Diffusion Approximation Models for Cloud Computations with Task Migrations

This paper proposes a novel analytical model for performance evaluation of cloud computations using G/G/m/m+K queue. The cloud computing infrastructure allows to execute the computations on virtual machines, which can be deployed, turned off or migrated between the physical servers. The performance evaluation of such systems is complicated, since the job arrival process as well as service time may have different distributions. The approximation approach allows us to make previously presented in literature models more general by assumption of general input flows and allowing analytical approach to transient analysis.