UNDERUTILIZED RESOURCES CAN BE EXPLOITED

UNDERUTILIZED RESOURCES CAN BE EXPLOITED

Grid computing can take advantage of underutilized computer resources, such as spare processing cycles, free or available disk storage, and even servers and many other types of resources. In most corporations, there are, at times, vast amounts of underutilized computer resources, such as the resources we mentioned above, particularly servers. A desktop PC is a good example of an underutilized resource; the average PC is used for 10 to 20 percent of an eight- hour day. When the workday is over, it stays idle for another 16 hours. This is resource that can be used in a corporate grid. Hundreds of PCs can be connected to the grid, forming a large virtual computer with substantial power and capacity.

Storage is another grid resource. Normally each processor, server, or computer in the grid will make a quantity of storage available. Two types are available: hard disk and dynamic storage. Sometimes this can be a very large amount of storage availability, allowing for very large databases to be stored in multiple locations, thus defeating some of the limitations imposed by certain operating systems. Temporary storage is also needed with some applications, and the grid management software can locate and schedule such storage.

Fast access storage can be used as additional memory for some applications, allowing for increased throughput and quicker processing times if needed.

Servers are another underutilized computer resource, with large amounts of idle time between peak processing times. Servers can be made available to the grid as needed and scheduled for use processing grid applications or jobs. To control priorities and management, rules can set up to control processing. If server utilization rises above a set limit—85 percent, for example—then the grid processing application is passed on to another section of the grid that is not so busy. Grid computing provides the framework to utilize these resources.

WHAT APPLICATIONS RUN ON A GRID?

Certain types of applications have already been very successful using grid computing. Scientific projects, such as the SETI@home project run by the University of Berkeley in California, have been a spectacular success. Here the Internet is used to send small packets of data to over 4 million PC users in a grid. The results are analyzed on the client processor and sent back to Berkley for further analysis, and another packet sent for processing. This project has been running since early 1999 and the statistics are astonishing.

• Since it's start in May 1999 over 4.8 million people have signed on from over 230 countries .

• At any given moment, 475 thousand machines are working fulltime on the problem.

• Results are being returned at a rate of over a 1 million files per day (34 trillion calculations per second, roughly equivalent to a 249-teraHz supercomputer). Each day over 1,000 years of CPU time is used.

• A new participant signs up on the Grid every 58 seconds.

• To date the total CPU time used is approaching 1,900,000 YEARS.

Some observers claim that it is the biggest computing project ever undertaken. No doubt, the success of the SETI@home project has spurred the development of grid computing and illustrated what can be accomplished.

Table 11.1 further illustrates other types of applications that can be used with grid computing. These applications have been successful for a number of years, and much has been learned about the systems design of effective grids. New applications are continuously evolving.

The commercial application of grid computing is in its early stages of development. It requires a different approach to conventional application or systems management. If the application can be split into one or more different subapplications, it can run on a grid. This is not easily achieved and requires considerable skill in system design and segmentation, as well as considerable programming skill to be successful.

Table 11.1. A Sample List of Grid Computing Applications

Many new software tools are needed to convert existing applications to a grid environment. This technology will accelerate as the need intensifies and vendors introduce the much-needed set of software tools to make grid computing applicable for commercial enterprises .

New computation- intensive applications should be designed and written for parallel execution so that they can be easily grid-enabled.