SNS collects copious statistics on all its applications using Mercury Interactive Corp.s Mercury Performance Center, watching for poor performance or troubling trends. After analyzing the data, SNS noticed that transaction growth and rising data volumes were straining its PeopleSoft ERP (enterprise resource planning) application. "We could see that we would have a bottleneck if we grew as we anticipated," Bingham said."The technology allows you to emulate production usage in a lab environment," said Rajesh Radhakrishnan, senior director of product marketing at Mercury, in Mountain View, Calif. "You can simulate thousands of users hitting your systems before you go live. The key benefit of that is risk mitigation. When you go live with [a solution], is it going to stay up and running, or is it going to crash?" Meantime, SNS tracked its changes with Mercurys TestDirector, an application for managing testing by recording what tests were done and by whom, and what the results were. The idea, said Radhakrishnan, is to ensure "quality control" during an organizations tests and changes. The tests verified SNS planned fixes. Once implemented, the company saw measurable results almost immediately. (The project ran for about six months, ending in May 2004.) SNS was able to get enough additional "juice" to handle another 20 percent increase in sales orders, said Bingham, without overstressing its systems. This bought SNS approximately a year because the company was able to push back the need for $750,000 in upgraded hardware. Among other benefits SNS has seen since putting its software through the paces are the following: The capacity to run more fulfillment cycles. Before fine-tuning its applications, SNS was running fulfillment cycles once per hour between 5 a.m. and 7 p.m. Now, Bingham said, it can handle twice as many. This gets more information to the distribution centers more quickly, improving workload management and customer service. Shorter batch processing times. SNS runs its billing cycles overnight between 7 p.m. and 5 a.m., when new order volume starts to pick up again. SNS has been able to shorten the time required to run its billing cycle by 2 hours, meaning theres less chance that the associated demands on the system will bump up against the demands of the order-taking system when customers start to log on around 5 a.m. Improved customer call center interactions. SNS tries to keep the customer experience smooth, which requires service representatives to have a responsive system at their disposal. With that in mind, SNS targets a maximum system transaction time of 3 seconds. Application tuning helped here: SNS now has 94 percent of its system transactions fall within that window, and an average of 1.88 secondsdown from 88 percent and 2.36 seconds beforehand. Having identified and fixed issues dealing with its applications, SNS could turn its attention to networking and other issues, finding even more opportunities to improve performance. All told, SNS was able to reduce downtime substantially"impacted user hours" fell from 15,000 in 2003 to 5,800 last yearand meet its goal of 99.9 percent system availability. In the end, SNS was able to generate meaningful results from performance testing while controlling labor costs, leading to better service for both customers and internal users. David P. Marino-Nachison is a free-lance writer in Washington. He can be reached at firstname.lastname@example.org. Check out eWEEK.coms for the latest news, reviews and analysis in programming environments and developer tools.
Using Mercurys LoadRunner performance-testing product, SNS was able to get a picture of how its software was underperforming. After identifying fixes, it used LoadRunner again to simulate the effects those fixes would have in real-life conditionsin short, whether its planned solutions would float or flounder once they reached users.