Andrea Carol Arpaci-Dusseau. Implicit coscheduling: Coordinated scheduling with implicit information in distributed systems. ACM Transactions on Computer Systems, 19(3):283--331, Aug 2001.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....system level workload. This approach has the appealing advantage that it does not require any changes to existing parallel software, and therefore is able to deal with existing legacy codes. For example, evaluation results show that coscheduling algorithms such as Implicit Coscheduling (ICS) [2] can potentially alleviate load imbalance and increase resource utilization. However, ICS is not always able to handle all job types because it cannot rely on global coordination. In this paper we show that it is possible to substantially increase the resource utilization in a cluster of ....

....cluster, using different schedulers. Specifically, we compare the performance achieved by FCS with that of first come first served batch scheduling (FCFS) gang scheduling (GS) local scheduling with busy waiting, and local scheduling with spin blocking (SB) which is very similar to ICS [2] in this context. Of these schedulers, different ones provide the best performance for the different mixes. However, in all the cases tested, FCS provided essentially the same performance as the best other scheduler. This testifies to its flexibility, its ability to identify the characteristics of ....

[Article contains additional citation context not shown here]

Andrea Carol Arpaci-Dusseau. Implicit Coscheduling: Coordinated Scheduling with Implicit Information in Distributed Systems. ACM Transactions on Computer Systems (TOCS), 19(3), 2001.

....a parallel job can be launched on a 1010 node cluster in about 15 seconds, depending on the binary size. Many recent research results show that job scheduling algorithms can substantially improve scalability, responsiveness, resource utilization and usability of a large scale parallel machine [2] [13] Unfortunately, the body of work developed in the last few years has not yet led to any practical applications implementations of such gang scheduling and coscheduling algorithms in parallel clusters. We argue that one of the main problems is the lack of flexible and efficient run time ....

Andrea Carol Arpaci-Dusseau. Implicit Coscheduling: Coordinated Scheduling with Implicit Information in Distributed Systems. ACM Transactions on Computer Systems (TOCS), 19(3), 2001.

....system level workload. This approach has the appealing advantage that it does not require any changes to existing parallel software, and therefore is able to deal with existing legacy codes. For example, evaluation results show that coscheduling algorithms such as Implicit Coscheduling (ICS) [2] can potentially alleviate load imbalance and increase resource utilization. However, it is not always able to handle all job types due to a complete lack of global coordination. In this paper we show that it is possible to substantially increase the resource utilization in a cluster of ....

....cluster, using different schedulers. Specifically, we compare the performance achieved by FCS with that of first come first served batch scheduling (FCFS) gang scheduling (GS) local scheduling with busy waiting, and local scheduling with spin blocking (SB) which is very similar to ICS [2] in this context. Of these schedulers, different ones provide the best performance for the different mixes. However, in all the cases tested, FCS provided essentially the same performance as the best other scheduler. This testifies to its flexibility, its ability to identify the characteristics of ....

[Article contains additional citation context not shown here]

Andrea Carol Arpaci-Dusseau. Implicit Coscheduling: Coordinated Scheduling with Implicit Information in Distributed Systems. ACM Transactions on Computer Systems (TOCS), 19(3), 2001.

....that, like STORM, scale logarithmically, not linearly, in the number of nodes. 5. 2 Process scheduling Many recent research results show that good job scheduling algorithms can substantially improve scalability, responsiveness, resource utilization, and usability of large scale parallel machines [1, 13]. Unfortunately, the body of work developed in the last few years has not yet led to many practical implementations of such coscheduling algorithms on production clusters. We argue that one of the main problems is the lack of flexible and efficient run time systems that can support the ....

Andrea C. Arpaci-Dusseau. Implicit coscheduling: Coordinated scheduling with implicit information in distributed systems. ACM Transactions on Computer Systems, 19(3), August 2001. Available from http://www.cs.wisc.edu/~dusseau/Papers/tocs01.ps.

....the behavior of certain software systems such as the file system can induce performance heterogeneity. 3.1.3 Workload Though we will not document this in detail, external workload fluctuations often result in a non uniform and dynamic execution environment. As discussed by A. Arpaci Dusseau in [8], the modern cluster environment will likely contain a mix of both parallel and sequential jobs. Unless physically separated, the likelihood that parallel applications will run on identically loaded machines is quite small. In disk systems, similar dynamic workload fluctuations occur; they are ....

....the system must constantly gauge the performance of underlying components. In designing such feedback driven systems, we must be careful to keep them as lightweight as possible; otherwise, the cost of adaptation will outweigh its benefits. Our distributed queue algorithm uses implicit feedback [8]: no explicit information probes are sent from producers to consumers. Instead, the producers track the responsiveness of consumers; those consumers that respond more frequently to data requests are sent a larger fraction of subsequent data requests. Feedback in graduated declustering is more ....

[Article contains additional citation context not shown here]

Andrea C. Arpaci-Dusseau. Implicit Coscheduling: Coordinated Scheduling with Implicit Information in Distributed Systems. PhD thesis, University of California, Berkeley, 1998. UCB/CSD-99-1052.

No context found.

Andrea Carol Arpaci-Dusseau. Implicit coscheduling: Coordinated scheduling with implicit information in distributed systems. ACM Transactions on Computer Systems, 19(3):283--331, Aug 2001.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC