Making Cloud-Computing Systems More Efficient

Grazed from NewYork Times.  Author: Quentin Hardy.

Computer scientists at Stanford have developed software that works like the recommendation engines for Amazon shoppers and Netflix movie subscribers, only for the distribution of workloads across large computing environments. The results, they say, can triple the efficiency of cloud-computing systems.

“Google, Microsoft, Twitter — everyone has a problem of underperforming servers,” said Christos Kozyrakis, a professor of electrical engineering and computer science at Stanford University, who led the creation of the management software, which is called Quasar. He has worked in a research capacity at Google and Microsoft, and he used some of Twitter’s resources in his research on Quasar…

In their traditional state, computer servers were typically run at about 20 percent of their capability to avoid overloading a machine if a lot of work suddenly showed up. For over a decade, a technique called virtualization has been used to merge the work of one server with several to avoid such load problems…

Read more from the source @ http://bits.blogs.nytimes.com/2014/03/06/a-computer-cluster-innovator/?_php=true&_type=blogs&_r=0