A method of allocating computers and supporting infrastructure (especially storage devices) to server components of distributed applications is disclosed. The method takes advantage of the portability of virtual machines between physical hosts. A genetic algorithm is used to generate a near optimal allocation of physical hosts to the server components of distributed applications. There is usually little difference between the near optimal allocation generated by the genetic algorithm than the optimal allocation which might be made other algorithms. Furthermore, the amount of time taken to execute the genetic algorithm is found to rise linearly with the number of allocation decisions to be made, making the allocation method more scalable than prior-art approaches.
展开▼