## Adaptive Allocation of Independent Tasks to Maximize Throughput

### Cached

### Download Links

- [idea.library.drexel.edu]
- [halcyon.usc.edu]
- DBLP

### Other Repositories/Bibliography

Citations: | 17 - 0 self |

### BibTeX

@MISC{Hong_adaptiveallocation,

author = {Bo Hong and Viktor K. Prasanna},

title = {Adaptive Allocation of Independent Tasks to Maximize Throughput},

year = {}

}

### OpenURL

### Abstract

www.library.drexel.edu The following item is made available as a courtesy to scholars by the author(s) and Drexel University Library and may contain materials and content, including computer code and tags, artwork, text, graphics, images, and illustrations (Material) which may be protected by copyright law. Unless otherwise noted, the Material is made available for non profit and educational purposes, such as research, teaching and private study. For these limited purposes, you may reproduce (print, download or make copies) the Material without prior permission. All copies must include any copyright notice originally included with the Material. You must seek permission from the authors or copyright owners for all uses that are not allowed by fair use and other provisions of the U.S. Copyright Law. The responsibility for making an independent legal assessment and securing any necessary permission rests with persons desiring to reproduce or use the Material.

### Citations

9061 | Introduction to Algorithms
- Cormen, Leiserson, et al.
- 2001
(Show Context)
Citation Context ... also easy to verify that such an f is a feasible solution for Base ProblemðG;sÞ and that it has the same throughput as f 0 . tu Problem 1 is the well-studied network flow problem. Several algorithms =-=[8]-=- can be used to solve this problem, e.g., the Edmonds-Karp algorithm, which has OðjVjjEj 2 Þ complexity, the Push-Relabel algorithm, which has OðjVj 2 jEjÞ complexity, and the Relabel-to-Front algorit... |

1340 | On Power-law Relationships of the Internet Topology
- Faloutsos, Faloutsos, et al.
- 1999
(Show Context)
Citation Context ...en the system topology is represented by a uniformly distributed random matrix. We also simulate more practical scenarios where computation resources are connected via the Internet. Empirical studies =-=[11]-=- have shown that the Internet topologies exhibit power law distributions. In this set of simulations, the graph representations of the systems were generated using Brite, which is a tool developed in ... |

601 | Parallel Computer Architecture: A Hardware/Software Approach - Culler, Singh, et al. - 1997 |

393 |
Parallel Programming with MPI
- Pacheco
- 1997
(Show Context)
Citation Context ...s of Cray XT3 [17]). Both techniques remove communication load from the CPU (at the cost of adding certain hardware resources). Examples of software techniques include message passing interface (MPI) =-=[26]-=-, which supports the overlap via nonblocking communication primitives. Some researchers have also pointed out that a certain cost is associated with the overlapping of computation and communication [1... |

249 | A comparison of eleven static heuristics for mapping a class of independent tasks onto heterogeneous distributed computing systems
- Braun, Siegel, et al.
(Show Context)
Citation Context ...complete [16] in its general form, designing heuristics and evaluating their performance become the key issues. For example, static scheduling heuristics for a set of independent tasks are studied in =-=[6]-=-, whereas the related dynamic scheduling problem has been studied in [22]. Other research efforts consider tasks with interdependencies. For example, a heuristic-based approach is studied in [31] to s... |

202 | On the origin of power laws in Internet topologies
- Medina, Matta, et al.
- 2000
(Show Context)
Citation Context ... have shown that the Internet topologies exhibit power law distributions. In this set of simulations, the graph representations of the systems were generated using Brite, which is a tool developed in =-=[23]-=- that generates networks with power law characteristics. We used the following parameters for Brite: each system has 20 nodes, the topology was generated using the “Barabasi-Albert” model at the Auton... |

195 | Heuristic algorithms for scheduling independent tasks on non-identical processors
- Ibarra, Kim
- 1977
(Show Context)
Citation Context ...mputing systems has received a lot of attention recently. Various research efforts have addressed the minimization of the makespan. Because the makespan minimization problem is known to be NPcomplete =-=[16]-=- in its general form, designing heuristics and evaluating their performance become the key issues. For example, static scheduling heuristics for a set of independent tasks are studied in [6], whereas ... |

172 | Dynamic mapping of a class of independent tasks onto heterogeneous computing systems
- Maheswaran, Ali, et al.
- 1999
(Show Context)
Citation Context ...heir performance become the key issues. For example, static scheduling heuristics for a set of independent tasks are studied in [6], whereas the related dynamic scheduling problem has been studied in =-=[22]-=-. Other research efforts consider tasks with interdependencies. For example, a heuristic-based approach is studied in [31] to schedule multicomponent applications in heterogeneous wide area networks. ... |

145 | The GrADS project: Software support for high-level Grid application development
- Berman, Chien, et al.
(Show Context)
Citation Context ...the performance of the system based on a snapshot of the current system status may not lead to optimal performance. Consequently, the task allocation needs to be adaptive to the changes in the system =-=[5]-=-, [27]. For the throughput maximization problem, we consider the system model in which 1) a computer can send and receive data to/from multiple neighboring computers concurrently and 2) computation an... |

88 | Scheduling Strategies for Master-Slave Tasking on Heterogeneous Processor Platforms
- Banino, Beaumont, et al.
- 2004
(Show Context)
Citation Context ...d to communicate with the root node only. When the application consists of a large set of equal-sized independent tasks, the throughput maximization in a general graph-structured system is studied in =-=[1]-=-. The problem is formulated as a linear programming problem, which is a well-studied problem with many available algorithms. However, these algorithms are centralized and are not suitable for distribu... |

86 | Folding@Home and Genome@Home: Using distributed computing to tackle previously intractable problems
- Larson, Snow, et al.
(Show Context)
Citation Context ...computation paradigm models a variety of research and commercial activities. Internet-based distributed computations are among the most well-known examples, which include SETI@home [18], Folding@home =-=[21]-=-, data encryption/decryption [10], and so forth. The computation paradigm also models various applications that are typically executed on tightly coupled systems (for example, Monte Carlo simulations ... |

84 |
Heterogeneous processing
- Freund, Siegel
- 1993
(Show Context)
Citation Context ... through local and/or wide area networks. By utilizing these distributed resources in a coordinated manner, a heterogeneous computing system can meet the computational demands of complex applications =-=[12]-=-. In this paper, we consider the computation of a large set of equal-sized independent tasks in a heterogeneous computing system. In particular, we consider the scenario where each task is to process ... |

55 |
Condor and the Grid", in Grid Computing: Making The Global Infrastructure a Reality (eds
- Thain, Tannenbaum, et al.
- 2003
(Show Context)
Citation Context ...in that we focus on the maximization of the system throughput. For computation in heterogeneous systems, maximizing the system throughput is not a new idea. A well-known example is the Condor project =-=[30]-=-. It develops a software infrastructure so that a distributed system with different ownerships can be utilized in a uniform manner to provide high throughput computation. The throughput maximization p... |

54 | E±cient Load Balancing for Wide-area Divide-and-Conquer Applications
- Nieuwpoort, Kielmann, et al.
- 2001
(Show Context)
Citation Context ...e multicomponent applications in heterogeneous wide area networks. In [7], a software-in-the-loop approach is proposed to design and implement task scheduling algorithms for heterogeneous systems. In =-=[25]-=-, an augmentation to Java is proposed to develop divide-and-conquer applications in distributed environments and several scheduling heuristics are experimentally studied. Note that the optimization ob... |

31 |
The alternator
- Gouda, Haddix
- 1999
(Show Context)
Citation Context ... shared between u andv. The problem of maintaining a consistent image of a shared variable has already been addressed by many researches (e.g., distributed consistency protocols have been designed in =-=[13]-=- and [20]). In summary, both “push” and “relabel” operations can be executed in a localized fashion. The adaptation operation changes fð^u; ^vÞ, which is local to ^u and ^v. The algorithm also increas... |

26 |
The police and the Community
- Carter
- 2002
(Show Context)
Citation Context ...6], which supports the overlap via nonblocking communication primitives. Some researchers have also pointed out that a certain cost is associated with the overlapping of computation and communication =-=[19]-=-, i.e., the computation capability of a computer may be reduced if the computer is involved intensively in communications. For the discussion in this paper, we do not consider the cost of overlapping.... |

24 | High-performance algorithm engineering for computational phylogenetics
- Moret, Bader, et al.
- 2001
(Show Context)
Citation Context ... [10], and so forth. The computation paradigm also models various applications that are typically executed on tightly coupled systems (for example, Monte Carlo simulations and Computational Phylogeny =-=[24]-=-). The computation paradigm reduces to the general problem of allocating or scheduling independent tasks in . B. Hong is with the Department of Electrical and Computer Engineering, Drexel University, ... |

22 | Assessing the impact and limits of steady-state scheduling for mixed task and data parallelism on heterogeneous platforms
- Beaumont, Legrand, et al.
- 2004
(Show Context)
Citation Context ...ping from tasks to computers (such that the throughput of data processing is maximized) is shown to be NP-complete, and a mapping heuristic is developed in [29]. A different scenario is considered in =-=[3]-=-, where the system topology is also graph structured but the application consists of a set of independent identical problems and each problem in turn consists of a set of interdependent tasks. Their r... |

20 |
SETI@home-Massively distributed computing for
- Korpela, Werthimer, et al.
(Show Context)
Citation Context ...tation nodes. This computation paradigm models a variety of research and commercial activities. Internet-based distributed computations are among the most well-known examples, which include SETI@home =-=[18]-=-, Folding@home [21], data encryption/decryption [10], and so forth. The computation paradigm also models various applications that are typically executed on tightly coupled systems (for example, Monte... |

9 |
Bandwidth-centric Allocation
- Beaumont, Legrand, et al.
- 2002
(Show Context)
Citation Context ...ormulated as a linear programming problem, which is a well-studied problem with many available algorithms. However, these algorithms are centralized and are not suitable for distributed execution. In =-=[4]-=-, a bandwidth-centric method has been obtained for the computation of equal-sized independent tasks when the computers are connected via a tree topology, which further leads to a localized autonomous ... |

9 | Dynamic adaptation to available resources for parallel computing in an autonomous network of workstations
- Rencuzogullari, Dwarkadas
- 2001
(Show Context)
Citation Context ...erformance of the system based on a snapshot of the current system status may not lead to optimal performance. Consequently, the task allocation needs to be adaptive to the changes in the system [5], =-=[27]-=-. For the throughput maximization problem, we consider the system model in which 1) a computer can send and receive data to/from multiple neighboring computers concurrently and 2) computation and comm... |

4 |
Scheduling Strategies for Master-Slave Tasking
- Banino, Beaumont, et al.
- 2004
(Show Context)
Citation Context ...When the nodes in the system are connected via a general graph topology, the problem of extracting a best spanning tree that has the highest throughput among all possible spanning trees is studied in =-=[2]-=-. The result in [2] shows that the achievable throughput of the optimal spanning tree, in the worst case, can be arbitrarily bad compared with that of the original graph-structured system. Similar to ... |

4 |
Parallel and Sequential Job Scheduling in Heterogeneous Clusters: A Simulation Study using
- Collins, George
- 2001
(Show Context)
Citation Context ...ther research efforts consider tasks with interdependencies. For example, a heuristic-based approach is studied in [31] to schedule multicomponent applications in heterogeneous wide area networks. In =-=[7]-=-, a software-in-the-loop approach is proposed to design and implement task scheduling algorithms for heterogeneous systems. In [25], an augmentation to Java is proposed to develop divide-and-conquer a... |

4 | Maximum Data Gathering in Networked Sensor 29 Int
- Hong, Prasanna
(Show Context)
Citation Context ... maximum data-gathering problem can be formulated as a flow problem with capacity constraints at the nodes (in addition to edge capacity constraints, which represents the communication bandwidth). In =-=[14]-=-, we show that the flow problem with node capacity constraints can be reduced to a standard network flow problem and, hence, a similar algorithm as mentioned above can be applied to data gathering pro... |

2 |
A Heuristic Algorithm for Mapping Communicating Tasks
- Taura, Chien
- 2000
(Show Context)
Citation Context ...ifferent ownerships can be utilized in a uniform manner to provide high throughput computation. The throughput maximization problem has also been studied from various algorithmic aspects. The work in =-=[29]-=- considers heterogeneous computing systems that are connected via a general graph topology. The application is of streaming style: A task continuously receives data from certain preceding tasks, proce... |

1 |
Cray XT3
- Inc
- 2007
(Show Context)
Citation Context ...er, we develop a distributed and adaptive task allocation algorithm for general graph-structured systems. the main processor via shared memory (see [9, Chapter 7.5] and the specifications of Cray XT3 =-=[17]-=-). Both techniques remove communication load from the CPU (at the cost of adding certain hardware resources). Examples of software techniques include message passing interface (MPI) [26], which suppor... |

1 |
Alternator in Read/Write Model
- Kulkarni, Bolen, et al.
- 2004
(Show Context)
Citation Context ...etween u andv. The problem of maintaining a consistent image of a shared variable has already been addressed by many researches (e.g., distributed consistency protocols have been designed in [13] and =-=[20]-=-). In summary, both “push” and “relabel” operations can be executed in a localized fashion. The adaptation operation changes fð^u; ^vÞ, which is local to ^u and ^v. The algorithm also increases hðsÞ b... |

1 |
Bo Hong received the MEng degree from Tsinghua University, Beijing, in 2000 and the PhD degree in electrical engineering from the University of Southern California in 2005. He is currently an assistant professor in the Electrical and Computer Engineering
- Weissman
- 2000
(Show Context)
Citation Context ...ed in [6], whereas the related dynamic scheduling problem has been studied in [22]. Other research efforts consider tasks with interdependencies. For example, a heuristic-based approach is studied in =-=[31]-=- to schedule multicomponent applications in heterogeneous wide area networks. In [7], a software-in-the-loop approach is proposed to design and implement task scheduling algorithms for heterogeneous s... |