What Is Packet Event?

Why Do We Need Packet Event?

Packet Event provides the following functions:

• Packet loss visualization: If packet loss occurs on a device, services are affected. Packet loss visualization supports the fast detection of packet loss on devices caused by abnormal forwarding, specified packet discarding rules, a full buffer, or a deny action in ACL rules, allowing for quick fault locating.
• Latency visualization: If the latency of packets exceeds a specified threshold, network congestion occurs. Latency visualization enables devices to promptly detect packets with a high latency, preventing network congestion.
• Packet loss self-healing: When packet loss occurs on a network due to specific causes, it can be detected and self-healed in real time, improving network reliability.
• Latency histogram: The device counts the number of packets in each user-defined latency segment for different queues on a port, and then presents these results in the form of a histogram to visualize the port-level latency. This helps network administrators analyze latency and adjust services rapidly.

How Does Packet Event Work?

On the network shown in the following figure, packets sent from a host and forwarded by devices may be discarded on a device due to the following causes:

• A forwarding exception: Packets are discarded in abnormal scenarios such as chip entry loss, chip entry delivery errors, and packet check errors.
• Specified packet discarding rules: The packet loss is predictable. For example, port isolation is configured to discard packets between two specified ports, a punishment action such as traffic suppression associated with MAC address flapping is configured, or split horizon is configured on ports.
• A full buffer: The buffer of a port or queue is fully occupied, resulting in the discarding of excess packets.
• ACL rule deny actions: Packets that match such ACL rules will be discarded.

After packet loss visualization or latency visualization is enabled on a device along the packet forwarding path, the device generates corresponding flow entries in its CPU whenever packets are discarded due to the above causes or their latency exceeds the threshold. The device periodically reports these flow entries to the collector using UDP packets in NetStream V9 format. The analyzer reads and analyzes the data from the collector to quickly identify causes and the locations of exceptions. This helps administrators locate faults promptly to prevent network congestion.

With packet loss self-healing enabled, the device can detect packet loss caused by specific factors and take corresponding self-healing measures based on cause IDs for improved network reliability.

Packet Event

The latency visualization function enables the device to only collect statistics on information such as the peak and minimum latency, number of packets, and inbound and outbound ports of packets whose latency exceeds the threshold. It cannot provide port-based statistics. The latency histogram function overcomes this limitation. It counts the number of packets in each latency segment for different queues on a port and presents the results in a histogram, offering a more intuitive view of port-level latency. The statistics can be reported to a collector through telemetry, helping network administrators analyze latency and adjust services rapidly. In the following latency histogram, the horizontal axis represents the latency segments. You can define the lower and upper thresholds for each segment as needed. After the latency histogram function is enabled on a port, the device counts the number of packets in each latency segment for every queue on that port. The results are then displayed on the vertical axis in the histogram.

Latency histogram