Bonn 2010 – wissenschaftliches Programm
HK 70.5: Vortrag
Freitag, 19. März 2010, 15:00–15:15, HG ÜR 6
Complexity management for heterogeneous computer clusters in case of the ALICE High Level Trigger cluster — Camilo Lara, •Timo Breitner, Udo Kebschull, Stefan Boettger, and Marian Hermann — Kirchhoff-Institut for Physics, Heidelberg University, Germany
Heterogeneous computer farms are used for processing data in the ALICE HLT. A high complexity of managing a heterogeneous environment is the result of the increasing number of hardware and software releases and the resulting number of needed management resources for their management. For system management using the SysMES framework two object-oriented and Common Information Model based models have been developed: One of them describes the cluster environment whereas the other one describes the resources used for managing the former, e.g. monitors and rules. The next step for handling the mentioned complexity was the definition of relationships between those two models e.g. between an object which describes a CPU and another one describing a monitor object controlling the CPU's temperature. Our implementation provides a Java API used internally by the SysMES server to access the models in order to identify pairs (device, management resource) and to deploy the resources to their target devices. A relational database was chosen as a back-end, due to the requirements of our management system like transactionality and data integrity. For its integration with the object-oriented model and API a specifically designed object-relational mapping formalism has been developed.