The death of distance
Computing grids use international networks to link computing resources from all over the world. This means you can sit in France and use computers in the U.S, or work from Australia using computers from Taiwan.
Such international grids are possible today because of the impressive development of networking technology. Ten years ago, it would have been stupid to send large amounts of data across the globe for processing on other computer resources, because of the time taken to transfer the data. Today, all this is possible and more!
Pushed by the Internet economy and the widespread penetration of optical fibers in telecommunications systems, the performance of wide area networks has been doubling every nine months or so over the last few years. That translates to a 3000x improvement in 15 years. Imagine if cars had made the same improvements in speed since 1985…you could easily go into orbit by pressing down hard on the accelerator!
FASTER! FASTER!
Some researchers have computing needs that make even the fastest connections seem slow: some scientists need even higher-speed connectivity, up to tens of gigabits per second (Gbps); others need ultra-low "latency", which means there is minimal delay when sending date to remote colleagues in "real time".
Other researchers want "just-in-time" delivery of data across a grid, so that complicated calculations requiring constant communication between processors can be performed. To avoid communication bottlenecks, grid developers also have to determine ways to compensate for failures, like transmission errors or PC crashes.
To meet such critical requirements, several high-performance networking issues have to be solved, including the optimization of Transport Protocols and the development of technical solutions such as high-performance Ethernet switching.
