On Saturday, August 23, 2014 8:53:55 PM UTC+5:30, Ben Noordhuis wrote:
>
> On Sat, Aug 23, 2014 at 10:09 AM, Ashish <aashee...@gmail.com 
> <javascript:>> wrote: 
> > 
> > This may not be directly related to LIBUV but like to know if at all 
> LIBUV 
> > can help here. 
> > 
> > I read on their blog that whatsapp guys have achieved 2 millions of 
> > concurrent connections to single server (running on FreeBSD): 
> > http://blog.whatsapp.com/170/ONE-MILLION! 
> > http://blog.whatsapp.com/196/1-million-is-so-2011 
> > 
> > Makes me think FreeBSD kernel supports those many connections (or can be 
> > tweaked). 
> > 
> > Is this possible on Windows server using LIBUV ? 
>
> Possibly.  A lot depends on the type of traffic and the configuration 
> of the server. 
>
> I speculate that whatsapp clients are idle most of the time and that 
> perhaps 1% is active at any one time.  So, with 2 million clients, 
> that means you have 2e6 * 0.01 = 20,000 clients that actually require 
> the program or the operating system do to anything in a given unit of 
> time.  The numbers are made out of whole cloth, of course, but the 
> point is that their clients are zero cost most of the time. 
>
> My Core i7 can execute 10 billion instructions per second per core 
> without breaking a sweat.  Let's assume half of that is wasted on 
> overhead (scheduling, bookkeeping, hardware interrupts, etc.)  That 
> still leaves you with 1e10 / 2 / 2e4 = 250,000 instructions/sec per 
> client - and that's when running on a single i7 core.  I will bet good 
> money that whatsapp a) employs beefier hardware, and b) fully exploits 
> every core in their systems. 
>



Well, but from operating system perspective, is Windows kernel capable to 
accommodate and handle those many connections
(Considering factors like resource allocations in kernel per connection)? 
Or do we need to tweak something in there? 

One opinion is that default WinSock cannot handle huge number of 
connections and we need install a custom socket stack that supports that 
many connections, but I am not sure about it.


-- 
You received this message because you are subscribed to the Google Groups 
"libuv" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to libuv+unsubscr...@googlegroups.com.
To post to this group, send email to libuv@googlegroups.com.
Visit this group at http://groups.google.com/group/libuv.
For more options, visit https://groups.google.com/d/optout.

Reply via email to