[Bloat] I feel an urge to update this

[Mikael Abrahamsson]

On Thu, 25 Sep 2014, David Lang wrote:

> well, even if you have a 10GE connection, you don't know how heavily 
> it's going to be used.

No, but I will have a hunch. I don't need to *know*, I need to have a 
decent probability of being right.

> what happens if the wrong hint is given? (either accidently or maliciously)

Then you get IW10 instead of IW4. Right now if I am correct, Google does 
IW10 all across the board.

> The approach of starting slow and ramping up works well, except in the 
> case where you have lots of very short connections. So I don't see a 
> benefit of trying to hint slowness. There may be some value in hinting 
> for faster ramp-ups, but what will that do to fairness with existing 
> systems?

In my world, core network congestion isn't something that is permanent and 
continous, but an anomaly. So since the only port that should be 
congesting is my access port, I don't care about fairness. Also, we're not 
talking about starting off with a 1M TCP window to blast a huge chunk of 
data down the wire, we're talking about perhaps being a bit more agressive 
in increasing the window from a slightly higher initial level, let's say 
IW10 and instead of doubling, tripling window size, if the client hints 
they have a fast connection, and if the client says they have a slow 
connection, IW4 and increase 1.5x instead?

Use same mechanisms as today, but tweak them to be a bit more or less 
agressive depending on what the client has hinted.

I mean, why do we send MSS in TCP? Couldn't we just have a generic probing 
mechanism to test what MTU works? Well, that would be inefficient, so MSS 
helps. Same here.

> But we have a number of companies who want to have things downloaded and 
> used as quickly as possible, so if this sort of thing really does help, 
> then I would expect that they would be experimenting with tweaking the 
> TCP stack to ramp up faster (although, you would also be expecting 
> outcrys about "how dare $bigcompany ignore the TCP defaults and be more 
> agressive, that doesn't compete fairly with $normaltraffic)

Errr, google and IW10? So, already done!?

-- 
Mikael Abrahamsson    email: swmike at swm.pp.se