학술논문

Operational Model of the ATLAS TDAQ Network
Document Type
Conference
Source
2007 15th IEEE-NPSS Real-Time Conference Real-Time Conference, 2007 15th IEEE-NPSS. :1-8 Apr, 2007
Subject
Computing and Processing
Switches
Ethernet networks
Distributed databases
Monitoring
Availability
Automatic control
Control systems
Statistical distributions
Throughput
Watches
Language
Abstract
The ATLAS TDAQ network consists of four separate Ethernet based networks which together total over 4000 ports with 200 edge switches and 6 multi-blade chassis switches at the core. System checks are invoked at every level of the installation. The full installation is described in different static databases. Tools are provided to automatically cross-check these for consistency. The configuration management is centralized: configuration files stored in a database are distributed to all devices and the actual settings are periodically verified. Monitoring systems are deployed to validate the connectivity, identify malfunctions and confirm the resources availability upon request from TDAQ control. Relevant operational statistics (e.g. port status and throughput) are continuously logged and made available to TDAQ control. Watches and alarms are set for dynamic threshold violations and the complete instantaneous status can be viewed at different levels of abstraction in a 3D flythrough. A tool-set has been developed to demonstrate aggregate achievable cross-sectional bandwidth for TDAQ-specific traffic profiles, as well as to analyze traffic flows and hot spot behaviour.