I’ve employed an RB450G in a datacenter recently, and experiencing strange 100% CPU usage spikes on the CPU graph. They always last a single tick only, and the SNMP monitoring tool sampling the same board (with more frequent, 2-minute intervals) shows no signs of them. There are no other kinds of resource spikes, no extra traffic or measurable suspicious activity. Below is a sample attached containing the present day’s daily graph. Should I create a support ticket? The router is otherwise functioning perfectly. Running on 5.14.
We had something similar happen with our monitoring tool. The tool would grab all SNMP values except the CPU first, which would load the CPU. We fixed this by changing the order of what SNMP values were polled, polling CPU first to avoid this issue.
Not sure if this is related, as your issue seems to be sporadic. Ours was fairly consistent.
i suggest use cacti for network monitoring 
Thanks for the replies. I’m using Zabbix for the moment, and it doesn’t influence the SNMP readings at all (CPU load or anything else). I’m suspecting RouterOS is misbehaving at certain periods. I might need to forget the built-in graphing completely if it’s so unreliable.