[Fwd: Re: [LARTC] RED/GRED implementation for InBound Traffic Control (from ISP)]

Ed Wildgoose lists@wildgooses.com
Sat, 10 Jul 2004 12:21:22 +0100


Sorry all, this is the missing email that I referred to previously.  I 
inadvertently sent it only to the original poster rather than the list.

Sorry for the wasted b/w for those who don't care...

Ed W

-------- Original Message --------
Subject: 	Re: [LARTC] RED/GRED implementation for InBound Traffic 
Control (from ISP)
Date: 	Fri, 09 Jul 2004 22:04:30 +0100
From: 	Ed Wildgoose <lists@wildgooses.com>
To: 	Ow Mun Heng <Ow.Mun.Heng@wdc.com>
References: 	<1089400271.19925.133.camel@neuromancer.home.net>



>I also want to know, just how efficient is this Algorithm. AFAIK,
>inbound traffic control can't really be achieved without losing
>bandwidth.
>In some of the howtos' I've read, one guy had to limit his downspeed to
>1/2 his bandwidth to actually control it. And he was saying that the
>only way to actually efficiently control inbound traffic control is to
>use TCP windowshaping which there is not an oss implementation of it.
>
>Can anyone please shed light on this?
>  
>

The issue is not as alarming as you think.  The point is that there is a 
buffer on the ISP end.  If this buffer fills up then the algorithm is 
usually first in first out.  ie you have no way to prioritise important 
stuff to go ahead of the regular traffic.  If you can control your ISP 
router then this isn't an issue of course

So the solution is to throttle incoming to 99.9% of total incoming 
bandwidth.  Well, actually since you have no control over who can send 
you data, this only works in steady state.  So perhaps you should make 
it 95% or 90%.  It depends whether you mind there being the odd blip 
where someone starts sending you traffic, but it takes a second or so 
while you instruct other senders to slow down.  In the meantime you will 
be overloaded.

Now the reason for dropping to really low numbers (50%) is because most 
of the throttle filters work on bandwidth consumed over a normal 
ethernet LAN.  However, you might be using ADSL. In this case you pay 
for NOT 512Kbit/s of IP bandwidth, but 512Kb/s of ATM bandwidth.  And 
unfortunately the relationship between the two is slightly complicated.  
First you need to add 10 bytes to every IP packet for PPP overheads, 
then some other overhead if you are on PPPoE, then you have to break it 
up into 48 byte chunks and add a 5 byte header.  That will tell you how 
much bandwidth you used.

To save you the headache of worrying about those calculations consider 
sending a 49 byte packet.  It will clearly need to be split into two 48 
byte packets (yes?), then each packet has a 5 byte header = 53 bytes 
each, so that 49 byte packet takes up 53*2 = 106 bytes of bandwitch on 
your ATM line.  On the other hand if you used really large IP packets 
then the overhead would be less, consider the effect of a wasted 53 byte 
packet when you are sending in chunks of 1500 bytes a time.

So big FTP transfers with large IP packets don't waste too much, but if 
you have a load of SSH users, or some P2P users, or something else which 
spits out tons of small packets then the IP bandwidth might be loads 
less than the ADSL bandwidth, hence some people really throttle back to 
be sure they have control of the inbound connection

However, if you scroll back a few weeks you will find an experimental 
patch from me which adds the correct calculations to HTB and other 
qdiscs.  At some point I will code it up in a much neater way, but in 
the meantime it works really well as it is.  So now you can say, 
throttle me to 500Kb/s and it throttles to that much ATM bandwidth, 
regardless of how much IP bandwidth that equates to.

Clear as mud?

Ed W