and deploy data marts to address short-term business intelligence (BI) demands, Kobielus writes in the report. Data warehouse appliances can be deployed in a fraction of the time it takes to get a data mart or enterprise data warehouse off the ground, he said -- and often with less customization.
Essentially prepackaged hardware and software, appliances are nothing new to IT. Firewall appliances and WAN optimization appliances, for example, have been around for years. The idea is that one convenient package helps companies with limited technical expertise quickly solve a particular problem.
Data warehouse appliances first hit the scene in 2002 with the emergence of data warehouse appliance vendor Netezza Corp., Kobielus said. The Framingham, Mass.-based vendor, and those that followed -- like San Mateo, Calif.-based Greenplum Inc. -- "stressed the whole quick deployment, quick time-to-value, preconfigured hardware and software bundle to process data warehouse workloads," he said.
Since then, many of the so-called mega-vendors have gotten into the data warehouse appliance game, most notably IBM, which re-launched its enterprise data warehouse portfolio as data warehouse appliances in March 2007. Oracle and Microsoft followed soon after with data warehouse appliances of their own, developed through partnerships with storage and hardware partners. As a result, Kobielus said, data warehouse appliances have come a long way in just a few short years.
"The recent embrace of the appliance go-to-market approach by the three top enterprise DBMS vendors -- IBM, Microsoft and Oracle -- shows that appliances, however defined, are becoming the dominant approach for delivering [data warehouse] functionality to customers of all sizes," the report states.
Data warehouse appliances maturing; can't yet match enterprise data warehouse functionality
But most data warehouse appliances are not as robust as traditional enterprise data warehouses, which are still needed for complex BI and analytics functions. Enterprise data warehouses are scalable to manage increasing data volumes and can process multiple data domains simultaneously, Kobielus said. Data warehouse appliances, on the other hand, are much less complex, designed to work with relatively stable data volumes and usually just one type of data domain -- such as customer, human resources or product data -- to address tactical data mart requirements. Because they are less customizable, appliances are not flexible enough to support mixed-query workloads, he said.
Also, data warehouse appliances are not truly plug-and-play, Kobielus said. They are significantly faster to deploy than enterprise data warehouses, sometimes implemented in a week or two, but some customization is inevitably needed with appliances.
"When you plop them down into your data center and connect them to your existing ETL code and your existing BI application … more likely than not, you're going to need to make some modifications to those ETL scripts. You're going to have to rewrite your SQL queries inside your BI application to be able to work most efficiently," Kobielus said. "It's not a huge issue, but nonetheless, nothing's really plug-and-play."
Kobielus expects data warehouse appliances to continue maturing into enterprise-grade data warehouse platforms, but he doubts they will replace enterprise data warehouses altogether.
"Basically, one can easily come at this whole topic with the misconception that it's an either/or thing, that you've got an enterprise data warehouse or you've got a data warehouse appliance, and the two shall never meet," Kobielus said. "In fact, that's far from the truth."
Data warehouse appliance evaluation and buying advice
The report offered the following recommendations for companies considering data warehouse appliances:
- Consider BI and data warehouse applications' query workload needs and whether a data warehouse appliance can truly meet them.
- Use the same criteria to evaluate data warehouse appliances and traditional enterprise data warehouses.
- Deploy data warehouse appliances first in tactical roles, such as in function-specific data marts, then roll them out enterprise-wide.
Kobielus also warned companies considering data warehouse appliances not to be confused by the marketing terminology. Some vendors, like IBM and Teradata, have embraced the term "appliance" while others -- Kobielus declined to name them -- have resisted, fearing that the term makes their products sound more lightweight or less robust than they truly are.
"[Either way] you'll know when an appliance is right for you when the appliance-based solution addresses all of your data warehousing requirements, irrespective of what they happen to go to market under in terms of banner or buzzword," Kobielus said.
Has your organization evaluated or implemented a data warehouse appliance? Email SearchDataManagement.com editors with your story!