Work Package V (AP-V): Resource Management for Grid Jobs
Within Work Package V we consider the distribution of incoming computing requests (jobs) in the grid onto the resources provided. These tasks could be seen as an intelligent generalisation of the functionality of batch or queue systems as they are usually provided on local high performance computers. However, in the grid the resource environment is generally much more complex and heterogeneous than for local resources. Hence there will be increased demands for user and programmer interfaces. This work package also integrates management of observing jobs at robotic telescopes, a special feature of astrophysics. There are three components or levels of job managament in the grid:
Description of requirements of grid jobs, such as expected resource consumption regarding compute time and memory, requests for special hardware, possible ways and requirements of parallel execution, informations for the start of a job (is there a preceding job which has to finish first?).
Job or Resource Brokerage
The job definition and its ensuing requirements have to be compared with resources existing in the grid and their momentary load. A strategy has to be defined and implemented into the software which distributes a number of jobs optimally among the resources.
Grid job descriptions, as they are passed via the resource broker to a local resource administration have to be translated into a locally understandable and submissable format.
Work Package Coordinator: Rainer Spurzem (ZAH)
Technical Communication Partners:
- R. Spurzem (ZAH)
- T. Brüsemeister (ZAH)
- D. Elstner (AIP)
- O. Wehrens (AEI)
- H.M. Adorf (MPA)
- W. Voges (MPE)
- A. Reiser (TUM)
- T. Roeblitz (ZIB)
- A. Beck-Ratzka (AEI)
- M. Hoegqvist (ZIB)
- F. Breitling (AIP)
- Specification of Requests and Design of Architecture
- Development of Grid Job Management
Grid Integration of Robotic Telescope
- Adaptation of User and Programmer Interfaces, Test