[p2p-hackers] Private

Henry Sinnreich hsinnrei at adobe.com
Wed Aug 13 18:08:01 EDT 2008


Dave,

Nice to see you online!

> As for scalability of the server, it's really not that bad.  A single,
> totally-standard $100/mo dedicated server can handle over 100K TCP
> connections.  The primary limit is CPU and RAM, and they get better all
> the time.

The issue is not the number of TCP connections but the bandwidth for 100k
media streams. Let's see:

At 100 kb/s per stream that is 100^3 times 100^3 = 100^6 = 100,000,000 kb/s
= 100 Gb/s for the servers.

Henry

On 8/13/08 3:50 PM, "David Barrett" <dbarrett at quinthar.com> wrote:

> iGlance has its NAT-penetration code based on UDP hole bunching open
> source here:
> 
> http://www.iglance.com/code.html
> 
> As for scalability of the server, it's really not that bad.  A single,
> totally-standard $100/mo dedicated server can handle over 100K TCP
> connections.  The primary limit is CPU and RAM, and they get better all
> the time.
> 
> Thus while you could (as the P2P-SIP folks are doing) completely
> decentralize the connection forwarding through a massive engineering
> effort, keep in mind that a dead-easy centralized service costs less
> than $0.001/mo/user to provide.
> 
> You might want to ask yourself if you're decentralizing it because it's
> really useful, or just because it sounds like fun.  If it's the latter,
> it might be more rewarding to find a problem that is both useful to
> solve *and* fun.
> 
> -david
> 
> shidong chen wrote:
>> 
>>   Hello,
>> 
>> 
>>   When I try to implement udp or tcp hole punching in real world (e.g.
>>   bittorrent), an scalability issue pops up.
>> 
>> UDP/TCP hole punching
>> 
>> 
>>   Hole punching requires peer A and B to sent packet to each other
>>   almost "at same time" to build the connection. It also requires server
>>   S keep a UDP/TCP connection with "every" peer to forward connection
>>   request. It could be very costly to build such a server S to serve
>>   millions of peers. Any solution?
>> 
>> It looks to me the connection request forwarding could be standardized
>> into a P2P control protocol (besides the data transfer protocols). could
>> anyone tell me if there is an effort to standardize this (a draft
>> version is great too)?
>> 
>> Greatly appreciated If you could point me to any document, source code,
>> library or examples on implementation of hole punching.
>> 
>> Thanks a lot,
>> 
>> Shidong
>> 
>> 
>> ------------------------------------------------------------------------
>> 
>> _______________________________________________
>> p2p-hackers mailing list
>> p2p-hackers at lists.zooko.com
>> http://lists.zooko.com/mailman/listinfo/p2p-hackers
> 
> _______________________________________________
> p2p-hackers mailing list
> p2p-hackers at lists.zooko.com
> http://lists.zooko.com/mailman/listinfo/p2p-hackers



More information about the p2p-hackers mailing list