Jinal Jhaveri on everything

Summary of Supercomputing 05 Things I attended 1) Tutorial on datawarehousing for Genomics and Proteomics (Marshal Peterson, CTO-Craig venter institute) - Talked about the Sargasso sea experiment. - produced 100 million of genetic code every 24 hours - 1800 new species - 1.2 million new genes - 12 complete genomes - Before this experiment we knew about only 120-180k genes - looked for something like: Take all genes which might be efficient in producing ethanol - generated 1 billion base pairs of dna - 85% of sequence was only seen once - Used it to better understand the functioning of natural ecosystem 2) Talked about Data Warehousing Netezza performance server system: Using Assymetric Massively Parallel Architecture: - combines symmetric multiprocessing with massively parallel architecture - Basically they tried to bring processing power to the data instread of sending data to the processing power - Also had facility of sql-blast i.e adding blast capacity in sql. - They have this con...

clawhmmer: blast: dynamic programming but relies on adhoc gap penalties, it still determines the type of guees HMMer:" probabilites model. Its robust, because it is a probabilisting it requires more flops new architectures: 1_ An order of magnitute more 2) Extremely parallel in nature 3) tiny processors 4) They share PC thousands of threads needed Brook of GPU: 1) Data is divided into streams 2) small function as KERNEL. fixed number of inputs and fix output 3) Encourages a data parallel type of programming challenges: 1) current software doesn't map well 2) large compute-> fetch ratio in the kernel 3)