The Dataflow System of the ATLAS DAQ/EF ”-1” Prototype Project G. Lehmann 1,3 , G. Ambrosini 3,9 , E. Arik 2 , H.P. Beck 1 , S. Cetin 2 , T. Conka 2 , A. Fernan- des 3 , D. Francis 3 , Y. Hasegawa 4 , M. Joos 3 , J. Lopez 3,10 , A. Mailov 2 , L. Mapelli 3 , G. Mor- nacchi 3 , Y. Nagasaka 7 , M. Niculescu 3,5 , K. Nurdan 2 , J. Petersen 3 , D. Prigent 3 , J. Rochez 3 , L. Tremblet 3 , G. Unel 3 , S. Veneziano 3,6 , Y. Yasu 8 1 Laboratory for High Energy Physics, University of Bern, Switzerland 2 Department of Physics, Bogazici University, Istanbul, Turkey 3 CERN, Geneva, Switzerland 4 ICEPP, University of Tokio, Tokio, Japan 5 Institute of Atomic Physics, Bucharest, Romania 6 I.N.F.N. Sezione di Roma, Roma, Italy 7 Nagasaki Institute for Applied Science, Nagasaki, Japan 8 High Energy Accelerator Research Organization (KEK), Japan 9 Now at Lightning Instrumentation S.A., Lausanne, Switzerland 10 Now at EDF, Grenoble, France Abstract In 1996 the Data Acquisition (DAQ) group of the ATLAS Collaboration started a project for the design and implementation of a full DAQ/Event Filter (EF) prototype, based on the Trigger/DAQ architecture described in the ATLAS Technical Proposal. The aim of this pro- totype was to allow for hardware and software technology investigations as well as their inte- gration aspects in order to reach maturity for the final ATLAS DAQ system design. Being a pre-design prototype it is referred to as ATLAS DAQ Prototype ”-1”. It consists of a ”vertical” slice of the ATLAS DAQ/EF architecture, including all the hardware and software elements of the data flow, its control and monitoring as well as all the elements of a complete DAQ system, from the detectors Read Out Driver to data recording. This paper describes the dataflow component of the prototype, its design, implementation and performance. More emphasis is given to the description of the Event Builder, the sub-system which merges data fragments coming from different detector parts into full events, since this is the critical element for the scalability to ATLAS sizes of the proposed architecture. Results of modelling studies based and tuned on the present implementation will be shown. Keywords: ATLAS, DAQ, event builder, LHC, switching network, VME bus 1 Introduction The data flow component of the ATLAS [1] DAQ-1 system [2] is responsible for moving the event data from the detector read-out links to the final mass storage, interacting with the three ATLAS trigger levels: level one (LVL1), level two (LVL2) and Event Filter (EF) [3]. It also provides event data for monitoring purposes and implements local control for the various data flow elements. A global view is shown in figure 1. Three main functions are provided by the data flow, namely: • the Front End DAQ which buffers the event data coming from the detector, possibly per- forms some data collection with the associated event formatting, and interfaces with the second level trigger • the Event Builder (EB), which merges data fragments coming from the different detector parts into full, formatted events, • the Farm DAQ which provides data flow support to the processor farms of the event filter and interfaces with the mass storage system.