Professor

Carsten Griwodz

Head of Department
Mobile: +47 926 84 380
Office: +47 67 82 83 83
E-mail: griff@simula.no

Carsten Griwodz leads the Networks and Distributed Systems Department at the Norwegian research company Simula Research Laboratory AS, Norway, and is Professor at the University of Oslo. He received his Diploma in Computer Science from the University of Paderborn, Germany, in 1993. From 1993 to 1997, he worked at the IBM European Networking Center in Heidelberg, Germany. In 1997 he joined the Multimedia Communications Lab at Darmstadt University of Technology, Germany, where he obtained his doctoral degree in 2000. His interests lie in the improvement of system support for interactive distributed multimedia, with heterogeneous parallel processing, operating systems and protocol support for streaming applications and multiplayer games in particular. He leads the StorIKT project Verdione that investigates system support for the World Opera, and is member of the Center for Research-based Innovation "Information Access Disruptions" that develops next generation search technology.


Research Projects

  • MiSMoSS: Middleware Services for Management of Shared State in Large-Scale Distributed Interactive Applications
  • ADIMUS: Adapted Internet Multimedia Streaming
  • HyStream: Hybrid Peer-to-Peer Technology for Scalable Commercial Video Streaming in the Internet
  • iAd: Information Access Disruptions
  • ROMUS: Robust Multimedia Streaming Services
  • Verdione: Virtually Enhanced Real-life synchronizeD Interaction - ON the Edge

Media distribution

Carsten is interested in several aspects of media distribution, where media has so far meant mainly audio and video. His interests are in

  • Distribution mechanisms and their interaction with replication mechanisms, especially simple ones such as caching.
  • Investigation of the optimal placement and dimensioning of replica or cache servers using analysis and simulation.
  • Operating system support for the efficient implementation of servers and caches, including the implementation of caches on embedded systems.
  • Protocol support that is efficient but compatible with existing real-world protocols.
  • Investigation of the influence of system- and network-level QoS support or the lack thereof.
  • The influence of existing AV-codecs on replication mechanisms. E.g., partial transfer of data could be implemented efficiently with a layered codec. MPEG4 or Quicktime may be packaging formats for alternative encodings. Investigation of the question which information a cache server must have at least to exploit such encoding formats.

Quality of Service and Perceived Consistency

Multimedia is all about faking the users' experience sufficiently well to give the expected impression. This has been exploited in compression, where JPEG and MP3 are well-known examples. Some investigations have been made to find the limits of the necessary synchronity of media streams. In distributed applications, a lot of research has gone into guaranteeing consistancy. If you consider distributed multimedia, though, there is no real need to aim at perfect consistency of data that is presented to the users. Rather, it is interesting to investigate the options to support the developer (or the user) in exploiting the limits of the users' perception.

  • Develop a transport system that supports application-specific requirements. The requirements are specified on a per-object basis. The first requirements are expressed as the reliability of the transfer, the delay of the transfer. The implementation should hide a distribution infrastructure that interconnects hosts across links that require TCP, allow UDP, or even support IP multicast.
  • Investigate the limits of users' perceptions. The idea is to use a distributed game for this.
  • Analyse the requirements of mapping the limits of perceived consistency to system- and network-level QoS-guarantees.
  • Develop a notation and build tools to support the specification and the test of perceived consistency limits at development time. The development should consider issues of object naming, location and deletion.
  •  

Personal tools