AMANDA - first running experiment to use GRID in production

TL;DR

This paper reports on the first operational use of GRID computing in production environments for physics experiments, demonstrating its feasibility and integration without reprogramming existing software.

Contribution

It presents a practical implementation of GRID technology in two large-scale physics experiments, highlighting methods to integrate existing software without modification.

Findings

Successful deployment in AMANDA and D0 experiments

Effective handling of software dependencies and job management

Demonstrated feasibility of GRID in production physics environments

Abstract

The Grid technologies are in ongoing development. Using current Grid toolkits like the Globus toolkit gives one the possibility to build up virtual organizations. Although these tookits are in still under development and do not feature all functionality, they can already now be used to set up an efficient computing environment for physics collaborations with only moderate work. We discuss in this paper the use of such a computing structure in two running experiments - the AMANDA (AMANDA = Antarctic muon and neutrino detector array) neutrino telescope and the D0 experiment at Tevatron, Fermilab. One of the main features of our approach is to avoid reprogramming of the existing software which is based on several programming languages (FORTRAN, C/C++, JAVA). This was realized with software layers around the collaboration software taking care about in- and output, user notification, tracking of running jobs, etc. A further important aspect is the resolution of library dependencies, which occur when a user runs self-compiled jobs on machines, where these libraries are not installed. These dependencies are also resolved with this layers.