HPC Cloud Bad; HPC in the Cloud Good

For most of the past two decades - since the rise of clustering in the mid-1990s - the HPC community has created its own IT components for deploying, managing, monitoring, and using horizontally-scaled resources to run its technical workloads. Historically, these tools have differed significantly from those used in commercial data centers, primarily because enterprise workloads and scaling requirements were different than those seen in HPC. But big changes are afoot in the commercial data center - changes that are being driven by several factors, including multicore, the rise of cloud computing, and the emergence of new and important workloads in the enterprise. Because of these and other trends, we are entering a new period of IT convergence in which the enterprise and HPC will share an increasing number of concerns and pain points. How this convergence unfolds and how actively the HPC community participates in this convergence will have far-reaching consequences for the future health of both enterprise computing and HPC.