Re: delay as a metric, aggregation

[Iljitsch van Beijnum <iljitsch@muada.com>]

On 6-okt-2005, at 21:51, Fred Baker wrote:

> Delay is a good metric as long as it is relatively stable. There is  
> ample experience to show that delays change in time, and when the  
> difference in two delays is comparable to the time implied by a  
> mean queue depth, can result in route oscillation. There are ample  
> research papers on the subject back to 1980-or-so, when delay was  
> in fact the primary metric used in ARPANET routing.

This is of course true, but I don't think it's as relevant today as  
it was 25 or even 10 years ago, because bandwidth has gone up so  
much. In order to incur 10 ms queuing delay on a 1 Gbps link, you  
need to have a queue of 833 packets of 1500 bytes (ignoring  
overhead). To arrive at an average queue depth of that size the line  
utilization has to be 99.88%. Although nothing is impossible, this is  
a very unlikely situation.

Also, I'm not sure many people even use this much buffer memory for a  
link. I believe Cisco uses a standard buffer of 75 packets, although  
this may be different for very high speed links.

Last but not least, if people use RED then TCP and other transports  
that use similar congestion control mechanisms will be slowed down  
anyway.

> What one wants to know is that one is getting among the best  
> service options available, but not that it is for sure the very  
> best available option at this instant.

Right.