On a customer setup, we looked at the mib2.HCbits graph (please see attached graph_datacollection.png) and at the corresponding NRT graph (please see attached graph_nrt.png). The values on the NRT graph are 10 times higher than on the other graph (20 MBit/s vs 2MBit/s). The value on the usual graph seems to be correct for this environment. I also checked the NRT graphing for other reports (e.g. cisco.memory) and could not find the error here.
If I change the NRT interval to 1sec, the values are equal to the ususal graph. If set the interval to 30secs, the values of the NRT graph were 30 times higher.
I set on the TRACE option for karaf.log and saw the values, that were collected by the SnmpProtocolCollector for ifHCInOctets.The collected values were correct and a diff of two values shows the value that was shown in the NRT graph. The diff of two values does not seem to be devided by the NRT collection interval.