arXiv subjects

M. Ellert

Publications and source records attributed to M. Ellert.

5 recordsLinked to original sources

arXiv2023

Intel Stratix 10 FPGA design for track reconstruction for the ATLAS experiment at the HL-LHC

The fast reconstruction of charged particle tracks with high efficiency and track quality is an essential part of the online data selection for the ATLAS experiment at the High-Luminosity LHC. Dedicated custom designed hardware boards and software simulations have been developed to assess the feasibility of a Hardware Tracking Trigger (HTT) system. The Pattern Recognition Mezzanine (PRM), as part of the HTT system, has been designed to recognize track candidates in silicon detectors with Associative Memory ASICs and to select and reconstruct tracks using linearized algorithms implemented in an Intel Stratix 10 MX FPGA. The highly parallelized FPGA design makes extensive use of the integrated High-Bandwidth-Memory. In this paper, the FPGA design for the PRM board is presented. Its functionalities have been verified in both simulations and hardware tests on an Intel Stratix 10 MX development kit.

physics.ins-det↗

arXiv2003

Atlas Data-Challenge 1 on NorduGrid

The first LHC application ever to be executed in a computational Grid environment is the so-called ATLAS Data-Challenge 1, more specifically, the part assigned to the Scandinavian members of the ATLAS Collaboration. Taking advantage of the NorduGrid testbed and tools, physicists from Denmark, Norway and Sweden were able to participate in the overall exercise starting in July 2002 and continuing through the rest of 2002 and the first part of 2003 using solely the NorduGrid environment. This allowed to distribute input data over a wide area, and rely on the NorduGrid resource discovery mechanism to find an optimal cluster for job submission. During the whole Data-Challenge 1, more than 2 TB of input data was processed and more than 2.5 TB of output data was produced by more than 4750 Grid jobs.

physics.comp-ph↗

arXiv2003

The NorduGrid architecture and tools

The NorduGrid project designed a Grid architecture with the primary goal to meet the requirements of production tasks of the LHC experiments. While it is meant to be a rather generic Grid system, it puts emphasis on batch processing suitable for problems encountered in High Energy Physics. The NorduGrid architecture implementation uses the \globus{} as the foundation for various components, developed by the project. While introducing new services, the NorduGrid does not modify the Globus tools, such that the two can eventually co-exist. The NorduGrid topology is decentralized, avoiding a single point of failure. The NorduGrid architecture is thus a light-weight, non-invasive and dynamic one, while robust and scalable, capable of meeting most challenging tasks of High Energy Physics.

physics.comp-ph↗

arXiv2002

An Overview of a Grid Architecture for Scientific Computing

This document gives an overview of a Grid testbed architecture proposal for the NorduGrid project. The aim of the project is to establish an inter-Nordic testbed facility for implementation of wide area computing and data handling. The architecture is supposed to define a Grid system suitable for solving data intensive problems at the Large Hadron Collider at CERN. We present the various architecture components needed for such a system. After that we go on to give a description of the dynamics by showing the task flow.

cs.DC↗

arXiv2002

Performance evaluation of the GridFTP within the NorduGrid project

This report presents results of the tests measuring the performance of multi-threaded file transfers, using the GridFTP implementation of the Globus project over the NorduGrid network resources. Point to point WAN tests, carried out between the sites of Copenhagen, Lund, Oslo and Uppsala, are described. It was found that multiple threaded download via the high performance GridFTP protocol can significantly improve file transfer performance, and can serve as a reliable data

cs.DC↗