When Green Mountain Energy Co. purchased U.S. PowerSolutions Corp. in July 2002, it needed to build a Microsoft Corp. .Net Web services bridge between Green Mountains CRM software and the U.S. PowerSolutions back-end billing database.
All was going well until Green Mountain officials encountered latency problems and found their monitoring tools couldnt get at the root cause. Thats when they started looking for a tool to examine the code inside Web services transactions.
Green Mountain, of Austin, Texas, is the largest retail provider of “green” electricity in the United States. It sells renewable energy from wind, water and solar sources in eight states to more than 600,000 customers. Its customer service operations are spread throughout the country; the customer relationship management application resides in its Austin headquarters; the billing database is housed in Waltham, Mass.
Chris Garrick, vice president of IT at Green Mountain, said the company took the Web services route because it wanted to provide a way to provide its customer service representatives with customer and billing information in real time. “Everything started great, then we started to get latency and timeout issues when people requested information from the billing system,” Garrick said.
Garrick had an arsenal of tools to monitor what was happening across Green Mountains network, but while the company could isolate the problem to the Web services application, it couldnt pinpoint the problem any closer because its monitoring tools wouldnt allow it to view the code behind the Web services transactions. Green Mountain wanted to avoid replicating the problem using dummy data because that would have added to network traffic and wouldnt have helped isolate the problem.
“We were looking to find a way to monitor the Web services application without creating synthetic transactions. We had a way to have people call in, and we had monitoring tools to show network bandwidth, but we wanted to look inside transactions, look at a Web services call and find out what was causing the slowdown,” Garrick said.
Green Mountain also needed to get the problem fixed quickly because the CRM system was high-profile. “We were time-critical on this. This is the main customer information system. All of [the] customer care [group] uses it. All of our sales reps use it. We had to find a solution as quickly as possible,” Garrick said.
Garrick also found that there were few vendors to help with the problem, and he lacked the in-house resources to build a homegrown solution. One potential source of help was Service Integrity Inc., of Newton, Mass. Service Integritys SIFT product is an XML-based solution that enables monitoring and analysis inside of Web services. As it turns out, Garrick had known Service Integritys founders when they worked at other companies; that, and Service Integritys SIFT product, made the company a good fit for Green Mountain.
In January of last year, Green Mountain began testing SIFT. After a short period, it decided to make the purchase. Garrick said it took only 10 to 15 minutes per server to install SIFT on each of its Web servers. The software automatically detected the Web services running on each node and presented information about what was happening. Garrick and his staff could choose which services they wanted to monitor.
The software required little training, and users were able to track down the source of their original latency problem very quickly. “Thanks to Service Integrity, it didnt take long at all to find the latency issue. We had one of the founders set it up and show us how to run reports. We got proficient within a day or two. If we ran into problems, we just called them, and they were right there,” Garrick said.
Later, Green Mountain had another problem: The customer care system was timing out every day at 1:30 a.m. The company used SIFT to track down the source of the problem. “We ended up using the Service Integrity tools to look at what happened during that time in the Web services products and were able to isolate a specific transaction that was coming from one of our batch applications that was calling a Web service, and the Web service was having an issue with a specific type of account. We were able to go in and make the code change, repair it, and we no longer have that issue,” Garrick said.
In May, Service Integrity released Version 2.0 of SIFT, and Green Mountain plans to upgrade to that version as soon as it upgrades its .Net installation across the enterprise. Garrick said he plans to use the new features in Version 2.0 to monitor Green Mountains IVR (interactive voice response) system and find ways to improve it.
Free-lance writer Ron Miller can be reached at [email protected]