Abacus

Story Details

Colorado-based Abacus relies on a mixture of Sun server and storage technologies to maintain the nation’s largest cooperative data warehouse of purchase history. Sun technologies also drive hundreds of business-intelligence applications that customers use to rapidly extract relevant information from terabytes of data. In 2008, the systems that supported the data warehouse had reached the ripe “old” age of five years and were at the end of life. Although server performance was not an issue, Abacus wanted to take advantage of next-generation I/O speeds and green technologies to reduce server footprint and cut utility costs.

Rather than engaging numerous vendors for bids, Abacus discussed its challenges directly with its account manager and service engineer, as well as reseller Advanced Systems Group. “We decided early on that we were going to establish a partnership with Sun and as long as it maintains company viability, quality of service, and quality of manufacturing, there’s no reason to change vendors,” explains Doug Tschudy, vice president of information technology operations, Abacus. “Sun hardware is commercial grade. HP and IBM manufacture similar products, but moving to another vendor with marginal performance gain would require an 18-month engineering cycle to migrate and test on a new platform. We've been successful in leveraging our long-term partnership with Sun, and the new higher performance M-Series servers are a testament to Sun's quality of products. With the Solaris 10 Operating System and applications' backward capability, we’re able to leverage new Sun technology with a great deal of confidence that our legacy code will port to it with little or no engineering investment.”

Sun representatives suggested that Abacus evaluate the Sun SPARC Enterprise M4000 and M5000 servers to support its data warehouse. Over the next 45 days, engineers at Abacus tested the systems at its datacenter and found that the Sun SPARC Enterprise configuration significantly improved the performance of data queries. Commenting on the testing process, Tschudy notes, “The fact that I can just drop in a brand new box and benchmark it against a five-year-old system running the Solaris 10 Operating System without creating any application issues is a big deal.”

In August 2008, Abacus deployed its data warehouse on Sun SPARC Enterprise M4000 and M5000 servers. Rather than storing purchase history on disk as in the previous configuration, the new Sun servers store the entire data warehouse in memory. As a result, business-intelligence applications simultaneously execute data queries more than 50% faster on the new configuration and support more jobs. In addition, the Sun SPARC Enterprise systems provide for greater data capacity, occupy less space (6 or 10 rack units versus standalone cabinets), and consume less power than the old Sun Fire 6800 and E6900 servers that previously supported the database.

Very pleased with its new systems, Abacus is evaluating Sun SPARC Enterprise M3000 servers to support smaller tasks in the datacenter. Company engineers are also working with representatives from Sun Support Services to redesign the datacenter so that it occupies 20% less space each year over the next five years — even though computing power will double annually. To realize its goal, Abacus is using a combination of innovative layout and virtualization techniques such as Dynamic Domains on the SPARC Enterprise M4000 and M5000 servers to create hardware partitions, and Solaris Containers in the Solaris 10 OS to provide for operating system virtualization. Administrators will also use Solaris Dynamic Tracing (DTrace) in the Solaris 10 OS to accelerate system troubleshooting.

“We leverage Sun’s experience and new technologies to gain insight into how we manage our facilities and how we plan for growth,” concludes Tschudy. “And really, that’s the biggest benefit we’ve realized related to our IT environment: the relationship and the trust that we’ve developed with Sun over the last ten years. They understand what our requirements are — everything from our I/O demands to just general knowledge about how data is handled as it comes in the door.”