Semantic-Based Scalable Decentralized Grid Resource Discovery

Resource Discovery (RD) is a key issue in Grid systems since resource reservation
and task scheduling are based on it. RD is about locating an appropriate resource type
that matches the user’s application requirements. Currently most Grid RDs adopt a
centralized or hierarchical model. However, this model is characterized by poor
scalability, dynamism and load-balancing features. Moreover, they do not support
semantic description and discovery. This paper proposes a new semantic based
scalable decentralized Grid RD model. Grid nodes are classified into classes based on
some criteria; each class has a head which is elected among its own class
nodes/members. Two Peer-to-Peer (P2P) network layers are used; the first one to
connect between the class heads and the second between each class members.
Ontology is used to describe the resources, applications and their relationships. We
introduce two kind of intelligent agents; Request Agent (RA) and Description Agent
(DA). Each node has both of the agents. DA describes resource capabilities, and RA
carries resource requests that are needed for some applications. We develop a RD
algorithm that optimizes the search of the resources on the network. Our discussion
shows how our model satisfies Grid RD features such as scalability, dynamism, fault
tolerance, and interoperability.