Weissman, Jon. Scheduling Parallel Computations in a Heterogeneous Environment. PhD thesis. University of Virginia, August 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....nature of the grid, as well as the differing demands of applications run on the grid, makes grid scheduling complicated. Existing schedulers in wide area heterogeneous systems require a large amount of information about the application and the grid environment to produce reasonable schedules [Lo88][Wei95][Wei00] However, this information may not be available, may be too costly to collect, or may increase the run time overhead of the scheduler such that the scheduler is rendered ineffective. We believe that no one scheduler is appropriate for all grid systems. Instead, grid systems require a ....

....and processing information with the related benefits of the resulting schedule. 5 1. 1 Thesis The general task placement problem for more than two processors has been shown to be NP complete [Ull75] Various sub optimal scheduling heuristics have been proposed in the literature [Lo88] Sal99][Wei95][Wei00] Zom01] Each of the solutions is appropriate in certain situations, but may not be appropriate in others. Clearly, no one scheduling algorithm will be equally appropriate for all grid systems and all applications. Instead, grid systems require a scheduling framework that produces ....

[Article contains additional citation context not shown here]

Weissman, Jon. Scheduling Parallel Computations in a Heterogeneous Environment. PhD thesis. University of Virginia, August 1995.

....for a user. This tool ensures that input files and output files are arranged such that the user can identify corresponding sets easily. 3.5 Scheduling In a metasystem, scheduling is the process of initiating runs on the best possible resources. The general scheduling problem is NP complete [12]. In addition, the parameters involved in making an optimal schedule are numerous and mutually dependent. Constructing a schedule may involve making decisions not limited to: a) the machine architectures for which a class has implementations, b) specific properties of a machine desired by the ....

Weissman, J., Scheduling Parallel Computations in a Heterogeneous Environment, Ph.D. Dissertation, University of Virginia, August 1995.

....be considered differently on each machine, depending on the architecture and on the local scheduling policy. Different groups have proposed systems that focus on scheduling of parallel applications on distributed systems in different ways (e.g. AppLeS 9 [5, 6] Legion [24] Globus [23] Prophet [51], MARS [22] Condor [10] SmartNet [27] and Nile [34, 35] In general, current scheduling systems assign tasks (or entire applications) to machines according to computation time and communication costs using several different performance measures (e.g. 25, 31] but few consider contention ....

J. Weissman, "Scheduling Parallel Computations in a Heterogeneous Environment," Ph.D. dissertation, University of Virginia, August 1995.

....be considered differently on each machine, depending on the architecture and on the local scheduling policy. Different groups have proposed systems that focus on scheduling of parallel applications on distributed systems in different ways (e.g. AppLeS 9 [5, 6] Legion [24] Globus [23] Prophet [51], MARS [22] Condor [10] SmartNet [27] and Nile [34, 35] In general, current scheduling systems assign tasks (or entire applications) to machines according to computation time and communication costs using several different performance measures (e.g. 25, 31] but few consider contention ....

J. Weissman, "Scheduling Parallel Computations in a Heterogeneous Environment," Ph.D. dissertation, University of Virginia, August 1995.

....certain security policies, file systems, resource accounting procedures, and most importantly, computing resources. It is unimportant to the wide area scheduling system how the sites themselves organize their resources. A model for site organization based on processor clusters is described in [14][16]. Each site runs a scheduling manager (SM) that has two components. One component is an interface to the local scheduling system (LS) and the other is the wide area interface between other SM s. The SM s and LS s are address space disjoint and communicate via message passing 1 . The SM s run the ....

....message size for communication. This is important because it must be possible to estimate these costs for heterogeneous machines. A model for data parallel programs demonstrated that this abstract information can be turned into very precise execution time estimates in a heterogeneous environment [16]. Some of the job information will require off line benchmarking. Scheduling histories If a job has been run previously by the wide area scheduling system, then the results of this previous schedule may be useful. The wide area scheduling system will return a record of the scheduling decision ....

J.B. Weissman, "Scheduling Parallel Computations in a Heterogeneous Environment," Ph.D. dissertation, University of Virginia, August 1995.

....information. Prophet has been integrated into Mentat [7] an object oriented parallel processing system developed at the University of Virginia, and significantly improved the scheduling quality for SPMD Mentat programs over the native Mentat scheduler in an ethernet network environment [15]. Previously, Prophet was limited to SPMD applications in local area ethernet networks, and now supports ATM and wide area networks, and other application types including task parallel pipelines 3 (PP) Although Prophet overhead was manageable for most applications, there was no way to ....

....LANs and WANs, with limited programmer effort. Prophet exploits the structure of the application and information about the system resources to produce reduced completion time for SPMD and PP application types. The performance of Prophet is superior to system schedulers such as the Mentat scheduler [15], but a direct comparison to hand tuned programmer scheduled applications 25 is needed. Our ultimate goal is show performance better than system schedulers and moderatelytuned applications, and within a few percent of highly tuned programmer scheduled applications. For the latter, Prophet will ....

J.B. Weissman, "Scheduling Parallel Computations in a Heterogeneous Environment," Ph.D. dissertation, University of Virginia, August 1995.

....programmer to specify the salient application details has been developed. This API is applicable to a large class of SPMD applications and is fairly easy to use. A system called Prophet that automates the scheduling process for SPMD applications to obtain reduced completion time has been developed [12]. Prophet uses network resource information and application information to guide the scheduling process, and has been integrated into Mentat, an object oriented parallel processing system developed at the University of Virginia [6] Prophet is unique in that it addresses the problems of ....

....The worst case order of this algorithm is O(mP) O(mlog 2 P) for cluster ordering, O(mlog 2 P) for phase 1, and O(mP) for phase 2, for m clusters and P total processors. For each cluster it locates the best number of processors by a form of binary search, the details of which are provided in [12]. In practice, the amount of computation performed for each configuration is very small and the linear dependence on P is mitigated to a certain extent. Future optimizations include setting a limit on the number of configurations explored based on the predicted overhead and run time length of the ....

[Article contains additional citation context not shown here]

J.B. Weissman, "Scheduling Parallel Computations in a Heterogeneous Environment," Ph.D. dissertation, University of Virginia, August 1995.

....a cost driven method for scheduling SPMD and PP applications in heterogeneous workstation networks. Prophet makes resource scheduling decisions that are predicted to produce reduced completion time for the application. Near optimal scheduling results have been obtained for SPMD applications [22]. Scheduling SPMD applications consists of two parts, partitioning and placement (Figure 13) In partitioning, the application is decomposed across a set of potentially heterogeneous processors. The number of processors selected depends on the computation granularity an application with a ....

Weissman, J.B., "Scheduling Parallel Computations in a Heterogeneous Environment," Ph.D. dissertation, University of Virginia, August 1995.

....scheduling techniques. Prophet has been integrated into Mentat [5] an object oriented parallel processing system, developed at the University of Virginia, and significantly improved the scheduling quality for SPMD Mentat programs over the native Mentat scheduler in an ethernet network environment [11]. Prophet had a number of shortcomings that have now been addressed. Prophet was limited to SPMD applications in local area ethernet networks, and now supports ATM and wide area networks [8] and other application types including parallel pipelines (PP) The system was not overhead aware and now ....

J.B. Weissman, "Scheduling Parallel Computations in a Heterogeneous Environment," Ph.D. dissertation, University of Virginia, August 1995.

No context found.

WEISSMAN95. Jon Weissman. Scheduling Parallel COmputations in a Heterogeneous Environment. PhD Thesis University of Virginia (August 1995).

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