* Sun previews the Sun Grid
endif; ?>Sun last week unveiled the latest incarnation of its grid strategy – one in which companies can contract with Sun and its partners to use idle compute power and storage capacity to perform tasks they would otherwise find expensive or prohibitive to do.
The Sun Grid will be available in the second quarter of 2005, according to Sun. Jonathan Schwartz, president and COO, says it will be available for $1 per CPU hour and $1 per gigabyte of storage capacity per month. Further Schwartz says the grid is not for everyone or every application. It is tailored to computation-intensive tasks such as geophysical simulations, 3D rendering and Monte Carlo simulations at financial institutions.
Sun has not worked all the details of the grid out yet. It still lacks a method for prioritizing jobs for customers that would run on the grid. Sun executives speculate that the Nauticus N2000 content switch that Sun acquired in September 2002 would be used for workload balancing and scheduling.
Consider, though, how workloads among companies competing in the same market would be handled. Say, for instance, two biotech companies are doing time-sensitive research in the same area. Analysts speculate that workload priority will once again depend on money and the type of service-level agreements customers work out with Sun.
It will be important for customers to carefully negotiate these contracts as they start to adopt grid computing.
Further, details of how Sun would secure workloads running on the grid were unspecific. While the company said application workloads could be protected and isolated from each other with Sun’s container technology, it did not specify how encryption would take place.
Sun expects the grid to relieve IT managers of tasks they don’t have the resources or inclination to perform. The company will work with other companies to collocate the grid across the U.S.
Sun did not elaborate on how businesses might return unused compute cycles or whether those customers would be charged for them. For instance, seasonal retailers may need more compute cycles during the holidays. The retailers may contract for a certain number of compute cycles – but when business falls off after the holiday rush, would they be able to return those unused compute cycles to the pool? And would they be able to sell them back to Sun or another customer and improve their bottom line?




