Cluster computer are the prevailing type of high performance computers. They are built of custom off-the-shelf processor boards that are connected by a high speed interconnection network. Although usually locally integrated, they are conceptually distributed systems with local operating system images. Their enormous potential, however, can only be exploited, if program code and data are optimally distributed across the nodes. Cluster management mechanisms also need to be scalable to be employed in systems with thousands of nodes. The lecture course gives an overview of the architecture of cluster computers and the related management problems for which algorithmic solutions are presented. It runs in parallel to the integrated course Parallel Programming. Both complement each other and form the module Parallel Systems.
Heiss, H.-U.: Prozessorzuteilung in Parallelrechnern, BI-Verlag, Mannheim, 1996
Andrews, G. A.: Foundations of Multithreaded, Parallel and Distributed Programming, Addison-Wesley, 2000
Pfister, G.: In Search of Clusters 2nd ed., Prentice Hall, 1998
Zomaya, A.: Parallel and distributed computing handbook, McGraw Gill, 1995
Buyya, R.: High Performance Cluster Computing, Vol. 1+2, Prentice Hall, 1999
Computer Science Master students
Experience with computers and software as well as programing skills.
The course language is German (or English if requested).
The exam will be formulated in German, but answers may be given in English, too.
Credits & Exams
The criteria for gaining credits are
active participation in the tutorials: regular preparation of assignements & presentation of results in the tutorials