Clusters of computers have emerged as mainstream parallel and distributed platforms for high-performance, high-throughput and high-availability computing. To enable effective resource management on clusters, numerous cluster managements systems and schedul
Libra: An Economy driven Job Scheduling System for Clusters
Jahanzeb Sherwani1, Nosheen Ali1, Nausheen Lotia1, Zahra Hayat1, and Rajkumar Buyya2
1Dept. of Computer Science and Engineering 2Grid Computing and Distributed Systems Laboratory
Lahore University of Management Sciences (LUMS) Dept. of Computer Science and Software Engineering
Lahore, Pakistan The University of Melbourne, Australia
{jahanzeb,nosheen,02020111,02020189}@lums.edu.pk raj@cs.mu.oz.au
Abstract: Clusters of computers have emerged as mainstream parallel and distributed platforms for high-performance, high-throughput and high-availability computing. To enable effective resource management on clusters, numerous cluster managements systems and schedulers have been designed. However, their focus has essentially been on maximizing CPU performance, but not on improving the value of utility delivered to the user and quality of services. This paper presents a new computational economy driven scheduling system called Libra, which has been designed to support allocation of resources based on the users’ quality of service (QoS) requirements. It is intended to work as an add-on to the existing queuing and resource management system. The first version has been implemented as a plugin scheduler to the PBS (Portable Batch System) system. The scheduler offers market-based economy driven service for managing batch jobs on clusters by scheduling CPU time according to user utility as determined by their budget and deadline rather than system performance considerations. The Libra scheduler ensures that both these constraints are met within an O(n) run-time. The Libra scheduler has been simulated using the GridSim toolkit to carry out a detailed performance analysis. Results show that the deadline and budget based proportional resource allocation strategy improves the utility of the system and user satisfaction as compared to system-centric scheduling strategies.
1 Introduction
Clusters of commodity computers (PCs) have emerged as mainstream parallel and distributed platforms for high-performance, high-throughput and high-availability computing [1]. They are presented together as a single, unified resource to the end users by middleware technologies such as resource management and scheduling system. Users submit jobs in batch to a resource management system queue and a centralized scheduler decides how to prioritize and allocate resources for jobs execution. To minimise the response time, the scheduling system strategy needs to prioritize competing user jobs with varying levels of priorities and importance and allocate resources accordingly. To perform these tasks effectively, schedulers require knowledge of how users value their computations and their quality of service requirements, which varies from time to time [9]; schedulers should be able to provide a feedback signal that prevents users from submitting unbounded amounts of work [11]. Unfortunately, the current approaches (e.g., Condor [2], PBS
[12], SGE [15], and LSF [16]) to batch scheduling provide limited means for users to expresses their valuation of resources and quality of service requirements, if any. Also, feedback signals provide no incentive for users to pay attention to and respond to them.
To overcome the shortcomings of traditional system-centric cluster management systems, we advocate the use of the computational economy metaphor [4][17][18][19]. Computational economy enables the regulation of supply and demand of resources. It allows the users to specify QoS parameters with jobs and also offers incentive to users for relaxing QoS requirements. This essentially means that user constraints such as deadline and budget are more important in determining the priority of a job by the scheduler, than system policies such as ordering jobs according to the basis of submission time.
The rest of this paper is organised as follows. In Section 2, we discuss related work and technologies that influence or aid the development of our Libra scheduling system. In Section 3, we present architecture of Libra and economic scheduling algorithm. In Section 4, we describe the implementation of Libra as scheduler for PBS system, the simulation of Libra using GridSim toolkit, and the development of Web interface. In Section 5, we mainly present simulation results that quantify the benefits of new deadline and
scheduling driven by computational economy in the emerging grid computing ...(such as traditional supercomputers, clusters [5], SMPs, or even powerful...
DYNAMIC RELIABILITY DRIVEN REAL-TIME SCHEDULING ON CLUSTERS Address for ...In our system model, parallel real-time jobs, which are modeled by ...
An Adaptive Scoring Job Scheduling algorithm for ...clusters with high speed networks, can achieve ...However, grid is a heterogeneous system. ...
An Economy Driven Resour... 暂无评价 9页 免费 Job Scheduling for Compu....CDNs ease the management of application distribution for clusters and grids. ...
Proterv-II An Integrated... 1页 免费 Libra An Economy-Driven ... 4页...106-113 Scheduling a Metacomputer by an Implicit Voting System F. Ramme ...
Scheduling for Workstation Clusters Under Windows NT...system and individual job efficiency [1, 5, 6]...DCS is driven directly by message arrivals: ...
An Economy Driven Resour... 暂无评价 9页 免费 An Economy Driven Resour.... Nimrod/G: An Architecture for a Resource Management and Scheduling System ...
Scheduling of Parallel A... 暂无评价 8页 免费 Understanding Virus Beha... 10页 免费 Windows NT Clusters for ... 暂无评价 6页 免费 Disk-to–disk ...
Scheduling of Parallel A... 暂无评价 8页 免费 Understanding Virus Beha... 10页 免费 Windows NT Clusters for ... 暂无评价 6页 免费 Disk-to–disk ...
Model-Driven Resource Ma... 暂无评价 6页 免费 Operating Systems for Re... 暂无评价 15页 免费 Chapter 1 Model-Driven D... 暂无评价 29页 免费 ...
热门文档