Sobalvarro PG, Pakin S, Weihl WE, Chien AA. Dynamic coscheduling on workstation clusters. Proceedings of the Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 1998 (Lecture Notes in Computer Science, vol. 1459). Springer: Berlin, 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....when multiprogramming a set of parallel jobs on a multiprocessor machine [22] Coscheduling strives to ensure that all processes belonging to a job are scheduled at the same time. Subsequent work has generalized and refined the coscheduling (now often called gang scheduling) concept [5, 6, 8, 9, 11, 14, 15, 19, 23, 24, 25, 26]. Gang scheduling schemes are a practical result of the multiprocessor scheduling community and have been adapted for inclusion in several production systems including the Intel Paragon [4] CM 5 [3] Meiko CS 2, multiprocessor SGI workstations [1] and the IBM SP2 [17, 18] Work has been done to ....

P. Sabalvarro, S. Pakin, W. Weihl, and A.A. Chien. Dynamic coscheduling on workstation clusters. In Proc. of the IPPS'98 Workshop on Job Scheuling Strageies for Parallel Processing, pages 231--256. Springer LNCS #1459, 1998.

....scheduling techniques strive to coschedule jobs that communicate frequently on massively parallel (MPP) systems conglomerated from single threaded processors. 21] describes a system that dynamically coschedules jobs that communicate frequently to increase system utilization and job response time. [25] 19 improves upon gang scheduling to dynamically produce emergent coscheduling of the processes constituting a parallel job. 20] improves upon gang scheduling to fill holes in utilization around gang scheduled jobs with pieces of work from jobs that do not require all resources in order to to ....

P. Sobalvarro, S. Pakin, W. Weihl, and A. Chien. Dynamic coscheduling on workstation clusters. In SRC Technical Note 1997-017, March 1997.

....methods with some simple schedulers, which are described in sections 2.5 2.7. 2. 2 Dynamic and Implicit Coscheduling Dynamic coscheduling (DCS) was proposed as a way to reduce the amount of global synchronization of coscheduling, and thus decrease the system s overhead and improve its sealability [19, 26, 27]. The philosophy behind DCS is demand based coscheduling: coordination is achieved by observing the communication between processes, and not by a master daemon, Communication between processes is used to deduce which processes should be coscheduled and to effect coscheduling. Thus, when a ....

....assumption: if a process receives an incoming message, its peer(s) must be currently running on other nodes, so the process should be prioritized for immediate execution. Several aspects and parameters can be tuned for different assumptions or workloads, e.g. when to make a preemption decision [26]. The design of the testbed should allow room for configuration and tuning of these parameters. Like DCS, Implicit coscheduling (ICS) makes local scheduling decisions based on monitoring communication activity [5, 6] Processes waiting for a communication action to complete use a spin block ....

[Article contains additional citation context not shown here]

Patrick Sobalvarro, Scott Pakin, William E. Weihl, and Andrew A. Chien. Dynamic Coscheduling on Workstation Clusters. In Dror G. Feitelson and Larry Rudolph, editors, Job Scheduling Strategies for Parallel Processing, volume 1459 of Lecture Notes in Computer Science, pages 231-256. Springer-Verlag, 1998.

....to resolve priority inversion in general and do not resolve priority inversions that arise due to process dependencies that are not explicitly identified in advance. Co scheduling mechanisms have been developed to improve the performance of parallel applications in parallel computing environments [14, 15, 16, 17]. These mechanisms try to schedule cooperating threads or processes belonging to the same parallel application to run concurrently. This reduces busy waiting and context switching overhead and improves the degree of parallelism that can be used by the application. Because many of these ....

P. G. Sobalvarro, S. Pakin, W. E. Weihl, and A. A. Chien, "Dynamic coscheduling on workstation clusters, " Lecture Notes in Computer Science, vol. 1459, pp. 231--257, 1998.

....version appeared in JPDPS 2001. Research supported by a Lady Davis Fellowship. with this approach is fragmentation: processors may be left idle if nothing is assigned to them in a certain slot. Several ideas for relaxations of gang scheduling that reduce this waste have therefore been proposed [11, 7, 22, 1]. Our work is also a relaxation of strict gang scheduling, which is called Paired Gang Scheduling [23] All of these improvements remove the requirement that all processes of a given job always run simultaneously on different nodes. There are also other types of improvements, e.g. the use of ....

....that this is achieved by a very simple device, based on data easily available directly to the scheduler. This distinguishes paired gang scheduling from other flexible gang scheduling schemes that are based on monitoring communication in order to deduce the characteristics of parallel jobs (e.g. [11, 7, 22, 1]) The idea can be extended if there are more I O devices in the system. One I O operation won t interfere with other I O operation, so a group of N 1 gangs can be dispatched, where N is the number of I O devices in the system. In particular, it is possible to overlap disk I O with ....

Sobalvarro P. G., Pakin S.,Weihl W. E. and Chien A. A., Dynamic coscheduling on workstation clusters. In Job Scheduling Strategies for Parallel Processing, Springer-Verlag, LNCS vol. 1459, pp. 231-- 256, 1998.

....ICS [1] This scheduling algorithm represents a new approach to dynamic coscheduling methods, since it can benefit both from scalable global scheduling decisions and local decision based on detailed process statistics. Furthermore, it differs from previous dynamic coscheduling methods like DCS [16] and ICS in that: 1. A CS process in FCS cannot be preempted before the time slot expires even if an incoming message arrives Algorithm 1: Context switch algorithm for FCS context switch: switch from one process to another process Invoked on each processing node by a global ....

P. Sobalvarro, S. Pakin, W. E. Weihl, and A. A. Chien, "Dynamic Coscheduling on Workstation Clusters ". In vol. 1459, pp. 231--256, Springer-Verlag, Mar 1998.

....in ICS [2] This scheduling algorithm represents a new approach to dynamic coscheduling methods, since it can benefit both from scalable global scheduling decisions and local decision based on detailed process statistics. Furthermore, it differs from previous dynamic coscheduling methods like DCS [14] and ICS in that: 1. A CS process in FCS cannot be preempted before the time slot expires even if an incoming message arrives for another process (processes classified as CS have proved that it is not worthwhile to deschedule them in their time slot) It should not yield the processor on ....

Patrick Sobalvarro, Scott Pakin, William E. Weihl, and Andrew A. Chien. Dynamic Coscheduling on Workstation Clusters. In Dror G. Feitelson and Larry Rudolph, editors, Job Scheduling Strategies for Parallel Processing, volume 1459 of Lecture Notes in Computer Science, pages 231--256. Springer-Verlag, 1998.

....in ICS [2] This scheduling algorithm represents a new approach to dynamic coscheduling methods, since it can benefit both from scalable global scheduling decisions and local decision based on detailed process statistics. Furthermore, it differs from previous dynamic coscheduling methods like DCS [14] and ICS in that: 1. A CS process in FCS cannot be preempted before the time slot expires even if an incoming message arrives for another process (processes classified as CS have proved that it is not worthwhile to deschedule them in their time slot) It should not yield the processor on ....

Patrick Sobalvarro, Scott Pakin, William E. Weihl, and Andrew A. Chien. Dynamic Coscheduling on Workstation Clusters. In Dror G. Feitelson and Larry Rudolph, editors, Job Scheduling Strategies for Parallel Processing, volume 1459 of Lecture Notes in Computer Science, pages 231--256. Springer-Verlag, 1998.

....coschedule jobs that communicate frequently on massively parallel (MPP) systems conglomerated from single threaded processors. Sistare et al. 22] describe a system that dynamically coschedules jobs that communicate frequently to increase system utilization and job response time. Sobalvarro et al. [26] improve upon gang scheduling to dynamically produce emergent coscheduling of the processes constituting a parallel job. Silva and Scherson [21] improve upon gang scheduling to fill holes in utilization around gang scheduled jobs with pieces of work from jobs that do not require all resources in ....

P. Sobalvarro, S. Pakin, W. Weihl, and A. Chien. Dynamic coscheduling on workstation clusters. In SRC Technical Note 1997-017, Mar. 1997.

....or batch, serial or parallel that are expected to be concurrently executing in the cluster. Recently, researchers have focused their attention and attempted to address exactly this scheduling problem. Emerging proposals include implicit scheduling [DAC96,ADC97] and dynamic coscheduling [SPWC97] In implicit scheduling, processes relinguish control of the processor if they block waiting for a message more than some interval, statically or dynamically determined. In dynamic coscheduling, the system manipulates the priority of the processes to effect coscheduling based on observed traffic ....

Patrick G. Sobalvarro, Scott Pakin, William E. Weihl, and Andrew A. Chien. Dynamic coscheduling on workstation clusters. Submitted for publication, March 1997.

....when multiprogramming a set of parallel jobs on a multiprocessor machine [22] Coscheduling strives to ensure that all processes belonging to a job are scheduled at the same time. Subsequent work has generalized and refined the coscheduling (now often called gang scheduling) concept [5, 6, 8, 9, 11, 14, 15, 19, 23 26]. Gang scheduling schemes are a practical result of the multiprocessor scheduling community and have been adapted for inclusion in several production systems including the Intel Paragon [4] CM 5 [3] Meiko CS 2, multiprocessor SGI workstations [1] and the IBM SP2 [17, 18] Work has been done to ....

P. Sabalvarro, S. Pakin, W. Weihl, and A.A. Chien. Dynamic coscheduling on workstation clusters. In Proc. of the IPPS'98 Workshop on Job Scheduling Strategies for Parallel Processing, pages 231--256. Springer LNCS #1459, 1998.

....related to local task control, as performed by the NLS. Depending on the level of inter NLS synchronization, different scheduling variations are possible. Explicit gang scheduling systems always run all the tasks of a job simultaneously [5 10, 17, 21] In contrast, communication driven systems [3, 19, 20] are more loosely coupled, and schedule tasks based on message arrival. Independent of the inter task scheduling approach, we address those cases in which all processes within a task are closely coupled, like the csh example of the first paragraph. In these situations, all processes of a task must ....

P. G. Sobalvarro, S. Pakin, W. E. Weihl, and A. A. Chien. Dynamic Coscheduling on Workstation Clusters. In IPPS'98 Workshop on Job Scheduling Strategies for Parallel Processing, March 1998.

....CVM problem. In Section 4, we then fully investigated implications from our results and the technique we used to reduce message delay. Finally, we presented our conclusion for similar polling structures and execution environments in Section 5. Previous studies of the message notification delay [2, 3] reduce its effect by applying special algorithms to re schedule and synchronize incoming messages with polling processes. Many target message passing parallel programs, in which synchronization delay is highly important. Significant improvements can be achieved with these approaches, but they ....

P.G. Sobalvarro, et al. Dynamic coscheduling on workstation clusters.in Proceedings of the Workshop on Job Scheduling Strategies for Parallel Processing. March 1998.

....to reduce messages waiting time and make good use of the idle CPU cycles when distributed applications are executed in a cluster or NOW system. Coscheduling decisions are made taking implicit runtime information of the jobs into account, basically execution CPU cycles and communication events [12, 13, 14, 15, 16, 22]. Our framework will be focused on an implicit coscheduling environment, such as scheduling the correspondents the most recent communicated processes in the overall system at the same time, taking into account both high message communication frequency and low penalty introduction into the delayed ....

....messages into account. For this reason, if the task to be inserted on the RQ has any incoming message in the RMQ queue (task h :rec 6= 0) and the main memory in such node is overloaded (Nk:mem Nk:M ) the inserted task is led to the top of the RQ (RQ[0] Thus, CSM applies a dynamic technique [15] to ensure that fine grained distributed applications are coscheduled. In this technique, the more scheduling priority is assigned to tasks the more the receiving frequency is. For the rest of the cases, the SCHED COND function together with the STARV COND one will establish the task ordering in ....

P.G. Sobalvarro, S. Pakin, W.E. Weihl and A.A. Chien. "Dynamic Coscheduling on Workstation Clusters". IPPS'98 Workshop on Job Scheduling Strategies for Parallel Processing, 1998.

....scheduling techniques strive to coschedule jobs that communicate frequently on massively parallel (MPP) systems conglomerated from single threaded processors. 22] describes a system that dynamically coschedules jobs that communicate frequently to increase system utilization and job response time. [25] improves upon gang scheduling to dynamically produce emergent coscheduling of the processes constituting a parallel job. 21] improves upon gang scheduling to fill holes in utilization around gang scheduled jobs with pieces of work from jobs that do not require all resources in order to to make ....

P. Sobalvarro, S. Pakin, W. Weihl, and A. Chien. Dynamic coscheduling on workstation clusters. In SRC Technical Note

....and other new distributed object technologies can only cause the importance of reducing notification delay to increase. Furthermore, our findings bode well for the use of non dedicated networks of workstations to run distributed applications, and complement recent work on emergent co scheduling [25]. 9. ....

P. G. Sobalvarro, S. Pakin, W. E. Weihl, and A. A. Chien, "Dynamic coscheduling on workstation clusters," in Proceedings of the Workshop on Job Scheduling Strategies for Parallel Processing, March 1998.

....which is tolerable even for such a short quantum. 5. Related Work Other clusters using user level communication have also dealt with the interaction between the user level communication and the process scheduling. FM specifically was used as the platform to investigate dynamic coscheduling [12]. The idea here is that instead of using gang scheduling, processes will be co scheduled on the different nodes only if this is warranted by the interactions between them. This was implemented based on a modification to FM so that incoming messages would trigger the scheduling of the processes to ....

P. G. Sobalvarro, S. Pakin, W. E. Weihl, and A. A. Chien, "Dynamic coscheduling on workstation clusters ". In Job Scheduling Strategies for Parallel Processing, pp. 231--256, Springer Verlag, 1998. LNCS vol. 1459. 9

....information is identical to a covert channel [24] in the field of security. Distributed systems that have leveraged this type of implicit information include the TCP congestion control algorithm [21] which observes packet loss to infer congestion, and the coordinated scheduling of parallel jobs [3, 35], which infers remote scheduling state from message arrivals and round trip time. The central advantage of implicit methods is that 3 they provide information for free no communication beyond the requisite data messages is required, only the ability to deduce remote behavior from a local ....

P. G. Sobalvarro, S. Pakin, W. E. Weihl, and A. A. Chien. Dynamic Coscheduling on Workstation Clusters. In Proceedings of the IPPS '98 Workshop on Job Scheduling Strategies for Parallel Processing, 1998.

....and then repeat. Each pattern in the figure is a different job, white space is packing loss. New jobs form new rows if they don t fit into existing ones, and departing jobs will cause rows to be collapsed together if possible. implement forms of time slicing in NOW environments [DAC96, ADCM98, SPWC98] Time slicing techniques can be distinguished by whether they use a global queue for the entire system, or local queues on each processor. Local queues are more common in distributed memory machines, because of the communication overhead involved in maintaining a shared global queue. A major ....

....large parallel computers. For this reason, they have been touted as the parallel computing platform of the future [ACP95] There is certainly evidence that parallel computing is feasible on this type of platform [CADAD 97] Much research [ADCM98, DAC96, ADC97, WZ97, KL97, LS93, ADV 95, SPWC98, DZ97, AS97, AFKT98] has been devoted to determining how to schedule parallel applications effectively in a NOW environment. In theory a NOW could be as powerful as a large, tightly coupled parallel machine, but there are some problems with the architecture that tend to limit performance. The ....

[Article contains additional citation context not shown here]

Patrick G. Sobalvarro, Scott Pakin, William E. Weihl, and Andrew A Chien. Dynamic Coscheduling on Workstation Clusters. In Proceedings of the IPPS '98 Workshop on Job Scheduling Strategies for Parallel Processing, volume 1459 of Lecture Notes in Computer Science, pages 231--256, New York, NY, USA, March 1998. Springer-Verlag Inc.

....methods with some simple schedulers, which are described in sections 2.5 2.7. 2. 2 Dynamic and Implicit Coscheduling Dynamic coscheduling (DCS) was proposed as a way to reduce the amount of global synchronization of coscheduling, and thus decrease the system s overhead and improve its scalability [19, 26, 27]. The philosophy behind DCS is demand based coscheduling: u coordination is achieved by observing the communication between processes, and not by a master daemon, u Communication between processes is used to deduce which processes should be coscheduled and to e ect coscheduling. Thus, when a ....

....assumption: if a process receives an incoming message, its peer(s) must be currently running on other nodes, so the process should be prioritized for immediate execution. Several aspects and parameters can be tuned for di erent assumptions or workloads, e.g. when to make a preemption decision [26]. The design of the testbed should allow room for con guration and tuning of these parameters. Like DCS, Implicit coscheduling (ICS) makes local scheduling decisions based on monitoring communication activity [5, 6] Processes waiting for a communication action to complete use a spin block ....

[Article contains additional citation context not shown here]

Patrick Sobalvarro, Scott Pakin, William E. Weihl, and Andrew A. Chien. Dynamic Coscheduling on Workstation Clusters. In Dror G. Feitelson and Larry Rudolph, editors, Job Scheduling Strategies for Parallel Processing, volume 1459 of Lecture Notes in Computer Science, pages 231256. Springer-Verlag, 1998.

....can then be used by the requestor to distribute the needed code, data, and control information to the various machines and to initiate processing. The actual execution of user jobs is not a part of this research. Thus issues such as process distribution mechanisms and the [co ]scheduling (e.g. [28, 4, 30]) of processes across cluster nodes are not discussed. An allocation of a set of machines to create a dynamic cluster that will execute a job at some future time represents a resource commitment. It is important to distribute such commitment information to all A.R.s that could schedule work onto ....

P. Sobalvarro, S. Pakin, W. E. Weihl, and A. A. Chien. Dynamic Coscheduling on Workstation Clusters. In Proc. Intl. Parallel Processing Symposium, 1998.

No context found.

Sobalvarro PG, Pakin S, Weihl WE, Chien AA. Dynamic coscheduling on workstation clusters. Proceedings of the Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 1998 (Lecture Notes in Computer Science, vol. 1459). Springer: Berlin, 1998.

No context found.

Patrick G. Sobalvarro, Scott Pakin, William E. Weihl, and Andrew A Chien. Dynamic Coscheduling on Workstation Clusters. In Proceedings of the IPPS '98 Workshop on Job Scheduling Strategies for Parallel Processing, 1998.

No context found.

Patrick G. Sobalvarro, Scott Pakin, William E. Weihl, and Andrew A. Chien. Dynamic Coscheduling on Workstation Clusters. Lecture Notes in Computer Science, 1459:231--257, 1998.

No context found.

P. G. Sobalvarro, S. Pakin, W. E. Weihl, and A. A. Chien. Dynamic coscheduling on workstation clusters. Lecture Notes in Computer Science, 1459:231--??, 1998.

First 50 documents

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