The predominant method for measuring router performance is via SNMP polling of the device. With routers, the top priority metrics in performance monitoring are interface rate (IN and OUT), interface discard rates, CPU and memory utilization. With these six key metrics, the majority of performance problems that persist on the router will be caught. Many tools exist on the market today to historically trend and analyze these statistics plus many more. For Linux routers, SNMP should be enabled on the server and the tool selected should support the right distribution MIB. Depending on budget, SNMP monitoring tools are available. What'sUpGold and BigBrother would probably be suitable for smaller environments but not for large scale enterprises.
Newer technologies exist in the marketplace to monitor traffic flows between devices. This allows visibility into the application data volume and users causing possible congestion on the circuits. With this information, isolating high volume users and spikes in virus activity quickly helps resolve issues and reduces the amount of unplugging and plugging of interfaces to identify a source. With RMON2 probes and/or NetFlow collectors, the data transferred between the interfaces can be measured to figure out when and who is communicating on a particular circuit. Linux also supports an open source NetFlow export and tools, which can provide useful insight into the traffic flows on the router.
For multiple-device networks, a combination of SNMP polling and NetFlow exports should help manage the performance of the Linux-based routers. Depending on budget, SNMP and NetFlow monitoring tools are available to assist with the collection and aggregation of SNMP data for easy reporting.
This was first published in March 2005