Troubleshooting Slow Networks: Identifying Routers with High CPU Utilization

Old No Comments on Troubleshooting Slow Networks: Identifying Routers with High CPU Utilization 349

In troubleshooting high CPU utilization on your network, start by identifying the root cause and confirming that the high load is due to network issues like dropped packets rather than system processes. CPU loads on a network often affect network performance—especially when you have routers and servers that are used for more than regular activities like business transactions and downloads. For example, when routers are used for VPN client/server activities, the chances of CPU load reaching its maximum are high. This means that it’s imperative to monitor CPU loads.

Unmonitored high CPU loads on core routers might slow down response times and ultimately lead to network downtime or failure. The most common symptoms of high CPU utilization are:

• Slow network performance
• SNMP timeouts
• Services on the router fail to respond and unable start to telnet or SSH sessions
• Failure to respond to normal requests like ICMP ping requests

Causes of High CPU Loads

  • Overloaded router
  • Deviation from the optimal routing path (i.e. BGP route flapping)
  • Background processes, such as data backup or virus scan, might be running
  • Spanning tree reconverges in the Layer 2 network
  • Input Queue drops
  • Interrupted processes
  • Kernal Process
  • Packet drops due to increased latency

When one of the routers tries to “hog” the CPU and significantly increases the load, data congestion occurs. This leads to irregular data transmission and latency. Identifying what causes CPU load hikes helps in identifying what causes data congestion and ultimately helps you fix slow network issues.

Generally, you could just do a telnet show processes cpu command to find out the CPU utilization. But, the result displays in a raw format.

CPU-telnet

To mitigate network downtime, it is crucial to monitor CPU utilization in real-time. Instead of reading raw CPU data, it’s much easier to visualize multiple CPU loads in multiple routers in a graphical format in real-time and narrow down on only those routers or switches with abnormally high CPU utilization—no need to stab in the dark!

5

Seeing the current CPU utilization and other stats like peak load level, minimum/maximum, and average CPU value in a graphical interface is much friendlier than the raw telnet data. Visual interpretation of the CPU metrics helps you see the high CPU loads at a glance, resulting in faster and easier troubleshooting of slow network performance.

Author

Vaishnavi Vijayaram

Technical Communicator with a penchant for improving usability and user experience and provide usable content. Author of Learning Curve, part of the book - Negotiating Cultures: Case Studies in Intercultural Engineering and Technical Communication - published by IEEE Wiley (Institute of Electrical and Electronics Engineers, Incorporated) for Kansas State University, US.

Related Articles

Leave a comment

Back to Top