[ULS-SIG] P2P Protocols (was: Decentralized, Smart Messaging)

[Duncan McGreggor]

Alex, these look like great resources -- once I'm settled in the new
house (we close tomorrow!), I'm gonna sit down and read through them
closely.

Thanks!

d

On Mon, Sep 21, 2009 at 10:16 AM, Alex Drahon <adrahon at gmail.com> wrote:
> On Sat, Sep 19, 2009 at 4:18 AM, Duncan M. McGreggor
> <duncan.mcgreggor at gmail.com> wrote:
>>
>>  * that would allow applications to auto-discover peers
>>
>>  * where a collection of nodes would segment into groups and elect a
>> peer to act as their communication proxy to managing or reporting
>> entities for that segmented group of nodes
>>
>>  * where the elected proxy would perform actions such as forwarding
>> messages or holding messages in a queue
>>
>>  * that has the flexibility to re-form node groupings and re-elect a
>> proxy based on various conditions
>>
>
> I mentioned P2P protocols like Bittorrent because they share part of these
> requirements (except the elected proxy) and they are proved to scale. I read
> a little more on the subject and found a few interesting things:
>  - There are a lot of docs and code to read if you're interested in overlay
> networks http://en.wikipedia.org/wiki/Overlay_network, which is part of what
> we're discussing here.
>  - The interesting (for our purpose) part of P2P protocols
> is generally based on a Distributed Hash Table, like Kademlia
>  http://en.wikipedia.org/wiki/Kademlia.
>  - Two interesting implementation in Python are Khashmir (which serves as
> the basis for Bittorrent DHT) http://khashmir.sourceforge.net/ and
> Entangled http://entangled.sourceforge.net/, which has more functionalities.
> Both use Twisted.
>  - A good article on the subject is "The Design of
> CoralCDN" http://sns.cs.princeton.edu/2009/04/the-design-of-coralcdn/
>  - The Coral Cache project also has developed a system that could be of
> interest for the "proxy election" system, OASIS http://oasis.coralcdn.org/
> Maybe we could include this in the SOOM experiments, as an alternative
> implementation, for instance.
> Other relevant things I stumbled upon:
>  - PSYC http://www.psyc.eu/whitepaper/
>  - Delay Tolerant Network Research Group http://www.dtnrg.org/wiki
>
>
> _______________________________________________
> ULS-SIG mailing list
> ULS-SIG at python.org
> http://mail.python.org/mailman/listinfo/uls-sig
>
>