Credit to the New Data Center

I t’s no surprise that financial-services giant JPMorgan Chase is a pioneer in the quest for automated, virtualized New Data Center technologies. With assets of $1.2 trillion and operations in 50 countries, the company is engaged in multiple initiatives all working toward the common goals of maximizing IT resources, reducing costs and speeding performance. Examples include grid computing, policy-based management of virtualized resources and automated application mapping and change control. As these projects evolve, they will ideally converge and enable the New York-based financial-services giant to conquer tomorrow’s IT challenges.

Shawn Findlan, a vice president responsible for the global credit-trading infrastructure, in early 2004 started to explore how to boost his department’s infrastructure. Pressed with performance demands and constricted by costs, Findlan realized he needed to revise his infrastructure rather than build new additions. At that time, the IT department had just completed a consolidation project that reduced costs by 25% to 30%. That was good, but not good enough. JPMorgan Chase predicted the business Findlan supported to grow by 300% in 2005 and needed its recently consolidated infrastructure to support that growth.

“Our challenge was to add 300% capacity and optimize the environment to enable less downtime and to failover more quickly for less money,” Findlan says.

While JPMorgan Chase had virtualized its compute resources via a grid, dubbed the Compute BackBone (CBB), Findlan was now looking to virtualize the application and database layers. Thus Findlan’s project, the Credit Derivatives Infrastructure Refresh (CDIR), was born. Instead of running scheduled jobs across resources using virtualization tools from VMware or Sun, Findlan wanted an application to be able to tap server, database and other components in an entirely virtualized environment that could be created on demand. When a trading application needed more server or database resources, this flexible infrastructure would create an end-to-end application environment on the fly to support the application’s latest needs.

“If the application can utilize the CBB for compute services, it will be sent to the CBB,” Findlan says. “If there was a failure in the primary data center, we wanted to be able to migrate an application to an [on-demand] infrastructure that doesn’t exist on a day-to-day basis.”

Virtualizing application resource pools

Findlan realized he’d need a tool that could create virtual pools of resources and then automatically distribute those resources when business needs fluctuated. His search led him to relative newcomer Enigmatec, which provides management software that can automatically distribute resources based on preset policies.

Dubbed Execution Management System (EMS), the software detects system failures and load changes on servers, and can fix problems using preset policies, Enigmatec says. The software also can separate an application from dedicated server resources and apply other available resources to the application.

EMS uses distributed agents to monitor system performance, measure actual performance against preset thresholds and take action when performance degrades. When action is required, Enigmatec will automatically, say, move CPU resources into an application environment to meet the demand for more capacity.

Enigmatec software doesn’t rely on a centralized management console to configure agents, take corrective action or store data. IT installs the agents on managed systems and use a Web interface to create policies, configure agents and monitor performance. The agents can interact via peer-to-peer networking. For example, a new agent installed on a server would instantly register itself with the closest neighbor agent and get updated with the policies configured in the neighboring agent.

When using Enigmatec, Findlan says he can disassociate the IT service, or the multiple services that make up an application, from dedicated hardware. This virtualization allows an application component to tap resources from various components in the infrastructure, he explains. Enigmatec allows the application to request services from any available hardware or software resources, the vendor says.

“Enigmatec brings a bit more intelligence to its automation and virtualization than some competitors,” says George Hamilton, director of enterprise computing and networking at The Yankee Group. “Policy-based management and virtualization are part of the bigger goal for overall data center automation, and Enigmatec addresses the problem with a disaster recovery/failure approach.”

Provisioning infrastructure components

In Findlan’s case, after working with application development and operations teams, he input multiple “what-if” scenarios into the Enigmatec software that will prompt it to take action when alerted by a monitoring system. Enigmatec, as well as a provisioning tool Findlan declined to name, is integrated with this monitoring system. The data collected from multiple third-party monitoring tools is aggregated into the centralized system.

Findlan clarifies that this effort was a collaboration of various tools. He says he needed to first break down applications to understand how they used the infrastructure. Then he needed to examine the processes involved to enable software to manage application performance automatically. For example, if the trading application required more server capacity, Findlan wanted to create a virtualized infrastructure that would provision a server to support the application load immediately.

After breaking down and understanding applications, he implemented a provisioning tool that would automatically build infrastructure components when prompted to by Enigmatec, which responds to actions kicked off by preset “if, then” scenarios Findlan defined in the software. Using a central management console, he can watch the process as it runs across application and infrastructure components.

“The monitoring system allows data to be collected and visualized in one place, and it allows us to act upon events happening across the environment from one area,” he says.

Findlan worked on this project throughout 2004 and it went live for the credit derivative application in 2005. He continues to work to expand the system to include more business units and their applications. The biggest challenge Findlan encountered was also the most critical, integration of the virtualized components.

“The biggest piece of this is getting all the parts working together in one deployment,” he says. “We needed to tie together the underlying interfaces so the Enigmatec software could, for example, tell the provisioning tool to build something out without manual intervention.”

Findlan would not share specifics about performance improvements, but says the company “experienced significant improvements in uptime” across several hundred nodes being managed through the CDIR initiative. While Findlan says the project is operational for the credit derivative business unit, and that he is implementing the CDIR solution at various stages in other lines of business, he also admits he has only partially achieved his goals with this project. The system using Enigmatec can automatically allocate resources, but he has yet to enable it to automatically reclaim resources, which would ease follow-up manual processes. This is part of the next phase, he says.

“We have achieved the capacity on demand,” Findlan says, “and now we are working to get it automated in real time in both directions.”

Previous: Security expert offers best practices | Next: New IT products