Xavier Grehant presented "A new approach to Grid execution environments" where he discussed ways of allowing users to have more control over the resource scheduling. The classical job scheduling proposes a two-step push model with little or no flexibility given to the users. A more flexible approach has already been established by experiments, such as ATLAS, who usesa glide-in approach and Xavier proposed that virtual machine technology may be used to improve the situation further.

Håvard Bjerke covered "Process Migration using Virtual Machines". He described the "OS Farm" development which allows images to be generated and stored. He went on to discuss the networking issues when processes are migrated and continued by describing "Content-Based Addressing", a scheme where only the blocks that are unique to the source machine and not available on the target, have to be copied. In certain cases more that 50% of the traffic can be avoided with this scheme.

Jose Dana presented the current situation with TYCOON at CERN. The user meeting in Manchester had increased interest in this market-driven approach, and this summer (with the help of a student) he will be performing scalability tests. The collaboration with HP Labs is good and he hopes to receive soon a version of TYCOON that works with our Scientific Linux distribution. This will then be used as a base for deployment tests that will also involve BalticGrid.

S. Jarp
CERN openlab CTO