nanog mailing list archives

Re: iOS 7 update traffic


From: Mark Andrews <marka () isc org>
Date: Fri, 20 Sep 2013 09:13:39 +1000


In message <64245AC1-BC00-4928-B2F7-F259E8632655 () puck nether net>, Jared Mauch 
writes:

On Sep 19, 2013, at 4:36 PM, "John Souvestre" <johns () sstar com> wrote:

Hi Jared.

The attitude in this email I have encountered elsewhere.  Apple pays
for bandwidth, customers pay for access. Not sure why their release
strategy is so highly critiqued.

Because it impacts other, non-Apple customers.  Or, it costs the ISP
more
(passed through to all customers) to add capacity to handle an
infrequent peak
load.

Question/suggestion:  Could Apple perhaps shift their release to a
Saturday
morning?  I would think that this would go a long way to diluting the
peak.

John

    John Souvestre - New Orleans LA - (504) 454-0899


I think there's a lot that could be done when looking at how to shift
this.

I've seen one other carrier privately talk to me about the impact and
possible impacts to their network.  Most of these are folks (along with
warren) who are worried about their RF budgets and these event traffic,
or even just the nightly traffic peaks.

I have advised some in the past to put up caches, but the content owners
also make it difficult to do this.  Apple sets very short expire values,
and you end up with lots of "bad" settings.  Apple devices don't honor
DHCP option 252 either.

Oh you mean that option that never made it past a internet-draft
that expired 13 years ago[1] and is in the private range[2] to boot.

If you want proxy discovery to work on all devices complete the
process of getting a code point allocated then get the OS vendors
to query for it.  252 is fine for experimenting / proof of concept
but it really is the wrong value for long term use.

Mark

[1] http://tools.ietf.org/html/draft-ietf-wrec-wpad-01
[2] http://www.iana.org/assignments/bootp-dhcp-parameters/bootp-dhcp-parameters.xhtml

This means you're stuck with a transparent proxy, (lets just say squid)
putting itself in all tcp/80 traffic, or worse with lots of settings
like: reload-into-ims override-expire etc..

This can solve some problems for those who have a 20-50Mb/s link to the
internet and 50-100 customers each getting 1Mb/s+ on their CPE.

The results I've always seen are you need to find the strategic location
to deploy these caches, capabilities or expand your network bandwidth,
etc..

Based on all the recent people asking for a fast link in "X" location
recently, I'm hoping there will be some better match-making happening
soon.

- Jared

-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka () isc org


Current thread: