Communications of the ACM
Cloud, Exascale Challenges Converge
Photo courtesy of HPC in the Cloud
Ioan Raicu, founder and director of the Illinois Institute of Technology's Data-Intensive Distributed Systems Laboratory, studies cloud computing, exascale computing, and the programming and efficiency challenges of multicore processors.
In a recent test, Raicu compared current supercomputer capacity that will power the coming age of exascale systems to major cloud computing providers such as Amazon. He found that in 2018 Amazon will look very similar to exascale supercomputers, with hundreds of thousands of nodes.
Amazon's current data centers are spread out across six locations and include about 40,000 servers, 320,000 cores, and consume an estimated $12 million a year in energy costs. By 2018, Raicu estimates that Amazon's data centers could grow anywhere from 100 to 1,000 times what they are today. However, more power-efficient technologies need to be developed to aid in the progress toward exascale computing, which can be used to power the growing mega-clouds, according to Raicu. There also needs to be applications that can take advantage of the huge number of cores available, he says.
From HPC in the Cloud
View Full Article
No entries found