Ch.D.V. Subba Rao and M.M. Naidu (India)
Checkpointing, Fault tolerance, Scalability, Performance
The two-level fault recovery scheme of distributed system inherits the merits of both disk-based and diskless checkpointing schemes. It makes use of diskless checkpointing strategy in its first-level recovery and follows disk-based checkpointing in second level recovery. This paper clearly specifies the conditions that lead for diskless checkpointing or disk-based checkpointing. The diskless checkpointing protocols available in literature have not addressed the problem of checkpointing applications with high locality of reference. Our paper addresses this issue and presents a solution to this problem. The earlier work on Two-level recovery scheme is based on analytical issues and not on experimental results. Our work includes the performance evaluation of proposed protocol in comparison with other schemes i.e. Neighbor based checkpointing scheme, N+1 parity and disk-based scheme and by means of simulation. Since the checkpoint overhead, checkpoint latency and overhead ratio is minimal in case of Enhanced Two-level Recovery Scheme, it has been concluded that the proposed scheme performs better than the other schemes available in the literature.
Important Links:
Go Back
