Thoughts on grid computing
January 5, 2006
This is another post from my old blog. There was a comment there too, but since I decided not to import my posts, I will have to leave the comment behind.
For many years people have talked about grid computing. In the scientific computing business, grid computing has been considered by many to be a panacea for various computing problems and there a number of projects are that try and tap into the power of grids. Examples include the world community grid from IBM and United Devices or the efforts from BOINC.
While community grids probably work well for the kinds of projects listed above, I have always wondered how useful they really are for more specific applications, e.g. a drug discovery effort at a pharma company. There are companies that are running grids, but do they really get better results than they might with a cluster. IMO, cluster computing has certain advantages in terms of efficiency that a grid cannot achieve in reality. While Prof. Charles Brooks, via Predictor@home, has demonstrated how multiscale modeling efforts can be applied to grids, I am still somewhat uncertain about their success in a commercial setting. Cluster computing, for now, would be my preference if I was an IT manager. Certainly there are applications, such as lower-priority, long term projects where a grid solution would be optimal (e.g. a routine, version controlled genome annotation project) as it can keep humming along in the background when resources are found within an organization. On the other hand, for a mission critical project (e.g. virtual screening for a specific kinase target), I would prefer to deploy a cluster, i.e. resources that I can control.
What do others think?
Further Reading
Sun’s Grid Flop
Cluster Computing
Clusters better than grids
Posted by Deepak Singh | Filed Under Computing
Affiliate link
bbgm recommends I Wish I'd Made You Angry Earlier
by Max Perutz
