While others talk about how utility computing and a services orientation can affect IT delivery, Wachovia Bank is a living demonstration.
For the financial services giant, which controls assets of about US$541 billion (NZ$822 billion), the utility project is the latest effort in an on-going virtualisation strategy. The project relies on DataSynapse’s FabricServer, which distances Java applications from application servers so they can be parcelled out to any available application server at run-time, says Tony Bishop, senior vice president and director of product management for North Carolina-based Wachovia.
The project flowed naturally from five years of advanced network work. Wachovia became one of the first commercial users to deploy business applications on a grid computing architecture with the implementation of DataSynapse GridServer (and, consequently, became one of the vendor’s institutional investors). By 2005, the grid had become Wachovia’s standard application server. Today, it operates on 3,000 dedicated engines, with an additional 5,000 CPUs that can be tapped as needed, Bishop says.
All this is the basis for Wachovia’s service-oriented infrastructure (SOI), which Bishop likens to an IT utility. For example, when a trader accesses an application, the SOI distributes, brokers and manages the various services involved in the application, ensuring that each service meets performance and business objectives. When traders are asking to price deals, “you want to give a higher-margin deal better service”, Bishop says. “GridServer and FabricServer become the mechanisms that let us do that across distributed and transactional applications.”
In all, for every dollar invested in the SOI, Wachovia reports a $3 — $4 return.
All virtual layers
The FabricServer deployment was a significant step in the evolution of Wachovia’s SOI because it virtualises Wachovia’s mission-critical Java 2 Platform Enterprise Edition (J2EE) applications. At run-time, FabricServer distributes an application to an available server. FabricServer handles the application’s configuration needs and eradicates conflicts, even if two or more applications are sharing the same server.
Because application servers can be better utilised, fewer are needed. This adds up to significant savings. While the bank didn’t perform a detailed cost-savings analysis (because it already has years of experience with savings achieved from virtualisation), Bishop estimates annual savings from this project will tally near the seven-digit mark — from both reduced hardware and maintenance expenses. Plus, he says application programming efficiency is up by 30%; application performance is up fivefold and throughput is 50% faster since FabricServer was implemented.
Grid computing and a virtualised Java application environment represent only a portion of Wachovia’s SOI, however. The bank also uses VMware for servers not part of the FabricServer project and has virtualised its data layer, relying on Tangosol’s Coherence for data access and Composite Software’s Composite Information Server (CIS) to virtualise the data-query process.
Coherence gives Wachovia’s SOI a mechanism to track where data is stored, so J2EE application developers do not have to worry about specific file locations when writing new services. CIS virtualises the query itself, draping a service in whatever custom application-interface formatting it needs to interact with other applications.
“So you have a data query that’s virtual; you have a data-query environment that’s virtual, and that ties right in with our processing virtualisation with GridServer and FabricServer. And VMware partitions and makes virtual different images of the operating environment,” Bishop says.
Next, the SOI will tie into a virtualised storage infrastructure that Wachovia engineers are working on now, he says. He sums up the goal: “Everything on top of the network — even the network — should be virtualised.”