Here comes The Sun (Grid)
Sun's massively parallel facility opens to non-US customers
In the past two decades, and until a few years ago, the computing power of commercially available computers has steadily increased, according to the so-called "Moore's Law". This has always been achieved in two ways:
Now, let's consider this scenario: suppose we need to perform a very demanding calculation, and we need the outcome as soon as possible. Possible applications include but are not limited to: astronomy, physics, meteorology, finance, etc. Being filtly rich, we could simply shell out the big bucks required to buy one of those multiprocessor monsters, and that's it; however this will most likely not be the case. Then the folks at Sun came up with a very clever idea: selling computing power from its multiprocessor machines at very cheap rates with a smart charging scheme (on a time and per- processor basis). This way, you only pay for the time and number of nodes needed to perform your task, with no system administration, configuration and training costs and hassles. It's really as easy as 1-2-3: "pack" your job in some form by using the plugins and tools provided by Sun, submit it by uploading its data and execution code, run it and get the outcome back.
Cool, huh? Sure. But there's a catch: up until a few weeks ago, you had to be living on US ground to take advantage of this facility. Fortunately the situation has recently changed, as Sun has made its technology available to 24 more countries throughout the world including Italy. Fabrizio Giudici and I have chosen to go along the parallel computing route for our open-source projects (Pleiades and Bluemarine) for about one year now: also, see the abstract of the presentation we gave at Javapolis 2006. Needless to say, this news opens up a wealth of new opportunities for us and many others.
Please check back often as more news is expected to come soon!
- By boosting clock frequencies
- By using multiple instruction pipelines, RISC-like architectures and bigger data/instruction caches.
Now, let's consider this scenario: suppose we need to perform a very demanding calculation, and we need the outcome as soon as possible. Possible applications include but are not limited to: astronomy, physics, meteorology, finance, etc. Being filtly rich, we could simply shell out the big bucks required to buy one of those multiprocessor monsters, and that's it; however this will most likely not be the case. Then the folks at Sun came up with a very clever idea: selling computing power from its multiprocessor machines at very cheap rates with a smart charging scheme (on a time and per- processor basis). This way, you only pay for the time and number of nodes needed to perform your task, with no system administration, configuration and training costs and hassles. It's really as easy as 1-2-3: "pack" your job in some form by using the plugins and tools provided by Sun, submit it by uploading its data and execution code, run it and get the outcome back.
Cool, huh? Sure. But there's a catch: up until a few weeks ago, you had to be living on US ground to take advantage of this facility. Fortunately the situation has recently changed, as Sun has made its technology available to 24 more countries throughout the world including Italy. Fabrizio Giudici and I have chosen to go along the parallel computing route for our open-source projects (Pleiades and Bluemarine) for about one year now: also, see the abstract of the presentation we gave at Javapolis 2006. Needless to say, this news opens up a wealth of new opportunities for us and many others.
Please check back often as more news is expected to come soon!
