Real Life experience of RAC scalability with a 6 Node Cluster

Map Unavailable

Date(s) - 19/02/2008
12:00 am UTC

Live Event
Eric Grancher (CERN)


CERN, the world’s largest particle physics laboratory has high requirements in term of database scalability with applications linked with its new accelerator, the LHC. Oracle RAC is being used for many applications to support the LHC. Obtaining scalability with Oracle RAC is a non trivial process for applications which are not ¿read-mostly¿. This paper describes, amongst other experiences with RAC scalability and pitfalls, the work which has been performed on the CERN control system ¿Oracle archiver¿ (the controls software used for LHC and its experiments). With this system, the performance of the event archiving module has been increased from 100 to 150000 event changes per second (x1500). This has been achieved with several architectural changes (core program separation from the data manipulation, data loading techniques, and database schema). The result is a near-linear scalability between the number of nodes in the cluster and the performance. This paper describes the methodology and the set of ideas developed and used in order to obtain almost linear scalability for some of the CERN typical database workloads.


Leave a Reply