Darmstadt 2016 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

HK: Fachverband Physik der Hadronen und Kerne

HK 54: Instrumentation XV

HK 54.5: Vortrag

Donnerstag, 17. März 2016, 15:00–15:15, S1/01 A2

The CBM First-level Event Selector, Timeslice Building and Availability Studies — •Helvi Hartmann, Jan de Cuveland, and Volker Lindenstruth for the CBM collaboration — Frankfurt Institute for Advanced Studies, Goethe University, Frankfurt, Germany

The Compressed Baryonic Matter (CBM) experiment is a fixed target high energy physics experiment collecting all produced data - no triggers are involved. This causes a very high data rate of 1 TByte/s. The First-level Event Selector (FLES) denotes a high performance computer cluster that will process all data and performs a full online event reconstruction. For this purpose the raw detector data is accessed in time intervals referred to as Timeslices. In the process of Timeslice building data from all input links are distributed via a high-performance Infiniband network to the compute nodes.

In order to ensure that the FLES is available the whole time while the detectors are running fault tolerance is inevitable. The questions is how often are incidents going to occur (Mean Time between Failure MTBF) and how long will they pause Timeslice building and hence physics analysis (Mean Time to Repair MTTR). These two factors make up the availability of the FLES. I would like to present a detailed analysis of possible sources of errors and their influence on the availability. Furthermore, I will discuss the development of Timeslice building on the basis of MPI with respect to the availability of the FLES. I will compare this approach to a low-level native Infiniband Verbs implementation combined with a socket-based error handling system.