Congestion

Bandwidth Utilization

High bandwidth utilization occurs when an interface or a link is nearing its capacity. Consistently high-bandwidth utilization might lead to congestion and degraded network performance. Large file transfers, backups, streaming services, peak usage time (business hours or scheduled tasks), and increased control traffic are some of the factors that increase the bandwidth utilization of an interface. A consistently high bandwidth utilization indicates possible congestion and signals the need for a bandwidth upgrade. Therefore, early detection of these utilization trends is critical to preventing performance issues such as packet drops, latency, and jitter.

To analyze an interface's bandwidth usage, Marvis collects bandwidth-utilization samples at short, periodic intervals for an hour and calculates the hourly average. Similarly, samples are collected for each hour throughout the day. For every hour, Marvis records the average bandwidth utilization. At the end of the day, the maximum hourly average bandwidth utilization is recorded.

Juniper Routing Assurance leverages a seven-day analysis period to assess the performance of a link. Marvis generates a congestion event when a link's bandwidth utilization exceeds the configured threshold of 70 percent for seven or more days. Marvis also recommends a bandwidth upgrade to ensure optimal performance.

The use of maximum hourly average instead of a simple daily average is insightful for proactive network capacity planning and troubleshooting performance issues. The goal of capacity planning is to avoid network performance issues. You must scale the link capacity to handle the predictable peak load. By monitoring the maximum hourly average bandwidth utilization that exceeds the threshold, you can determine the additional bandwidth that the interface requires to prevent congestion during peak traffic periods.

The Congestion event provides details such as the site, router, date and time when the event was recorded. Click View More to view the list of interfaces that have high bandwidth utilization.

In Figure 2, Marvis has reported congestion event on two interfaces of router A:

• xe-0/0/5:0─The current average hourly bandwidth utilization recorded on the receive (Rx) link of the xe-0/0/5:0 interface during the previous day is 86.7 percent. The highest hourly average bandwidth utilization recorded on the link since the issue was first detected is 87.1 percent. Since the link's current average hourly bandwidth utilization exceeds the configured threshold of 70 percent, Marvis recommends administrators to upgrade the bandwidth to ensure optimal performance.

• xe-0/0/11:0─The current average hourly bandwidth utilization recorded on the transmit (Tx) link of the xe-0/0/11:0 interface during the previous day is 86.7 percent. The highest hourly average bandwidth utilization recorded on the link since the issue was first detected is 87.1 percent. Since the link's current average hourly bandwidth utilization exceeds the configured threshold of 70 percent, Marvis recommends administrators to upgrade the bandwidth to ensure optimal performance.

See Table 1 for field descriptions.

Figure 2: Congestion: Bandwidth Utilization

Packet Drops

Congestion in a network occurs due to several factors, including traffic bursts, speed mismatches between links, and insufficient buffer depth. When buffers become full due to congestion, some packets must be dropped to avoid network delays. Marvis detects and reports such packet drop events. Marvis reports a congestion event when an interface drops packets for a continuous duration of 60 minutes.

The Congestion event provides details such as the queue number, priority, forwarding class and the number of packets that were dropped since the issue was first detected. Marvis reports the top three affected queues, sorted by their QoS priority level.

When an interface queue drops packets, Marvis recommends administrators to optimize the queue by adjusting the queue's QoS settings. Optimizing the output queue properties can significantly reduce the number of drops. Output queue properties include the amount of interface bandwidth assigned to the queue and the size of the memory buffer allocated for storing packets. You can configure the output queue properties from the router's CLI.

Click the router name to navigate to the router Insights page to view the Total Interfaces Drop Packets Rate graph, Queued or Transmitted Bytes Rate graph and Depth Percentage graph for the interval reported by the event. You can correlate the data in these graphs to specific congestion events that might have occurred during the same interval.

You can mark the status of an event as In Progress to indicate that an administrator is working on resolving the event. Resolving a specific event will only resolve that occurrence of the event and will not resolve future occurrences of the same event. If a congestion event has occurred and remains inactive for seven days, Marvis moves its status to AI Validated.

In Figure 3, you can see that Marvis has detected packet drops on Router B's interface. Queue 0 of the interface xe-0/0/5:0 has dropped a total of 124047049786 packets since the issue was first detected. Since the hourly average bandwidth utilization of the interface is below the threshold, Marvis recommends administrators to only adjust the queue settings to prioritize critical traffic.

For field descriptions, see Table 1.

Figure 3: Congestion: Packet Drops

Congestion Event Evaluation Criteria

Marvis recommends a bandwidth upgrade or queue optimization based on the occurrence patterns of high-bandwidth utilization and packet drops. Marvis analyzes bandwidth utilization and packet drops as two independent events.

• If an interface’s bandwidth utilization exceeds the 70 percent threshold consistently for seven days, Marvis recommends a bandwidth upgrade.

• If an interface queue drops packets at any time within the seven-day window period, Marvis recommends administrators to optimize the queue by adjusting the queue's QoS settings.

• If an interface’s bandwidth utilization exceeds the 70 percent threshold consistently and its queue drops packets simultaneously, Marvis recommends that administrators upgrade the bandwidth.