Task Allocation for Service Provider using Difference Equation Work Load Status Consideration

Abstract

Task allocation algorithm topic in distributed computing system, such as cloud computing, is now an interesting issue. Service Manager (SM) has to consider which Service Provider (SP) is suited to service data processing or even data storage for a Service Request (SR). Normally, SP also has to handle its own tasks thus new allocated tasks should not consume more CPU’s workload status which could make SP decrease its base performance in processing. This research tries to invent the task allocation algorithm by studying the status of the CPU usage according to the time period then transforming it to a Transition Probability (TP) matrix. This TP should be used to predict the SP workload status, using the Markov chain technique at a period of given time. These information were transformed to difference equations in order to simplify calculation handlings. Another information was gradient of SP’s work load which should trajectory the increasing or decreasing of SP’s workload. These information were used to make a decision by SM, whether this SP is suited to service request from SR or not. The experiment is presented in practical calculations. It could point out that some request of SR is more suitable for sending to which one of SPs.