On Wed, 2008-11-19 at 17:45 +1300, Joel Wiramu Pauling wrote:

BTW IIRC, automated caching of copyrighted content is protected under fair harbour provisions of DMCA. Although that is an American perspective, I believe under NZ copyright law it could be considered "Technical Expert/Archivist" exemption.

IANAL however.

http://www.legislation.govt.nz/act/public/2008/0027/latest/DLM1122643.html Check 92E..

Maybe I've missed something here, but does implementing traffic shaping on content that is sitting in a local cache seem a little bit silly...? And is it really fair to count data that is merely being pushed from inside their own network towards people's (rather insignificant) usage caps? Any ISP type people care to enlighten me? Or is it just that accounting departments have taken what could be a great benefit to the user and turned it into a way to make easier money? Scott Howard wrote:

I've spoken to a number of ISPs over the past year who are using P2P caching products, and they all stated that they were seeing benefits from them.

You're correct in that there's probably a lot of data which is only requested once (just like a HTTP cache), but there will also be files which could be downloaded dozens of hundreds of more of times - and 100 downloads of a 60 min TV shows (350MB) is 35GB of saved traffic - not a huge amount in the scheme of things, but multiple it out to a month or a year and the numbers get significant enough.

Most of these products also cache in both directions. ie, they will pull from the cache not only when one of your clients downloads a file from elsewhere on the Internet, but also when a host elsewhere on the Internet downloads from one or your clients. Depending on how you pay for your tail to the client the caching can have a saving not only on bandwidth out to the Internet but even on domestic tail costs.

As others have said, the copyright issue isn't really any different to that for a web cache. Sure, you might suspect that there'd be a higher percentage of "illegal" material transfered over P2P than HTTP, but that doesn't mean it's all illegal...

Most/all of the caching products also give the ability to shape P2P traffic, which can be a good thing or a bad thing depending on how you manage it. From memory Exetel in Australia openly admit that they shape P2P traffic during peak hours, but let it run free during quiet times, and that hasn't caused them any issues. Comcast in the US on the other hand got taken to court for doing basically the same thing. Most of the ISPs I've spoken to are doing some form of P2P shaping, although generally not signifant, and normally they don't make that fact known.

I'd say the majority of ISPs are still not doing P2P caching, although more and more are at least using it on part of their network. If you'd asked me a few months ago I would have said that this number will increase over time, but at the moment in the US at least there seems to be a move away from P2P and back to HTTP/streaming protocols. Why bothered downloading a P2P of a TV show when you can watch it on Hulu, InnerTube or YouTube? Why bothered downloadind a movie when you can stream it form NetFlix?

Scott.

On Tue, Nov 18, 2008 at 6:42 PM, Cameron Kerr mailto:ckerr(a)cs.otago.ac.nz> wrote:

I'm updating my labnotes for my network management paper regarding the use of proxy caches, and previously I have pointed to the possibility of caching peer-to-peer traffic as shown in the paper /Deconstructing the Kazaa Network/

http://portal.acm.org/citation.cfm?coll=GUIDE&dl=GUIDE&id=837393 http://portal.acm.org/citation.cfm?coll=GUIDE&dl=GUIDE&id=837393

I would like to know whether or not caching of peer-to-peer is something that ISPs actually do today, and if not why not. I imagine there are probably some interesting legal interactions, but I'm eager to find out how industry is moving in this field.

-- Cameron Kerr mailto:ckerr(a)cs.otago.ac.nz> Teaching Fellow, Computer Science, University of Otago

_______________________________________________ NZNOG mailing list NZNOG(a)list.waikato.ac.nz mailto:NZNOG(a)list.waikato.ac.nz http://list.waikato.ac.nz/mailman/listinfo/nznog

------------------------------------------------------------------------

_______________________________________________ NZNOG mailing list NZNOG(a)list.waikato.ac.nz http://list.waikato.ac.nz/mailman/listinfo/nznog

Drew Calcott wrote:

Maybe I've missed something here, but does implementing traffic shaping on content that is sitting in a local cache seem a little bit silly...? And is it really fair to count data that is merely being pushed from inside their own network towards people's (rather insignificant) usage caps?

Any ISP type people care to enlighten me? Or is it just that accounting departments have taken what could be a great benefit to the user and turned it into a way to make easier money?

But providing bits to the customer is what an ISP gets paid for. If they can find a more efficient way of doing that, good on them. We rely on competition to prevent excessive ISP profit-taking, which is a different discussion. - Donald Neal -- Donald Neal | "It's surprising how effective politicians Research Officer | can be when they stop listening to their WAND | spin droids." The University of Waikato | - The Doctor

Hi Drew,

-----Original Message----- From: Drew Calcott [mailto:drew.calcott(a)auckland.ac.nz] Sent: Thursday, 20 November 2008 8:43 a.m. To: nznog(a)list.waikato.ac.nz Subject: Re: [nznog] Caching peer-to-peer traffic: a done thing?

Maybe I've missed something here, but does implementing traffic shaping on content that is sitting in a local cache seem a little bit silly...? And is it really fair to count data that is merely being pushed from inside their own network towards people's (rather insignificant) usage caps?

Last mile delivery and exchange/cabinet backhaul links are typically highly oversubscribed. Tails can now be the most expensive part of running an ISP (aside from support, if one happens to offer something resembling support.) Witness Probe-era DSLAMs backhauled by E1. (Can we say Conklin?) If you don't shape everything, including content in local cache, you won't have much of a service left. -JB

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Ask a few on-topic questions and lots of people jump down your throat - except Aj. Who'da thunk it? He's normally pretty quick to get a couple of jabs in on me when the opportunity arises. The only possible explanation I can think of is that he has a word-of-the-day calendar on his desktop - and today's word was "facetious". - --- Drew Calcott Linux System Administrator Science IT University of Auckland (p) +64 9 373 7599 x84269 Alastair Johnson wrote:

Drew Calcott wrote:

...

Maybe I've missed something here, but does implementing traffic shaping on content that is sitting in a local cache seem a little bit silly...? And is it really fair to count data that is merely being pushed from inside their own network towards people's (rather insignificant) usage caps?

As other people have mentioned, shaping towards the subscriber access loop may be required to avoid congestion in the access network. Shared spectrum networks in particular have this requirement especially for the upstream direction (i.e. DOCSIS style).

On the transit-ingress side of the equation, the cache appliance provides a nice easy place to provide aggregate contention - and if it has logic in it that can do it itself based on some rules (time-of-day, number of subscribers, content type, whatever) then it makes it a really good place to ensure that the cache itself isn't congesting your upstream transit.

aj

-----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.8 (Darwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iEYEARECAAYFAkkkpPgACgkQD+yXTWfduLHS/gCgjDSemBp9K7bSUgMCN+cQJOEr 3i8An29ZYfm7wVwTO1CuXYNQDPjj3aHR =b0Fs -----END PGP SIGNATURE-----

On Wed, Nov 19, 2008 at 11:42 AM, Drew Calcott wrote:

Maybe I've missed something here, but does implementing traffic shaping on content that is sitting in a local cache seem a little bit silly...?

Not at all. For starters much/most of the data isn't going to be in the cache anyway, so shaping that traffic makes sense as it's real bandwidth - both from the Internet as well as to the client. For content that is in the cache, the odds are that you're going to be able to send it to the client far faster than they would get it if the cache didn't exist. Now, if the purpose of the cache was to speed up client access this would be a good thing - but that's generally not why ISPs are using P2P caches. ie, should the customer benefit from something the ISP has done to save themselves money - especially if in doing so it reduces the amount of money the ISP saves? Ask that to 10 people and I'm sure you'll get 12 or more differing answers. There's plenty of claims you can make here, from "the more the ISP increases the speed of the P2P download, the more likely the user is to download even more content, and thus use more bandwidth and negate the benefit of the cache for the ISP" to "the faster the content downloads the faster the user will turn their computer off and go watch what they just downloaded". I don't think anyone has done enough research to show what actually occurs in these situations. To tell you the truth, I'm not sure if most ISPs shape traffic that hits the cache or not - It's not a question I've ever asked, but I will start asking it now! And is it really fair to count data that is merely being pushed from

inside their own network towards people's (rather insignificant) usage caps?

This is a country-biased perspective. Many countries don't do caps, so it becomes more relevant. Even for countries that do caps, the expectation is that every user won't download as much as their cap allows, only some of them will. Here in the US caps generally don't exist. Comcast has recently introduced a 250GB/month cap on some users, and you wouldn't believe the screaming... Scott

On Wed, Nov 19, 2008 at 2:38 PM, Alastair Johnson wrote:

That said - as I mentioned to Cameron offlist - I have seen a reasonably large number of midsize ISPs in Asia deploying P2P Caches. It has not been so common in the telco-size ISPs yet, which I suspect is because the lawyers have heart attacks at the idea.

There's at least one telco-size ISP running P2P caching in a country very near to you - I was talking to them about it only a month or so back :) Scott.