The purpose of using this paper cloudlet is remotely reduces the cloud communication delay and the system load. However, since the powerful system performance of the remote cloud is not in the cloudlet, as the no. of user’s changes, the cloud provides resources for each user change. Moreover, as the cloudlet's service security is low, cloudlet's service for user, when of work activation is eliminated from coverage. In this case, users will not get the computational results, so the cloud compute fails. This paper provides a real-time resource allocation framework for users to allocate sufficient and necessary resources. This paper proposes the Movement Record-based Particle Swarm Optimization (MRPSO) method to solve the problem of work failure and real-time resource allocation. The method proposed by this paper provides a better powerful result than the real PSO method