|
|
|
Research is both enlightening and fun. In my experience, best results are obtained when working in a team:
With only one exception (NegaScout), I did all of my research jointly with international colleagues and/or
PhD students of my research group.
Ongoing
-
Distributed Data Management
Our attributed data management system ZIB-DMS is the core project of my research group. Headed by
Florian Schintke, several PhD and MSc students work on the design and implementation of algorithms to support
data distribution, replication, placement, attributation, availability, partial remote data access, etc.
-
Self Managing Overlay Networks
With Thorsten Schütt and Florian Schintke I work on high-level services and functions for inherently distributed systems
to widen the application domain of P2P computing beyond the naive file sharing.
Our two patented routing algorithms support multi-dimensional, non-rectangular range queries in peer-to-peer overlays.
-
Hardware Accelerators for High Performance Computing
We believe that future supercomputers will be heterogeneous with co-processors (FPGA, GPU, Cell, Clearspeed, ...) attached to multi-purpose processors.
We investigate high-level programming tools that allow HPC users to write code for their daily production runs on FPGAs.
This is joint work with Thorsten Schütt and Thomas Steinke (co-founder of OpenFPGA
).
-
Co-Reservation of Resources in the Grid
Large-scale scientific simulations often involve complex workflows with multiple resources. Together with Thomas Röblitz,
we devise methods for advance reservation and co-scheduling of multiple resources in the Grid. This work was partly
funded by an IBM Faculty Award in 2003.
-
Alignment and Threading
Thomas Steinke (head of our junior research group) leads the work on parallel threading algorithms.
We investigate high-level schemes for implementing core functions on the FPGA co-processors of our Cray XD1.
-
Miscellaneous
We also work on autonomic computing, transactions in the Grid, log-based data reconciliation,
erasure codes for WAN distributed data storage, ...
Previous
-
Scheduled Transfer on Gigabit Ethernet
We ported the low-level system-area protocol onto Gigabit Ethernet. Joint work with Jens Simon and Sebastian Heidl.
-
Scalable Coherent Interface
Together with Dolphin and Scali we worked on the SCI interconnect protocol.
It is used, for example, in Fujitsu-Siemens' hpcLine clusters. Book on SCI:
-
Computing Center Software CCS
We designed and implemented the cluster management software CCS. Joint work with Axel Keller et al. at Paderborn's PC2.
-
Solving Interval Constraint Problems
The algebraic method that I developed with Peter Ladkin in 1991 is - according to the
Handbook of Temporal Logic in Artificial Intelligence (2005) - still the fastest known method for
handling constraints in the full interval algebra. Here is one of our papers:
-
NegaScout
A minimax algorithm faster than Alpha-Beta which is still used in many game playing programs. First published in my MSc thesis in 1982.
|
|