Parallel and Distributed Systems
Research Group

Faculty

Thomas M. Stricker (thomas.stricker@inf.ethz.ch)
Assistant Professor, Ph.D. and M.Sc. Carnegie Mellon University Pittsburgh, PA, USA.
Now a research staff member at the Google European Engineering Centre, Zürich.
Curriculum Vitae as Adobe PDF

Ph.D. students and recent graduates

• Christian Kurmann (christian.kurmann@inf.ethz.ch)
  Dr. sc. techn. (12/2002), Dipl.Informatik-Ing. (1996), ETH Zürich.
  Now a research staff member at headquarters of Google Inc. in Mountain View CA.
• Michela Taufer (michela.taufer@inf.ethz.ch)
  Dr. sc. techn. (10/2002) ETH Zürich, Laurea in Ingegneria Informatica (1996), Universita di Padova, Italia.
  Moved to the University of California / Scripps Institute on 1/1/2003. Holds a faculty position at the University of Texas since 1/1/2005.
• Felix Rauch (felix.rauch@inf.ethz.ch)
  Dr. sc. techn. (11/2003), Dipl.Informatik-Ing. (1997), ETH Zürich
  works with NICTA and moved to the University of New South Wales, Sydney, Australia
• Irina Chihaia (irina.chihaia@inf.ethz.ch), co-advised with Thomas Gross.
  Dipl. Ing. (1999), Techn. Univ. of Cluj-Napoca, Romania.
  Now a research staff member at the Google European Engineering Centre, Zürich.

About our Research

Our research interests are in parallel and distributed computing. Our current plans include research in computer architecture (memory systems modeling) and in system software (an infrastructure for parallel and distributed applications). We are currently building a system based on commodity PC hardware connected by a Gigabit networking technology with good scalability for several applications in scientific computing, distributed databases and everyday computing.

About our Vision on the Future of Computing

While the computer hardware itself is getting faster and faster, steadily and according to a well predictable technology curve, the overall performance of a system remains dominated by system software overheads. Solving the problem of software overheads of a large system is more complex and challenging, than increasing clock speed, the capacity of memory or the instruction issue rates of the microprocessors. Commodity hardware components found in high performance desktop computers (currently based 1-8 Intel Pentium III) and new standard network technologies (currently Gigabit Ethernet, 1GB SCI/Myrinet and 622 MByte/s ATM) are becoming commodities and are available at an excellent price/performance ratios. Such systems will give us an cost effective hardware test bed for our investigation of overall system performance. Optimizing a parallel or distribute system as a whole still requires many research contributions of better architectural models, better software structures and more efficient programming at the system software level.

PDS Lectures

• 37-836 Informatik II, Algorithms, Data structures and Problem solving for EE students (IIIb), SS99, homepage.
• 37-235 Computer Systems Performance Analysis and Benchmarking, WS01/02 homepage.
• 37-023 Systemprogrammierung (low level programming),WS01/02, homepage.

Public inauguration lecture of Prof. Thomas M. Stricker in the Auditorium Maximum, HG F30, April 1997 at ETH Zürich.
The transparencies for this lecture are available for download: (stricker-antritt.pdf, [1.5 MB]).

PDS Research projects The  - Project   A Cluster of PCs, interconnected by a Gigabit/s Network : Our CoPs-Project investigates architectural and operating system support for parallel and distributed computing on PC clusters interconnected with a Gigabit/sec network. Recent projects unter the umbrella of CoPs are the following: CoPs Related Projects:

• ECT memperf - Extended Copy Transfer Characterization
  ECT memperf is a Memory Performance Characterization Toolkit and Benchmark. The method allows to characterize the performance of memory systems by capturing two aspects of the memory hierarchy. First its behavior with temporal locality by varying the working set size and second the spatial locality by varying the access pattern. (Memory Performance Characterization Toolkit and Benchmark)
• Patagonia CloneSys and Dolly
  CloneSys is a free tool to install multi-boot environments by partition cloning. Using the streaming tool dolly, which implements a multi-drop chain, the tool provides a very efficient method to distribute OS-images. The CloneSys tool is especially usefull for Cluster Installations. It is used at our CS-Departement to maintain more than 100 Multiboot Student-PCs as well as the Xibalba- and CoPs-Cluster.
• Zero Copy Enabled Gigabit Ethernet
  In order to reach the full Gigabit/s speed on normal PCs with their typically weak memory subsystems requires either additional hardware for protocol processing or alternatively a highly efficient software system that circumvents data copies. We successfully introduced speculation techniques into system software design and managed to implement a clean zero-copy solution entirely in software that runs with commodity network interface cards (NICs).
• Xibalba, our recent work in computer architecture
  Xibalba is a 128 x 1GHz multiboot cluster, which is used for a variety of research within the departement of computer science. The systems is based on commodity PC hardware connected by a Gigabit networking technology with good scalability for several applications in scientific computing, distributed databases and everyday computing.
• A Comparative Study of Gigabit Networking Technologies for PCI Bus Based PCs
  A fair comparison of different Gigabit/sec network technologies needs a common denominator as a basis. We propose to carry out a comparison at three different levels: (1) simple remote load/store operations, (2) message passing libraries and (3) standard IP over LAN networking. Our comparison includes the two Gigabit Interconnects for the PCI bus that are targeted at clusters of PCs, Myrinet and Dophin SCI. For a reference the interconnect technology of the Cray T3D is evaluated with the same benchmarks.
• Opam - Optimizing Parallel Applications despite Black-Box Middleware
  This project presents a novel performance method to reduce the complexity and improve the quality of performance evaluation on clusters of PCs. A single framework structure for workload characterization and performance predictions is introduced and deals with different application domains (e.g. scientific computation, distributed database) and with different standard middleware packages (e.g. open source middleware packages such as MPI and PVM, black-box DBMSs such as ORACLE).
• On the Migration of the Scientific Code Dyana from SMPs to Distributed Platfroms
  In this project, we present a systematic method to study the viability of a code migration from SMP platforms to commodity and high-end clusters of PCs and the grid well in advance and demonstrate it with the highly parallel molecular biology code, Dyana.
• Parallelization of CHARMM on a LINUX-PC Cluster
  Analysis of the resource requirements of a CHARMM molecular dynamics simulation on PC clusters with a particle mesh Ewald (PME) treatment of long-range electrostatics, and investigation of the scalability.

PDS Publications

• Papers in refereed conferences and journals
• Technical reports

Student Masters Theses (Diplomarbeiten) or Projects (Semesterarbeiten)

Students of all experience levels are welcome to take part in our research and development!

Independent Research (eigene Semester and Diplomarbeiten):

Our group also welcomes student proposals for independent research. If you think, that you might have a good idea for a project, simply send a one page proposal to me for a review. We look forward to coach any promising student research project within the area of our interest and expertise (computer architecture, benchmarking, high speed networking, certain aspects of compiler construction).

Abstracts of the latest student projects done in the PDS group.

PDS Lab Facilities

The CoPs Lab was once in RZ J3 and had a beautiful view - but it did not help the productivity much, so we relocated within 30 yards of the capuccino machine... After refurbishing the view is back for general PC work in the J3 Lab.

• Hardware- and Software-Infrastructure of our Computer Cluster.
• Take a look into our Computer Lab and out of its windows.
• A few HOWTOs for the usage of the CS-ZZ-Cluster (ETH intranet only).

ETH Zürich: Department of Computer Science
Last Update: December, 2003, (fr)
Hrsg. Thomas Stricker,< thomas.stricker@inf.ethz.ch>