I represent a reasonably popular game server community here in Australia. We rent a dedicated server that has limited bandwidth. In the last couple of days we have seen a huge spike in our bandwidth consumption, despite outsourcing our download server to the US.
We are very interested in figuring out what exactly is using all this bandwidth suddenly. Each of our servers is assigned a specific port, so it makes sense to monitor bandwidth by port. We'd like to do this monitoring over a reasonably long time period so that we can see accurate results.
Here's a list of things I've tried and the problems I've had with them.
nethogs - Monitors bandwidth per process and therefore doesn't seem to properly count bandwidth for services using child processes (for example, I'm pretty sure Apache has used more than 0.8MB over the last 2 days)
iftop - Monitors bandwidth by host, data seems pretty useless considering the amount of users we have. Our load is most likely not caused by one person but by all of them.
iptraf - Monitors bandwidth by connection instead of by port, so the data is useless to us.
darkstat - Groups connections by host instead of by port.
vnstat - Monitors bandwidth by interface not by port.
bandwidthd - Measures by host or subnet not by port.
Is there a tool that exists that would help me out?