Faster networks and inexpensive computing power provide the technological foundations of grid computing, but ease of use is the key to successful implementation.
Even at the National Research Council, where scientists are eager to master new skills to create supercomputing tools, a great deal of effort is spent to put grid computing just a few keystrokes away.
Even though security features are built in to the Globus Toolkit, the open source software resources that support many major grid computing projects, organizations must still build their own infrastructures to issue and maintain trustworthy certificates for their users.
“From the user point of view,” said the NRC’s Roger Impey, “there is a barrier. The end user, a physicist or a chemist, has to be somewhat aware of this technology so it is another burden for them. They have to be aware of their certificate, aware of what resources may or may be available, so for some people that overhead is too much. We are trying to hide as much of the complexity from the end user as possible.”
From the system administrator’s perspective, grid computing means another set of installation and management chores on their networks: At the very least, programs to deploy and monitor end users’ computers, as well as controller software on their own consoles.
At the senior management level within a department, the benefits of this extra layer of complexity must far outweigh the considerable costs of its implementation.
Given the discrete nature of government computing, just the negotiations involved in implementing grid computing across departments may make that prohibitive.
To date, grid computing has mostly been confined to the hobbyist, the skilled and the highly motivated, the individuals and organizations who can devote the time and tinkering to the creation of high performance, widely dispersed networks.
Until and unless it becomes more a product and less a process, most government-supported grid computing will likely be confined to research laboratories, wherever they may be.
