The new online program is already generating some 4,000 retirement plans per week, compared with the 1,500 to 2,500 total projected by the firm, said John Dalton, a principal at the IT division of Valley Forge, Pa.-based Vanguard Group.
The mutual fund and retirement planning company completed a pilot program to test the new SOA-based Web interface to its Vanguard Financial Plan (VFP) service in March. The program automatically generates customized 10-to-20-page investment plans based on data entered by a consumer. Capacity Challenges
"That volume creates lots of capacity challenges," Dalton said. The company's original plan called for the service to use up to 2% of the capacity in Vanguard's Unix servers, but the unexpectedly hefty use of the system requires 10% to 15% of system capacity, he noted.
"We are in the process of increasing the number of and speed of CPUs and will vertically scale out and run services on different boxes over time," Dalton said.
The service previously required that consumers provide data to financial planners over the telephone. The planners then entered the data into the VFP application and sent the resulting plan back to the client. That service had generated about 15,000 reports a year, Dalton said.
The new online interface is built on top of the J2EE-based VFP application that runs on IBM WebSphere application servers. The VFP application uses an underlying advice engine that was built as a service and uses asynchronous messaging to generate a retirement plan based on input from users, said Jeff Dowds, a principal at Vanguard's retail systems division.
Because the online service reduces the number of phone calls between clients and planners, the company now can provide plans to five times the number of clients at a tenth of the cost, Dowds added.
"We've been able to take our advice platform and truly scale it," he said. "We've been able to expose it across multiple channels without having to rewrite the platform."
The program is available without charge to those with more than US$250,000 in assets and for $1,000 to those with $100,000 to $250,000 in assets, Dowds said. The service isn't available to consumers with less than $100,000 in assets.
Jason Bloomberg, an analyst at ZapThink LLC in Baltimore, said capacity is becoming an issue for many companies using customer-facing, SOA-based applications.
The need for policies that outline for the architecture team how many users may be accessing a service and how many may access it at one time becomes even more important with these external applications, he added.
"You don't want to hard-code your capacity planning," Bloomberg said, "because if you do have to put in new servers, you have to rewrite your service."