as utility: computation jobs can be scheduled on-demand in Grid hosts based on available computation capacity. In this paper, we study another emerging usage of Grid utility: the hosting of application services. Different from a computation job, an application service such as e-Laboratory or on-line shopping has longer lifetime, and performs multiple jobs requested by its clients. A service Hosting Utility Platform (HUP) is formed by a set of servers in the Grid, and multiple application services will be hosted on the HUP. We present the design and implementation of SODA, a Service-On-Demand Architecture that enables on-demand creation of application services on a HUP. With SODA, an application service will be created in the form of a set of virtual service nodes; each node is a virtual machine which is physically a `slice' of a real host in the HUP. SODA involves both OS and middleware level techniques, and has the following salient capabilities: (1) on-demand service priming: the image of an application service as well as the OS on which it runs will be created on-demand and bootstrapped automatically; (2) better service isolation: services sharing the same HUP host are isolated with respect to administration, faults, attacks, and resources; (3) integrated service request management: for each service, a service switch will be created to direct client requests to appropriate virtual service nodes. Moreover, the application service provider can replace the default request switching policy with a service-specific policy.

We describe the porting, redesign, and tuning of a high performance disk-to-disk parallel sort on a general purpose Myrinet connected PC cluster running Windows NT. This cluster employs the high speed communication of Fast Messages (from the HPVM system). The study exposes the performance limitations and key architectural issues in achieving high performance. Key distinguishing features of our effort include the use of MPI, Windows NT and the HPVM software. Our effort builds on the Berkeley NOWSort, adapted and tuned for an HPVM NT-cluster. Using 60 nodes, we sorted 10.3 GB in 56.51 seconds breaking the fastest published Minute sort record of which we are aware 1 . Each machine in our cluster consisted of dual processor, 300 MHz Pentium II's, with 512MB of memory, making possible the use of a one pass sort, and a single 4.5GB 10000 RPM disks. 1 Introduction Clusters of commodity PC's or workstations have achieved success and popularity in the academic world and most recently in in...

In order to measure database performance of computer systems, a collection of companies and universities defined three benchmarks in 1985, including a disk-to-disk sort of one million records, which became a standard for database systems. After the first official result for this benchmark in 1986, performance steadily improved each year thereafter. In 1996 the best systems for the DatamationSort were multiprocessors, and since 1997 clusters of workstations have dominated these disk-to-disk sorting benchmarks. We explore the benefits, performance limitations, and key architectural issues involved in achieving high performance in disk-to-disk sorting. The testbed for this work was the NCSA NT-Supercluster, a cluster of Pentium-II workstations running Windows NT and the HPVM software suite. We selected the MinuteSort benchmark, which requires the sorting of as many records as possible in one minute, as the performance benchmark because it offered the clearest opportunity for compari...

Clusters of workstations are becoming popular platforms for parallel computing, but performance on these systems is more complex and harder to predict than on traditional parallel machines. Hence, performance monitoring and analysis is important for understanding application behavior and improving performance. We present a performance monitor for HPVM, a high-performance cluster running Windows NT. The novel features of our monitor are: an integrated approach to performance information, a (software) global clock to correlate performance information across cluster nodes and leverage of Windows NT performance monitoring facilities. We discuss the design issues for this tool, and present results of using this tool to analyze the performance of a cluster application. Keywords: cluster performance monitoring 1 Introduction Clusters of workstations or PCs are now an attractive alternative to massively parallel processors (MPPs) for parallel computing. They have good price/performance char...

My research interests are in building secure systems. I am especially interested in the interactions between theoretical approaches to security, the principles of systems design, and human factors that affect security in real-world systems. As we entrust ever more of our lives to the correct operation of complex interconnected systems, the security of these systems in an increasingly important concern. However, we still do not understand how to build large systems that are secure. While low-level security mechanisms are reasonably well-understood, security failures are often the result of inappropriate composition of these elements. As computing becomes more pervasive, new classes of users and applications are emerging, giving rise to a new range of security and privacy concerns. My Ph.D. research focused on access control for a ubiqutious computing environment. Such environments differ from traditional networked systems in many ways, and require a re-examination of the security assumptions for distributed systems. The main contributions of the thesis research were an access control model for a class of ubiquitous computing environments, and a prototype implementation within a research “ubiquitous operating system”. The novel features of this model were support for the various types of collaboration common in such spaces, and the ability to represent system contextual information in the policies to enable support for the heavily context-dependent and dynamic nature of such systems. The model supports both discretionary and mandatory access controls, to