Main menu >> Computer Science >> Operating Systems

Computer Science

Operating Systems
^ ACCORDS (Automatically Configurable, COmponent-based, Reflective Distributed Systems)
R. H. Campbell,* M. D. Mickunas
Cooperative project with Departments of Computer Science of the University of Sao Paulo and the University of Campinas

The goal of this project is to apply knowledge about component requirements and dependencies to permit the automatic configuration of software systems with little or no intervention. The research will develop formal models and tools for reifying the interactions between components, producing an infrastructure for automatic configuration of distributed component-based systems and a framework for collecting, analyzing, and visualizing information about component interaction. Quantitative information about component interaction will help to decide when the system must be reconfigured. The infrastructure for automatic configuration will guarantee that the new configuration will be achieved affordably, reliably, and with the best quality of service.

^ LegORB
R. H. Campbell,* M. D. Mickunas,* F. Kon, M. Roman
University of Illinois

LegORB is a component-based reflective ORB that uses a microkernel-like approach, where only essential components are offered. The core of LegORB is a component configurator that provides the glue to hold all the components together. LegORB permits connection of different components belonging to different categories. Each category defines a specific functionality and provides different implementations (components). LegORB defines a standard set of categories that can be extended to provide new functionality. Components can be changed on-the-fly. The number of components and the way they are interconnected defines the behavior, functionality, and size of the ORB.

^ Renaissance UNIX: An Architecture for Robust UNIX System
R. Campbell,* P. Apu, S. Chetan, B. He, C. Li, D. Raila
Hewlett-Packard
roy@cs.uiuc.edu

This project is designed to research an architecture for a robust UNIX system. The UNIX kernel is a monolithic structure that makes debugging, error containment, resiliency, information assurance, security, and dynamic reconfiguration difficult. These problems are exacerbated for systems with large numbers of processors, peripherals, storage, and complex applications. Goals of this research are to provide a platform for debugging, error containment, resiliency, information assurance, and integrity effectively and efficiently within enterprise class systems. The approach is to create a management service and interface that provide local and remote access and control to system management and system configuration, debugging, and analysis. Researchers are building a lightweight run-time environment that virtualizes key UNIX components, subsystems, and hardware. Existing UNIX subsystems are supported by and interact with the framework for debugging, error containment, resiliency, information assurance, and security. The run-time environment will assure continuous operation of the management service, irrespective of the behavior of the kernel.

^ Application/System QoS Interface Capabilities
K. Nahrstedt,* W. J. Jeon, B. Kalter, J. H. Seo
National Aeronautics and Space Administration, NAG 2-1250

Researchers are investigating application-system QoS interface capabilities for visual tracking distributed applications. The interface between the application and the underlying QoS-aware resource management system must provide several important functionalities: application QoS application programming interface, translation between the application QoS into the system QoS parameters, integrated reservation coordination policies and protocols to avoid and prevent deadlock situations, adaptation policies and their application enforcement, and others. These functions will reside in the end-system management entity called the QoS Broker, which represents the application/system interface for provision of end-to-end QoS guarantees.

^ End-to-End QoS over Heterogeneous Networks
K. Nahrstedt,* R. H. Campbell,* Y. Zhou
Motorola Inc.

This research focuses on investigating Quality-of-Service (QoS) translations in edge devices, connecting heterogeneous networks, and end-to-end setup protocols over heterogeneous networks. The current topology consists of Ethernet and cable LAN networks that are interconnected via an ATM backbone. The LAN networks run RSVP/IP network protocols which need to be translated into ATM service classes and the corresponding protocols once they arrive at the ATM edges. Researchers investigate the QoS translation between RSVP, ATM, and cable networks, the end-to-end connection setup between the receiver-oriented RSVP protocol and the sender-oriented ATM signaling protocol, and impact on QoS between the asymmetric cable channels and symmetric data network channels.

^ QoS Issues in the 2K Distributed Operating System
K. Nahrstedt,* D. Wichadakul, D. Xu
National Science Foundation, EI S98-70736

This project investigates 2K, a Quality-of-Service (QoS) model for a distributed operating system. The QoS model will be encapsulated in a QoS proxy, which will provide proper QoS translation and contracting between the application and the 2K system as well as provide proper reservation capabilities to access distributed services. Issues such as service coordination, resource coordination, reservation protocols, and service adaptation must be present to allow for QoS and guarantees services in the distributed operating system.

^ QoS Routing
K. Nahrstedt,* J. Qian, L. K. Shan
Defense Advanced Research Projects Agency, F30602-97-2-0121

The task of Quality-of-Service (QoS) routing is to find a path in the network that satisfies constraints on such metrics as bandwidth, delay jitter, and cost. This study focuses on QoS routing algorithms and their design within routers. The problem of finding a path with constraints on two or more additive metrics (delay and delay jitter) is NP-complete. This research concentrates on heuristic algorithms and study of the family of distributed and hierarchical routing algorithms to solve the multiconstrained routing problem. The QoS routing solutions are applied to point-to-point as well as multicasting scenarios.

^ QoS-Aware Resource Management
K. Nahrstedt,* K. Kim, A. K. Viswanathan, J. Wang
Partnerships for an Advanced Computational Infrastructure

Operating systems and communication systems need new algorithms, services, and protocols to support processing of audio/visual streams according to QoS specification. This project concentrates on the CPU brokerage service with advanced reservation, admission, scheduler, and adaptation control for soft real-time and nonreal-time tasks. At the communication level researchers provide IntServ bandwidth brokers in the edge networks and DiffServ brokers within the backbone routers to provide end-to-end guarantees.

^ Unstructured Mesh Generation for ASCI's Center for Simulation of Advanced Rockets
S. H. Teng*
DOE Lawrence Livermore National Laboratory

Numerical methods approximate a continuous physical problem by discretizing the geometric domain into single elements. The most versatile type of mesh is the unstructured triangular mesh where each element is a simplex. Not all discretization serves equally well, however. The approximation errors depend on the geometric quality of the mesh, which is often measured by the aspect ratio of elements. The simplest mesh is the regular Cartesian grid, as used in finite-difference methods. However, regular grids are limited to problems with simple domains and smooth changes in solution. For problems with complex geometry boundaries and solutions that change rapidly, researchers are studying algorithmic issues for unstructured three-dimensional mesh generation.


Summary of Engineering Research