Networks are becoming more sophisticated and more extensive with an increase in business requirements. The advent of virtualization and software defined networks has furthered the complexity in troubleshooting network issues. To troubleshoot the networks effectively, superior visibility into the network is needed.

The advantages of using advanced telemetry include:

• Better network visibility
• Ease of managing scalable networks
• Assistance in real-time network issue solving
• Faster response time to network issues
• More in-depth insight into the root cause of the problem

These advantages provide administrators significant insight and ease of management of network in large enterprise data centers and cloud. The complexity and effort needed to troubleshoot packet drops, congestions, and other network issues can be difficult without the appropriate insight into the network. The administrator might be interested in finding the bottle-neck in the network or congestion of network traffic. Another use case includes tracking the latency in the network. Another use case is the troubleshooting of dropped packets and the causes associated with it. These use cases provide the incentive to use advanced telemetry.

For example, consider a fast-growing network in an enterprise data center or cloud, as shown in the following figure. With the increase in business requirements, the scaling of network equipment results in numerous switches to be managed. The use cases that are listed above can be useful to the network administrator.

Figure 17: Leaf-spine topology example in a scalable data center

SONiC provides granular, real-time insights from advanced merchant silicon-based telemetry to enable rapid provisioning and remediation of network issues. These include:

• Inband flow analysis
• Congestion monitoring
• Mirror on drop

Inband Flow Analysis provides packet-level visibility for granular latency and path tracking at the flow level. Proactive port congestion detection using buffer monitoring is carried out using congestion monitoring. Notification for packet drops with the reason and the flows being impacted is monitored using Mirror on Drop. The collector gets the information from the switches using REST APIs.

Inband flow analysis

Inband flow analysis allows scalable flow monitoring solution using Inband Telemetry. It provides packet-level visibility for granular latency, path tracking, and congestion analysis. The collector receives UDP encapsulated info to provide a network-level solution.

Some of the critical metadata parameters include:

• Ingress and egress timestamp
• Ingress and egress port
• Device ID
• Queue ID and congestion status
• IP TTL value

Congestion monitoring

The congestion monitoring provides real-time visibility into the network. It offers scalable tracking of buffer occupancies and the ability to monitor peak or current occupancy. The granular details can be accessed at a port level, port group level, or service pool level counters. It also provides the advantage of monitoring various types of drop counters.

Mirror on drop

Mirror on drop captures the drops in both ingress and egress pipelines. It captures the first dropped packet. It provides details on the packet drop by generating an event. The generate event contains the drop reason and is sent to the collector for a mirror on drop flow. The details include the first dropped packet with drop reason and drop reason change.

Inband flow analysis, congestion monitoring, and mirror on drop provide several advantages and solutions in network management. The information obtained from granular, real-time insights from advanced merchant silicon-based telemetry will help with rapid provisioning, troubleshooting, and resolving network issues.