symmetric vs. asymmetric [was: Verizon Policy Statement on Net Neutrality]

Jack Bates jbates at paradoxnetworks.net
Fri Feb 27 21:27:50 UTC 2015


On 2/27/2015 2:47 PM, Miles Fidelman wrote:
> Folks,
>
> Let's not go overboard here.  Can we remember that most corporate and 
> campus (and, for that matter home) networks are symmetric, at least at 
> the edges.  Personally, I figure that by deploying PON, the major 
> carriers were just asking for trouble down the line. It's not like 
> carrier-grade gigE switches are that much more expensive than PON gear.
>

I'll disagree on the home part. I doubt that most homes are symmetric.

Of course, what needs to happen is for standards bodies to start 
thinking more dynamic when they build their protocols where possible. 
Passive splitters obviously have the limitation of limiting frequencies, 
but our xDSL technologies and cable technologies do not have the 
restriction to my knowledge. Future protocols ideally would have a 
signaling band, recognition of frequency support bidirectionally and 
perhaps support dynamic allocation of those channels as-needed.

If an end node is saturating the upload but not using the download, why 
shouldn't the system shift the frequency usage? If only 10mb/s is being 
used out of a 50mb/s circuit for download, why not allow that extra 
capacity to be used for upload, temporarily shifting it's direction?

My 2 cents. I don't design these things, but you'd think people would 
start realizing that static allocation is kind of limiting. Giving 
someone 50mb/s with 20mb/s waste is annoying when they are saturating 
3mb/s the opposite direction. Wouldn't it be cool if your backup at 
night could use 50mb/s upstream and drop your downstream to 5mb/s 
because you aren't downloading anything?

For that matter, is there a reason we don't dynamically adjust 
frequencies on Ethernet? My servers would definitely love 1.8gb/s 
transmit since they receive very little.

Jack





More information about the NANOG mailing list