Any way to configure PER APP network usage?

[resetreset]

---

If I'm listening to a radio station in one of the browser tabs, and surfing in another, is there a way to tell Linux to set the audio to stream without skipping, and leave only the rest for the other tab?

Thanks,

[voidpointer69]

What is your problem exactly?

I listen to streamed radio all the time whilst I use other tabs for other things. I do experience breaks and skips from time to time but that is only because of the bandwidth to the outside world.

[resetreset]

It's the principle of it. I just want to know how I set each program to get so much.

[docbop]

Throttling is usually on the network level not within a workstation. It could probably be done, but way more work than its worth.

[voidpointer69]

"...It's the principle of it." principle of what exactly?

Are you trying to give "x" amount of bandwidth to a given tab in your browser?

I doubt your browser has support for that.

[resetreset]

The principle of how such a thing would be handled in the kernel, if it was possible to configure per app usage.

OK, assume that it's not a web site radio station, assume I'm listening to something streamed in one of the audio players, and surfing while I'm doing that.

[docbop]

The principle of how such a thing would be handled in the kernel, if it was possible to configure per app usage.

OK, assume that it's not a web site radio station, assume I'm listening to something streamed in one of the audio players, and surfing while I'm doing that.

Throttling like you're talking about is done at the application layer of the TCP/IP stack. It is done by inspecting packets.

[voidpointer69]

What you seem to be aiming at is some kind of bandwidth control inside the box. Not really doable at that level however...

As doc hints at above, if you have control of the TCP/IP stack, you could try to arrange to schedule interrupts (or responses depending on how the application is managing the socket) so that you achieved some kind of "load-balance" paradigm among all the applications involved.

However, what you have no control over is the responses from the other (ie remote) end of the connection. You have no control on the bandwidth used by any given application.

In an earlier life, I spent much effort on attempting to achieve "load-balancing" at the network/application level. This involved having multiple "servers" delivering the same information classes. This information was asynchronously delivered in a "broadcast" pattern. What I needed to achieve was to get the loads on each server as close to each other as possible.

This always requires a "brokerage" layer of some sort since it is necessary for something, either the servers themselves acting in consort, or a third-party monitoring the server loads to have a view of all load over all servers.

In my case, simply switching the requests on a "round-robin" basis was inadequate since some requests resulted in far more traffic being generated than others so one request of a certain type could put as much load into the mix as ten requests of another type.

Tricky to say the least.

[resetreset]

Well, at least you tried

Thanks for your input. Was this some kind of research project in a college lab or something?

Also, what do you mean by "control of the TCP/IP stack"? Does that mean programming the kernel?

[docbop]

Here's where to start to understand the IP stack in order to program for it. You can see layer 7 of the OSI model that is where you would start trottle at the app level.

The Linux socket, TCP/IP protocols and client-server network fundamental concepts and model background story