How does the SNMP polling work...

[FlyingDuffman]

This is just a general question about how the graphing works with Cacti...

By default, it polls SNMP data every five minutes, right?

Is that data that it returns the current state of the interface at that moment or is it an average of that interface over a five minute period or some other third thing?

The reason I ask is that I've done some speed tests with our DSL users which max out their connection but I'm not seeing those spikes when I look on my graphs.

[rony]

Concerning interface counters, what you are asking about, it's important to note that the data being collected is a counter. This is a increasing number, denoting the number of octets of data that have passed on the interface.

Polling every 5 minutes, Cacti default, a value is gotten, then when polled again another value is gotten. RRDtool then takes these 2 values and subtracts the latest from the current to figure out the difference. This is the amount of traffic that has passed on the interface in those 5 minutes. When RRDtool displays this on the graph, based on the default interface graph, this is shown as either bits/s or bytes/s, depending on applied CDEF.

Values on the graphs are calculated bits or bytes per second that occurred over the 5 minute collection window.

[gandalf]

While all of Tony's statements are correct, I would like to add, that rrdtool deals with time jitter when updating. That means: You cannot assure to poll each interface exactly at 300 sec intervals. You may have 299 or 301 or other intervals as well. rrdtoll will correct all values by re-calculating them to the "step" base which is 300 sec by default for cacti settings
Reinhard

[Netcooler]

Just want to ensure I am on track here as i had the same questions:

I am polling an interface using the inherent interface graphs bits/s

current query yeilds counter numbers: 3591401668
query before that yeilds counter numbers: 3590947901
Delta between the two numbers is: 453767

On the graph this is represented as CURRENT: 1.58k and I am struggling with HOW that is calculated. Not questioning whether it is right or not but I do not understand the calculations getting to that number. Can someone enlighten me? Tony's comments above were great but I need to validate what I am seeing. Not sure if i fully understand what is going on behind the scenes of the tool.

Again this is with the inherent interface graphs with the cdef turn bytes into bits (current_data_source,8,*)

[Netcooler]

OK so this fundamental calculaion was really bothering me so I read the rrdtool documentation...

if I understand this correctly:
bps = (counter_now - counter_before) / (time_now - time_before)**in seconds** X 8

bps = (3591401668 - 3590947901) / 300 = 1512.5566666666666666666666666667

That give me the approximate 1.58k (ofcourse polling sycles are not exactly 300 seconds and could be normalization. Not sure I understand the X 8 because if I do that I am off but am I on the right track??