Loading...

A New Fault Tolerant Task Scheduling Algorithm in Cluster Environments

Tabbaa, Nabil | 2011

661 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: English
  3. Document No: 42029 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Movaghar Rahimabadi, Ali
  7. Abstract:
  8. Emerging technologies enable a set of distributed resources across a network tobe linked together and used in a coordinated fashion to solve a particular parallel applicationat the same time. Such applications are often abstracted as directed acyclicgraphs (DAGs), in which vertices represent application tasks and edges represent datadependencies between tasks.The problem of scheduling in a distributed environment is that ofdeciding where to place tasks, and when to start execution of the tasks. Effective scheduling mechanisms for DAG applicationsare essential to exploit the potential of computational resources. Fault tolerance is an important requirement in distributed computing systems when the applications need a technique to continue execution even though some system components are subject to failure. In this thesis, a fault tolerant task scheduling algorithm is proposed for mapping task graphs to processing nodes with different capabilities in cluster computing systems. The starting point of the algorithm is a DAG representing an application with information about the tasks. This information consists of the execution time of the tasks on the target system processors, communication times between the tasks having data dependencies, and the number of the processor failures (npf) which should be tolerated by the scheduling algorithm. The algorithm uses the list scheduling heuristic, and is based on active replication toschedule npf+1 replicas of each task on different processors to achieve the required fault tolerance. Simulation results show the efficiency of the proposed algorithm in spite of its implementation simplicity compared to the other similar algorithms
  9. Keywords:
  10. Fault Tolerance ; Distributed Computing ; Cluster ; Task Scheduling Algorithm ; Directed Acyclic Graph (DAG)

 Digital Object List

 Bookmark

No TOC