A model for resource discovery among grids

Abstract

Resource discovery plays an important role in Grid. The key of Resource Discovery is organization of all information of resources and the status about systems. In currently deployed Grid systems, the information services used for resource discovery usually have been organized into centralized structure or hierarchical structure, such as China Grid, so such information model is efficient. The reason may include two: most resources in such Grids are traditional clusters with relatively stable high-performance computers; the rules to govern the availability of the participating resources are strict. But with the QoS constraints that currently govern most Grids being loosening up, Grids will move towards more popular and diverse application scenarios. Moreover, currently deployed Grids are all isolated from each other and almost each Grid are statically over-provisioned to handle peak loads so that resources utilization rates within such applications are typically low, which means that sometimes in Grid there are available resources can be used by other Grids. At the same time, with the development of computers such as the computing power of PCs and Supercomputers, increasing bandwidth and storage, the available resource inside Grids will increase, so resource utilization rates will be lower and lower. Hence pooling resources together to form a virtual resource house seems natural to enhance resource utilization as well as to reduce costs. But it is nature for all kinds of resource not to be shared by all Grids, especially special data source or devices. So we categorize Grids into communities according to application domain, such as Health or Petroleum domain. Here we proposed a model for resource discovery among Grids based on the community categorized by application domain. © 2008 IEEE.