CODEWITHSUNDEEP
linuxubuntuprocessnetwork-monitoring
John Elaine
John Elaine

Reputation: 389

Monitor network usage of a process

This question might sound fairly repetitive, but there are subtle details which make it a bit different.

I am looking for a simple tool (for Ubunut/Linux) to monitor network usage such that it gives the min, max, average, and time-plot of network usage by 1) a single process; and, 2) the system; only during the time when the process was running. The major requirement is that I am not looking for a GUI (or terminal GUI like top) based tool but I want this monitoring information to be pushed to a file so that I can perform some post-processing over that.

I come across the following link which lists various options: http://www.binarytides.com/linux-commands-monitor-network/. However, most tools are GUI based and ones which are not do not provide above information.

Any help would be much appreciated.

Upvotes: 2

Views: 2193

Answers (1)

twm
twm

Reputation: 1458

Wireshark might work, depending on how far you're willing to relax the non-GUI requirement and whether locating your target processes is simple. Wireshark is of course a GUI app, but the tshark command which comes with it is headless and can be used to capture packets to a file. After capturing all packets on an interface, you can run tshark again on the pcap file to filter the file using Wireshark "Display filters" and extract just the packets for your process. That's one part that may or may not be simple, depending on whether you can identify your process from network traffic content, port(s), or by adding some sentinel dummy data. You'll then have two pcap files, one for the whole network interface and one for just your process.

The capinfos command will report the average throughput. Wireshark can be used to generate a time-plot of the traffic with millisecond (or other) granularity via the menu "Statistics >> IO Graph". As for min and max, you can either eyeball that from the time-plot or use editcap to split the pcap files into chunks, run capinfos on each chunk, and calculate the min and max over all chunks.

That might not be the simplest approach, it's just what occurred to me off the top of my head.

Upvotes: 2

Related Questions