A Multiple Fault Tolerant Approach with Improved Performance in Cluster Computing
|
Full Text |
Pdf |
|
Author |
Sanjay Bansal, Sanjeev Sharma, Rajiv Gandhi Prodhyogiki Vishwavidya
|
|
ISSN |
2079-8407 |
|
On Pages
|
116-121
|
|
Volume No. |
2
|
|
Issue No. |
3
|
|
Issue Date |
March 01, 2011 |
|
Publishing Date |
March 01, 2011 |
|
Keywords |
Message Passing Interface (MPI), Distributed Scheduling, Interprocess Communication (IPC), Multiple Faults, Failure Detection, Failure Recovery.
|
Abstract
In case of multiple node failures performance is very low as compare to single node failure. Failures of nodes in cluster computing can be tolerated by multiple fault tolerant computing. In this paper, we propose a multiple fault tolerant technique with improved failure detection and performance. Failure detection is done by improved adaptive heartbeats based algorithm to improve the degree of confidence and accuracy. Failure recovery is based on reassignment of load with a rank based algorithm Performance is achieved by distributing the load among all available nodes with dynamic rank based balancing algorithm. Dynamic ranking algorithm is low overhead algorithm for reassignment of tasks uniformly among all available nodes. Message logging is used to recover message loss
Back