Re: [alto] Problems with the Alto problem statement

[Laird Popkin <laird at pando.com>]

I just spoke with Paul, and I'd like to apologize publicly if my email came across as insulting. I think that we're having a very productive and important discussion here, and I wouldn't want to derail that through clumsy wording.

- Laird Popkin, CTO, Pando Networks
  mobile: 646/465-0570

----- Original Message -----
From: "Laird Popkin" <laird at pando.com>
To: "Y. Richard Yang" <yry at cs.yale.edu>
Cc: alto at ietf.org
Sent: Saturday, November 15, 2008 1:42:57 PM (GMT-0500) America/New_York
Subject: Re: [alto] Problems with the Alto problem statement

I think I see quite a few incorrect assumptions about ALTO that you're making in this discussion. I'll try to make a list so that we can address them:

1) You think that P2P companies won't use ALTO guidance because there is a 'tradeoff' between performance and resilience, where slower performance benefits the networks by forcing people to remain in the p2p swarm longer. I believe (based on discussions with numerous p2p companies) that they want to improve performance, and use other mechanisms to encourage contribution to the network. I'd suggest that the strong interest by p2p networks in P4P and now ALTO is evidence that they view this guidance as beneficial to their networks.

2) You believe that ALTO requires p2p networks to expose personally identifiable information of their users and their download activity to ISPs. This is not the case - ALTO specifically has a requirement that it not require P2P networks to expose any user or content identity to the ALTO server.

3) As a result of (2), you believe that neither ISPs not P2P networks will make use of ALTO, except for 'legit' downloads. Since (2) isn't the case, it shouldn't be a reason to prevent implementation of ALTO. ISPs may choose to provide ALTO guidance selectively, of course, but it seems premature to assume that they will all do so.

4) You assume that 'legit' p2p traffic is not worth optimizing. While a few years ago this was probably the case, there are many 'legit' p2p platforms distributing content (e.g. BBC iPlayer, Joost, NBC Direct, almost all large game downloads, e.g. World of Warcraft) that are moving significant volumes of traffic, which is certainly worth optimizing. And the projections are that 'legit' p2p data volumes will continue to grow. So while ALTO as a technology isn't limited to 'legit' content, even if it were only used that way the impact would be significant and growing. 

5) You assume that ALTO applies only to tracker-based p2p networks, and thus that ISPs would be required to approve specific Trackers (i.e. not where the pirated content is). This is not the case; as Richard emailed, ALTO is applicable to both tracker-based and trackerless p2p networks (as well as any other application where network locality is of value). So if p2p client X implements ALTO, it would benefit from ALTO for all of its downloads (if it was running at an ISP that provided ALTO guidance to it).

In addition, I'll point out that (if P4P is any guide) implementing ALTO should be extremely easy for ISPs to implement, so even if the benefit were small, it's almost certainly worth doing anyway.

- Laird Popkin, CTO, Pando Networks
  mobile: 646/465-0570

----- Original Message -----
From: "Y. Richard Yang" <yry at cs.yale.edu>
To: "Paul Hoffman" <paul.hoffman at vpnc.org>
Cc: alto at ietf.org
Sent: Friday, November 14, 2008 10:26:12 PM (GMT-0500) America/New_York
Subject: Re: [alto] Problems with the Alto problem statement

Dear Paul,

>> The P4P WG participation includes many major p2p networks, including Pando, BitTorrent, Solid State, Grid Networks, LimeWire, Abacast, Azureus, Joost, etc.
>>     
>
> Just to be clear here, "BitTorrent" in your list is a single BitTorrent tracker which happens to be run by BitTorrent.com. It is not representative of the BitTorrent communities of which I spoke, namely the ones that are inundating ISPs.
>
>   
Just to add some information. Vuze (formerly called Azureus) is one of 
the most popular BT clients. I assume that users using Vuze is a part of 
the BT community that you are referring to. Please correct me if I 
misunderstand here. Vuze is also a member of P4P. They are working 
pretty actively lately to improve both network efficiency and 
application performance. For example, their most recent release adds 
features to try to use network information through a network coordinate 
system. We are also preparing to launch a client-based plugin for Vuze 
to utilize P4P information (one of the ALTO framework). The objective of 
the plugin is that, through this plugin, either tracker-based or 
trackerless swarms will be able to start to use network information. We 
hope that this will be a good starting point to start to have affects on 
the large amount of P2P traffic that you are referring to, if I 
understand you correctly. There are sure other BT clients/trackers that 
we need to work on to better move the tide, and to maximize effects, we 
need to enter through both trackers and clients. But I do believe that 
ALTO can make a difference here. Maybe I am too optimistic here...
 
Richard

>> The summary is that in exactly the scenario you assert that ALTO "does nothing" P4P (think of it as pre-ALTO) provide huge performance and efficiency benefits.
>>     
>
> That is quite possibly true; it is also not relevant to what I said. I was speaking about the BitTorrent communities that are currently the bandwidth problem for ISPs; you are speaking about other p2p communities whose current bandwidth usage is probably at least an order of magnitude less than the ones of which I am speaking.
>
> Of course, in the future, the bandwidth market may shift towards the kinds of systems you are talking about. If so, Alto could very well be helpful. People clearly want to work on that future problem, and Alto might be a good place to do it. My expressly-stated concern was that the problem statement document has wording that makes it sound like the Alto proposals might have an affect on the current large amount of traffic, which I believe it won't.
>
> <rest of information about Pando and P4P elided>
>
> --Paul Hoffman, Director
> --VPN Consortium
> _______________________________________________
> alto mailing list
> alto at ietf.org
> https://www.ietf.org/mailman/listinfo/alto
>   


_______________________________________________
alto mailing list
alto at ietf.org
https://www.ietf.org/mailman/listinfo/alto
_______________________________________________
alto mailing list
alto at ietf.org
https://www.ietf.org/mailman/listinfo/alto
_______________________________________________
alto mailing list
alto at ietf.org
https://www.ietf.org/mailman/listinfo/alto