IDG News Service (06/14/11) Joab Jackson
Investigating new strategies in scheduling workloads in the computer cloud that could work to the advantage of both cloud providers and users was the focus of the first round of papers presented at the Usenix HotCloud 2011 Workshop. Under discussion by researchers was how to define cloud computing jobs by the required hardware, and how to instill more flexibility into pricing. "The physical resources to actually implement a cloud are incredibly expensive, and so, having built it, you want to pack in as much work as possible into that infrastructure," says Microsoft Research's David Maltz. "Hence scheduling becomes really important." The researchers proposed that two types of machines should be assigned to each cloud--a set of core node servers committed to basic tasks, and a set of accelerated nodes that can be added on a temporary basis to assist with performing the more computationally heavy workloads. Users would submit jobs to a cloud driver that would apportion the appropriate number of nodes to the chore. Institute of Science and Technology Austria researcher Damien Zufferey says that cloud computing services should account for other factors, such as if the client wants a job conducted quickly or at the lowest possible cost, or some combination in between.
Investigating new strategies in scheduling workloads in the computer cloud that could work to the advantage of both cloud providers and users was the focus of the first round of papers presented at the Usenix HotCloud 2011 Workshop. Under discussion by researchers was how to define cloud computing jobs by the required hardware, and how to instill more flexibility into pricing. "The physical resources to actually implement a cloud are incredibly expensive, and so, having built it, you want to pack in as much work as possible into that infrastructure," says Microsoft Research's David Maltz. "Hence scheduling becomes really important." The researchers proposed that two types of machines should be assigned to each cloud--a set of core node servers committed to basic tasks, and a set of accelerated nodes that can be added on a temporary basis to assist with performing the more computationally heavy workloads. Users would submit jobs to a cloud driver that would apportion the appropriate number of nodes to the chore. Institute of Science and Technology Austria researcher Damien Zufferey says that cloud computing services should account for other factors, such as if the client wants a job conducted quickly or at the lowest possible cost, or some combination in between.
No comments:
Post a Comment
The Open Source Web Hosting company