Towards SLA-based optimal workload distribution in SANs

Storage area networks (SANs) connect storage devices to servers over fast network interconnects. We consider the problem of optimal SAN configuration with the goal of meeting service level agreements (SLAs) for server processes while retaining flexibility for future changes. Our approach proceeds in two stages, by setting up pseudo-Boolean constraint problems and solving them with an off-the-shelf solver. First, we give an algorithm for assigning storage devices to applications running on the SANpsilas hosts. This algorithm tries to balance the workload as evenly as possible over all storage devices. Our second algorithm takes these assignments and computes the interconnections (data paths) that are necessary to achieve the desired configuration while respecting redundancy (safety)requirements in the SLAs. Again, this algorithm tries to balance the workload of all connections and devices in proportion to their capacity. Thus, our network configurations respect all SLAs and provide flexibility for future changes by avoiding bottlenecks on storage devices or switches.