Hi, All !

I have some gigabit links loaded for 300-400Mbps. The Dude correctly display the hour graphs for them. But while hour link history graph for gigabit links is quite right the corresponding day, week, month and year's graph never surpass 100-120Mbps and look oddly.

When the load on some Gigabit links fall below 100Mbps all graphs become corrects.

How to fix this ?

Thanks in advance.

you can manually select the link type and max speed. apparently your device didn't report this information correctly

Thank you for the answer, Normis.

I think my devices (cisco catalyst c3560 and cisco router c7604) report interface speed correctly. For example, snmpget give me this information:

1.3.6.1.2.1.2.2.1.2.10126 (iso.org.dod.internet.mgmt.mib-2.interfaces.ifTable.ifEntry.ifDescr.10126) octet string GigabitEthernet0/26

1.3.6.1.2.1.2.2.1.5.10126 (iso.org.dod.internet.mgmt.mib-2.interfaces.ifTable.ifEntry.ifSpeed.10126) gauge 1000000000

Nevertheless I tried to set interface speed manually with no luck.

I think the problem arise when data is transforming from hour to day set because the hour graph is always correct and the graph of the day, week, month and year are not when the real interface load of the gigabit link exceeds 100-120Mbps.

When the graph period get longer (hour --> day --> week --> month --> year) the displayed data is averaged out more and more. Let's say each data point in the hour graph is a 20 second average of the network usage. This then may shows burst of a few hundred megabits that lasts a few seconds. But each data point in the day graph is an average of 24 hour graph points. So there you won't see the bursts anymore. Etc. etc. Each successive graph becomes flatter and flatter...

Couldn't this simply account for your situation?

(Just trying to help, my apologies if I'm way off...)

I have similar problem.
It seems some type of counter overflow when summarizing data.
I seen this problem with MRTG too.

I have the same problem:

Hour's graph is ok always:
Day's graph is not while throughput exceed 100-120Mbps:

Looks like a problem with wrapping 32-bit counters. A 32-bits counter will wrap if the bandwidth is over 100 Mbit, and the time between polls is about 5 minutes...

You can try to poll the device more often, or maybe the dude could support 64-bit counters instead...

Replace Cisco and use RouterOS vs SNMP to graph ?

SNMP v1 (and v2 I think) are indeed limited to 32-bit counters and can therefor quite easily overflow. SNMP v3 (or was it already in v2?) introduces some 64-bit counters, so that should work better...

I don’t think this is SNMP version issue because hour graph is correct and daily graph do not depends on SNMP’s counters size but rather on the averaging-out function.

I don’t think this is SNMP version issue because hour graph is correct and daily graph do not depends on SNMP’s counters size but rather on the averaging-out function.

You are correct. If the hourly graph is correct, it can't be overflowing SNMP counters and it must be an internal Dude problem...