Task execution in multi-agent environments may require cooperation among agents. Given a set of agents and a set of tasks which they have to satisfy, we consider situations where each task should be attached to a group of agents that will perform the task. Task allocation to groups of agents is necessary when tasks cannot be performed by a single agent. However it may also be beneficial when groups perform more efficiently with respect to the single agents' performance. In this paper we present several solutions to the problem of task allocation among autonomous agents, and suggest that the agents form coalitions in order to perform tasks or improve the efficiency of their performance. We present efficient distributed algorithms with low ratio bounds and with low computational complexities. These properties are proven theoretically and supported by simulations and an implementation in an agent system. Our methods are based on both the algorithmic aspects of combinatorics and approximat...

In many optimization problems, one seeks to allocate a limited set of resources to a set of individuals with demands. Thus, such allocations can naturally be viewed as vectors, with one coordinate representing each individual. Motivated by work in network routing and bandwidth assignment, we consider the problem of producing solutions that simultaneously approximate all feasible allocations in a coordinate-wise sense. This is a very strong type of "global" approximation guarantee, and we explore its consequences in a range of discrete optimization problems, including facility location, scheduling, and bandwidth assignment in networks. A fundamental issue --- one not encountered in the traditional design of approximation algorithms --- is that good approximations in this global sense need not exist for every problem instance; there is no a priori reason why there should be an allocation that simultaneously approximates all others. As a result, the existential questions concerning such g...

. This paper surveys, analyses and establishes new polynomial-time reductions among scheduling problems that connect identical parallel machines with unit-time shops and the preemption facility with chain-like precedence constraints in equal machine environments. The reductions turn out to be unexpectedly fruitful in clarifying the complexity status of many scheduling problems that were open before. New complexity results in the paper are devoted to identical parallel machines, ow shops and open shops. 1. Introduction Recently found mass polynomial-time reductions between scheduling problems on identical parallel machines and shop scheduling problems with nonpreemptive unit processing time operations proved to be very eective in studying their complexity. The latter problems we simply call unit-time shops. Reductions of unit-time shops to identical parallel machines we call parallelizings, and the inverse reductions, i.e., of identical parallel machines to unittime shops, we c...

: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : xii CHAPTERS 1 INTRODUCTION : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 1 1.1 Topic Statement : : : : : : : : : : : : : : : : : : : : : : : : : : : : 2 1.2 Overview : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 2 2 BACKGROUND AND RELATED EFFORTS : : : : : : : : : : : : : : 5 2.1 LU Factorization : : : : : : : : : : : : : : : : : : : : : : : : : : : 5 2.2 Algorithm Stability and Error Analysis : : : : : : : : : : : : : : : 6 2.3 Sparse Matrix Concepts : : : : : : : : : : : : : : : : : : : : : : : 11 2.4 Multifrontal Methods : : : : : : : : : : : : : : : : : : : : : : : : : 19 2.5 Factorization Sequences of Matrices : : : : : : : : : : : : : : : : : 24 2.6 Parallel Matrix Computations : : : : : : : : : : : : : : : : : : : : 28 2.7 Multiprocessor Scheduling : : : : : : : : : : : : : : : : : : : : : : 32 3 IMPLEMENTATION PLATFORM : : : : : : : : : : : : : : : : : : : : 37 3.1 Hardware Ar...

To benefit from grids, scientists require grid workflow engines that automatically manage the execution of interrelated jobs on the grid infrastructure. So far, the workflows community has focused on scheduling algorithms and on interface tools. Thus, while several grid workflow engines have been deployed, little is known about their performance-related characteristics, and there are no commonly-used testing practices. This situation limits the adoption of the grid workflow engines, and hampers their tuning and their further development. In this work we propose a testing methodology for grid workflow engines that focuses on five characteristics: overhead, raw performance, stability, scalability, and reliability. Using this methodology, we evaluate in a real test environment several middleware stacks that include grid workflow engines, including two based on DAGMan/Condor and on Karajan/Globus. 1

The problem of scheduling precedence graphs for which every task has to be executed in a non-uniform interval is considered, with interprocessor communication delays. For the following classes of graphs we will present a polynomial time algorithm that finds minimumlateness schedules. 1. Outforests on two processors. 2. Series-parallel graphs and opposing forests with the least urgent parent property (to be defined) on two processors. 3. Inforests with the least urgent parent property on m processors.

Abstract Grid computing promises to enable a reliable and easy-to-use computational infrastructure for e-Science. To materialize this promise, grids need to provide full automation from the experiment design to the final result. Often, this automation relies on the execution of workflows, that is, of jobs comprising many inter-related computing and data transfer tasks. While several grid workflow execution tools already exist, not much is known about their workload. This lack of knowledge hampers the development of new workflow scheduling algorithms, and slows the tuning of existing ones. To address this situation, in this work we present an analysis of two workflow-based workload traces from the Austrian Grid. We introduce a method for analyzing such traces, focused on the intrinsic and on the environment-related characteristics of the workflows. Then, we analyze the workflows executed in the Austrian Grid over the last two years. Finally, we identify six categories of workflows based on their intrinsic workflow characteristics. We show that the six categories exhibit distinctive environmentrelated characteristics, and identify the categories that are difficult to execute for common workflow schedulers.

WedevelV a constant time transposition"oraclo for the set of perfectelectd-V-U orderings of chordal graphs. Using thisoracl; we can generate a Gray code ofal perfectelectd-;;- orderings in constant amortized time using knownresul: about antimatroids. Using clngd trees, we show how theinitialdW:A-: of theald:::Eq can be performedinlerfo time. Weal- develB two new characterizations of perfectelectd::qq orderings: one in terms ofchordlqd paths, and the other in terms u-v separators.

This paper is concerned with the problem of scheduling preemptive tasks subject to precedence constraints in order to minimize the maximum lateness and the makespan. The number of available parallel processors is allowed to vary in time. It is shown that when an Earliest Due Date first algorithm provides an optimal nonpreemptive schedule for unitexecution -time (UET) tasks, then the preemptive priority scheduling algorithm, referred to as Smallest Laxity First, provides an optimal preemptive schedule for real-execution-time (RET) tasks. When the objective is to minimize the makespan, we get the same kind of result between Highest Level First schedules solving nonpreemptive tasks with UET and the Longest Remaining Path first schedule for the corresponding preemptive scheduling problem with RET tasks. These results are applied to four specific profile scheduling problems and new optimality results are obtained. Keywords: Preemptive Scheduling, List Schedule, Priority Schedule, Variable P...

The RISC revolution has spurred the development of processors with increasing degrees of instruction level parallelism (ILP). In order to realize the full potential of these processors, multiple instructions must continuouslybe issued and executed in a single cycle. Consequently, instruction scheduling plays a crucial role as an optimization in this context. While early attempts at instruction scheduling were limited to compile-time approaches, the current trends are aimed at providing dynamic support in hardware. In this paper, we present the results of a detailed comparative study of the performance advantages to be derived by the spectrum of instruction scheduling approaches: from limited basic-block schedulers in the compiler, to novel and aggressive schedulers in hardware. A significant portion of our experimental study via simulations, is devoted to understanding the performance advantages of run-time scheduling. Our results indicate it to be effective in extracting the ILP inh...