A comparison of clustering heuristics for scheduling directed acyclic graphs on multiprocessors

Gerasoulis, Apostolos; Yang, Tao

doi:10.1016/0743-7315(92)90012-c

Cited by 322 publications

(141 citation statements)

References 4 publications

Supporting

Mentioning

133

Contrasting

Unclassified

Order By: Relevance

“…In general, an application can be represented by a weighted directed-acyclic task graph. The algorithm for finding optimal solutions for the multiple-processor scheduling problem has been shown to be NP-complete [3,4,13]. Various heuristic algorithms that have been proposed are capable of finding sub-optimal solutions [6,7,8,10,17,18,19,20].…”

Section: Introductionmentioning

confidence: 99%

“…These heuristics are categorized into several classes, such as list-based algorithms [5,6,12,17,20], clustering algorithms [4,9,11,19], and duplication-based algorithms [1,18,16,18]. Among these algorithms, list-based scheduling algorithms are generally regarded as having a good costperformance trade-off because of their low cost and acceptable results [2,4,13,14].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Improving Static Task Scheduling in Heterogeneous and Homogeneous Computing Systems

Yang

Lee

Chung

2007

2007 International Conference on Parallel Processing (ICPP 2007)

View full text Add to dashboard Cite

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Improving Static Task Scheduling in Heterogeneous and Homogeneous Computing Systems

Yang

Lee

Chung

2007

2007 International Conference on Parallel Processing (ICPP 2007)

View full text Add to dashboard Cite

show abstract

“…Duplication is already considered in this case since one task can be assigned to multiple processors, which means that the task is duplicated multiple times. With grouping, the problem becomes more complicated and resembles a traditional clustering problem [22] [23]. Just as a variation of a traditional bin packing problem, our problem attempts to maximize the throughput for each task and also packs maximal PPFs into each task to minimize the number of tasks.…”

Section: Optimal Partitioning and Mappingmentioning

confidence: 99%

A Throughput-Driven Task Creation and Mapping for Network Processors

Liu¹,

Li²,

Chen

2007

High Performance Embedded Architectures and Compilers

View full text Add to dashboard Cite

Abstract. Network processors are programmable devices that can process packets at a high speed. A network processor is typified by multithreading and heterogeneous multiprocessing, which usually requires programmers to manually create multiple tasks and map these tasks onto different processing elements. This paper addresses the problem of automating task creation and mapping of network applications onto the underlying hardware to maximize their throughput. We propose a throughput cost model to guide the task creation and mapping with the objective of both minimizing the number of stages in the processing pipeline and maximizing the average throughput of the slowest task simultaneously. The average throughput is modeled by taking communication cost, computation cost, memory access latency and synchronization cost into account. We envision that programmers write small functions for network applications, such that we use grouping and duplication to construct tasks from the functions. The optimal solution of creating tasks from m functions and mapping them to n processors is an NP-hard problem. Therefore, we present a practical and efficient heuristic algorithm with an O((n + m)m) complexity and show that the obtained solutions produce excellent performance for typical network applications. The entire framework has been implemented in the Open Research Compiler (ORC) adapted to compile network applications written in a domain-specific dataflow language. Experimental results show that the code produced by our compiler can achieve the 100% throughput on the OC-48 input line rate. OC-48 is a fiber optic connection that can handle a 2.488Gbps connection speeds, which is what our targeted hardware was designed for. We also demonstrate the importance of good creation and mapping choices on achieving high throughput. Furthermore, we show that reducing communication cost and efficient resource management are the most important factors for maximizing throughput on the Intel IXP network processors.

show abstract

“…Speci cally, w e prove the following bounds in the next two sections. : (2) 3 The Upper Bounds in the Tradeo…”

Section: The Real Control-memory Tradeomentioning

confidence: 99%

Scheduling Tree-Dags Using FIFO Queues: A Control–Memory Trade-Off

Bhatt

Chung

Leighton

et al. 1996

Journal of Parallel and Distributed Computing

View full text Add to dashboard Cite

Abstract. We study here a combinatorial problem that is motivated by a genre of architecture-independent s c heduler for parallel computations. Such schedulers are often used, for instance, when computations are being done by a cooperating network of workstations. The results we obtain expose a control-memory tradeo for such s c hedulers, when the computation being scheduled has the structure of a complete binary tree. The combinatorial problem takes the following form. Consider, for each i n teger N = 2 n , a family of n algorithms for linearizing the N-leaf complete binary tree in such a w ay that each nonleaf node precedes its children. For each k 2 f 1 2 : : : n g, t h e kth algorithm in the family employs k FIFO queues to e ect the linearization, in a manner speci ed later (cf., 1], 5] -7]). In this paper, we expose a tradeo between the number of queues used by e a c h o f t h e n algorithms | which w e view as measuring the control complexity of the algorithm | and the memory requirements of the algorithms, as embodied in the required capacity o f t h e largest-capacity queue. Speci cally, w e p r o ve that, for each k 2 f 1 2 : : : n g, the maximum per-queue capacity, call it Q k (N), for a k-queue algorithm that linearizes an N-leaf complete binary tree satis es

show abstract

A comparison of clustering heuristics for scheduling directed acyclic graphs on multiprocessors

Cited by 322 publications

References 4 publications

Improving Static Task Scheduling in Heterogeneous and Homogeneous Computing Systems

Improving Static Task Scheduling in Heterogeneous and Homogeneous Computing Systems

A Throughput-Driven Task Creation and Mapping for Network Processors

Scheduling Tree-Dags Using FIFO Queues: A Control–Memory Trade-Off

Contact Info

Product

Resources

About