Grid Computing 101 - Page 2

Grids can share more than just processors; they also can share drive space for both greater storage room and more robust data availability. Usually this is done with mountable networked file systems, such as AFS (Andrew File System), that old Unix distributed storage favorite; NFS (Network File System); or DFS (Distributed File System). The grid software, in turn, can provide virtual storage with an overreaching file system that spans both drives and file systems.

A grid also can be used to set multiple computers to work on a single problem. In such cases, grids must support IPC (interprocess communication) between programs running on different systems. Typically, grids that support such activity borrow MPC (massively parallel computing) message-passing models

Common supercomputer message-passing models that have been borrowed by grids are MPI (Message Passing Interface) and PVM (Parallel Virtual Machine).

Another advantage of this borrowing is that application developers dont have to reinvent programs that can make use of a grids parallel computing resources.

To date, most grid systems, such as Sun Microsystems Inc.s N1 Grid Engine and DataSynapse Inc.s GridServer, have been proprietary designs. Recently, however, open-source approaches based on Linux systems have gained popularity.

The most influential of these efforts, which has IBMs backing, is The Globus Alliance. This consortium creates open-source tools, the Globus Toolkit 3.2, for building grids. Globus uses Java and Web services to help developers create grid-capable applications.

Globus supporters are not the only ones working toward an open source-based grid. Dell, EMC, Oracle and Intel are working on the Linux-based "Project MegaGrid," which will run on Dells PowerEdge servers.

The business case driving all of these efforts is the same: Provide customers with a utility model for their computing needs. Hewlett-Packards Adaptive Enterprise, IBMs On Demand Business and Suns N1 take different takes on grids central themes, but are all designed to provide low-cost computing power to customers.

This commercial aspect to grid is also relatively new. Traditionally, grids have been used in scientific and academic environments, where they shared the same jobs as its cousins, supercomputing and grids.

Now, however, as the technology has matured and open source has brought the price of grid development down, companies are taking it to the marketplace. In particular, financial companies—with their vast need for real-time processing—have become important grid customers.

So it is that as Microsoft prepares to launch its Bigtop, the Redmond giant will face several opponents with mature technologies.

IBM is currently the grid leader, according to financial services magazine Waters, with Oracle and DataSynapse following. Thus, this is one market battle where Microsoft will face a stern test.


Check out eWEEK.coms for the latest news, views and analysis on servers, switches and networking protocols for the enterprise and small businesses.