RESULTS FROM THE MACRAME 1024 NODE SWITCHING NETWORK
Paper: 168
Session: F (talk)
Speaker: Heeley, Roger, CERN, Geneva
Keywords: network monitoring, special architectures, switches (eg ATM, trigger algorithms, large systems
RESULTS FROM THE MACRAME 1024 NODE
SWITCHING NETWORK
R.W. DOBINSON, S. HAAS, R.HEELEY and M. ZHU
CERN, 1211 Geneva 23, Switzerland and
University of Liverpool, Liverpool, UK
B. MARTIN
CERN, 1211 Geneva 23, Switzerland
D.A. THORNLEY,
University of Kent, Canterbury, UK, and
CERN, 1211 Geneva 23, Switzerland
ABSTRACT
It is recognised that the data acquisition and trigger systems for LHC
experiments will depend critically on the use of high speed point to point
links and packet switches. To date, practical experience has been confined to
relatively small systems and there are no experimental results on how the
performance of such systems will scale up to several hundred or even several
thousand nodes.
The work reported involves the construction and testing of a large modular
setup using IEEE 1355 DS link and switch technology. Up to one thousand nodes
can be interconnected by a switching fabric based on the ST C104 packet switch
which interconnects 32 bi-directional ports. The system has been designed and
constructed in a modular way to allow a variety of different network
topologies to be investigated.
Network throughput and latency has been studied for various traffic conditions
as a function of the packet length, topology and network size. In particular
we present results of studies carried out with traffic patterns expected for
the ATLAS second and third level trigger systems.
The work is carried out within the framework of the European Union's Esprit
program as part of the MACRAME project (Esprit project 8603).