Monitoring networks using SNMP requires that the NMS poll each managed device on a periodic basis to determine its status. Frequently polling many devices or MIB variables on a device across a network to a central NMS might result in performance issues, including congestion on slower links or at the NMS connection, or overwhelming the NMS resources when processing all the collected data. The following are recommended polling guidelines:
■ Restrict polling to only those MIB variables necessary for analysis.
■ Analyze and use the data collected; do not collect data if it is not analyzed.
■ Increase polling intervals (in other words, reduce the number of polls per period) over low-bandwidth links.
■ For larger networks, consider deploying management domains, a distributed model for deploying an NMS. Management domains permit polling to be more local to the managed devices. As a result, they reduce overall management traffic across the network and the potential for one failed device or link to interrupt management visibility to the remaining network. Aggregated management data might still be centralized when management domains are used. This model is particularly appropriate for networks that already have separate administrative domains or where large campuses or portions of the network are separated by slower WAN links.
■ Leverage nonpolling mechanisms such as SNMP traps, RMON, and syslog (as described in later sections of this chapter).
Was this article helpful?