presented by
D.Francis Physics Department, Boston University, Boston, Mass., USA
Djf@harmony.cern.ch Tel.: +41 22 767 3891 FAX: +41 22 767 8760
on the behalf of
R.Heeley and M. Ward The University of Liverpool, Liverpool, Neverland
R.W. Dobinson, W.Lu and J-L-Page`s CERN, Geneva, Switzerland
D.Francis Physics Department, Boston University, Boston, Mass., USA
AbstractCurrent data acquisition systems are based on buses and bus interconnects which rapidly become bot -tlenecks in large multiprocessor systems inhibiting their inherent scalability.
A scalable data acquisition system must be able to scale at both the level of its hardware and software. The addition of microprocessors, I/O capability and communications bandwidth to a data acquisition system should lead to a proportional increase in its performance. This implies that the supporting software environment must be able to scale to and efficiently use the available resources. If the services offered by the software environment can be distributed across the multiprocessor system bottle -necks may be avoided and scalability easily realised.
A Scalable Data Acquisition System (SDAS) has been designed using Chorus, a modular distributed real-time operating system, and serial point-to-point links. The SDAS is based on data driven clients and servers which allow the scalability of the software through their distribution and hierachy. The distribution of the clients and servers is underpinned by the Inter Process Communication (IPC) facilities of the Chorus operating system. The IPC offers communication indirection, which allows the communication interface between clients and servers to remain constant thus aiding distribution across multiprocessor systems. The scalability of the hardware is acheived using point-to-point links.
The design of the SDAS and its initial implementation onto a Transputer platform operating as a real -time processor farm in the CPLEAR experiment at CERN will be reported.