Network Monitoring Thresholds

Every 60 seconds, the generator agent sends the network metrics measurements to the Obkio Cloud. When received, the measurements are analyzed and Network Issues can be generated based on the thresholds. These thresholds can be configured in the Advanced Parameters section of the Network Monitoring Template.

For the Packet Loss, Latency and Jitter thresholds, two severity levels are available: Warning and Error. For the Session Timeout threshold, the severity level is always Critical.

When a network issue is raised, a notification (usually by email) can be sent to the users based on the Organization Notification Settings and the User Notification Settings. In the Network Monitoring Graphs, the color yellow is used for warning and red for errors.

The session timeout threshold is the number of seconds to wait until a Session Down network issue is raised. This network issue is raised directly by both the agents (generator and reflector), and never during the analysis of the measurements.

The packet loss thresholds are straightforward: if the packet loss for the 60-sec interval is higher than the threshold, a network issue is raised. Two thresholds are available, one for severity level warning and the other one for severity level error.

Since the thresholds are configured on Network Monitoring Templates, it is not ideal to use absolute values because the latency between agents depends on multiple factors such as the distance and the connection medium (DSL, Cable, Fiber, Wireless). This is why the thresholds are relative to the historical value, which is the median of the last 48 hours, updated every hour.

If the current value (latency or jitter) is higher by more than X percent of the historical value, a network issue could be raised if the difference is a minimum of Y milliseconds.

The minimum difference is there to avoid network issues if the latency or jitter are very low. For example, a latency of 2ms between two offices connected with optical fiber that increases to 3ms, which is a 50% increase, is only 1ms increase. In that case, a minimum difference of 10ms for example would require the latency to increase to 12ms to raise a network issue.