The peak of the graph topped out at 140Mbit while I’d been transferring data at around 600Mbit. It turns out that this is a limitation of 32-bit counters that overflow after 5 minutes.
So I decided to switch to 64-bit counters in Cacti. I ran my tests again and this time my graphs were a total mess. According to the graph I’d transferred multiple terrabytes of data in just 5 minutes.
So I’ve sent an email to HP to make them aware of this bug, I even received a response:
Thank you for using Hewlett Packard Customer Service.
This info has been send to the division and in a later Firmware update this issue will be adressed
unfortunatly there is no timeframe known for this fix yet.
Hewlett Packard Customer Service
Doesn’t sound very convincing, but I hope the next firmware release fixes this annoying bug.