Component reuse depends not only on the functionality they implement, but also the performance characteristics they exhibit. For example, the same functionality of solvers may be implemented in a variety of ways, with differing performance profiles over a range of parameters. Some may be more efficient for small-sized matrices, while some may be faster for large matrices. Some may be optimized for small number of processors, while some may be designed to be scalable to large number of processors. Some may perform better on shared memory machines than in distributed memory machines. In the Faucets project [48], we are working on characterizing performance profiles for applications. Similar performance profiling could be done for components, and this profile could be stored and continuously updated in the component repositories. This will enable the application composer to choose from among components implementing the same functionality based on their performance profiles.