Results 1  10
of
68
Application scheduling and processor allocation in multiprogrammed parallel processing systems
 Performance Evaluation
, 1994
"... ..."
Parallel Query Processing
 ACM Computing Surveys
, 1993
"... With relations growing larger and queries becoming more complex, parallel query processing is an increasingly attractive option for improving the performance of database systems. The objective of this paper is to examine the various issues encountered in parallel query processing and the techniques ..."
Abstract

Cited by 57 (0 self)
 Add to MetaCart
With relations growing larger and queries becoming more complex, parallel query processing is an increasingly attractive option for improving the performance of database systems. The objective of this paper is to examine the various issues encountered in parallel query processing and the techniques available for addressing these issues. The focus of the paper is on the join operation with both sortmerge join and hash joins being considered. Three types of parallelism can be exploited, namely intraoperator, interoperator, and interquery parallelism. In intraoperator parallelism the major issue is task creation, and the objective is to split a join operation into tasks in a manner such that the load can be spread evenly across a given number of processors. This is a challenge when the values on the join attribute are not uniformly distributed. Interoperator parallelism can be achieved either through parallel execution of independent operations or through pipelining. In either case,...
Multidimensional Resource Scheduling for Parallel Queries
, 1996
"... Scheduling query execution plans is an important component of query optimization in parallel database systems. The problem is particularly complex in a sharednothing execution environment, where each system node represents a collection of timeshareable resources (e.g., CPU(s), disk(s), etc.) and c ..."
Abstract

Cited by 37 (4 self)
 Add to MetaCart
Scheduling query execution plans is an important component of query optimization in parallel database systems. The problem is particularly complex in a sharednothing execution environment, where each system node represents a collection of timeshareable resources (e.g., CPU(s), disk(s), etc.) and communicates with other nodes only by messagepassing. Significant research effort has concentrated on only a subset of the various forms of intraquery parallelism so that scheduling and synchronization is simplified. In addition, most previous work has focused its attention on onedimensional models of parallel query scheduling, effectively ignoring the potential benefits of resource sharing. In this paper, we develop an approach that is more general in both directions, capturing all forms of intraquery parallelism and exploiting sharing of multidimensional resource nodes among concurrent plan operators. This allows scheduling a set of independent query tasks (i.e., operator pipelines) to be seen as an instance of the multidimensional bindesign problem. Using a novel quantification of coarse grain parallelism, we present a list scheduling heuristic algorithm that is provably nearoptimal in the class of coarse grain parallel executions (with a worstcase performance ratio that depends on the number of resources per node and the granularity parameter). We then extend this algorithm to handle the operator precedence constraints in a bushy query plan by splitting the execution of the plan into synchronized phases. Preliminary performance results confirm the effectiveness of our scheduling algorithm compared both to previous approaches and the optimal solution. Finally, we present a technique that allows us to relax the coarse granularity restriction and obtain a list scheduling method that is provably nearoptimal in the space of all possible parallel schedules.
Fairness in Parallel Job Scheduling
, 2000
"... This paper introduces a new preemptive algorithm that is well suited for fair online scheduling of parallel jobs. Fairness is achieved by selecting job weights to be equal to the resource consumption of the job and by limiting the time span a job can be delayed by other jobs submitted after it. Fur ..."
Abstract

Cited by 31 (5 self)
 Add to MetaCart
This paper introduces a new preemptive algorithm that is well suited for fair online scheduling of parallel jobs. Fairness is achieved by selecting job weights to be equal to the resource consumption of the job and by limiting the time span a job can be delayed by other jobs submitted after it. Further, the processing time of a job is not known when the job is released. It is proven that the algorithm achieves a constant competitive ratio for both the makespan and the weighted completion time for the given weight selection. Finally, the algorithm is also experimentally evaluated with the help of workload traces.
A Convex Programming Approach for Exploiting Data and Functional Parallelism on Distributed Memory Multicomputers
, 1994
"... Compilers have focussed on the exploitation of one of functional or data parallelism in the past. The PARADIGM compiler project at the University of Illinois is among the #rst to incorporate techniques for simultaneous exploitation of both. The work in this paper describes the techniques used in the ..."
Abstract

Cited by 29 (8 self)
 Add to MetaCart
Compilers have focussed on the exploitation of one of functional or data parallelism in the past. The PARADIGM compiler project at the University of Illinois is among the #rst to incorporate techniques for simultaneous exploitation of both. The work in this paper describes the techniques used in the PARADIGM compiler and analyzes the optimality of these techniques. It is the #rst of its kind to use realistic cost models and includes data transfer costs which all previous researchers have neglected. Preliminary results on the CM5 show the e#cacy of our methods and the signi#cant advantages of using functional and data parallelism together for execution of real applications. 1. INTRODUCTION Distributed memory multicomputers such as the Intel Paragon, the IBM SP1 and the Thinking Machines CM5 o#er signi#cant advantages over shared memory multiprocessors in terms of cost and scalability. Unfortunately,to extract all that computational power from these machines, users have to write e#...
Scheduling parallelizable tasks to minimize average response time
 In 6th Annual ACM Symposium on Parallel Algorithms and Architectures
, 1994
"... ..."
Lineartime Approximation Schemes for Scheduling Malleable Parallel Tasks
 Proceedings of the 10th Annual ACMSIAM Symposium on Discrete Algorithms
, 1999
"... A malleable parallel task is one whose execution time is a function of the number of (identical) processors alloted to it. We study the problem of scheduling a set of n independent malleable tasks on a xed number of parallel processors, and propose an approximation scheme that for any xed > 0, comp ..."
Abstract

Cited by 28 (12 self)
 Add to MetaCart
A malleable parallel task is one whose execution time is a function of the number of (identical) processors alloted to it. We study the problem of scheduling a set of n independent malleable tasks on a xed number of parallel processors, and propose an approximation scheme that for any xed > 0, computes in O(n) time a nonpreemptive schedule of length at most (1 + ) times the optimum. 1 Introduction In this paper, we study the following scheduling problem. Suppose there is given a set of tasks T = fT 0 ; : : : ; T n 1 g and a set of identical processors M = f1; : : : ; mg. Each task T j has an associated function t j : M ! Q + that gives the execution time t j (`) of task T j in terms of the number of processors ` 2 M that are assigned to T j . Given j processors alloted to task T j , these j processors are required to execute task T j in union and without preemption, i.e. they all have to start processing task T j at some starting time j , and complete it at j + t j (...
Resource Scheduling for Parallel Database and Scientific Applications
 in Proceedings of the 8th Annual ACM Symposium on Parallel Algorithms and Architectures
, 1996
"... We initiate a study of resource scheduling problems in parallel database and scientific applications. Based on this study we formulate a problem. In our formulation, jobs specify their running times and amounts of a fixed number of other resources (like memory, IO) they need. The resourcetime trade ..."
Abstract

Cited by 28 (5 self)
 Add to MetaCart
We initiate a study of resource scheduling problems in parallel database and scientific applications. Based on this study we formulate a problem. In our formulation, jobs specify their running times and amounts of a fixed number of other resources (like memory, IO) they need. The resourcetime tradeoff may be fundamentally different for different resource types. The processor resource is malleable, meaning we can trade processors for time gracefully. Other resources may not be malleable. One way to model them is to assume no malleability: the entire requirement of those resources has to be reserved for a job to begin execution, and no smaller quantity is acceptable. The jobs also have precedences amongst them; in our applications, the precedence structure may be restricted to being a collection of trees or seriesparallel graphs. Not much is known about considering precedence and nonmalleable resource constraints together. For many other problems, it has been possible to find schedule...
Polynomial Time Approximation Schemes for ClassConstrained Packing Problems
 Proc. of Workshop on Approximation Algorithms
, 1999
"... . We consider variants of the classic bin packing and multiple knapsack problems, in which sets of items of different classes (colors) need to be placed in bins; the items may have different sizes and values. Each bin has a limited capacity, and a bound on the number of distinct classes of items ..."
Abstract

Cited by 27 (6 self)
 Add to MetaCart
. We consider variants of the classic bin packing and multiple knapsack problems, in which sets of items of different classes (colors) need to be placed in bins; the items may have different sizes and values. Each bin has a limited capacity, and a bound on the number of distinct classes of items it can hold. In the classconstrained multiple knapsack (CCMK) problem, our goal is to maximize the total value of packed items, whereas in the classconstrained binpacking (CCBP), we seek to minimize the number of (identical) bins, needed for packing all the items. We give a polynomial time approximation scheme (PTAS) for CCMK and a dual PTAS for CCBP. We also show that the 01 classconstrained knapsack admits a fully polynomial time approximation scheme, even when the number of distinct colors of items depends on the input size. Finally, we introduce the generalized classconstrained packing problem (GCCP), where each item may have more than one color. We show that GCCP is APX...
Smart SMART bounds for weighted response time scheduling
 SIAM Journal on Computing
, 1998
"... Consider a system of independent tasks to be scheduled without preemption on a parallel computer. For each task the number of processors required, the execution time, and a weight are known. The problem is to nd a schedule with minimum weighted average response time. We present an algorithm called S ..."
Abstract

Cited by 25 (5 self)
 Add to MetaCart
Consider a system of independent tasks to be scheduled without preemption on a parallel computer. For each task the number of processors required, the execution time, and a weight are known. The problem is to nd a schedule with minimum weighted average response time. We present an algorithm called SMART for this problem that produces solutions that are within a factor of 10.45 of optimal. To our knowledge this is the rst polynomialtime algorithm for the minimum weighted average response time problem that achieves a constant bound. In addition, for the unweighted case (that is, where all the weights are unity) we describe a variant of SMART that produces solutions that are within a factor of 8 of optimal, improving upon the best known bound of 32 for this special case. 1