In the 1990s the largest machines had a few thousand processors and PVM and MPI were key tools to making these machines useable. Now with the growing interest in Internet computing and the design of cellular architectures such as IBM’s Blue Gene computer, the scale of parallel computing has suddenly jumped to 100,000 processors or more. This talk will describe recent work at Oak Ridge National Lab on developing algorithms for petascale virtual machines and the development of a simulator, which runs on a Linux cluster, that has been used to test these algorithms on simulated 100.000 processor systems. This talk will also look at the Harness software environemnt and how it may be useful to increase the scalability, fault tolerance, and adaptability of applications on large-scale systems.