Grid-Federation: Decentralized Resource Brokering and Resource Allocation

Introduction

The Grid resource brokering ( or superscheduling ) problem is defined as: " scheduling jobs across the grid resources such as computational clusters, parallel supercomputers, desktop machines that belong to different administrative domains". Brokering in computational grids is facilitated by specialized application schedulers such as Nimrod-G, Condor-G, Work-Flow Engines. Brokering activity involves (i) querying grid resource information services (GRIS) for locating resources that match the job requirements; (ii) coordinating and negotiating Service Level Agreements; and (iii) job scheduling. The grid resources are managed by their local resource management systems such as Condor, Portable Batch System, Sun Grid Engine and Alchemi. These systems manage job queues, initiate and monitor their execution.

Existing approaches to resource brokering in a Grid environment are non-coordinated and are based on centralized information services. In this case, application schedulers such as Nimrod-G, Condor-G perform scheduling related activities independent of the other schedulers in the system. They directly submit their applications to the underlying resources without taking into account the current load, priorities, utilization scenarios of other application level schedulers. Clearly, this can lead to over-utilization or bottleneck of some valuable resources while leaving others largely underutilized. Furthermore, these brokering systems do not have a co-ordination (or cooperative) mechanism, hence this exacerbates the load sharing and utilization problems of distributed resources because of the sub-optimal schedules that are likely to occur.

To overcome this, we propose federating these distributed brokers as part of one decentralized grid system. The resulting grid system is referred to as Grid-Federation. Our Grid-Federation system is defined as a large scale decentralized resource sharing system that consists of a coordinated federation of distributed computational resources. The key features of our proposed Grid-Federation includes: (i) a market-based grid scheduling technique; (ii) decentralization via a shared federation directory that gives site autonomy and scalability; (iii) ability to provide admission control facility at each site in the federation; (iv) incentives for resources owners to share their resources as part of the federation; and (v) access to a larger pool of resources for all users.

Project Team Members

           
Grid Computing and Distributed Systems (GRIDS) Laboratory
Department of Computer Science and Software Engineering
The University of Melbourne, Australia