Main menu >> Computer Science >> Distributed Systems

Computer Science

Distributed Systems
^ Actor Coordination Abstractions, Semantics, and Implementation
G. Agha,* C. Varela
University of Illinois

This research focuses on the complexity of expressing interaction and coordination in Web-based computing. Researchers are working on providing high-level mechanisms to manage the complexity of scaling up computations over the Web, piggy-backing on the availability of Java byte-code for portability. The project defines several actor-based abstractions (casts, directors, messengers) to effectively harness the power of the World Wide Web as a global computing infrastructure. Groups of actors, or casts, represent an abstraction unit for naming, synchronization, migration, composition, and load balancing. Each cast contains a director, and intercast communication is performed via special actors named messengers.

^ Software Architectures for Distributed Systems
G. Agha,* M. Astley
University of Illinois; National Science Foundation, CCR 9619522

The term middleware describes a set of services for integrating components of a distributed application, such as coordination and communication mechanisms. Recently, middleware services have been developed which support fault-tolerance, security, and other high-level policies. Such services have a fixed semantics, their implementation being influenced by the semantics of the application and the nature of the execution environment. The goal of this research is to provide a modular framework for developing middleware services. The project is formulating theoretical, linguistic, and run-time support for developing the needs of a particular application. Particular attention is paid to placement and mobility issues and vertical integration requirements.

^ Specifying and Deriving Mobile Systems
G. Agha,* P. Thati, R. Ziaei
U.S. Army, JHU 8812-48151

This research is focused on studying formal methods for specifying and verifying distributed software systems. The objective is to use automated deduction tools to reason about certain properties of mobile agents in open distributed systems. More specifically, security issues in authentication protocols and agent design are being studied. The project is formalizing an appropriate semantic framework that captures the fundamental properties of mobile computing and simplifies the task of reasoning. A specification language and logic will be developed based on the semantic framework. Finally, automated reasoning environments will be explored to find a suitable platform to implement the reasoning system.

^ Studies in Distributed Database Systems
G. G. Belford*
University of Illinois

Distributed databases can provide enhanced data availability to support applications such as office automation and computer-aided design and manufacturing. But first, further work is needed on problems such as maintaining the consistency of replicated data in an efficient, reliable manner. This research addresses these and related problems by the development of provably correct, reliable protocols and by the analysis and simulation of competing protocols and algorithms.

^ 2K-A Component-based Network-centric Operating System
R. H. Campbell,* M. D. Mickunas,* K. Nahrstedt,* J. Al-Muhtadi, D. Carvelho, B. Gill, C. Hess, F. Kon, S. Kowshik, P. Liu, L. Magalhaes, M. Roman, P. Viswanathan, D. Wichadakul, T. Yamane
National Science Foundation, EIS 98-70736

This research investigates an adaptable operating system architecture based on network-centric components. The architecture encompasses a framework for architecture awareness which provides the basis for adaptable QoS, dynamic security, and self-configuration. This approach adopts an application-oriented and user-oriented service model in which the system customizes itself to better fulfill the user and application requirements. In this model, a user is no longer an entity local to one specific machine—each user is a machine-independent entity that exists in the network and can transparently carry its profile across different hardware and software platforms. The operating system manages and allocates the resources to support the user on the platforms selected.

^ Customizable ORBs
R. H. Campbell,* F. Kon, M. Roman
University of Illinois; IBM Fellowship

Middleware is an important distributed computing solution that bridges legacy applications and new technologies. While middleware supports application evolution, the evolution of its own architecture is problematical. As technology, economics, and trade-offs change, middleware may become obsolete if it is difficult to retrofit with new technology. Such problems are already being encountered with CORBA extensions for real-time, group communication, and fault-tolerance. Therefore, researchers are investigating middleware solutions that can be evolved and customized with technology. Research includes customizable, reflective ORBs that support application specific customization to overcome these problems.

^ Optimizations and Compilation Techniques for Networks of Workstations
R. H. Campbell,* A. Dave
University of Illinois

Networks of workstations (NOW) provide an inexpensive environment for running parallel applications. However, performance results indicate that generic network protocols like TCP which guarantee reliable message delivery add unnecessary overheads that degrade the performance of parallel applications. Communication compilation is an automatic technique for generating efficient application-specific network protocols. We investigate algorithms for communication compilation which use knowledge of a parallel application's communication and computation behavior to generate optimized network protocols that guarantee the reliable delivery of messages. The resulting network protocols are more efficient than conventional communication libraries used in current NOW environments.

^ Hybrid Adaptive Algorithms for End System Middleware
K. Nahrstedt,* B. Kalter, B. Li
Defense Advanced Research Projects Agency, F30602-97-2-0121

Current distributed multimedia applications demand Quality-of-Service (QoS) from the supporting system. However, within the QoS demands, lower level transport facilities may not constantly provide guaranteed QoS without perturbation. In this scenario, researchers are investigating hybrid adaptive algorithms in the middleware level of end systems to perform QoS adaptation on a critical QoS metric. The research concentrates on analysis of QoS adaptation in dependence of system resource availability changes by applying theories from digital control systems.

^ Enabling Network-Aware Middleware for the Next Generation Internet: Uniform Detection, Notification, and Adaptation Mechanisms for Grid Environments
D. A. Reed,* M. Medina, S. Peechu
U.S. Department of Energy, DEFC02-99ER25401

To catalyze the development of both network-aware middleware and sophisticated, network-aware applications, this project is exploring uniform detection, notification, and adaptation mechanisms that can be applied across layers. The project aims to exploit the substantial technology base embodied in the Pablo Scalable Performance toolkit and the Globus "Grid" services toolkit. The goal is to use these technologies to create network-aware middleware that can sustain performance both by adapting network protocol and middleware parameters to minimize the adverse effect of transitory resource demand and availability and by adapting itself to evolving network behavior.

^ Human-Computer Interactions and Interfaces Collaboration Support for the Development and Evolution of Complex Systems
D. A. Reed,* S. Kaplan,* H. George, Y. Huang, C. Lu, B. Schaeffer, E. Shaffer, S. Whitmore, S. Wu
Defense Advanced Research Projects Agency, F30602-96-2-0264

Complex systems are developed and evolved by groups, often separated in time and space and using disparate tools and techniques. No circumscribed tool suite can encompass all future needs. Instead, software support and collaboration environments must themselves evolve and grow to meet the needs of collaborative groups. As a solution to this problem, this research is focused on developing an extensible collaboration framework that supports the full range of activities undertaken by current team members, integrates the wide variety of tools and artifacts they use to realize their goals, and supports the evolution of the development team and its processes over time.

^ Intelligent Information Spaces: A Testbed to Explore and Evaluate Intelligent Devices and Augmented Realities
D. A. Reed,* R. H. Campbell, R. Kravets, D. Mickunas, K. Nahrstedt, L. Sha
National Science Foundation, EIA 99-72884

To support information environments where ubiquitous, intelligent devices unobtrusively share data, preferences, and contexts about users and their movement among environments, this project is developing interoperable component architectures for device coordination, seamless object communication for user quality of service, and adaptive user context and modality management. The goal is to define a software architecture capable of enabling a mobile, responsive, and contextual information environment where a broad collection of high-end data display and visualization systems, low-power mobile devices, and "smart" devices with widely varying capabilities are seamlessly integrated using dynamically tailored software components.

^ Virtual Environments for Direct Software Manipulation
D. A. Reed,* D. Cox, J. Estabrook, M. George, Y. Huang, C. Lu, B. Schaeffer, S. Whitmore, W. Wu
Defense Advanced Research Projects Agency, DABT63-96-C-0027

Current software tools provide an increasingly myopic, unnecessarily limited window on the complex world of large-scale heterogeneous metacomputing, with few options to study application software structure or to modify behavior during execution. To address this problem, researchers are developing virtual environments that allow software developers to directly manipulate software components and their behavior while immersed in scalable, hierarchical representations of software structure and real-time performance data. Such a virtual environment will allow users to explore 3-D representations of executing software, interactively replace software components via physical manipulation, and modify software parameters.

^ Wide-Area, Adaptive I/O Systems for Data and Visualization Corridors
D. A. Reed,* F. Rothganger, E. Shaffer, S. Whitmore
U.S. Department of Energy, LLNL B505214

Working with the DoE's Accelerated Strategic Computing Initiative (ASCI) laboratories, this effort is extending and deploying the PPFSII toolkit to support DVC visualization, focusing on key issues in distributed caching and prefetching, access pattern classification, and distributed access. Further, to focus research on I/O problems most likely to benefit DVC visualization and immersive display systems, it will integrate this toolkit with other advanced visualization software and hardware, including CAVERNsoft, Virtual Director, and PC-based Scalable Display Walls, for collaborative visualization of grid-based data sets. Additionally, a series of extended PPFSII prototypes will be tested and deployed.


Summary of Engineering Research