Cloud/Grid/Virtualization Architecture for Air Force Weather

ABSTRACT: Advances in the basic enterprise computing model have been proposed over the past several years with the stated aim of improving enterprise flexibility and resiliency. Despite serious advances in areas such as cloud and grid computing, creating a system-wide infrastructure capable of supporting a distributed enterprise requires solving the problem of achieving flexibility over long geographic distances. This is a significant challenge since the design of modern IT systems does not generally contemplate operation in a high bandwidth-delay product environment. Solving this problem requires us to examine all elements of the distributed enterprise ranging from IPC/RPC communication among distributed processes, to middleware and including compute resources, storage, file systems, communications systems and so on. These elements are all critical to the fundamental goal of allocating workloads among a set of distributed sites. We propose to begin by characterizing the workloads unique to a distributed enterprise. From here we develop a strawman architectural model and begin analysis of various technologies and products used to distribute these workloads, leading to the selection of components such as file systems, capable of operating in a WAN environment. Finally, we integrate these"WAN-capable"components with the strawman model and analyze the resulting architecture. BENEFIT: Although certainly not trivial, the problem being addressed by this solicitation, the requirement for flexibility across broad geographical distances, is by no means unique to the DoD. For example, members of the financial services industry have a compelling need for systems that are able to operate in real time across WAN distances. This is not generally possible using present data center architectures, even though the key technologies, such as high performance WANs, are generally available today. Other sources of demand for such flexibility and agility include any global enterprise, or any enterprise with round-the-clock operations or for whom a disruption of service to its clients is unaccceptable. These all demand the types of flexibility over long distances, contemplated by this solicitation. Therefore, solving the problem of flexible allocation of resources over long distances is likely to accelerate adoption of cloud computing and other techniques designed to improve the efficient use of enterprise resources. Our discussions with members of the financial services industry, and with network providers confirm that demand for such services exists, provided we can describe a complete system capable fo providing such services.