I requested some direction to address a couple of years ago that the graph just shows the “delta” metric and while some people liked the post, no one even answered my question, which I thought was a pretty easy question at the time.
I think using the delta would solve the SNMP issue in that you aren’t looking at the history of errors on a particular interface, just the amount of errors seen in the last polling period. It honestly seems like a simple fix of just adding a couple of columns for the error delta (in/out), but I guess it’s either more complicated than it appears or very few people have an issue with it.
I do agree it’s helpful to have the total errors of the interface displayed. I like being able to see it’s an interface with a history of low or high interfaces on the initial list as a way to prioritize the entries on the list. The one that saw 100 errors in the delta but has less than a thousand in the past month isn’t as big a deal as the one that has a dozen errors in the delta but seen 100,000 errors in the past month.
However, I don’t think displaying “errors per second” is particularly helpful. Anything less than 300 errors in one 5 minute polling period will show up as less than 1 error per second, which confuses and annoys many people. At least in a graph the spikes can convey the problem in a visual context, that raw numbers can’t quite convey… but it’s hard for anyone to prioritize 6 errors per microsecond.