Hierarchical Scheduling in Grid Systems

Khaldoon Al-Zoubi

Abstract

This research proposes hierarchal scheduling to manage resources in the grid-computing systems. The “Grid” is a parallel and distributed system that enables large collection of geographically distributed heterogeneous systems that may span over several organizations to share a variety of resources dynamically at runtime depending on their availability, capability, user’s requirements and any other predefined rules by systems and resources owners.

In general, Grid scheduling consists of three stages: Resource discovery, resource selection and job execution. Our research focuses on the first two scheduling stages with the addition to other related issues such as:

• Rescheduling jobs when better resources become available,
• Rescheduling jobs that had their resources taken away from the Grid before the actual resource allocation (note that Grid scheduling is tentative until the actual allocation of the resources), and
• Balancing load among resources.

We’ve built grid model that consists of 2400 nodes to run a series of experiments on the proposed scheduling scheme. The model is written in C++ using the CD++ tools. We compare our simulation results against the Peer-to-Peer (P2P) scheduling system.

In summary, this presentation outlines the following:

• Introduction to grid computing.
• Introduction to hierarchal scheduling.
• Proposed scheduling schemes and algorithms for grid systems.
• Simulation model.
• Discussion of results.