Grid computing provides a platform for users easily accessing worldwide distributed resources. To meet the time limit and quality requirement posted by an application running on a grid, the computing nodes allocated to the application need to be periodically redeployed. The grid which periodically redeploys computing resources to meet the needs of an ongoing application is named a PR2 grid. An M-task is a type of workflow application which performs work stage by stage. In this paper, three algorithms MaxH-LP, MaxHrLP and MaxH-rCLP are proposed for scheduling an M-task application onto the PR2 grid. The algorithms are compared with the well known algorithms Minmin and Sufferage. Our simulation results show that the average performances achieved by MaxH-rLP and MaxHrCLP outperform Minmin and Sufferage for applications of various communication computation ratios.