A framework for economic scheduling in Grid computing using tender/Contract-net model Mohammad Bsoul, Iain Phillips and Chris Hinde {M.A.R.Bsoul, I.W.Phillips, C.J.Hinde}@lboro.ac.uk Research School of Informatics, Department of Computer Science, Loughborough University Abstract— In Grid computing, the resources of many machines are used to execute jobs submitted by users. One of the research issues in Grid is the allocation process; the mapping of jobs to the various resources. This paper addresses the concept of applying strategies to the scheduling task. To this end a framework for economic scheduling in Grid computing using Tender/Contract-net model is presented. The performance of each user strategy in terms of job success rate, cost per MI (Million Instruction) and average job satisfaction rate and the performance of each resource strategy in term of profit are evaluated by simulation. The evaluations show the performance of each strategy for different numbers of jobs with both static and dynamic submission of jobs. Keywords— Grid computing, Economic scheduling, Framework, Tender/Contract-net model. I. I NTRODUCTION In Grid computing, resource providers supply their processing resource to the Grid, and jobs are allocated to these resources for their execution. The aim of resource providers is to make profit, and the aim of users is to reduce cost. This situation creates an economic market, which is investigated in this paper. Much work has been done on finding an optimal allocation. Some of the projects use conventional techniques [5], [8], [9] that concern the system performance in general and don’t consider the economics (prices) for allocating jobs to resources. On the other hand, a large number of projects study economic strategies for allocating jobs. Economic techniques [11], [6], [4], [7] take into consideration the price of the resources when it needs to allocate jobs to resources and that price usually reflects the value of the resource to the user. Furthermore, economic techniques give an incentive for resource providers to supply their resources, and enforce the job owners to make cost-effective use of the resources. There are other advantages for using economic strategies [12]. The competing aims of maximising profit and reducing cost lead to negotiations of price. This is determined by the used economic model. Some of the models that are applicable to Grid environment are auction [2], [3], [15], [14], [10], commodity [2], [3], [14], [13], [1] and tender/contract-net [4] models. The Tender/Contract-net model is discussed in this paper’s proposed framework. In this model, the user issues a request for job execution with specific requirements. Then, the various resource owners send their offers/bids to the user which selects the most appropriate offer based on its strategy, and contacts that resource owner in order to start the execution in the case the resource owner has accepted the award. The Tender model is an appropriate model to employ in the Grid because users begin the negotiation for price. Additionally, the users specify their job requirements with their requests and the resources reply with bids that are based the the cost of taking on the job. This paper begins with proposing a framework that supports economic scheduling in Grid computing. The users and resources employ tender/contract-net model to negotiate the prices and deadlines. The negotiated users and resources aim to maximise their performance which is measured by a number of metrics. The rest of this paper is organised as follows: Section II gives a description of the framework. Section III describes the metrics for measuring the strategies’ performance. Section IV discusses the simulation results. Section V concludes the paper and mentions some possible future work. II. FRAMEWORK DESCRIPTION The framework consists of a number of entities that use Tender/Contract-net model. The entities are classified into resources and users. Fig. 1 shows the negotiations that take place between users and resources, while Table I shows the parameters which are sent through the negotiations that occur between users and resources. For each negotiation, the job parameters such as job characteristics and requirements are sent from the user to resources that the user wants to negotiate with. Then, the resources send their bids to the user that in turn accepts one of the sent bids or keeps negotiating. Resource 2 Resource 1 Resource m Job Parameters (JP) Bid Parameters (BP) JP JP JP JP JP JP BP BP BP BP User 1 User 2 User n BP BP Fig. 1. Interaction between users and resources ISBN: 1-9025-6013-9 © 2006 PGNet