Integrating Trust into Scheduling Algorithms in Grid System

Recently, Trust has been recognized as important factor for Grid computing security. In this paper, we propose trust model in Grid system. It consists of Application Domain (AD), Client Domain (CD), Resource Domain (RD), and Trust Manager (TM). TM controls the relationship between RD and CD depending on trust level value of each client and classification of each resource. Performance criteria are makespan and utilization. We evaluated our trust model in six scheduling algorithms in three scenarios. The simulation results show that the proposed trust model improves the performance in all scheduling algorithms. The security problem is a hot topic in Grid research due to the dynamics and uncertainty of Grid system. There are three entities defined as users, applications and resources in Grid environment. In such situation, users are vulnerable to risk because of potential incomplete or distorted information provided by malicious resources, and as Grid system grows tremendously in size, the possibility of users to attack the network by providing aggressive or vicious applications will increase greatly. Trust management is an effective method to maintain the credibility of the system and keep honesty of entities (8). Trust (1-3) is the firm belief in the competence of an entity to act as expected such that this firm belief is not a fixed value associated with the entity but rather it is subject to the entity's behavior and applies only within a specific context at a given time. Trust management (TM) is collecting, codifying, analyzing, and evaluating evidence relating to competence, honesty, security, or dependability with the purpose of making assessments and decisions regarding trust relationships (9). Trust management systems (TMS) must support analysis of trust and recommendation specifications to detect conflicts and inconsistencies and support trust queries related to decision making. We believe that the trust management system has a wider role in Grid environment as: It can help automate security decisions rather than replacing the existing Grid Security Infrastructure (GSI), It has broader scope for interaction and it may impact the whole Grid administration, and the implementation TMS should have minimal overhead in terms of computation, infrastructure, storage, and complexity as in Grids, the data needed for the operation of TMS is already available through other services. In this paper, we focus on trust management in Grid computing. The proposed model is based on the trust model proposed by Azzedin and Maheswaran (1-5). They measured the performance of Grid system applied their trust model in resource management system. They worked with Minimum Completion Time heuristic, Min-min heuristic, and Sufferage heuristic algorithms. In this paper, we examined six scheduling algorithms by our trust model. The new point in the proposed trust model is the computing and evaluating trust. Trust manager's operations in the proposed trust model make more control in the system than Trust agent's operations in the conventional trust model. II. RELATED WORK There is a lot of research on the trust in distributed systems. Here we just mention some works that are deeply related to our paper. Abdul-Rahman and Hailes (10) proposed a trust model for computing the trust for an agent in a specific context based on the experience and recommendations. They applied and implemented their trust model in P2P networks. Trust can have only four possible values; very trustworthy, trustworthy, untrustworthy, and very untrustworthy. Each agent stores the trust values for the agents with him/her interacts and the recommender trust with respect to another agent. So each agent has to store all history of past experiences and received recommendations.