- Sunday, February 22, 2009
- Fault Tolerant Mobile Computing
- Published at:Not Found
Fault Tolerant Mobile Computing
Mobile (infrastructured or Ad-hoc) systems merge as a substitute for regular wired distributed systems. There are many issues that have to be resolved before the mobile system becomes a valid and acceptable substitute to the original wired distributed system, such as routing protocols, limited bandwidth and mobility issues. Research methodologies are classified into simulation, qualitative and quantitative and analytical approach. Quantitative and analytical approach is the best because it is used to get a closed form solution after defining the model of the system and system parameters clearly and correctly. Therefore, in this paper we have developed an analytical model to study the effect of different system parameters when using the infrastructured mobile system for distributed computing. In our model, we considered all system parameters such as host mobility, host speed, communication overhead, host distance from the originating node, checkpointing rate and checkpointing time. The infrastructured mobile system is considered as a lossy system. That is, it is very possible that a host disappears from the system due to many reasons such as battery loss, leaving the cell and loses connection. Thus the system is considered faulty system. Fault tolerant approaches are characterized into preservative and curative. Therefore, the preservative approach used in our system is the checkpointing-rollback recovery mechanism, where each mobile host (MH) is forced to periodically save its state into a mobile support station (MMS). Each MMS is assumed highly reliable (has backup). The curative approach is achieved in our system by assuming that the unprocessed load by failure MHs will be redistributed over the arriving MHs. In this system a new checkpointing algorithm is introduced, which concerns about the bandwidth and consistency at the same time. Accordingly it is suitable for mobile computing environment. The results show that our model woks and one can use it to tune different system parameters and study the effect of each individual parameter on the overall system performance.