Before there was cloud computing, there was grid computing. Instead of sending your jobs to the cloud, you’d send them to the grid. Instead of provisioning big banks of on-premise computers to do your calculations, you’d send them to the grid.
Virtualization, services and ever cheaper hardware paved the way from grid to cloud. It may be too early to say, but cloud does seem to be a bit of a better play for a wider variety of jobs. Part of the reason grid didn’t get too far off the mark was that its poster stories were usually scientific applications…something of a niche.
Still, it is interesting to conjecture on how science – both big and small science – will be done on the cloud. Relational data bases are a tried and true way of dealing with data, scientific and other. The cloud at this point is highly influenced by Google’s flatter MapReduce/Hadoop ways of handling data. Still, Amazon just augmented its SimpleDB with a straightahead RDB (MySQL).
Some view on the nature of that possible transformation of science data and cloud this we comes via Michael Schatz. For the biology community, moving to MapReduce/Hadoop-style architecture would be a challenging undertaking. SOAP and XML made some serious inroads there. Schatz is working on adapting important bioinformatics tools to the cloud paradigm. He has discussed the issues on SourceForge where he documents some project work. Check out Haddop for Computational Biology.