The virtue of virtualization

Virtualization is a well-known concept in networking, from virtual channels in ATM, to virtual private networks, virtual LANs and virtual IP addresses. However, an even more fundamental type of virtualization is achievable with today’s ubiquitous networks: machine cycle and storage virtualization through the auspices of grid computing and IP storage.

Also known as utility computing or on-demand computing, grid computing is a virtualization technology that was talked about in the 1980s and ’90s and entered scientific computing in the last 10 years. It has received a lot of press and market activity over the last couple of years, and a number of proponents see major penetration in the immediate future. Big players in this market include AT&T, IBM, Oracle and Sun.

Grid computing cannot exist without networks (the grid), because the user is requesting computing or storage resources that are located miles or continents away. A user need not be concerned about the specific technology used to deliver the computing or storage power; all a user wants and gets is the requisite service. One can think of grid computing as middleware that shields a user from the raw technology. The network delivers job requests anywhere in the world and returns the results, based on an established service-level agreement.

The advantages of grid computing are its ability to mix and match different hardware in the network; its lower cost, from better, statistically averaged utilization of underlying resources; and its greater availability: If a processor fails, another one is automatically switched into service.

Grid computing is intrinsically network-based: Resources are distributed on an intranet, extranet or the Internet. Users also can get locally based virtualization by using middleware such as VMWare that lets a multitude of servers in the corporate data center be utilized more efficiently. Typically corporate servers are utilized for less than 30% to 40% of their available computing power. With virtualization, a company can improve utilization, increase availability, reduce costs and make use of a plethora of mix-and-match processors.

Security is a key consideration in grid computing. The user wants to receive services in a trustworthy and confidential manner. There is the desire for guaranteed levels of service and predictable, reduced costs. There also is the need for standardization, so that a user with appropriate middleware client software can reach any registered resource in the network transparently. Grid computing supports the concept of the service-oriented architecture, in which clients obtain services from loosely coupled, service-provider resources in the network. Web services based on the Simple Object Access Protocol and Universal Description, Discovery and Integration protocol are now key building blocks of a grid environment.

To grid or not to grid? Try it; you’ll like it. But first make sure you have an effective corporate network infrastructure in place.

Minoli is an adjunct professor at the Stevens Institute of Technology’s graduate school and author of A Networking Approach to Grid Computing. He can be reached at minoli@att.net.