Optimized Deployment Plans for Platform as a Service Clouds

Abstract

We approximately solve the problem of computing deployment plans of multiple cloud services with soft real-time constraints. We recognize that this is a generalized bin packing problem with fragmentable items. We formalize the problem domain and develop an autonomous deployment planning system with soft real-time constraints. The system incorporates a genetic algorithm with quadratic worst-case time complexity for approximately solving the packing problem, providing a service deployment plan with an optimal number of servers and an approximately optimal number of service instances.