Truth to be told, I already had a mind for putting up all my computational research on the cloud next time because I intent to fly light on my bootstrap styled startup, yet aim to deliver high-performance computational biology services. The cloud offers me the freedom, scalability, power that I can possibility need.
What Simone gave me instead was a few realization that were invaluable:
· most high costing, on-demand computation services would be moved on to the cloud simply because the cloud will operate much like a public utility because it simply makes economic sense
· In order to leverage on the cloud, my programs have to written in a parallelization and scalable(distributed computing architecture) – like the Apache Hadoop architecture
· Huge bioinformatics databases like Genbank operates on the cloud, which I never knew. I guess the whole bioinfo sector would start moving into the could with the first movers already present. This is a good thing which would reduce cost and increase performance.
· He cautions that network resources and performance cannot scale in accordance to Moore’s law. And I wonder why. Few ideas that I can think of which supports this hypothesis: But I am really not sure if this is true.
o Exponential complexity of network operations with a lot of callbacks and crossed talks
o Limitations of physical infrastructure
o No possibility improvement beyond the optimized network algorithm/resource sharing (p2p, virtualization, caching…) May be there are better algorithms/architectures to be discovered.
I think this might be a fraction of what Prof Tan (Bioinfomatics) saw in cloud computing. (=
1 comment:
whoa.. I think you're on the something great here.. Will be good to see how cloud computing works out (haha, last night's lecture was a taad too technical for me)..
Post a Comment