This paper presents a two-stage service migration decision method which combines business workload forecasting with real-time load sensing, and thus adds business forecasting to previous load balancing approaches that rely solely upon real-time load sensing. The migration decision procedure and the detailed causal analysis algorithms based on Bayesian networks are also given. After the critical business indicators have been obtained from causal analysis, business fluctuation related with the critical indicators can be forecasted by using Markov chain method. And then, the migration decision can be made based on the forecasting results and the real-time load information together. We evaluate the migration decision method through three sets of experiments. We found that by migrating service on a shared multi-tenant service environment, the QoS requirement can be assured dynamically and the capability of workloads increases under same resource cost, which is helpful in optimised deploying for multi-tenant applications. include service computing and data engineering. This paper is a revised and expanded version of a paper entitled 'Probabilistic-based workload forecasting and service redeployment for multi-tenant services' presented at The

No abstract

No abstract

No abstract

No abstract