Abstract: An optimizer optimizes the configuration of a computer system, such as a web farm. The optimizer makes use of a performance simulator that simulates a cluster of machines running a given workload. A modular architecture allows different optimization modules to be plugged into the optimization system. In addition, the modular architecture also interfaces with different simulators through a simulator interface module. The simulator is decoupled from the optimization module. Simulation data may be filtered to provide and store only relevant objective and parameter values. A unique comparison module also guides the optimization module towards the optimum solution, even if no solutions have yet satisfied the objective completely. By caching simulation and/or optimization results (i.e. as opposed to raw simulation output data from the simulator), the simulations from previous simulation runs and different designers may be reused to reduce the time required for each optimization.

Abstract: An optimizer optimizes the configuration of a computer system, such as a web farm. The optimizer makes use of a performance simulator that simulates a cluster of machines running a given workload. A modular architecture allows different optimization modules to be plugged into the optimization system. In addition, the modular architecture also interfaces with different simulators through a simulator interface module. The simulator is decoupled from the optimization module. Simulation data may be filtered to provide and store only relevant objective and parameter values. A unique comparison module also guides the optimization module towards the optimum solution, even if no solutions have yet satisfied the objective completely. By caching simulation and/or optimization results (i.e. as opposed to raw simulation output data from the simulator), the simulations from previous simulation runs and different designers may be reused to reduce the time required for each optimization.