A method of distributing load amongst two or more computer system resources. The method includes distributing load to the system resources in accordance with their modeled response times and a predetermined Service level Agreement (SLA) associated with each system resource. By modeling the response time of each resource, load can be distributed with a view to maintaining response times within a predetermined Service level Agreement (SLA). The response time may be modeled by analytical modeling, which uses a queuing network model for predicting the response time, typically along with other parameters such as, utilization, throughput and queue length.