Grid Computing: A Practical Guide to Technology and Applications (Programming Series)

Rating:
Summary: Fairly good overview
Review: The interest in grid computing has increased dramatically in scientific circles in the last decade, due in part to the need for high performance scientific computing and the availability of operating systems that make the deployment of distributed computing more painless for the scientific investigator. With the exception of financial firms, grid computing has not made inroads into the business community. Private industry has expressed concerns about the security of grid computing and various psychological barriers have prohibited it from being incorporated even in business LAN environments. This book gives a fairly comprehensive overview of grid computing for those interested in obtaining knowledge as to its efficacy in high performance computing or for those who are investigating whether it can indeed be practical for business. It is readily apparent that the author wants private industry to take grid computing more seriously, and he gives ample discussion of just how this could be done.

Part of this discussion involves the relation between Web services and grid computing. Those readers who deal with Web services would expect a connection between grid computing and Web services, and the `Open Grid Services Architecture', spearheaded by IBM and the Globus team, is an attempt to unify the two. The author points out the main difference between the two architectures, namely that Web services support "persistent" services while grid architectures must also support "transient" services, such as video conferencing. Web services is in place in many different industries at the present time, but it remains to be seen whether it will remain so in years to come, due in part to the conflicts between the different standardization efforts.

The different types of grids that can be configured are also discussed in the book. These include departmental grids, designed for a group of people within an enterprise, enterprise grids which cover all users within an enterprise, and extraprise grids, which can be established within companies. Grid computing has had some reported successes, particular the SETI grid project and the FOLD grid project for calculating protein folding. Both are popular with the public and have GUI interfaces that are very pleasing from an aesthetic point of view.

One of the biggest reasons for not being able to do grid computing in a business environment is the reluctance of management to allow many or all of the machines in the organization to be dedicated to the grid, even if done when the machines are offline. This is true even for the `desktop' grids that are discussed in this book. Subjective factors, such as privacy issues (even if they are not valid) and imagined interference come into play when approaching grid computing in a professional business environment. The presence of distributed software on the various machines in the organization may cause many to believe it is the cause of an outage or other problems when they occur. Trust in grid computing has to develop before it will be used routinely in a business environment. The author does address these concerns in the book.

He also discusses the need for an easier transition to grid computing in a business context, if the decision to deploy it has been made. The time taken to make grid computing a reality in this context must be minimal, considering the great amount of investment that has already been made in designing, implementing, and maintaining existing applications. Such a transition can be handled by using the approach of Web services, or what he calls Grid services. He outlines a few different ways in which the existing code can be wrapped. If the source code is not available, one can wrap the executables for example. If it is, it can wrapped and additional code overlaid on it in order to interface properly with any existing applications. A WSDL (Web Service Description Document) is then generated and placed in a registry service, in order that other applications can make use of the service. The Universal Description, Discovery and Integration (UDDI) registry is the one that is advocated by the author.

Several applications of grid computing are discussed in the book, each having various degrees of ease in actual deployment. Numerical applications using Monte Carlo are viewed as the easiest ones to be "grid-enabled", and this is born out in experience. Financial and biotechnology firms in particular are heavy users of grid-enabled applications that utilize Monte Carlo simulations. The author discusses a rudimentary test, called the `compute intensity ratio' to check whether an application is suitable for deploying on a grid. If this ratio is greater than one, then the application is deemed to be well suited for distributed processing on a grid. Applications in desktop grid computing such as risk management and financial derivatives, molecular docking for drug discovery, and architectural rendering are briefly discussed.

As an example of a cluster grid, the famous Beowulf cluster, which is heavily used in scientific computing, is discussed in the book. Scientific computing is the major driver behind grid computing, as is readily apparent throughout the book. Discussion of high performance grid computing occupies an entire chapter of the book in fact. Production High Performance Computing via the use of the Message Passing Interface (MPI) has allowed scientists to develop grid applications more effectively, without having to worry too much about architectural issues.

The author has included several examples of how grid computing is used in the business community, such as in telecommunications and bioinformatics. There are more examples than he discusses, but they are usually not made public because of considerations of propriety. Businesses that have used grid computing to further their success are usually not vocal in their approaches. The book would have been better if the author had included actual benchmarking studies of how businesses have improved their financial positions by using grid computing, with in-depth figures that illustrate quantitatively the power of grid computing. The inclusion of such studies would definitely assist those who are seriously considering grid computing.

Rating:
Summary: Good Overview
Review: The newest, highest performance PC's are more powerful than the biggest supercomputers of the early 1990's. But with the higher computational capacity the bigger the tasks assigned to be computed. Note that these are not your typical business applications. The biggest computational tasks today lie in four areas: atomic energy research, weather prediction, cryptography and bioinformatics.

Some of these tasks lend themselves well to being split up amongst a bunch of computers. Cryptography, as an example lends itself well to taking a message and assigning each computer to attack the message with a different key. The processing is independent of the results from other computers attacking the message. What one computer could do eventually, 30,000 or 3 million can do just that much faster.

This book is obviously on using a collection of computers, not necessarily co-located to handle such complex computing tasks. A full time practitioner writes it with collaboration from leaders in the field.