Improving the Success Rate of Concurrent Mobile Transactions by Predicting Time for Execution

Abstract

In mobile database environment maintaining consistency of shard data is a challenging problem. This is because mobile devices may be prone to disconnections and mobility. As a result the transactions are aborted, delayed or frequently restarted. This not only reduces the success rate of transaction but also reduces the overall system throughput. In the proposed strategy attempt is made to increase the success rate of transactions by evaluating expected time for execution of a transaction. The expected time for execution is evaluated using K-Means and KNN (K-Nearest Neighborhood) algorithms. The simulated results show that the proposed strategy works better if sufficient data logs of previous successful transactions are available.