The idea behind "Gossip" protocols is that when you talk to another peer, you piggy back information about what other peers you know (hence you gossip about them).

The idea behind SWIM is that you can't possibly ping each and every one of them to see if they are alive (if you did you would have N*N messages every few seconds, and with N in the order of thousands you might find millions or billions of messages floating around). So what you do is randomly pick one, and ping it. Along with the ping and the response to the ping to piggy back messages about the state of the peers that you know.

It can be shown that this model will converge to the idea "heart beat" style of pinging and have considerably less traffic.

Chaz.

David Barrett wrote:
I'm not familiar with GOSSIP/SWIM -- can you give a quick summary?

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:p2p-hackers-
[EMAIL PROTECTED] On Behalf Of Chaz.
Sent: Sunday, June 17, 2007 2:30 AM
To: theory and practice of decentralized computer networks
Subject: Re: [p2p-hackers] p2p, mDNS and scaling

David,

   I like mDNS as well but for the specific application I built I wrote
a membership module (in Python) using a GOSSIP like approach (called
'SWIM'). It works well for thousands of peers but I would have liked to
switched to mDNS. But the concern was the scaling; and I was hoping
someone had already tried to simulate it and to tell me it did or didn't
work.

Chaz.

David Barrett wrote:
I'm really fascinated by mDNS.  I think it's an excellent design, and
now
that Safari is out for Windows (with Bonjour support built in, I'm told)
then it'll probably start to gain greater acceptance.

Indeed, I think mDNS + dynamic DNS is the recipe for a really
interesting,
completely decentralized naming system that works seamlessly both on and
off
the internet:

Users register a domain name using any of the many registrars out there.
They transfer it to the care of any of the many dynamic DNS providers,
and
then configure their client to use that name and update the dynamic DNS
server whenever the client's IP address changes.  Finally, the client
listens on UDP port 53 and responds via mDNS to any request for that
domain
name.

The net result is a domain name that always resolves to that client's
IP,
whether you do the resolution using the regular DNS system on the
internet,
or whether you do it using mDNS over a disconnected LAN / ad-hoc
network.
But more back to your question, I'm not sure mDNS is designed to or
capable
of scaling up to that level.  As far as I know (but I might be way off),
mDNS works by broadcasting totally standard DNS requests using UDP
broadcast
and, if anybody requests your name, you respond.  This requires that
everyone is within UDP broadcast range of each other, which typically
means
they're on the same physical LAN or VLAN.

Does anyone now if this is inaccurate?  If so, it doesn't seem suited to
scale to the level you need.

-david

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:p2p-hackers-
[EMAIL PROTECTED] On Behalf Of Chaz.
Sent: Saturday, June 16, 2007 6:17 AM
To: theory and practice of decentralized computer networks
Subject: [p2p-hackers] p2p, mDNS and scaling

Right now I have a p2p application that uses a proprietary membership
protocol. This protocol has been shown to scale to thousands of peers.
I
was thinking of changing over to something like Bonjour. After a little
research I can't seem to find anything that indicates if Bonjour can
scale to thousands or tens of thousands of peers. I was wondering if
any
one has any experience or knows of any research on the topic.

Thanks and regards,

Chaz.
_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers
_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers

_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers

_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers


_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers

Reply via email to