Advanced monitoring techniques for a large‐scale data‐processing network

Abstract

Purpose

The aim of this paper is to describe the methods used to monitor and measure the performance of the data acquisition network deployed in the ATLAS detector at the LHC in CERN, Geneva.

Design/methodology/approach

Strict inventory and connectivity control was employed through cross‐correlation of multiple databases. Basic monitoring uses a commercial product and, in addition to that, 2‐D graph display tools were developed to compare and contrast multiple instances of peer performance in terms of through‐put and error. Traffic analysis and diagnostics were improved by displaying at carefully customized aggregate plots. By overlaying geographical, logical and traffic information a 3‐D, system‐wide, visualization system was developed with fly‐through navigation and zoom‐based levels of detail to monitor the 6,000‐port system in quasi‐real time.

Findings

The difficulties of automatically detecting unacceptable traffic patterns are presented and both signal‐processing techniques and rules‐based expert systems are suggested as possible solutions to the problem.

Originality/value

It is believed that the scope of the 3‐D visualization program is unique in the field of network monitoring. These techniques should be of value to any manager of a complex network situation.

Citation

Martin, B., Al‐Shabibi, A., Batraneanu, S.M., Ciobotaru, Darlea, G.L., Ivanovici, M., Leahu, L. and Stancu, S.N. (2008), "Advanced monitoring techniques for a large‐scale data‐processing network", Campus-Wide Information Systems, Vol. 25 No. 5, pp. 287-300. https://doi.org/10.1108/10650740810921448