Tucker A, Gupta A. Process control and scheduling issues for multiprogrammed shared-memory multiprocessors. Proceedings of the 12th ACM Symposium on Operating System Principles, December 1989; 159--166.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....with little or no modification to existing multiprocessor operating systems [17, 35] Unfortunately, for some job mixes, coscheduling is not appropriate. For example, a job mix consisting of one parallel computation and one serial computation cannot be coscheduled efficiently. With process control [36], processors are dynamically partitioned among the running computations so that each computation runs on a set of processors that grows and shrinks over time, and each computation creates and kills processes so that the number of processes matches the number of processors. We are not aware of any ....

Andrew Tucker and Anoop Gupta. Process control and scheduling issues for multiprogrammed shared-memory multiprocessors. In Proceedings of the Twelfth ACM Symposium on Operating Systems Principles, pages 159-- 166, Litchfield Park, Arizona, December 1989.

....wakes up and applies the scheduling policy distributing processors among parallel applications. Once the processor allocation has been decided, the scheduler enforces it by suspending or resuming the applications processes. Our scheduler implements the process control approach proposed in [Tucker89]. The scheduler informs applications about the number of processors assigned to each one and applications are in charge of adapting their parallelism to their current allocation. The arrival of the applications is controlled by a long term scheduler, the queueing system. This queueing system ....

A. Tucker, A. Gupta, "Process control and scheduling issues for multiprogrammed shared-memory multiprocessors", in 12th Symp. Operating Systems Principles. pp. 159-166, dec. 1989.

.... a multiprocessor scheduler should solve a similar problem how to coschedule threads on different processors to maximize efficiency in the face of bottlenecks on shared system resource (such as main memory or communication fabric) Chapin [7] emphasizes load balancing, as does Tucker and Gupta [30]; the idea is migrate threads to under utilized processors. Others have concentrated on keeping the cache warm by favoring the mapping of threads to processors where they have executed before [6] 29] 33] Coscheduling on traditional single threaded architectures often leads to increased ....

A. Tucker and A. Gupta. Process control and scheduling issues for multiprogrammed shared memory multiprocessors. In Symposium on Operating Systems Principals, December 1989.

....on the overall demand for processors. A server with this policy can notify applications to reconfigure when it changes the number of processors available. Researchers at Stanford are pursuing this approach and have implemented a server for this scheduling policy under Mach with good initial results[51]. Programs that use more virtual processors (e.g. Mach threads) than physical processors need support for managing the resulting system concurrency. An important problem faced by such programs is interacting (communicating or synchronizing) with a virtual processor that isn t currently executing ....

.... benefits of limiting the parallel execution demanded by applications to the available hardware parallelism suggests that a multiprocessor operating system like Mach can expect to spend much of its time in a condition where the number of threads and the number of processors roughly correspond [51, 54]. This makes the performance of the run queue per processor solution unacceptable for a multiprocessor operating system. This and the performance benefits of the idle processor queue (when there are fewer threads than processors) suggest that the global run queue plus idle queue alternative used ....

[Article contains additional citation context not shown here]

Andrew Tucker and Anoop Gupta. Process Control and Scheduling Issues for Multiprogrammed Shared Memory Multiprocessors. In Proceedings of the Twelfth ACM Symposium on Operating Systems Principles, pages 159--166, December 1989.

....the same problems occur with multiprogramming as with I O and page faults. If there is only one job in the system, it can receive all of the machine s processors; if another job enters the system, the operating system should preempt some of the first job s processors to give to the new job [Tucker Gupta 89] The kernel then is forced to choose which of the first job s kernel threads, and thus implidtly which of the user level threads running in their context, to run on the remaining processors. The need to preempt processors from an address space also occurs due to variations in parallelism within ....

Tucker, A. and Gupta, A. Process Control and Scheduling Issues for Multipro- grammed Shared Memory Multiprocessors. In Proceedings of the 12th A CM Symposium on erating Systems Principles, pages 159-166, De- cember 1989. 108

....in message passing systems migrate tasks from nodes with high processor load to neighboring nodes with lower processor load [2] 3] Migration decisions in these systems ignore the issue of data locality. Some migration strategies in shared memory multiprocessors, such as central ready queues [7], also focus solely on processor load. Others, however, incorporate the benefits of co located data into scheduling decisions. These approaches explore the benefits of not migrating threads whose data is resident in the local cache [6] 8] and or local memory [1] 4] Research in data migration ....

A. Tucker and A. Gupta, "Process Control and Scheduling Issues for Multiprogrammed, Shared Memory Multiprocessors," in Proc. of the 12th Symp. on Operating Systems Principles, Litchfield Park, AZ, pp. 159166, Dec. 1989.

....find practical algorithms that have good analytic properties. We study the simple Equi partition algorithm, for which an equal number of processors is assigned to every job. The approach of Equi partition was first introduced to parallel scheduling by Tucker and Gupta as a process control policy [24], and later modified by Zahorjan and McCann [27] to dynamically adjust processor allocations as job requirements for processors change. This algorithm is known as dynamic equi partition (DEQ) We show that the Equi partition algorithm (which performs at most n preemptions) achieves a performance ....

....can be achieved with few reassignments of processors. Of course, much more research is required to make this theoretical understanding useful in a practical setting. In fact, performance in such systems has been already studied using simulation, experimental, and queuing theoretical approaches [2, 13, 17, 18, 24, 27, 1]. In this perspective, our research constitutes a theoretical confirmation of these efforts. 1.1 Modeling Job Execution In our model, all jobs arrive at time zero. That is, we adopt a batch job processing model. It would be more general to allow jobs to arrive at arbitrary times. However, this ....

A. Tucker and A. Gupta. Process control and scheduling issues for multiprogrammed shared-memory multiprocessors. In Proceedings of the Twelfth ACM Symposium on Operating Systems Principles, pages 159--166, 1989.

....or when all the jobs are fully parallelizable, or when the scheduler has considerable knowledge about the jobs. First let us consider the results on batch jobs, i.e. all jobs arrive at time zero. Motwani et al. 18] prove that on fully parallelizable jobs, the scheduler Equi partition algorithm [24], which partitions the processors evenly between the unfinished jobs and preempts only when jobs complete, has mean response time within two of optimal. They also prove that no non clairvoyant scheduler has a better competitive ratio. Turek et al. 25] consider jobs with a single phase of an ....

....) This gives ffi ffiT , which is the length of time for EQUI s to complete the new fully parallelizable phase with work fi ffiT . 11 Open Problems The performance of Equi partition has been studied extensively using simulation, experimental, and queuing theoretical approaches [3, 13, 16, 17, 24, 28, 2]. Our research constitutes a theoretical confirmation of these efforts. The main open problem is to close the gaps between the lower bounds on the competitive ratio known for general non clairvoyant schedulers and those known for the specific schedulers Equi partition and Balance. This gap is ....

A. Tucker and A. Gupta. Process control and scheduling issues for multiprogrammed shared-memory multiprocessors. In Proceedings of the Twelfth ACM Symposium on Operating Systems Principles, pages 159--166, 1989.

....or when all the jobs are fully parallelizable, or when the scheduler has considerable knowledge about the jobs. First let us consider the results on batch jobs, i.e. all jobs arrive at time zero. Motwani et al. 18] prove that on fully parallelizable jobs, the scheduler Equi partition algorithm [24], which partitions the processors evenly between the unfinished jobs and preempts only when jobs complete, has mean response time within two of optimal. They also prove that no non clairvoyant scheduler has a better competitive ratio. Turek et al. 25] consider jobs with a single phase of an ....

....n) However, the batch model assumes that the number of jobs n is at most the number of processors p because all these jobs arrive and are executed at once. Because our jobs arrive at arbitrary times, n is assumed to be much bigger then p. 28 13 Open Problems and queuing theoretical approaches [3, 13, 16, 17, 24, 28, 2]. Our research constitutes a theoretical confirmation of these efforts. The main open problem is to close the gaps between the lower bounds on the competitive ratio known for general non clairvoyant schedulers and those known for the specific schedulers Equi partition and Balance. This gap is ....

A. Tucker and A. Gupta. Process control and scheduling issues for multiprogrammed shared-memory multiprocessors. In Proceedings of the Twelfth ACM Symposium on Operating Systems Principles, pages 159--166, 1989.

....jobs may tolerate the perturbations caused by I O, that I O bound jobs suffer under gang scheduling, and therefore argue in favor a flexible gang scheduling. Several studies have revealed that EQUI does very well, even when some moderate charge for the overhead of frequent preemptions is made [86,48]. Squillante [84] provides an analysis of the performance of dynamic partitioning. Deng et al. show that EQUI is optimally competitive [11] Dussa et al. 14] compares space slicing against no partitioning, and finds that space partitioning pays off. Knowledge Based Scheduling Majumdar, Eager and ....

A. Tucker and A. Gupta, "Process control and scheduling issues for multiprogrammed shared-memory multiprocessors". In 12th Symp. Operating Systems Principles, pp. 159--166, Dec 1989.

....thesis 9 olent Bandit Laboratory (BBL) 40] a PC based system. The BBL system architecture is closely related to Cilk NOW s. The Prospero resource manager [78] also employs a similar system architecture. Runtime systems for the programming language COOL [28] running on symmetric multiprocessors [101, 102] and cache coherent, distributed, shared memory machines [26, 69] use process control to support adaptive parallelism. These systems rely on special purpose operating system and hardware support. In contrast, Cilk NOW supports adaptive parallelism entirely in user level software on top of ....

Andrew Tucker and Anoop Gupta. Process control and scheduling issues for multiprogrammed shared-memory multiprocessors. In Proceedings of the Twelfth ACM Symposium on Operating Systems Principles (SOSP 12), pages 159--166, Litchfield Park, Arizona, December 1989.

.... a multiprocessor scheduler should solve a similar problem how to coschedule threads on different processors to maximize efficiency in the face of bottlenecks on shared system resource (such as main memory or communication fabric) Chapin [7] emphasizes load balancing, as does Tucker and Gupta [31]; the idea is to migrate threads to underutilized processors. Others have concentrated on keeping the cache warm by favoring the mapping of threads to processors where they have executed before [6] 30] 34] Coscheduling on traditional single threaded architectures often leads to increased ....

A. Tucker and A. Gupta. Process control and scheduling issues for multiprogrammed shared memory multiprocessors. In Symposium on Operating Systems Principals, Dec. 1989.

....time, processes may be blocked while waiting for a preempted process or may be required to context switch after every synchronization operation. Several studies have shown that this effect can lead to severe performance degradation [Leutenegger, 1990; Lo and Gligor, 1987a; Lo and Gligor, 1987b; Tucker and Gupta, 1989] Many operating systems support a single centralized ready queue from which processes are dispatched according to their priority. This approach is very popular on small scale UMA machines. However, Squillante and Lazowska [1990] have shown that by ignoring the affinity that may have been ....

....may communicate without an in tervening context switch. There are also disadvantages to coscheduling. If there are several applications in the system, the machine must cycle through each of them, during which time the caches can be expected to lose any contents related to an earlier execution [Tucker and Gupta, 1989]. Second, utilization may suffer if applications have a variable amount of parallelism, or if processes cannot be evenly assigned to time slices of the machine. When hardware partitions are used, no two applications share a processor. A set of processors may be dedicated to an application for a ....

[Article contains additional citation context not shown here]

A. Tucker and A. Gupta, "Process Control and Scheduling Is- sues for Multiprogrammed Shared-Memory Multiprocessors," In Proceedings of the 12th Symposium on Operating Systems Principles, pages 159-166, December 1989.

....dispatching implies attaching a user thread to a kernel thread, and blocking implies detaching. This solution is clearly superior to changing the kernel since such a threads package can provide several other benefits such as a mechanism for allocating processors in a multiprogrammed environment [Tucker89], a large number of user threads [Golub90] and light weight threads [Powell91] We also made measurements using two real parallel applications: A thinning program for digital patterns, and a parallel merge sort program. For the thinning program, the use ofmultiplexed threads reduced the running ....

....doubleword of this thread s footprint; Wait for barrier synchronization. END repeat loop END parallel code Print out time spent in parallel section. Figure 3.0. The synthetic benchmark program. Several papers have described the use and implementation of multiplexed threads packages: [Tucker89] for efficient scheduling of nmltiprogranmed, multi threaded programs; Golub90] to provide a large number of threads per application; Powell91] for controlling concurrency and for light weight. Here we use the multiplexed user threads for successful exploitation of affinity. 3 Kernel level ....

A. Tucker and A. Gupta, "Process Control and Scheduling Issues for Multiprogrammed Shared- Memory Multiprocessors," Proc. of the 12th ACM Syrup. on Operating Systems Principles, December 1989.

.... (in which the scheduler attempts to ensure that all processes of a given application run simultaneously) and processor partitioning (in which each application is given exclusive use of a subset of the processors of the machine for an extended period of time) A growing body of evidence [2, 5, 10, 15, 16] suggests that throughput is maximized by a processor partitioned environment in which each application runs exactly one process per processor. In such an environment, or in one that employs coscheduling, all the processes that share a given lock will always run simultaneously. Unfortunately, this ....

A. Tucker and A. Gupta. Process Control and Scheduling Issues for Multiprogrammed SharedMemory Multiprocessors. In PROC of the Twelfth SOSP, pages 159-166, Litchfield Park, AZ, 3-6 December 1989.

....of various scheduling algorithms. 1 Introduction In recent years massively parallel processors (MPP) computers have made a significant presence. With this growth in MPPs, a number of researchers have developed and are continuing to develop various job scheduling subsystems for these MPPs [1, 2, 3, 4, 5, 6]. During the development of these schedulers and their related algorithms, it is important to have an accurate characterization of the workloads experienced by the MPPs. It is extremely advantageous to have these workloads characterized by a compact model that is representable by a few parameters, ....

....these moments are given in table 3. rives at an analytical expression for the cdf of the models used in this paper, namely cdf(x j ) 1 Gamma 2 X i=1 p i e Gamma i x j in Gamma1 X k=0 (x j i ) k k j (20) From this expression of the cdf and a uniform random number generator, U[0,1], the random variables fx j g that are commensurate with equation 5 can be obtained by solving for x j in equation 19. Since a cdf(x) is a monotonic function that is bounded by 0 pmin pmax 1 2 3 of jobs 1 322 9.126e 04 2.601e 11 2.255e 18 100.00 1 1 1.537e 04 6.051e 08 2.941e 13 40.43 2 2 ....

A. Tucker and A. Gupta, "Process control and scheduling issues for multiprogrammed shared-memory multiprocessors," in Proceedings of the 12th. ACM Symposium on Operating Systems Principle, pp. 159--166, 1989.

....wakes up and applies the scheduling policy distributing processors among parallel applications. Once the processor allocation has been decided, the scheduler enforces it by suspending or resuming the applications processes. Our scheduler implements the process control approach proposed in [Tucker89], which assumes that applications perform better when the number of running threads is the same as the number of processors. The scheduler informs applications about the number of processors assigned to each one and applications are in charge of adapting their parallelism to their current ....

....The only information provided by the application is the maximum number of processors that it can use. Reallocations are done at job arrival and completion. Equipartition is considered a quasi static policy. The equipartition is based on the process control policy proposed by Tucker and Gupta in [Tucker89]. The equip algorithm has been implemented following the algorithm showed in Figure 2. 3.2 Equal efficiency The goal of the equal efficiency [NguyenZV96] is to maximize the system efficiency. The idea is to allocate more processors to those applications that have better efficiency and less ....

A. Tucker, A. Gupta, "Process control and scheduling issues for multiprogrammed shared-memory multiprocessors", in 12th Symposium Operating Systems Principles. pp. 159-166, dec. 1989.

....oblivious to your random choices) can match your moves all the time. The question then is: Which is the best randomized scheduling policy 3.7. 4 Re Allocating PEs An oft suggested technique to break the adversary s stranglehold is to change job PE allocations in the middle of execution [MVZ93, TG89] This allows the algorithm to adapt to changing load conditions and to recover from poor decisions. 66 For example, if a job terminates, other jobs can be expanded to use the free PEs. Conversely, if a new job arrives, existing jobs can be squeezed to free some PEs. We did not consider ....

.... fact, what if nothing is known about them In that case, most papers from academia suggest dynamic space slicing: jobs are space sliced, and as soon as a job terminates the remaining jobs are re allocated to make use of the free PEs [NSS93, R 94, Sev89, ST93, EZL89, SHCV94, LV90, MVZ93, MZ95, TG89, D 95] Unfortunately, in most cases, this result is obtained by facilely neglecting the exorbitant cost of re allocating jobs (see Section 3.7.4) Industry practices gang scheduling: On the other hand, the industry has always leaned towards gang scheduling. Most vendors already support gang ....

A. Tucker and A. Gupta. Process control and scheduling issues for multiprogrammed shared-memory multiprocessors. In Symposium on Operating System Principles, pages 159--166, December 1989.

.... a multiprocessor scheduler should solve a similar problem how to coschedule threads on different processors to maximize efficiency in the face of bottlenecks on shared system resource (such as main memory or communication fabric) Chapin [7] emphasizes load balancing, as does Tucker and Gupta [30]; the idea is migrate threads to under utilized processors. Others have concentrated on keeping the cache warm by favoring the mapping of threads to processors where they have executed before [6] 29] 34] Coscheduling on traditional single threaded architectures often leads to increased ....

A. Tucker and A. Gupta. Process control and scheduling issues for multiprogrammed shared memory multiprocessors. In Symposium on Operating Systems Principals, Dec. 1989.

No context found.

Tucker A, Gupta A. Process control and scheduling issues for multiprogrammed shared-memory multiprocessors. Proceedings of the 12th ACM Symposium on Operating System Principles, December 1989; 159--166.

No context found.

A. Tucker and A. Gupta. Process control and scheduling issues for multiprogrammed shared-memory multiprocessors. In Proceedings of the twelfth ACM Symposium on Operating systems principles, pages 159-166. ACM Press, 1989. , TR-001-03, 2003 Jan Hungershofer et al.

No context found.

A. Tucker, A. Gupta, "Process control and scheduling issues for multiprogrammed shared-memory multiprocessors", in 12th Symp. Operating Systems Principles. pp. 159-166, dec. 1989.

No context found.

A. Tucker, A. Gupta, "Process Control and Scheduling Issues for Multiprogrammed Shared-Memory Multiprocessors", in 12th Symp. Operating Systems Principles. pp. 159-166, dec. 1989.

No context found.

A. Tucker and A. Gupta. "Process Control and Scheduling Issues for Multiprogrammed SharedMemory Multiprocessors". In Proceedings of the 12th Symposium on Operating Systems Principles, pages 159-166, December 1989.

No context found.

A. Tucker and A. Gupta. "Process Control and Scheduling Issues for Multiprogrammed Shared- Memory Multiprocessors". In Proceedings of the 12th Symposium on Operating Systems Principles, pages 159-166, December 1989.

First 50 documents  Next 50

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