Recovery protocol for mobile checkpointing

Abstract

Information systems consist of multiple mobile stations and fixed stations communicating with each other. Mission critical applications are required to be executed fault tolerantly in these systems. However, mobile stations support neither enough volume of storage and processing power nor enough capacity of battery to do reliable communication for a long period. Moreover, wireless communication channels are less reliable. Hence, the communication channels with mobile stations are often disconnected. Therefore, it is difficult for multiple mobile stations to take checkpoints synchronously since the communication channels with the mobile stations may be disconnected even while taking the checkpoints. We have proposed hybrid checkpointing where checkpoints are taken asynchronously by the mobile stations and synchronously by the fixed stations. In addition, the mobile stations record messages for getting local states consistent with the checkpoints taken by the fixed stations. We propose a method for how the mobile stations record the messages, gather the messages stored in the stable storage distributed in multiple mobile stations, and recompute the messages in a consistent order.