M. Cierniak, M.J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, 40(6):356--372, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....target static schemes, we briefly discuss a few important references in the field of dynamic approaches. Simple paradigms are based upon the idea use the past to predict the future , i.e. use the currently observed speed of computation of each machine to decide for the next distribution of work [11, 12, 4]. Several authors [25, 24, 30, 20] propose a mapping policy which dynamically minimizes system degradation (including the cost of remapping) for each computation step. Other papers [31, 15] advocate local schemes where data is exchanged only between neighbor processors. Generally speaking, there ....

M. Cierniak, M.J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, 40(6):356--372, 1997.

....past ten years, load balancing problems for homogeneous parallel computing systems have been thoroughly studied. A large number of results have been gained with the help of simulations, theoretical investigations or experimental applications [1] 2] 3] 4] 5] Other studies [6] 7] 8] [9] have dealt with heterogeneous systems; these studies only considered the relative processing power of workstations or dedicated network systems. In this research we formally define the load balancing problem in section 2 then we explore non dedicated HNOW, which is the case of many existing ....

Michal Cierniak, Mohammed Zaki, and Wei Li, "Compile-time Scheduling Algorithms for Heterogeneous Network of Workstations," The Computer Journal, Automatic Loop Parallelization, Vol. 40, No. 6, December 1997.

....in various contexts. It ranges from data partitioning for parallel video processing [1] to finding the optimal number of processors in linear algebra algorithms [3] Some works are closer to ours. The distribution of loops for heterogeneous processors so as to balance the work load is studied in [6] and, in particular, the case of independent iterations, which is equivalent to a scatter operation. However, computation and communication cost functions are affine. A load balancing solution is first presented for heterogeneous processors, only when no network contentions occur. Then, the ....

M. Cierniak, M. J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, special issue on Automatic Loop Parallelization, 40(6):356--372, Dec. 1997.

....which highlights the pro le information in an easy to use manner. All of this can be done over the Internet using a client server approach and is independent of the underlying architecture machine and human intervention. More details can be found in [67] Other Rochester publications, see [36, 39, 38, 37, 35, 58, 34, 80, 81, 79]. 31 8 Florida University Report Florida became involved in the project after Sanjay Ranka, one of the original co PIs, moved there from Syracuse. We evaluated the High Performance Fortran (HPF) language as a candidate for implementing scienti c, engineering and computer science software on ....

Michal Cierniak, Wei Li, and Mohammed Javeed Zaki. Compile-time scheduling algorithms for heterogeneous network of workstations. To appear The Computer Journal, 1997.

....cannot perform any computation. Most related work only considers a single level of heterogeneity. The work of [12] uses a combined compile time runtime approach to create customized load balancing strategies for an application to minimize its execution time on a single cluster of workstations. In [11], a purely static, architectureconscious approach is considered for load balancing parallel loops which may have di erent amount of work in each iteration, on a (possibly) heterogeneous network of workstations. This work also considers the e ects of limited memory size. In Andonie et al. 2] a ....

Michal Cierniak, Mohammed J. Zaki, and Wei Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, 40(6):356372, 1997.

....which highlights the profile information in an easy to use manner. All of this can be done over the Internet using a client server approach and is independent of the underlying architecture machine and human intervention. More details can be found in [67] Other Rochester publications, see [36, 39, 38, 37, 35, 58, 34, 80, 81, 79]. 8 Florida University Report Florida became involved in the project after Sanjay Ranka, one of the original co PIs, moved there from Syracuse. We evaluated the High Performance Fortran (HPF) language as a candidate for implementing scientific, engineering and computer science software on ....

Michal Cierniak, Wei Li, and Mohammed Javeed Zaki. Compile-time scheduling algorithms for heterogeneous network of workstations. To appear The Computer Journal, 1997.

....results in the assignment A 0 = 0; 3; 6; 9, A 1 = 1; 4; 7 and A 2 = 2; 5; 8. The work load is now given as W 0 = 9 6 3 = 18, W 1 = 8 5 2 = 15 and W 2 = 7 4 1 = 12. The load imbalance is much smaller, however it is still present. Bitonic Partitioning (Single Equivalence Class) In [15] we propose a new partitioning scheme, called bitonic partitioning, for load balancing that can be applied to the problem here as well. This scheme is based on the observation that the sum of the workload due to itemsets i and (2P i 1) is a constant: w i w 2P i 1 = n i 1 (n (2P i 1) 1) 2n ....

....the workload due to itemsets i and (2P i 1) is a constant: w i w 2P i 1 = n i 1 (n (2P i 1) 1) 2n 2P 1 We can therefore assign itemsets i and (2P i 1) as one unit with uniform work (2n 2P 1) If n mod 2P = 0 then perfect balancing results. The case n mod 2P 6= 0 is handled as described in [15]. The nal assignment is given as A 0 = f0; 5; 6g,A 1 = f1; 4; 7g, and A 2 = f2; 3; 8; 9g, with corresponding workload given as W 0 = 9 4 3 = 16, W 1 = 8 5 2 = 15 and W 2 = 7 6 1 = 14. This partition scheme is better than the interleaved scheme and results in almost no imbalance. ....

M. Cierniak, M. J. Zaki, and W. Li. Compile-time scheduling algorithms for a heterogeneous network of workstations. The Computer Journal, 40(6):356-372, December 1997.

....partition, which results in the assignment A 0 = 0; 3; 6; 9, A 1 = 1; 4; 7 and A 2 = 2; 5; 8. The work load is now given as W 0 = 9 6 3 = 18, W 1 = 8 5 2 = 15 and W 2 = 7 4 1 = 12. The load imbalance is much smaller, however it is still present. Bitonic Partitioning (Single Equivalence Class) In (Cierniak, Zaki, Li 1997) we propose a new partitioning scheme, called bitonic partitioning, for load balancing that can be applied to the problem here as well. This scheme is based on the observation that the sum of the workload due to itemsets i and (2P Gamma i Gamma 1) is a constant: w i w 2P Gammai Gamma1 = n ....

.... Gamma i Gamma 1 (n Gamma (2P Gamma i Gamma 1) Gamma 1) 2n Gamma 2Pt Gamma 1 We can therefore assign itemsets i and (n Gammai Gamma1) as one unit with uniform work (2n Gamma 2Pt Gamma 1) If n mod 2P = 0 then perfect balancing results. The case n mod 2P 6= 0 is handled as described in (Cierniak, Zaki, Li 1997). Processors a) Work Load per Itemset Itemset Processors 0 b) Bitonic Partitioning c) Interleaved Partitioning 2 4 6 0 8 2 1 0 2 1 0 Work Load 0 2 4 6 8 0(9) 1(8) 3(6) 4(5) 5(4) 2(7) 14 15 16 18 15 12 0(9) 2(7) 1(8) 3(6) 4(5) 5(4) 6(3) 7(2) 8(1) 9(0) 6(3) 7(2) 8(1) ....

Cierniak, M.; Zaki, M. J.; and Li, W. 1997. Compile-time scheduling algorithms for a heterogeneous network of workstations. The Computer Journal 40(6):356-- 372.

No context found.

M. Cierniak, M.J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, 40(6):356--372, 1997.

No context found.

M. Cierniak, M. J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, special issue on Automatic Loop Parallelization, 40(6):356372, Dec. 1997.

No context found.

M. Cierniak, M. J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. Comput. J, special issue on Automatic Loop Parallelization, 40(6):356-372, Dec. 1997.

No context found.

M. Cierniak, M.J. Zaki, and W. Li, "Compile-Time Scheduling Algorithms for Heterogeneous Network of Workstations," The Computer J., vol. 40, no. 6, pp. 356-372, 1997.

No context found.

M. Cierniak, M.J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, 40(6):356--372, 1997.

No context found.

M. Cierniak, M. J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, special issue on Automatic Loop Parallelization, 40(6):356372, Dec. 1997.

No context found.

M. Cierniak, M.J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, 40(6):356--372, 1997.

No context found.

M. Cierniak, M.J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, 40(6):356--372, 1997.

No context found.

M. Cierniak, M.J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, 40(6):356--372, 1997.

No context found.

M. Cierniak, M. Zaki, and W. Li. Compile-time scheduling algorithms for a heterogeneous network of workstations. In The Computer Journal, volume 40, pages 356--372.

No context found.

M. Cierniak, M. J. Zaki, and W. Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, special issue on Automatic Loop Parallelization, 40(6):356372, Dec. 1997.

No context found.

M. Cierniak, M. Zaki, and Wei Li. Compile-time scheduling algorithms for a heterogeneous network of workstations. In The Computer Journal, volume 40, pages 356--372.

No context found.

Michal Cierniak, Mohammed J. Zaki, and Wei Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, 40(6):356-372, 1997.

No context found.

Michal Cierniak, Mohammed J. Zaki, and Wei Li. Compile-time scheduling algorithms for heterogeneous network of workstations. The Computer Journal, 40(6):356372, 1997.

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