Matt W. Mutka

Condor-a hunter of idle workstations

The design, implementation, and performance of the Condor scheduling system, which operates in a workstation environment, are presented. The system aims to maximize the utilization of workstations with as little interference as possible between the jobs it schedules and the activities of the people who own workstations. It identifies idle workstations and schedules background jobs on them. When the owner of a workstation resumes activity at a station, Condor checkpoints the remote job running on the station and transfers it to another workstation. The system guarantees that the job will eventually complete, and that very little, if any, work will be performed more than once. A performance profile of the system is presented that is based on data accumulated from 23 stations during one month.<<ETX>>

Service discovery in pervasive computing environments

Pervasive computing environments pose unique service discovery challenges. We have developed a taxonomy of existing protocols as a basis for analyzing their approaches and identifying problems and open issues relative to service discovery in pervasive computing environments.

Prophet address allocation for large scale MANETs

A mobile device in a MANET must be assigned a free IP address before it may participate in unicast communication. This is a fundamental and difficult problem in the practical use of any MANET. Several solutions have been proposed. However, these approaches have different drawbacks. A new IP address allocation algorithm, namely prophet allocation, is proposed in the paper. The proposed scheme may be applied to large scale MANETs with low complexity, low communication overhead, even address distribution, and low latency. Both theoretical analysis and simulation experiments are conducted to demonstrate the superiority of the proposed algorithm over other known algorithms. Moreover, the proposed prophet allocation is able to solve the problem of network partition and merger efficiently.

The available capacity of a privately owned workstation environment

Abstract Powerful workstations interconnected by networks have become widely available as sources of computing cycles. Each workstation is typically owned by a single user in order to provide a high quality of service for the owner. In most cases, an owner does not have computing demands as large as the capacity of the workstation. Therefore, most of the workstations are underutilized. Nevertheless, some users have demands that exceed the capacities of their workstations. In order to effectively share the capacity of workstations, there must be algorithms that allocate available capacity and long periods when owners do not use their stations. To understand the profile of station availability, we analyzed the usage patterns of a cluster of workstations. The workstations were available more than 75% of the time observed. Large capacities were steadily available on an hour to hour, day to day, and month to month basis. These capacities were available not only during the evening hours and on weekends, but during the busiest times of normal working hours. A stochastic model was developed which was based on an analysis of the relative frequency distribution and the correlation of available and non-available interval lengths. A 3-stage hyperexponential cumulative distribution has been fitted to the observed cumulative relative frequency of available periods. The fitted distribution closely matches the observed relative frequency distribution. This stochastic model is important as a workload generator for the performance evaluation of capacity sharing strategies of a cluster of workstations. The model assists in the design of resource management algorithms that take advantage of the characteristics of the usage patterns.

Splendor: A secure, private, and location-aware service discovery protocol supporting mobile services

In pervasive computing environments, powerful handheld devices with wireless connections create opportunities for many new nomadic applications. We propose a new service discovery model, called Splendor, supporting nomadic users and services in public environments. Splendor emphasizes security and supports privacy. Location awareness is integrated for location dependent services discovery and is used to lessen service discovery network infrastructure requirements. We analyze the Splendor system performance and provide our experimental results.

A private, secure, and user-centric information exposure model for service discovery protocols

Service Discovery as an essential element in pervasive computing environments is widely accepted. Much research on service discovery has been conducted, but privacy and security have been ignored and may be sacrificed. While it is essential that legitimate users should be able to discover services, it is also necessary that services be hidden from illegitimate users. Since service information, service providers information, service requests, user presence information, and users identities may be sensitive, we may want to keep them private during service discovery processes. There appears to be no existing service discovery protocols that solve these problems. We present a user-centric model, called Prudent Exposure, which exposes minimal information privately and securely. Users and service owners exchange code words in an efficient and scalable form to establish mutual trust. Based on the trust, secure service discovery sessions are set up. The model is further improved to counter attacks. We analyze the mathematical properties of our model, formally verify our security protocol, and measure the performance of our prototype system.

PrudentExposure: a private and user-centric service discovery protocol

Service discovery as an essential element in pervasive computing environments is widely accepted. Much active research on service discovery has been conducted, but privacy has been ignored and may be sacrificed. While it is essential that legitimate users should be able to discover services of which they have credentials, it is also necessary that services be hidden from illegitimate users. Since service information, service providers information, service requests, and credentials to access services via service discovery protocols may be sensitive, we may want to keep them private. Existing service discovery protocols do not solve these problems. We present a user-centric model, called Prudentexposure, as the first approach designed for exposing minimal information privately, securely, and automatically for both service providers and users of service discovery protocols. We analyze the mathematical properties of our model and formally verify our security protocol.

Priority based real-time communication for large scale wormhole networks

As advances are made in parallel processing technology, an increasing number of real-time applications are being developed for large-scale parallel processors. Since the wormhole network is a popular communication system used in the new generation of large-scale parallel multiprocessors, real-time communication support on wormhole networks becomes an important issue. We evaluate a priority mapping scheme, a priority adjustment scheme and a message dropping method for large-scale real-time wormhole networks. The priority mapping scheme embeds the timing property of a message into a priority for flow control decisions. The priority adjustment scheme dynamically modifies the priority of a message as the timing property of the message changes. The tardy messages, which miss their deadlines, are removed from the network by the message dropping method. Simulation studies show that the schemes outperform the conventional flow control scheme implemented in general purpose wormhole networks, and the schemes provide desirable performance when the size of the parallel system increases.<<ETX>>

Estimating capacity for sharing in a privately owned workstation environment

The author analyzes workstation patterns in order to understand opportunities for exploiting idle capacity. This study is based on traces of users workstation activity in a university environment. It identifies two areas where enhancements can be made. One area is the ability of a manager of the shared capacity of a workstation cluster to schedule jobs with deadline constraints. This opportunity is the result of the ability to make good predictions of the time-varying amount of capacity that is available for sharing. A prediction strategy is developed that is shown to have only a small amount of error. For the second area of enhancement, it is shown that it is feasible to allocate partitions of workstations for specific periods. This aids those users who on occasion need exclusive access to several machines. The author examines the profile of periods during which exclusive access to partitions can be given, the rate that owners preempt users of partitions, and the distribution of interpreemption intervals. >

Dependency removal for transport protocol header compression over noisy channels

Van Jacobson (VJ) TCP header compression significantly reduces TCP protocol overhead in a noiseless environment. Unfortunately, when the algorithm operates in error prone environments similar to wireless, the performance of the compressor drops to below that of not using header compression. This work describe modifications to the VJ algorithm that allow it to better tolerate errors. By removing many of the dependencies that are implicitly transmitted on the link, one is able to generate a noise tolerant VJ compressor. On noiseless links, the new algorithm shows similar performance to the original algorithm. On random or burst error channels, the new algorithm significantly outperforms the original.