[Wylug-discuss] BBC speeds up TCP video streaming on Linux (paraphrase)

[Andrew 'Leny' Lindley]

From: Graham Whaley <graham.whaley at gmail.com>
Subject: [Wylug-discuss] BBC speeds up TCP video streaming on Linux (paraphrase)
Date: Thu, 15 Oct 2015 13:45:28 +0100

> RobS, is this you and yours?
> 
> http://www.theregister.co.uk/2015/10/12/linux_networking_api_showing_its_age/
> 
> something we can discuss at the next meet maybe. I'd be interested if
> it is just the buffer share/zero copy that gets the major speedup, or
> the buffer batching, or the reduction in syscalls and context switches
> - or a combination of all three.

I'm unlikely to make it to the next meeting, so:

ISTR all three being in the 1st Edition of Tanenbaum... (3rd Ed [1])

The article is, like most things trade rag, somewhere different to
what I understand to be the reality.

Firstly the Teclo stuff it mentions is to do with mobile data.  Where
AIUI from someone in the industry the problem is essentially common or
garden mobile app developers program their apps to treat mobile data
like fixed wire.  So once again various people are scratching their
heads on how best to cope with programmers with stiff wrists. :)

The BBC thing is the networked-age old problem of there not being
enough capacity in affordable CPUs to switch / route / drive high
speed links unless you doctor a protocol or take extreme measures.
Take a look at MPLS[1] to see how in the 90s a solution to the problem
of having to switch/route fast links on chip was engineered.

Why El Reg should see the doom of TCP/IP in two apples and oranges
uses of the most hackable kernel on the planet is, of course, merely a
matter of hit counts meaning money and good news not selling.

Leny

[1] http://www.amazon.co.uk/Operating-Systems-Implementation-Prentice-Software/dp/0131429388/ref=sr_1_6?ie=UTF8&qid=1444921481&sr=8-6&keywords=operating+systems+tanenbaum

[2] https://en.wikipedia.org/wiki/Multiprotocol_Label_Switching