The amount of traffic is dependent on all the characteristics of the managed devices, and so will differ from installation to installation and is not simply dependent on the numbers of ports being managed.
However, from a practical perspective, port numbers (and therefore port data) are the most significant metric upon which to base estimates of bandwidth requirements.
We will estimate how much data is being polled for each port (based on the eyepoller process' 5 minute streams) and then add another 50% to account for all other data. The eyepoller process collects for each port the following 4 datasets every 5 minutes:
- port traffic attributes.
- port packet attributes.
- port status attributes.
- port fault attributes.
Assuming each dataset requires a single SNMP response totalling 200 bytes, across 4 datasets this will be 800 bytes every 300 seconds for every port. This totals 2.7 bytes per second per port. Add 50% to account for all other data, and the estimated bandwidth requirement is 4 bytes per second per port.
Note, traffic rate is an average value and not the maximum rate. There will be bursts of traffic when collectors run, and the more devices under management, the more even this traffic load will become. Bandwidth utilization for flow traffic (Netflow, sFlow, jFlow, IPFIX) is dependent upon the number of flows being exported by the device and the flow export rate. It is not possible to provide a completely accurate estimate of bandwidth utilization, because each network implementation is different.