GEANT2 Workshop - Rome - Session 10

From GEANT2-JRA1 Wiki

Hades alert system (rather old)

• Features:
  • Independent software on measurement boxes (This is a disadvantage!)
  • SNMP
  • Nagios on central server
• Problems:
  • Not working ;)
  • So far only simple IPPM and traceroute analysis
  • Not really integrated into Hades system
  • NOT a perfSONAR system

Alert system touches various components

• Data analysis
  • Talk by Thomas Holleczek on statistical analysis of Hades data (Hades_Data_Analysis-Rome_2008.pdf)
  • Where to analyse what?
  • Manual thresholds? Automatic threshold detection? Semi-automatic?
  • Different amount of data available, especially for new or changed links

• Real time
  • Different requirements for different metrics
  • Analysis problematic on boxes (interfering with measurements)
  • Different places of data analysis. What about thresholds?

• Out of band communication
  • Alerting channel: SNMP vs. perfSONAR

  perfSONAR most likely the better choice

  • Configuration channel: Threshold propagation

  Use perfSONAR would be a good solution

• User interface
  • e-mail creation
  • (Graphical) User interface
  • Threshold configuration
  • Nagios has its specialities (e.g. everything is a service, Web-Interface is only a viewer)
  • Integration in other visualisation tools useful
  • Most likely a full integration into perfSONAR should be achieved.

  Some sort of perfSONAR alerting support.