[AusNOG] Why not Symmetric ingress and egress?

Paul Brooks pbrooks-ausnog at layer10.com.au
Mon Jun 21 12:39:57 EST 2010


On 18/06/2010 8:40 AM, Mark Smith wrote:
> Just quickly,
>
> On Thu, 17 Jun 2010 22:02:21 +1000
> Paul Brooks<pbrooks-ausnog at layer10.com.au>  wrote:
>
>    
>> On 17/06/2010 8:09 PM, Mark Smith wrote:
>>      
>>> My guess is that ADSL was chosen as the broadband technology to use,
>>> rather than a symmetrical DSL technology (not in Australia, I'm
>>> talking by the broadband groups who standardise it i.e. the Annex M
>>> people), because it sounded right for the way people were using the
>>> Internet at the time (consumers rather than producers), rather than
>>> understanding that the Internet protocols have operated over symmetric
>>> links for most of their life and it is therefore an unstated design
>>> assumption. If that wasn't the case, I don't think the above RFC would
>>> exist and be a Best Current Practice RFC.
>>>
>>>        
>> Nice try Mark, but your guess would be wrong!. ADSL was developed as a
>> technology for video delivery
>>      
>                   ^^^^^^^^^^^^^^
>
> Exactly - a purely asymmetric traffic profile. There'll never be a
> reason in a video delivery system for a high bandwidth backchannel -
> because it is only for video signalling and nothing else.
>    

Actually, I was responding to your guess that ADSL asymmetry was 
developed due to "the way people were using the Internet at the time".
Yes, the history lesson was a bit long-winded (it was almost Friday!), 
but the point was that ADSL pre-dates "people using the Internet" on a 
large scale by at least half a decade or more. It is only serendipity, 
not design, that usage patterns happened to match the bandwidth profile 
of ADSL and ADSL2.
I'll grant you that it was probably a factor in pushing for the 
extension to ADSL2+ - which made it even more asymmetric.

> there) of link technologies that have been used to deploy Internet
> protocols, asymmetric links, such as 1200/75, cable, ADSL, 56K analog
> modems, are exceptions, not the common case. There would be no reason
> for the RFC I pointed to to exist if TCP's operation wasn't negatively
> impacted by asymmetry.
>    
Well, with FTTP we're talking about access networks , not backbone/trunk 
networks, so most of the technologies you list aren't really applicable 
in an access network context.

Mark, if you look at the RFC closely, you'll see that the main issue 
they are discussing is asymmetric transfer one-way latency, not 
asymmetric bandwidth links as such. The problem is caused by differences 
in TIMING, not bandwidth.  Asymmetric bandwidth links certainly can 
cause asymmateric latency, if the smaller direction is congested 
(queuing delay), as can asymmetric packet paths in each direction caused 
by hot-potato routing or other things.

The technologies you are concerned about - ADSL, and asymmetric FTTP 
connections through NBN - are access network issues, so multiple paths 
in each direction are not an issue (in the segment from the premise to 
the DSLAM or OLT). On an UNCONGESTED link (and this is the key point) no 
matter how asymmetric the DSL or FTTP connection might be, the packet 
transfer time in each direction is close enough to the same - so the 
problems described in RFC 3449 are not applicable.

Note - I'm NOT saying that NBN connections should not be symmetric, and 
I'm not saying RFC 3449 is wrong.
What I am saying is that RFC 3449 is not a valid reason why NBN 
connections should have symmetric bandwidth.

In practical terms - most residential users will get a better Internet 
experience out of a 75M down /25M up connection than a 50M/50M 
connection,  and RFC 3449 effects won't rear its ugly head on either of 
them, unless you transmit so much you congest the upstream path.

[...removed...]
>
> (I didn't mention the ADSL video story, but my reference is Computer
> Networks 4th edition)
>    
Here is an interesting if a little self-absorbed take on the early 
history from John Cioffi  - 
http://www.eetimes.com/disruption/essays/cioffi.jhtml

Paul.

-- 
Paul Brooks               |         Mob +61 414 366 605
Layer 10 Advisory         |         Ph  +61 2 9402 7355
-------------------------------------------------------
Layer 10 - telecommunications strategy&  network design




More information about the AusNOG mailing list