V. Lo, K. Windisch, W. Liu, and B. Nitzberg. Non-contiguous processor allocation algorithms for mesh-connected multicomputers. IEEE Transactions on Parallel and Distributed Computing, 8(7), 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....time. There has been considerable prior research into each of the two topics of i) scheduling of parallel jobs [1,2,6,10,14,19,20,23,24,26] and ii) contiguous node allocation strategies [3,4,5,11,13,27] There have also been a few studies that have considered both these issues in combination [12,15,18]. However, only [15] addresses the impact of contiguous node allocation schemes in conjunction with a job scheduling policy that takes fairness into consideration by use of a FCFS (First Come First Served) scheduling policy. In [15] contiguous and non contiguous node allocation schemes for ....

.... considerable prior research into each of the two topics of i) scheduling of parallel jobs [1,2,6,10,14,19,20,23,24,26] and ii) contiguous node allocation strategies [3,4,5,11,13,27] There have also been a few studies that have considered both these issues in combination [12,15,18] However, only [15] addresses the impact of contiguous node allocation schemes in conjunction with a job scheduling policy that takes fairness into consideration by use of a FCFS (First Come First Served) scheduling policy. In [15] contiguous and non contiguous node allocation schemes for meshconnected systems ....

[Article contains additional citation context not shown here]

V. Lo, K.J. Windisch, W. Liu and B. Nitzberg, "Noncontiguous Processor Allocation Algorithms for Mesh- Connected Multicomputers", IEEE Trans. Parallel and Distributed Systems, Vol.8 pp.712726, Jul. 1997.

....in parallel, resulting in a reduced allocation time. 2.2. Noncontiguous allocation Current communication technologies like wormhole routing enable us to consider noncontiguous allocation schemes, since the number of hops between nodes is not the dominant factor determining message latency [2]. The use of small partitions of free processors scattered in the machine to form larger non contiguous partitions decreases the external fragmentation significantly. However, noncontiguous allocation introduces potential problems due to message contention because the messages occupy more links, ....

V. Loet al. Noncontiguous processor allocation algorithms for mesh-connected multicomputers. IEEE Transactions on Parallel and Distributed Systems, 8(7), July 1997.

....directions allocate processors in parallel, resulting in a reduced allocation time. Current communication technologies like wormhole routing enable us to consider noncontiguous allocation schemes, since the number of hops between nodes is not the dominant factor determining message latency [2]. The use of small partitions of free processors scattered in the machine to form larger non contiguous partitions decreases the external fragmentation significantly. However, noncontiguous allocation introduces potential problems due to message contention because the messages occupy more links, ....

V. Loet al. Noncontiguous processor allocation algorithms for mesh-connected multicomputers. IEEE Transactions on Parallel and Distributed Systems, 8(7), July 1997.

....Hypercube computer, processor allocation, queue delay, topological delay, space sharing 1 Introduction Processor allocation has been an active research area for directly interconnected parallel computers. There proposed allocation algorithms for hypercubes [1] 2] 3] 4] and for meshes [5] 6] 7] 8][9]. They partition the interconnection topology and the corresponding processors and assign subsets of processors to requesting jobs. Multiple jobs share the topological space of the system. The multitasking or space sharing runtime environment gain more attention recently as parallel systems are ....

V.Lo, K.J.Wndisch, W.Liu and B.Nitzberg, "Noncontiguous Processor Allocation Algorithms for Mesh- Connected Multicomputers", IEEE Trans. Parallel and Distributed Systems, Vol.8 pp.712726, Jul. 1997.

....have been proposed for many architectures. For meshconnected parallel computers, these algorithms include the Frame Slide[1] the Two DimensionalBuddy[2] the First Fit[3] the Best Fit[3] the Adaptive Scan[4] the Busy List[5] the Quick Allocation[6] and the non partitioning algorithm[7, 8]. However, such partitioning algorithms have a typical drawback, namely, the blockade situation. Figure 1 illustrates the blockade situation occurring on a mesh connected parallel computer. Since such partitioning algorithms have a first come firstserved (FCFS) policy, their incoming tasks(1 and ....

....and allocates a job to the processor which has the smallest load. We must consider the new partitioning algorithm for this purpose. 7. 3 New partitioning Algorithm We have proposed a new partitioning algorithm[20] which is a combination of a contiguous and a non contiguous partitioning algorithm[7,8]. The new algorithm compensates for the weak points of contiguous and non contiguous partitioning algorithms. It could achieve high processor utilization and quick response. We plan to implement this algorithm on a real machine. However, there may be difficultiies implementing multiple tasks ....

V. Lo, K. Windish, W. Liu, and B Nitzberg. Non-contiguous Processor Allocation Algorithms for Mesh-connected Multicomputers. IEEE Trans. on PARALLEL AND DISTRIBUTED SYSTEMS, 8(7):712--726, 1997.

....have been proposed for many architectures. For meshconnected parallel computers, these algorithms include the Frame Slide[1] the Two DimensionalBuddy[2] the First Fit[3] the Best Fit[3] the Adaptive Scan[4] the Busy List[5] the Quick Allocation[6] and the non partitioning algorithm[7, 8]. However, such partitioning algorithms have a typical drawback, namely, the blockade situation. Figure 1 illustrates the blockade situation occurring on a mesh connected parallel computer. Since such partitioning algorithms have a first come firstserved (FCFS) policy, their incoming tasks(1 and ....

....and allocates a job to the processor which has the smallest load. We must consider the new partitioning algorithm for this purpose. 7. 3 New partitioning Algorithm We have proposed a new partitioning algorithm[20] which is a combination of a contiguous and a non contiguous partitioning algorithm[7,8]. The new algorithm compensates for the weak points of contiguous and non contiguous partitioning algorithms. It could achieve high processor utilization and quick response. We plan to implement this algorithm on a real machine. However, there may be difficultiies implementing multiple tasks ....

W. Liu, V. Lo, K. Windish, and B Nitzberg. Non-contiguous Processor Allocation Algorithms for Distributed Memory Multicomputers. Supercomputing, pages 227-- 236, 1994.

....Zhang and Yan also use a probabilistic model for contention in their performance prediction model. However, both these works assume that each workstation is shared by at most one compute intensive task and one or more local tasks that alternate idle with busy cycles. 1.5. 2 Contention in MPPs In [32], Liu et al. show that non contiguous allocation algorithms for distributed memory multicomputers perform better overall than a contiguous allocation algorithm, even when contention effects on interprocessor communication are considered. Their results indicate that contention for the ....

....3 I O I O S E 31 1 3 2 2 2 2 2 2 222 111 3 48 tion communication, we performed a set of experiments. We wanted to determine the effects of traffic in a link on the time to transfer messages through the same link. We developed a benchmark (similar to the one discussed by Liu et al. in [32]) to show these effects. The strip benchmark executes on a strip with an even number of nodes that communicate in a pairwise way. A strip of nodes is a contiguous partition formed by a set of nodes in the same row or column of the mesh. Figure 3 8 illustrates the strip concept by showing a ....

[Article contains additional citation context not shown here]

W. Liu, V. Lo, K. Windisch, and B. Nitzberg, "Non-Contiguous Processor Allocation Algorithms for Distributed Memory Multicomputers ", in Proceedings of Supercomputing 94, pp. 227-236, 1994.

....Zhang and Yan also use a probabilistic model for contention in their performance prediction model. However, both these works assume that each workstation is shared by at most one compute intensive task and one or more local tasks that alternate idle with busy cycles. 1.5. 2 Contention in MPPs In [32], Liu et al. show that non contiguous allocation algorithms for distributed memory multicomputers perform better overall than a contiguous allocation algorithm, even when contention effects on interprocessor communication are considered. Their results indicate that contention for the ....

....3 I O I O S E 31 1 3 2 2 2 2 2 2 222 111 3 48 tion communication, we performed a set of experiments. We wanted to determine the effects of traffic in a link on the time to transfer messages through the same link. We developed a benchmark (similar to the one discussed by Liu et al. in [32]) to show these effects. The strip benchmark executes on a strip with an even number of nodes that communicate in a pairwise way. A strip of nodes is a contiguous partition formed by a set of nodes in the same row or column of the mesh. Figure 3 8 illustrates the strip concept by showing a ....

[Article contains additional citation context not shown here]

W. Liu, V. Lo, K. Windisch, and B. Nitzberg, "Non-Contiguous Processor Allocation Algorithms for Distributed Memory Multicomputers ", in Proceedings of Supercomputing 94, pp. 227-236, 1994.

....does not become a bottleneck. Another approach is to coordinate the PE allocation using a hierarchical structure. For example, a buddy system can be used where a partition can be composed of a number of blocks from different levels, such that the sum of the allocated PEs is the requested number [375, 606]. Alternatively, one can use a virtual hierarchy of control on the PEs themselves. Such a structure forms the basis of the so called wave scheduling mechanism developed for the MICROS distributed operating system [595, 596] It guarantees that the allocated PEs are close to each other in the ....

.... allocations are also squares with sides that are powers of two, leading to significant internal fragmentation [364, 363] These problems are solved by using the buddy system only to identify free submeshes, and allocating a number of free submeshes of different sizes to satisfy each request [375, 606]. Such a scheme is used by NQS to pack batch jobs on the Intel Paragon. The price is that the allocation is not necessarily a rectangle, and may even be non contiguous. Another interesting modification is to use a precise buddy system, in which buddy sizes are not predefined powers of two, but ....

[Article contains additional citation context not shown here]

W. Liu, V. Lo, K. Windisch, and B. Nitzberg, "Non-contiguous processor allocation algorithms for distributed memory multicomputers". In Supercomputing '94, pp. 227--236, Nov 1994.

.... allocations are also squares with sides that are powers of two, leading to significant internal fragmentation [219, 218] These problems are solved by using the buddy system only to identify free submeshes, and allocating a number of free submeshes of different sizes to satisfy each request [224, 352]. Such a scheme is used by NQS to pack batch jobs on the Intel Paragon. The price is that the allocation is not necessarily a rectangle, and may even be non contiguous. Another interesting modification is to use a precise buddy system, in which buddy sizes are not predefined powers of two, but ....

....job cause a degradation in the network performance of another. For example, this has been known to happen on the Intel Paragon, when one job is allocated a set of PEs that form a concave or disjoint group, and another job runs in the middle, effectively surrounded by PEs belonging to the first job [224, 251, 246]. Traffic belonging to the first job then passes through the links in the partition belonging to the second job (Fig. 20) 51 6.2 Impact on System Performance The way in which competing applications are handled affects the amount of resources that are lost to fragmentation, system overheads, ....

[Article contains additional citation context not shown here]

W. Liu, V. Lo, K. Windisch, and B. Nitzberg, "Non-contiguous processor allocation algorithms for distributed memory multicomputers". In Supercomputing '94, pp. 227--236, Nov 1994.

....messages from one partition to another (such as I O traffic) can pass through partitions that are in the way, using the same links that are used by the application running in that partition. This can cause noticeable congestion and degradation in communication performance for such applications [42]. Other systems, such as the IBM SP2 and Meiko CS 2, do not necessarily concentrate all the I O nodes into an I O partition. These machines are based on a 5 Intel Paragon design: job A job B I O partition I O traffic from job A does not interfere with job B job A job B I O I O traffic from ....

W. Liu, V. Lo, K. Windisch, and B. Nitzberg, "Non-contiguous processor allocation algorithms for distributed memory multicomputers". In Supercomputing '94, pp. 227-- 236, Nov 1994.

....and partitioned may be formed using arbitrary subsets of processors. A rigid job is submitted for execution along with a specification of the number of processors that it requires. The scheduler then creates a partition of that size and schedules the job to execute within that partition [53,32,20,1,9,31,33]. With moldable jobs, it is the scheduler that selects the partition size [44] Evolving and malleable jobs require partitions that are not only flexible but can also change dynamically at runtime. This places an added burden both on the programmer, who must write application code that requests ....

W. Liu, V. Lo, K. Windisch, and B. Nitzberg, "Non-contiguous processor allocation algorithms for distributed memory multicomputers". In Supercomputing '94, pp. 227--236, Nov 1994.

No context found.

V. M. Lo, W. Liu, B. Nitzberg, and K. Windisch, "Noncontiguous Processor Allocation Algorithms for Mesh-Connected Multicomputers," IEEE Transactions on Parallel and Distributed Systems, July 1997.

No context found.

V. M. Lo, W. Liu, B. Nitzberg, and K. Windisch, "Noncontiguous Processor Allocation Algorithms for Mesh-Connected Multicomputers," IEEE Trans. on Par. and Dist. Sys., vol. 8, no. 7, July 1997.

No context found.

V. M. Lo, W. Liu, B. Nitzberg, and K. Windisch, "Noncontiguous Processor Allocation Algorithms for Mesh-Connected Multicomputers," IEEE Trans. on Par. and Dist. Sys., vol. 8, no. 7, July 1997.

No context found.

V. M. Lo, W. Liu, B. Nitzberg, and K. Windisch, "Noncontiguous Processor Allocation Algorithms for Mesh-Connected Multicomputers," IEEE Transactions on Parallel and Distributed Systems, July 1997.

....on the performance evaluation of several classes of scheduling and allocation strategies. Included were two scheduling algorithms: First Come First Served and ScanUp [16] a multi level queuing algorithm, and three static allocation strategies: First Fit [24] Frame Sliding [3] and Paging [17]. The real traces were captured from four production machines in use for scientific computing at research labs and supercomputer sites around the world (two IBM SP 2s, an Intel Paragon, and a Cray T3E) The synthetic models include naive models and those developed by Downey [6, 5] and Feitelson ....

....non contiguous, based on whether or not the set of allocated processors are directly connected by links in the interconnection network. Research in job scheduling and processor allocation is thoroughly surveyed in [8] more recent work in this area includes that reported in [10] as well as our own [17, 22, 18]. This project is distinguished from our previous work in that we evaluate the experimental method, not the scheduling techniques themselves. Some of the first researchers to use real traces from production machines to drive their simulations include [19, 1, 18] At the same time, analysis of ....

[Article contains additional citation context not shown here]

V. M. Lo, K. Windisch, W. Liu, and B. Nitzberg. Non-contiguous processor allocation algorithms for mesh-connected multicomputers. IEEE Transactions on Parallel and Distributed Systems, 8(7):712--726, July 1997.

....class of contiguous allocation strategies restricts the nodes allocated to a given job to form a convex shape. Fig. 1 shows an example. Performance suffers significantly due to processors being wasted because of internal and external fragmentation. Utilizations of only 34 to 66 are reported [9, 18, 8, 10]. In contrast, the class of noncontiguous allocation strategies allocates nodes that are dispersed throughout the system. Fig. 3 shows examples. They experience no fragmentation and thus outperform contiguous strategies reaching utilizations of up to 78 [10, 15, 14] To further improve the ....

....only 34 to 66 are reported [9, 18, 8, 10] In contrast, the class of noncontiguous allocation strategies allocates nodes that are dispersed throughout the system. Fig. 3 shows examples. They experience no fragmentation and thus outperform contiguous strategies reaching utilizations of up to 78 [10, 15, 14]. To further improve the performance of noncontiguous strategies, it is necessary to select allocations that cause minimal message passing contention. External contention is due to job interference: communication between dispersed nodes of a given job may require a communication link that is ....

[Article contains additional citation context not shown here]

W. Liu, V. Lo, K. Windisch, and B. Nitzberg. Non-contiguous processor allocation algorithms for distributed memory multicomputers. In Proceedings of Supercomputing '94, pages 227--236, 1994. Best student paper award.

....on the performance evaluation of several classes of scheduling and allocation strategies. Included were two scheduling algorithms: First Come First Served and ScanUp [15] a multi level queueing algorithm, and three static allocation strategies: First Fit [23] Frame Sliding [3] and Paging [16]. The real traces were captured from four production machines in use for scientific computing at research labs and supercomputer sites around the world (two IBM SP 2s, an Intel Paragon, and a Cray T3E) The synthetic models include naive models and those developed by Downey [6, 5] and Feitelson ....

....non contiguous, based on whether or not the set of allocated processors are directly connected by links in the interconnection network. Research in job scheduling and processor allocation is thoroughly surveyed in [8] more recent work in this area includes that reported in [10] as well as our own [16, 21, 17]. This project is distinguished from our previous work in that we evaluate the experimental method, not the scheduling techniques themselves. Some of the first researchers to use real traces from production machines to drive their simulations include [18, 1, 17] At the same time, analysis of this ....

[Article contains additional citation context not shown here]

V. M. Lo, K. Windisch, W. Liu, and B. Nitzberg. Non-contiguous processor allocation algorithms for meshconnected multicomputers. IEEE Transactions on Parallel and Distributed Systems, 8(7):712--726, July 1997.

....the runtime of the application, and then release the subcubes back to the operating system. The algorithm used by the operating system to handle the requests and relinquishments of subcubes is the processor allocation algorithm and has been the target of of intensive research for the past decade [14, 12, 11, 17, 3]. The key idea to remember is that a subcube is equivalent to a block of addresses that is expressible using a single address expression or mask. This equivalence means that subcube recognition techniques can be applied to the problem of multicast address allocation. Note, however, that while ....

Virginia Lo, Kurt J. Windisch, Wanqian Liu, and Bill Nitzberg. Noncontiguous processor allocation algorithms for mesh-connected multicomputers. IEEE Transactions on Parallel and Distributed Systems, pages 712-726, 1997.

No context found.

V. Lo, K. Windisch, W. Liu, and B. Nitzberg. Non-contiguous processor allocation algorithms for mesh-connected multicomputers. IEEE Transactions on Parallel and Distributed Computing, 8(7), 1997.

No context found.

V. Lo, K. Windisch, W. Liu, and B. Nitzberg. Non-contiguous processor allocation algorithms for mesh-connected multicomputers. IEEE Transactions on Parallel and Distributed Computing, 8(7), 1997.

No context found.

V. Lo, K. Windisch, W. Liu, and B. Nitzberg. Non-contiguous processor allocation algorithms for mesh-connected multicomputers. IEEE Transactions on Parallel and Distributed Computing, 8(7), 1997.

No context found.

W. Liu, V.M. Lo, K. Windisch, and B. Nitzberg, "Non-Contiguous Processor Allocation Algorithms for Distributed Memory Multicomputers, " Proc. Supercomputing 1994.

No context found.

V. Lo, K. J. Windisch, W. Liu, and B. Nitzberg. Noncontiguous Processor Allocation Algorithms for Mesh-Connected Multicomputers. IEEE Trans. on Parallel and Distributed Systems, 8(7):712--726, 1997.

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