Narrow Results

Traditional message passing based checkpointing and rollback recovery algorithms perform well for tightly coupled systems. In wide area distributed systems these algorithms may suffer from large overhead due to message passing delay and network traffic. Mobile agents offer an attractive option for designing checkpointing schemes for wide area distributed systems. Network topology is assumed to be arbitrary. Processes are mobile agent enabled. When a process wants to take a checkpoint, it just creates one mobile agent. Concurrent initiations by multiple processes are allowed. Synchronization and creation of a consistent global state (CGS) for checkpointing is managed by the mobile agent(s). In the worst case, for k concurrent initiations among n processes, checkpointing algorithm requires a total of O(kn) hops by all the mobile agents. A mobile agent carries O(n/k) (on the average) size data.

The inherent shortcoming of the conventional Sender-Based Message Logging (SBML) protocols is to require additional control message interactions per application message to satisfy the always-no-orphans condition in case of sequential failures. In this paper, a scalable SBML protocol is introduced to lower the communication overhead by handling a sequence of messages consecutively received by each process before sending as a party. The protocol enables the process to delay the update of their receive sequence numbers to their senders until there comes out the first message it is willing to send, and then perform the collective filling out task with each sender requiring only one control message exchange. Experimental results show that our protocol outperforms the previous one in terms of the number of control messages generated.

Checkpoint policies have been studied by several researchers. In this paper, we propose a new model for evaluating the database recovery actions and for determining an optimum checkpoint interval. Taking account of unsuccessful rollback recovery after system failure, we derive the stationary availability as a function of the checkpoint interval under the assumption that the failures occur at a Poisson process. We further discuss the optimum checkpoint interval which maximizes the availability above. The influence of unsuccessful rollback recovery on the system is illustrated by numerical examples.