04:00 PM
Wall Street Firms Form Enterprise Clouds
Like the image of clouds themselves, the concept of cloud computing is soft and fluffy. The term "cloud" tends to be applied to most software-as-a-service (SaaS) offerings, although Amazon's broad service consisting of server and storage access over the Internet is the classic example of a computing cloud. The McKinsey Group recently defined clouds as hardware-based services offering compute, network and storage capacity where hardware management is highly abstracted from the buyers, buyers incur infrastructure costs as variable operating expense and infrastructure cost is highly elastic (up and down).
Within a firm the use of the cloud model as a way to deliver internal IT resources sometimes is called an "internal cloud," an "enterprise cloud" or a "private cloud." The definition here is even looser; as far as WS&T knows, no formal definition has yet been adopted in the IT community. Seemingly, however, internal clouds are characterized by widespread server, desktop, storage, input/output (I/O) and memory virtualization; easy and quick provisioning on demand; scalability; and a sophisticated management layer that makes these features possible.
These elements all are appealing to Wall Street firms. "In the U.S. we've been a serious virtualization optimization organization for some time," notes Jim Carney, senior executive in Citi's technology infrastructure group. "We deploy only virtual servers under a strict process that we see as an internal cloud."
Merrill Lynch is building an enterprise cloud based on a stateless computing initiative it started a few years ago (the firm declined requests for an interview but has commented publicly on its project at several trade shows). And hedge fund AQR Capital Management is leveraging an enterprise cloud to improve system availability and disaster recovery.
Cloud Building Blocks
Any internal cloud starts with virtualization, according to Bernard Golden, CEO of HyperStratus, a virtualization and cloud computing vendor. "There's no doubt that the first step is to be fully virtualized," he says. "You've got to have virtualization spread across whatever portion of your internal IT infrastructure is going to become cloud."
Golden believes most organizations will start small with cloud initiatives and build them gradually over time. "We'll see more organizations say, 'I'm not going to cloudify everything; I'm just going to put 10 percent or 20 percent of my resources on a cloud,' " he says. "The vision is of a more agile IT organization, one that's able to respond more quickly and scale applications, so when somebody wants to get a server up and running they don't have to go to a committee meeting, submit a request in triplicate, and wait a month and a half to get it."
Another necessary step toward cloud computing requires creating or buying an automated infrastructure or orchestration layer, such as HP's Adaptive Infrastructure or VMware's vSphere, Golden says. This technology enables a cloud to automatically assign processing, network and storage resources so that a user could, for example, go to an internal Web site, request a server with four processors, 20 gigabytes of storage and 54 gigabytes of RAM, hit a button, and have access to that virtual machine flow through his desktop, Golden explains.