A low-overhead checkpointing protocol for mobile networks

The mobile wireless networks present challenges in designing fault-tolerant systems because of the mobile host (MH) mobility, limited bandwidth on wireless links, limited MH local storage, limited battery power, and the handoff. Checkpointing and rollback techniques can be used to provide fault-tolerance in a distributed application consisting of a set of cooperating processes in which some of them run on MHs. Most of the checkpointing protocols proposed for general distributed systems are not efficient with respect to the limited mobile host resources. This paper presents a low-overhead protocol which uses system resources efficiently, and requires minimal amount of information to be exchanged over the network. The protocol uses a "filtering" process at the base stations to reduce the number of "forced checkpoints" from other MHs. A comparative study with some existing protocols is made.