More time for analytics

With SAS^® Grid Computing, SunTrust analysts spend less time moving data

"Analysts should spend their time doing analysis," says Dudley Gwaltney, Manager of Analytical Modeling at SunTrust. "They shouldn't spend time copying files from one place to another because of capacity constraints or be forced to wait for a batch job to be completed or a server to come back online from maintenance. They should be working in a centrally managed computing environment that most effectively addresses workload management, high availability and performance."

SunTrust uses SAS Grid Computing for precisely those reasons. The grid-enabled computing architecture enables Gwaltney and his colleagues to work in an environment where SAS applications and services are guaranteed to be available, resources are prioritized according to demand and jobs are accelerated through parallel processing.

Bottom line: Gwaltney can more effectively do what he's paid to do.

All data in one place

Gwaltney leads the analytical modeling team within SunTrust's Client Information Group – a team that supports all lines of business. Its predictive models for upsell, cross-sell, attrition and other marketing activities are data and process intensive. Although Gwaltney's staff is first rate, the computing environment in which they've been working has been lacking.

Under SunTrust's old configuration, servers were allocated based on business function, resulting in sufficient computing resources for some and not others, and an overall lack of flexibility in reassigning servers when more space was needed. Depending on which initiative they’re supporting, Gwaltney's analysts might have to retrieve data from multiple locations and transfer it to other servers for processing.

Gwaltney, a longtime SAS Enterprise Guide and SAS Enterprise Miner user, says he regularly deals with files that exceed 10 gigabytes. "It takes quite a while to move that much data between servers, and we're constantly having to move data around." In SunTrust's new grid-enabled computing architecture and cluster file sharing, all the data is in one place. "We also won't be wasting space with duplicate files," Gwaltney says. "That will be a real advantage to both IT and my analytics team."

Answers today, not tomorrow

The previous server configuration also meant time lost waiting for a processing resource. Gwaltney and his colleagues regularly have jobs that take seven or eight hours to run. Narrowing that to three or four hours, Gwaltney says, "is the difference between providing an answer today instead of tomorrow. And in our business, that makes quite a difference."

SAS Grid Computing addresses the issue of slow response times by providing a tool for standardized workload management and centralized administration of servers. The workload for SAS programs can be distributed across multiple CPUs automatically. And the ability to prioritize jobs based on business rules ensures that necessary resources are freed up when they are needed.

"Jobs in the grid environment will now run more efficiently," Gwaltney says. "Processors will be better managed and distribute the work based on available capacity and priority."

Gwaltney believes that grid computing will allow more time for in-depth analytics. "The things that used to take me several hours now take an hour," he says. "They happen so much faster. The data is all in one place, and it's on a system that processes things faster. It gives analysts more time to ask questions and do analytics, which is what we’re paid to do."

Doing work differently, more efficiently

Gwaltney says it will be interesting to see how the end-user perspective will evolve as the SAS Grid Computing environment is rolled out. He believes SAS users at SunTrust will follow a natural learning curve from "How can I do what I've always done?" to "How can I do it more efficiently?" to "How can I take advantage of this new system to find new opportunities?"

"Once users get past that point," Gwaltney predicts, "they'll start thinking about the analysis they could never do before and start taking advantage of the new capabilities." All because there's additional time for analysts to conduct analytics.