STOC: Secure, Tactical On-Demand Cloud

ABSTRACT: High performance computing (HPC) is moving from the data center to the tactical edge of military networks, an unprecedented change driven by the need for information superiority in every environment in which the Air Force operates. Harmonia extends this trend to the most challenging of environments ad hoc wireless tactical networks through the development of a novel architecture called Secure, Tactical On-Demand Cloud (STOC). Two of the most powerful new HPC technologies GP-GPUs (general purpose, graphical processing units) and the MapReduce programming paradigmare mostly limited to use in data centers and server rooms today. Our proposed STOC prototype will reduce the time required and make it easier for software developers to write application programs to take advantage of performance improvements arising from coupling CPU and GP-GPU computational power, provide instrumentation interfaces for use by third parties to collect and analyze data about jobs running in STOC, engage other researchers who may want to use STOC to experiment with their algorithms, and ultimately make this level of computationally-intensive work possible on ad-hoc tactical wireless networks. BENEFIT: This work on STOC adapts clouds to a tactical edge environment, where commercial clouds (e.g., from Amazon, Google) are not designed to work. Through the use of general purpose Graphical Processing Unit (GP-GPU) chips, a benefit is enabling real-time response to decentralized tactical users by exploiting more massive parallelism for a given size, weight, and power limit than conventional Central Processing Units (CPUs) can achieve. GP-GPU chips are approaching a thousand or more stream cores (e.g., 6 GFLOPS [giga floating point operations per second] per watt for one chip). We also allow cloud computing with seamless distribution of computation over heterogeneous GP-GPU/CPU nodes, which is ideal for a tactical setting that may combine various types of hardware devices with and without GPUs. We also simplify the MapReduce framework for end users to allow users with lower expertise and programming skills to configure computations; this allows faster deployment of new capabilities on our cloud architecture.