Re: Microsoft's participation in World IPv6 day
--- Valdis.Kletnieks@vt.edu wrote: What's special about Sunday peaks and Friday lows on that graph? I think I asked that once before, with no firm conclusions. But there's a definite sawtooth there, big enough that we probably want to understand it. ----------------------------------------- There're about 52 peaks in a year on the timeline... :-) scott
On Fri, 03 Jun 2011 15:20:22 PDT, Scott Weeks said:
There're about 52 peaks in a year on the timeline... :-)
Right. But why is Google seeing noticeably higher IPv6 loads on Sunday and lower loads on Friday? I'd buy a "different traffic pattern for home/office", but then you'd expect Friday to be about the same as M-Th, and Sat/Sun to be about even.
On 3 June 2011 23:24, <Valdis.Kletnieks@vt.edu> wrote:
On Fri, 03 Jun 2011 15:20:22 PDT, Scott Weeks said:
There're about 52 peaks in a year on the timeline... :-)
Right. But why is Google seeing noticeably higher IPv6 loads on Sunday and lower loads on Friday? I'd buy a "different traffic pattern for home/office", but then you'd expect Friday to be about the same as M-Th, and Sat/Sun to be about even.
I wonder if there is a disproportionately large amount of IPv6 usage in the Middle East where a number of countries have their weekend on Friday and Saturday, with Sunday being the first day of their working week? UAE and Israel as examples. Tony
On Sat Jun 04, 2011 at 12:04:42AM +0100, Tony McCrory wrote:
I wonder if there is a disproportionately large amount of IPv6 usage in the Middle East where a number of countries have their weekend on Friday and Saturday, with Sunday being the first day of their working week? UAE and Israel as examples.
Interestingly, providing access services to students in the UK, I see Friday and Saturday as my quiet days, with Sunday being as busy as Monday - Thursday. I always just put it down to students going out drinking on Fridays and Saturdays. Simon
On Jun 3, 2011, at 3:24 PM, Valdis.Kletnieks@vt.edu wrote:
On Fri, 03 Jun 2011 15:20:22 PDT, Scott Weeks said:
There're about 52 peaks in a year on the timeline... :-)
Right. But why is Google seeing noticeably higher IPv6 loads on Sunday and lower loads on Friday? I'd buy a "different traffic pattern for home/office", but then you'd expect Friday to be about the same as M-Th, and Sat/Sun to be about even.
Everyone is out interacting with Humans on Friday nights. Sunday, everyone is home trying to avoid dealing with their families. (Mostly tongue in cheek) Owen
On Jun 3, 2011, at 4:13 PM, Owen DeLong wrote:
On Jun 3, 2011, at 3:24 PM, Valdis.Kletnieks@vt.edu wrote:
On Fri, 03 Jun 2011 15:20:22 PDT, Scott Weeks said:
There're about 52 peaks in a year on the timeline... :-)
Right. But why is Google seeing noticeably higher IPv6 loads on Sunday and lower loads on Friday? I'd buy a "different traffic pattern for home/office", but then you'd expect Friday to be about the same as M-Th, and Sat/Sun to be about even.
Everyone is out interacting with Humans on Friday nights.
Sunday, everyone is home trying to avoid dealing with their families.
Note that from Geoff's published experiment presented in IETF v6ops the success rate of v6 connection attempts particularly auto-tunneled is higher on the weekends than during weekdays, you can thank corporate firewall policy for that particular phenomena. http://tools.ietf.org/agenda/80/slides/v6ops-22.pdf
(Mostly tongue in cheek)
Owen
Note that from Geoff's published experiment presented in IETF v6ops the success rate of v6 connection attempts particularly auto-tunneled is higher on the weekends than during weekdays, you can thank corporate firewall policy for that particular phenomena.
Indeed... Unfortunately, this means that LSN is going to _REALLY_ suck for such tunnel users. Owen
On Jun 4, 2011, at 12:09 PM, Owen DeLong wrote:
Note that from Geoff's published experiment presented in IETF v6ops the success rate of v6 connection attempts particularly auto-tunneled is higher on the weekends than during weekdays, you can thank corporate firewall policy for that particular phenomena.
Indeed... Unfortunately, this means that LSN is going to _REALLY_ suck for such tunnel users.
The smart money is on there being no-saving the auto-tunneling users. The situation is not that good now and it will get worse.
Owen
In message <D90C3982-9321-4BC2-A37C-8DBF2321AD2A@bogus.com>, Joel Jaeggli write s:
On Jun 4, 2011, at 12:09 PM, Owen DeLong wrote:
=20 Note that from Geoff's published experiment presented in IETF v6ops = the success rate of v6 connection attempts particularly auto-tunneled is = higher on the weekends than during weekdays, you can thank corporate = firewall policy for that particular phenomena. =20 http://tools.ietf.org/agenda/80/slides/v6ops-22.pdf =20 Indeed... Unfortunately, this means that LSN is going to _REALLY_ suck = for such tunnel users.
The smart money is on there being no-saving the auto-tunneling users. = The situation is not that good now and it will get worse.
There really is no reason that everyone has to be behind a LSN. ISP's offer residential customers differentent levels of service today. See the web pages to re-open port 25 for examples of this. There is no reason that they can't do a similar thing to move customers who are doing things that break with LSN out from behind the LSN. Also when Microsoft and Apple have shipped fixed versions of IE and Safari that have sub-second failover most of the visible issues with broken 6to4 tunnels will disappear. Mark -- Mark Andrews, ISC 1 Seymour St., Dundas Valley, NSW 2117, Australia PHONE: +61 2 9871 4742 INTERNET: marka@isc.org
2011/6/6 Mark Andrews <marka@isc.org>:
There is no reason that they can't do a similar thing to move customers who are doing things that break with LSN out from behind the LSN.
Oh, you're right, they'll surelly do that. But not in time, and not for free. LSN is beeing actively implemented in the core network of several ISPs, and most didn't yet consider it as optional. Nor are ready for v6 connectivity to residential customers, though. For users behind a forced NAT (no way to disable it on the CPE) or LSN, the only way out is still tunneling. Talking about bandwidth and infrastructure waste... -- Jérôme Nicolle
In message <BANLkTimGkuL7ycrYG6kTC1U7OWis9dOA+YaV-YHwr+5C8=0Pxw@mail.gmail.com> , =?UTF-8?B?SsOpcsO0bWUgTmljb2xsZQ==?= writes:
2011/6/6 Mark Andrews <marka@isc.org>:
There is no reason that they can't do a similar thing to move customers who are doing things that break with LSN out from behind the LSN.
Oh, you're right, they'll surelly do that. But not in time, and not for fre= e.
Well here in Australia I would be calling the ACCC is a ISP tried to charge extra for a address that is not behind a LSN. As for in time it should be in place before they turn on LSN. If you can adjust port 25 filters whenever a customer gets a new address you can also ensure that they get address from the correct pool when they connect to the network. This really isn't rocket science. It's updating the provisioning database from a web form and generating new configs based on that database. Yes there is some work required to ensure that this gets done properly and there needs to be checks that address pools are appropriately sized. If I were doing it I would also have checkboxes for some of the more common reasons and include IPv6 connectivity as one then have a 6 month grace period once the ISP offers IPv6 connectivity before removing that as a valid reason for needing a address that is not behind the LSN.
LSN is beeing actively implemented in the core network of several ISPs, and most didn't yet consider it as optional. Nor are ready for v6 connectivity to residential customers, though.
For users behind a forced NAT (no way to disable it on the CPE) or LSN, the only way out is still tunneling. Talking about bandwidth and infrastructure waste... -- Mark Andrews, ISC 1 Seymour St., Dundas Valley, NSW 2117, Australia PHONE: +61 2 9871 4742 INTERNET: marka@isc.org
2011/6/6 Mark Andrews <marka@isc.org>:
Well here in Australia I would be calling the ACCC is a ISP tried to charge extra for a address that is not behind a LSN.
On France, our bigger ISP charges extra for a fixed IP. Its network beeing rather old-fashioned, every DSL (and residential fiber) line is terminated on a LNS through a PPP session. Assigning a fixed IP is technically done by adding a RADIUS parameter to force the termination LNS to those having a static pool. The same method could be applied to get a user out of the LSN, but as their LSN isn't yet in place, we have no clue of what they'll do. We just know their CEO just announced ongoing discussions with CDNs (including google) about service differenciation and charging users for priority bandwidth.
If you can adjust port 25 filters whenever a customer gets a new address you can also ensure that they get address from the correct pool when they connect to the network. This really isn't rocket science.
Well, you can't open port25 on Orange's ADSL service ;) -- Jérôme Nicolle
In message <BANLkTik+qgTPXOwaSsHseYQbP0MBJw25Tb2bO6b3kyrKvhGj=Q@mail.gmail.com>, = ?UTF-8?B?SsOpcsO0bWUgTmljb2xsZQ==?= writes:
2011/6/6 Mark Andrews <marka@isc.org>:
Well here in Australia I would be calling the ACCC is a ISP tried to charge extra for a address that is not behind a LSN.
On France, our bigger ISP charges extra for a fixed IP. Its network beeing rather old-fashioned, every DSL (and residential fiber) line is terminated on a LNS through a PPP session. Assigning a fixed IP is technically done by adding a RADIUS parameter to force the termination LNS to those having a static pool. The same method could be applied to get a user out of the LSN, but as their LSN isn't yet in place, we have no clue of what they'll do. We just know their CEO just announced ongoing discussions with CDNs (including google) about service differenciation and charging users for priority bandwidth.
Which just reinforces the point that it is not technically hard. Remember when you introduce LSN you are degrading the service not adding to it. I can seen consumer bodies saying thay you need to compensate your customers unless there is a free path to get into the exception pool.
=C2=A0If you can adjust port 25 filters whenever a customer gets a new address you can also ensure that they get address from the correct pool when they connect to the network. =C2=A0This really isn't rocket science.
Well, you can't open port25 on Orange's ADSL service ;)
--=20 J=C3=A9r=C3=B4me Nicolle -- Mark Andrews, ISC 1 Seymour St., Dundas Valley, NSW 2117, Australia PHONE: +61 2 9871 4742 INTERNET: marka@isc.org
From nanog-bounces+bonomi=mail.r-bonomi.com@nanog.org Fri Jun 3 17:25:39 2011 To: surfer@mauigateway.com Subject: Re: Microsoft's participation in World IPv6 day From: Valdis.Kletnieks@vt.edu Date: Fri, 03 Jun 2011 18:24:42 -0400 Cc: nanog@nanog.org
--==_Exmh_1307139882_2680P Content-Type: text/plain; charset=us-ascii
On Fri, 03 Jun 2011 15:20:22 PDT, Scott Weeks said:
There're about 52 peaks in a year on the timeline... :-)
Right. But why is Google seeing noticeably higher IPv6 loads on Sunday and lower loads on Friday? I'd buy a "different traffic pattern for home/office", but then you'd expect Friday to be about the same as M-Th, and Sat/Sun to be about even.
Possibly traffic from the 'wrong side' of the International Date line??
participants (9)
-
Joel Jaeggli -
Jérôme Nicolle -
Mark Andrews -
Owen DeLong -
Robert Bonomi -
Scott Weeks -
Simon Lockhart -
Tony McCrory -
Valdis.Kletnieks＠vt.edu