Responsive Systems
Responsive systems are single or multiprocessor systems which deliver a
satisfactory service on time under given load and fault hypotheses. Our
goal is research in responsive computing with real-world applications to
process control, transportation systems, and other areas of significant
economic impact and relevance.
Objectives
-
Development of foundations and paradigms for responsive computing, scheduling,
consensus and decision making
-
Design and implementation of a novel microkernel to support responsiveness
in new and existing systems (CORE - Consensus for Responsiveness)
-
Dynamic, responsive, and customized reconfiguration of application software
in presence of faults
-
Modeling and simulation of responsive systems
-
Application demonstrators for process control, transportation systems,
and transaction processing
Parallel Computing in Distributed Environments
Despite the existence of a wide variety of parallel computer architectures,
high costs, poor programmability, and insufficient software support remain
to be a problem. Furthermore, high computing power can also be obtained
from already installed networked workstations. Therefore we investigate
network-based parallel computers which integrate workstations into multicomputer-systems
interconnected by fast local area networks.
Objectives
-
Investigation of new paradigms for parallel programming
-
Tools for user-defined configuration and dynamic reconfiguration of parallel
programs
-
Design of a programming system for execution of parallel programs in distributed
environments
-
Development of new approaches for memory management (object-based, dynamically
configurable, distributed)
|
Communication in Responsive Systems
The service of responsive systems is determined by fault, load, and time
criteria. They have impact on design and function of a given communication
system. We examine the necessary conditions to reach an efficient communication
in responsive systems.
Objectives
-
Investigation of protocol structures in LAN/MAN/WAN, protocol design, switching
algorithms, LAN emulation
-
Network and system management in distributed computer systems
-
Design and implementation of communication software
-
Development of a testbed for responsive systems
-
Specific projects:
-
Join OSI IP networks, CLNS pilot net in GRN (German Research Net)
-
ATM and IP next generation for research nets
Modeling and Simulation
The objective is modeling and simulation of parallel/distributed computing
systems with emphasis on fault tolerance, timeliness, configurability effectiveness,
and load/capacity evaluation.
Objectives
-
Graph models, Petri Nets and other computation and communication models
-
Modeling and simulation of responsive computer systems
-
Modeling and simulation of memory management
-
Modeling and simulation of computer networks
-
Software/hardware configuration evaluation
-
Composability
-
Fault injection in distributed systems
Application-specific Systems
The goal is the specification, design and implementation of application-specific
computer/communication architectures, which are distinguished by their
outstanding performance and/or dependability in solving specific
problems.
Objectives
-
Application semantics, specification, algorithms
-
Hardware/software configuration for specific applications, e.g., multimedia
or video-on-demand
-
Tabu search engine (special hardware to accelerate search strategies in
computing-intensive permutations problems)
-
Hardware/software/algorithm interoperability
|
