Download presentation
Presentation is loading. Please wait.
Published byJessie Emery Park Modified over 8 years ago
1
CERN, 14.03.2006 Artur Barczyk, CERN/PH-LBC 1 Network Monitoring and Control in LHCb A (very) brief introduction
2
2CERN, 14.03.2006Artur Barczyk, CERN/PH-LBC Some facts Network monitoring folklore SNMP “heavy weight”, slow, but flexible Unreliable for counter real-time update Counters are pulled by the management station Internal update time hard coded in the device‘s OS Traps: a network device sends an asynchronous message to a monitoring station in case of a predefined event Link status change Cold/warm start enterprise defined events SFLOW Newer industry standard for performance monitoring More reliable for real-time updates Push method: a network device configured for monitoring sends updates to the monitoring station on a regular basis Update time configurable, typically 15 s Linked to SNMP in a light way, e.g. Port index is given as SNMP oid Configuration folklore CLI A de-facto standard way of configuring a switch The only method to allow complete configuration SNMP The RFC defines some objects as „read-only“, most vendors are compliant, so configuration possibilites are limited
3
3CERN, 14.03.2006Artur Barczyk, CERN/PH-LBC Best combination For a management system: Configuration through CLI, using automated telnet sessions SNMP to gather information about the device, e.g. Port names and indices, and counter updates on devices which have no sflow support (yet), e.g. PCs SFLOW for counter updates
4
4CERN, 14.03.2006Artur Barczyk, CERN/PH-LBC Proposed Structure PVSS HOST MANAGEMENT STATION MANAGEMENT STATION DB SWITCH SFLOW counter updates DIM TELNET (CLI) for configuration SNMP traps (asynchroneous) signal status change SNMP counter updates
5
5CERN, 14.03.2006Artur Barczyk, CERN/PH-LBC Management Station A PC, running the monitoring and configuration software Will use DIM for comunication with the controls system Publish current counters and switch states Execute client commands, e.g. Reboot a switch Disable a link Down/upload configuration Upon reception of a trap message, notify the client of status change (dis_update_service) Configuration of each switch should be stored in the confDB The first implementation (no dim interface yet) implements SNMP for information gathering SFLOW for counter updates SNMP trap collector The configuration part, using telnet will make use of libexpect
6
6CERN, 14.03.2006Artur Barczyk, CERN/PH-LBC Demo
7
7CERN, 14.03.2006Artur Barczyk, CERN/PH-LBC Outlook The curent implementation is in prototype phase Need to include Dim interface PVSS (client) side CLI access to the switch, using telnet sessions (libexpect) A student from ISIMA (Clermont-Ferrand) will arrive in April to work on this project Primary goal is to monitor the DAQ network, but could extend it also to ECS
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.