I'm using zenoss to monitor edge switches at my WAN locations. Beyond knowing if the thing goes down I really need to be able to keep the bandwidth utilization data. This way if we need to change providers in the future I will have a good record of how much each site is actually using. I know I can see the information in the graphs but I really need the data behind the graph so I can calculate 95th percentile, averages and standard devs.
Also if I want to increase the resolution of the graphs can I simply shorten the time between polling periods? I believe that the throughput is calculated based on the deltas of interface rx/tx octets taking into account the time since last poll. Is there any reason I shouldn't poll more often?
The data that backs the graphs is stored in the RRD files ($ZENHOME/perf/Device/<deviceName/ip>. You have quite a few options to extract that data, but they all boild down to working with the RRD data. A google search for dumping RRD to CSV should yield more options than I could list here. You also have the ability to run custom Zenoss reports, but that is not a trivial undertaking.
As far as the resolution, simply shortening the polling interval won't exactly work due to the way RRD leverages STEP. You'll need to also alter the default RRD Create command (found under the collector configuration) if you want that too work.
Thanks for the info! I think I can write a cron job to ftp those rrd files off to the SAN. I did find that there is a builtin export command with rrd but I think it exports to xml. I don't know it that would be easier to work with.
As far as the RRD create command. I have found it, but I really don't understand the syntax do you have anywhere you can point me on this?