Goals

  • Ease of administration.
    • Adding, altering and removing monitors for large sets of machines should be trivial.
    • Simple to ascertain what is monitored and whether or not the monitoring system itself is meeting its contracts.
    • Small, tight and deployable on any unix variant.
    • All reconfiguration can be performed in-flight.
  • Efficiency in scale.
    • Monitors should support thousands of machines on reasonably tight intervals:
      • e.g. ICMP 10 pings over 10 seconds, repeat every 5 minutes -- over several thousands of IPs.
      • HTTP request several URLs checking for content and/or HTTP code over thousands of machines.
  • Delegated deployment.
    • centralized configuration management.
    • decentralized configuration manipulation.
  • Applying policies (such as notification and escalation) to large groups of service