Area of Research: Massively Parallel Software Systems

The research area Massively Parallel Software Systems is divided into three streams:

Software engineering methodologies for massively parallel architectures: The performance of single processing units has been stagnating for over a decade. Software parallelism has thus been steadily gaining relevance, beyond numerical simulations, across all computation-intensive domains of computer science. Efficient usage of parallel -- possibly even heterogenous -- architectures is however still far from being trivial. Efficient yet consistent access to shared data is particularly challenging in parallel software.

This profile area is hence concerned with developing programming methodologies and tools aiming at significantly simplifying the tasks of developing efficient parallel programs and verifying their correctness. The research activities therein focus both on high performance systems such as the Lichtenberg computer, as well as on broadly applicable server systems and embedded systems.

Software engineering methodologies for massively distributed systems: Our modern society is characterized by a growing demand for electronically assisted means of direct and indirect commmunication. The resulting interaction of devices inevitably leads to massively dynamic distributed systems, exacerbating existing bottlenecks in performance as well as in correctness, fault tolerance, and security.

The goal of this profile area consists in providing programmers with tools and techniques that enable the development of robust, reliable, and secure software for massively distributed systems. The twofold focus on fundamental issues in distributed systems and the programming level yields access to a variety of areas, e.g., sensor networks, cloud computing, and big data.

Static analysis, semantics, and formal verification: Even program libraries tested most thoroughly can harbor, over many years, unexposed faults and security vulnerabilities affecting millions of end devices. Parallel and distributed systems in particular quite commonly contain defects which are subtle and hard to find.

The correctness of concurrent programs is particularly challenged by the advent of new execution models (e.g., in many-/multicore architectures), beside the inherent conceptual intricacies of such programs. The researchers in this focus area thus develop a series of static analysis, test generation, and verification methods. Scalability, ease of use, and automation are particularly emphasized in the process.

Dean´s Office

TU Darmstadt
S2|02 Piloty-Building, Room D103
Hochschulstraße 10
64289 Darmstadt


A A A | Drucken Print | Impressum Impressum | Sitemap Sitemap | Suche Search | Kontakt Contact | Website Analysis: More Information
zum Seitenanfangzum Seitenanfang