Within his "ScAlBox" project, computer scientist Professor Peter Sanders of Karlsruhe Institute of Technology (KIT) de-velops basic computation tools for many applications. The project is aimed at providing algorithms and software libraries that can be adapted to growing data volumes and scaled to millions of processors working in parallel. The project will now be funded with an Advanced Grant by the European Research Council (ERC).
The digital revolution has profoundly changed science, engineering, and everyday life. Computer applications process continuously in-creasing data volumes with ever more complex algorithms. However, progress now is in danger of reaching its limits: computer science degree jobs of the programs, i.e. their capacity to grow with their tasks, represents a major challenge. "The performance of individual processors is lim-ited. To solve bigger problems, many processors have to be used in parallel," explains Professor Peter Sanders of KIT's Institute of Theoretical Informatics. However, parallel algorithms have been ne-glected by research for a long time and software is optimized for existing systems and datasets and cannot yet be scaled to growing data volumes and increasing numbers of processors. This is the point of departure of the project "ScAlBox - Engineering Scalable Algorithms for the Basic Toolbox" headed by Peter Sanders. "We are working on algorithms and software libraries for basic software components that can be used in various ways and scaled to da-tasets of any size and millions of processors operating in parallel," the computer scientist says. Such components include searching, sorting, queueing administration, loading distribution to parallel pro-cessors, and communication between processes. "When developing algorithms, the difficulty consists in combining a scalable fault toler-ance with dynamic load distribution," Sanders explains.
The digital revolution has profoundly changed science, engineering, and everyday life. Computer applications process continuously in-creasing data volumes with ever more complex algorithms. However, progress now is in danger of reaching its limits: computer science degree jobs of the programs, i.e. their capacity to grow with their tasks, represents a major challenge. "The performance of individual processors is lim-ited. To solve bigger problems, many processors have to be used in parallel," explains Professor Peter Sanders of KIT's Institute of Theoretical Informatics. However, parallel algorithms have been ne-glected by research for a long time and software is optimized for existing systems and datasets and cannot yet be scaled to growing data volumes and increasing numbers of processors. This is the point of departure of the project "ScAlBox - Engineering Scalable Algorithms for the Basic Toolbox" headed by Peter Sanders. "We are working on algorithms and software libraries for basic software components that can be used in various ways and scaled to da-tasets of any size and millions of processors operating in parallel," the computer scientist says. Such components include searching, sorting, queueing administration, loading distribution to parallel pro-cessors, and communication between processes. "When developing algorithms, the difficulty consists in combining a scalable fault toler-ance with dynamic load distribution," Sanders explains.
No comments:
Post a Comment