Verizon Policy Statement on Net Neutrality

Owen DeLong owen at delong.com
Sun Mar 1 06:00:49 UTC 2015


>> causes enough interference to prevent reverse adsl (i.e. greater bandwidth
>> from customer to exchange) from working well.
> 
> So SDSL didn't exist? Anyhow, *DSL is falling so far behind it's
> difficult to analyze what could have been.
> 

SDSL existed, but every bps upstream that you get in SDSL is a bps that you
can’t use for downstream. Usually in 128kbps chunks. So, for example, if you
have a line that will support ADSL/VDSL at 1536/768, it can also work at
2048/256 or 1280/1024. For SDSL, you’d get a maximum of 1152/1152 on
that same line.

Looking at my usage patterns, 1536/768 is probably the best balance. However,
I’d bet that the vast majority of residential users would be happier with 2048/256
than 1536/768. I know that 1152/1152 definitely wouldn’t be desirable in my case
or in most cases.

Now, as speeds get higher and the downstream speed starts to be adequate or
exceed adequate, getting close to symmetry looks more appealing because
stealing from the downstream channel to provide faster uploads on the occasions
when the traffic pattern shifts is less painful most of the time.

Still, that doesn’t mean that symmetrical is the best choice in all cases.

>> Some operators used and continue to use asymmetric bandwidth profiles and
>> bandwidth caps as methods for driving up revenue rather than anything else
>> in particular.  International cellular roaming plans come to mind as one of
>> the more egregious example of this, but there are many others.
> 
> Sure. once it became institutionalized and the market got used to it
> why not sell tiered bandwidth services at different price points, but
> that could have been true of symmetrical service also.

And, indeed, it is even at the large end of the spectrum, even today.

A 100Mbps circuit will cost you a little more than 10% of a 1G circuit.
A 1G circuit will cost a little more than 10% of a 10G circuit, etc.

Oddly, however, at the residential side, this often is inverse. Often,
a 10M service will cost you less than 1/2 of a 20M service which is
again significantly less than 1/2 the price of a 50M service.

For example:

10M/1M		$45/month
20M/5M		$120/month
50M/10M		$300/month

I don’t know if this is still an accurate reflection of any actual provider’s
pricing, but it is adequately exemplary of several that are out there today.

Owen




More information about the NANOG mailing list