Ousterhout JK. Scheduling techniques for concurrent systems. Proceedings of the 3rd International Conference on Distributed Computer Systems, 1982; 22--30.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the addition of only a few lines of simple code, hence this work should have an immediate practical impact. 1 Introduction Ousterhout proposed the notion of co scheduling to efficiently support interprocess synchronization 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 ....

....all classes by reporting mean slowdowns and response times. We provide numerous simulation results to make a case for proper quantum allocation. In addition to considering the DHC algorithm, we consider the effect of different quantum allocations with simpler gang scheduling schemes such as Matrix [22] and LRS [5] These simpler algorithms are especially relevant since many current production level schedulers use variants of these simpler algorithms. Finally, our work also provides a more exhaustive comparison of DHC with Matrix and LRS by considering more workloads than considered in [5] and ....

[Article contains additional citation context not shown here]

J. Ousterhout. Scheduling techniques for concurrent systems. In Proc. of Distributed Computing Systesms Conference, pages 22--30, 1982.

....but it fails on multiprogrammed systems, in which a neighbor thread may be preempted, and thus unable to cooperate. Theproblem of preemption in critical sections has received considerable attention over the years. Alternative strategies include avoidance [6, 10, 14] recovery [3, 4] and tolerance [9, 16]. The latter approach is appealing for commercial applications because it does not require modification of the kernel interface: if a thread waits too long for a lock, it assumes that the lock holder has been preempted. It abandons its attempt, yields the processor to another thread (assuming ....

....in scalable queue based locks makes spin locks a viable mechanism for user level synchronization on large multiprogrammed systems. In future work we hope to evaluate our algorithms in the context of commercial OLTP codes. We also plan to develop variants that block in the scheduler on timeout [9, 16], cooperate with the scheduler to avoid preemption while in acritical section [6, 10] or adapt dynamically between test and set and queue based locking in response to observed contention [11] In a related vein, we are developing a tool to help verify the correctness of locking algorithms by ....

J. K. Ousterhout. Scheduling Techniques for Concurrent Systems. In Proceedings of the Third International Conference on Distributed Computing Systems,pages 22--30, Miami/Ft. Lauderdale, FL, October 1982.

....nodes at different time due to independent scheduling, communication time and process response time will be more than scheduling them in same time. Literature merely provides many algorithms because in difficulty to dynamically determine the interprocess communication patterns. Ousterhout [12] proposed several algorithms based on the concept of Co scheduling, which takes takes interprocess communication patterns into account while scheduling to ensure that all members of a group run at the same time. It puts the processes in different group by considering that intragroup communication ....

J. K. Ousterhout*, "Scheduling techniques for concurrent systems", Proc. 3rd International Conference on Distributed Computing Systems, IEEE, pp. 22-30, 1982.

....for scientific computations. In practice, a NOW system is heterogeneous and nondedicated. These two unique factors make scheduling policies on multiprocessor multicomputer systems not suitable for NOWs. Since the nature of parallel processing on NOWs does not change, the coscheduling [13] principle is still an important basis for parallel process scheduling on NOWs. Thus, heterogeneity, job interactions, and coscheduling are three major concerns in our design. Many research groups currently are using homogeneous NOWs as experimental platforms. In practice, more and more ....

....execution times of the process was an additional parameter that was considered. We assumed the system calls were exponentially distributed in the lifetime of each local process. The scheduling policies in the simulator included the SVR4 local scheduling, coscheduling using the matrix scheme in [13], and the self coordinated local scheduling scheme based on SVR4 which is discussed in Section 4. 5.2. Precision of Power Preservation Using the simulator, we first studied the effects of the quantum in PJ1 and maxwait in PJ0 on the precision of power preservation of a workstation. We first ....

Ousterhout, J. Scheduling techniques for concurrent systems. Proceedings of the 3rd International Conference on Distributed Computing Systems. Oct. 1982, pp. 22--30.

....or particularly exact one. The interested reader is referred to the original papers wherever applicable. Note: during the rest of this paper, the terms CPU (central processing unit) and PE (processing element) are used interchangeably 2. 1 Introduction to Gang Scheduling Gang scheduling [8, 15, 20] was introduced as an efficient way to multiprogram jobs on a parallel or distributed computer by coscheduIing processes (running all processes of a parallel job concurrently) The basic idea behind gang scheduling is that application can not only space share the compute resources, but also ....

....perform the communication, and possible to read the next line of the workload file, before going back to sleep. It can also be awaken by a message from one of the NMs, announcing the termination of a process. For initial allocation of resources, we intend to use the buddy tree allocation algorithm [20]. The ParPar scheduler has a software module that can be use for this puprose with relative ease. It s generic design allows for plugging into different schedulers and use various allocation schemes. It is important though that we keep its internal data structures updated by forwarding to it ....

J. K. Ousterhout. Scheduling Techniques for Concurrent Systems. Proceedings of Third International Conference on Distributed Computing Systems, 1982.

....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 ....

J. K. Ousterhout, "Scheduling techniques for concurrent systems," International Conference on Distributed Computing Systems, pp. 22--30, 1982.

....nodes to be left idle when waiting for more nodes to become available. This can be reduced by using backfilling, in which jobs may be selected out of order provided this will not delay the execution of jobs which have a higher location in the queue [13] Another approach, called gang scheduling [14], slices the time in a round robin manner. The jobs are packed into a matrix. Each column in the matrix represents a node and each row represents a time slot. Jobs assigned to different rows are executed one after the other. One of the problems A preliminary version appeared in JPDPS 2001. ....

....framework for paired gang scheduling is depicted in Figure 2. We use a centralized gang scheduler, as is used in many conventional implementations [4, 9, 5, 8] This is a user level daemon process that runs on a distinguished node in the cluster. It maintains information in an Ousterhout matrix [14], in which rows denote scheduling slots and columns represent processors. In strict gang scheduling, one row is selected each time. Each node scheduler is then directed to schedule the process in the respective cell of the chosen row. This is typically done by making only this process runnable, ....

[Article contains additional citation context not shown here]

Ousterhout J. K., Scheduling techniques for concurrent systems. 3rd Intl. Conf. Distributed Comput. Syst., pp. 22--30, Oct 1982.

....(heartbeat messages) that are implemented by an efficient hardware multicast. For the second goal, the daemons were designed so that modules for different scheduling algorithms can be plugged into them. In this paper, we focus on one of the most popular of these algorithms, gang scheduling (GS)[8, 20]. GS employs both space sharing and time sharing to allocate resources to jobs. All the processes of a given job run in the same allotted time slot, for the duration of the timeslice quantum, and are then context switched to a different job in a cyclic manner, at the end of each time slot. 2.1 ....

J. K. Ousterhout. Scheduling Techniques for Concurrent Systems. Proceedings of Third International Conference on Distributed Computing Systems, pages 22--30, 1982.

....framework for paired gang scheduling is depicted in Figure 2. We use a centralized gang scheduler, as is used in many conventional implementations [8, 11, 9, 10] This is a user level daemon process that runs on a distinguished node in the cluster. It maintains information in an Ousterhout matrix [13], in which rows denote scheduling slots and columns represent processors. In strict gang scheduling, one row is selected each time. Each node scheduler is then directed to schedule the process in the respective cell of the chosen row. This is typically done by making only this process runnable, ....

....with p end p start = p start 1 Figure 3: The matching algorithm (indentation is used to delimit blocks) calculates the average CPU utilization of all processes in the job. This is calculated anew at the end of each quantum in which the job ran. Scheduling is based on an Ousterhout matrix [13], with the jobs in successive slots scheduled in turn in a round robin manner. But as the job in the next slot is considered, the master will look for a match for this job, and schedule both of them. Let us assume the job to be scheduled now (the one in the next slot) has a predicted x CPU ....

Ousterhout J. K., Scheduling Techniques for Concurrent Systems. 3rd Intl. Conf. Distributed Comput. Syst., pp. 22--30, Oct 1982.

....is used, with all the processes in a gang being preempted and rescheduled at the same time. Time slicing is obtained using a coordinated multi context switch, which occurs at regular intervals of time, called the timeslot quantum. The following are important issues regarding gang scheduling [28]: Effect of timeslice on overhead Smaller timeslices yield better response time at the cost of decreased throughput (due to scheduling overhead that cannot be amortized) In Section 3.2.1, we show that STORM s scheduling overhead is so low that it can support workstation time quanta with ....

....Gigabit Ethernet [36] 46 log n Not available Myrinet [5, 8, 9] 20 log n 15n Infiniband [25] 20 log n Not available QsNET (see Section 3. 3) 10 150n BlueGene L [18] 2 700n Generality of Mechanisms Currently, STORM supports batch scheduling with and without backfilling, gang scheduling [28], and implicit coscheduling [2] However, we believe that STORM s mechanisms are sufficiently general as to be used for an efficient implementation of a variety of schemes. We initially plan to implement a number of additional coordinated scheduling algorithms (such as buffered coscheduling ....

J. K. Ousterhout. Scheduling techniques for concurrent systems. Proceedings of the Third International Conference on Distributed Computing Systems, pages 22--30, October 1982.

....of service to all the system s users. This allocation of resources is done by a central operating system which runs on a host that manages the whole system. The system scheduler determines when and on which nodes a job will be executed. One way of scheduling parallel jobs is gang scheduling [10, 5]. The idea is to map the threads of a parallel job to distinct processors, and then schedule them to be executed simultaneously on their respective processors. Time slicing is used for interactive response times, and this is coordinated across the processors. Most studies find gang scheduling to ....

....idea is that the processes of a parallel job are mapped to distinct processors, and are then scheduled to execute simultaneously on their respective processors. The mapping of processes to processors uses Ousterhout s matrix algorithm, where columns represent nodes and rows are scheduling slots [10]. Packing of jobs into slots is done using a buddy system, as in the Distributed Hierarchical Control scheme [5, 6] This uses a tree of controllers that preside over nested groups of powerof two processors. After we estimated the memory size of a job, we need to add this information to our ....

J. K. Ousterhout, "Scheduling techniques for concurrent systems ". In 3rd Intl. Conf. Distributed Comput. Syst., pp. 22-- 30, Oct 1982.

....logs, the original estimates provided by users are used, whereas in models the actual runtime is used as an estimate. Gang scheduling is a preemptive scheme in which jobs are assigned to rows of a scheduling matrix, where columns represent the nodes of the system and rows represent time slots [13]. The jobs in each row are scheduled in turn using coordinated context switching across all the nodes. The packing of the matrix is based on a variant of the Distributed Hierarchical Control scheme [5] which uses a buddy system to allocate processors in blocks that are powers of 2. Seven ....

J. K. Ousterhout, "Scheduling techniques for concurrent systems ". In 3rd Intl. Conf. Distributed Comput. Syst., pp. 22--30, Oct 1982.

....in current cluster environments that employ user level messaging (to reduce latencies and improve bandwidth) wherein the operating system is unaware of the task waiting for a message. Multiprocessors have traditionally addressed this problem using a technique called Coscheduling Gang Scheduling [18], wherein tasks of a job are scheduled on their respective nodes during the same time quantum. In this paper, we use the term tasks to refer to the processes (as defined by the native operating system) constituting an application to avoid any ambiguities with stochastic processes. However, gang ....

J.K. Ousterhout. Scheduling techniques for concurrent systems. Proceedings of International Conference on Distributed Computing Systems, 22--30, 1982.

No context found.

Ousterhout JK. Scheduling techniques for concurrent systems. Proceedings of the 3rd International Conference on Distributed Computer Systems, 1982; 22--30.

No context found.

Ousterhout, J. K. Scheduling Techniques for Concurrent Systems. In Third International Conference on Distributed Computing Systems, pp. 22--30, May 1982.

No context found.

J. K. Ousterhout. Scheduling techniques for concurrent systems. 3rd Intl. Conf. on Distributed Computing Systems, Oct. 1982.

No context found.

Ousterhout, J.: Scheduling techniques for concurrent systems. In: Proceedings of the Conference on Distributed Computing Systems. (1982)

No context found.

J. K. Ousterhout, "Scheduling techniques for concurrent systems ". In 3rd Intl. Conf. Distributed Comput. Syst., pp. 22--30, Oct 1982.

No context found.

J. K. Ousterhout. Scheduling Techniques for Concurrent Systems. International Conference on Distributed Computing Systems, pages 22--30, 1982.

No context found.

J. K. Ousterhout, "Scheduling Techniques for Concurrent Systems", In Third International Conference on Distributed Computing Systems, pp. 22-30, 1982.

No context found.

J. K. Ousterhout. Scheduling Techniques for Concurrent Systems. In Third International Conference on Distributed Computing Systems, pages 22--30, 1982.

No context found.

J. K. Ousterhout, Scheduling techniques for concurrent systems, Proceedings of Third International Conference on Distributed Computing Systems, May 1982, pp.20-30.

No context found.

J. K. Ousterhout, Scheduling techniques for concurrent systems, Proceedings of Third International Conference on Distributed Computing Systems, May 1982, pp.20-30.

No context found.

J. K. Ousterhout. Scheduling Techniques for Concurrent Systems. In Proceedings of Third International Conference on Distributed Computing Systems, pages 22--30, 1982.

No context found.

J. K. Ousterhout, "Scheduling Techniques for Concurrent Systems ", In Third International Conference on Distributed Computing Systems, pp. 22-30, 1982.

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