Hi.
At tuenti, we have found a performance degradation on UDP that can be
solved using a range of ports instead of a single one.
More info on the problem and the solution can be found here:
http://blog.tuenti.com/dev/using-udp-in-memcached/
And the patch with the implementation:
https://g
On 28. feb. 2011, at 15.32, Jaime Medrano Navarro wrote:
> Hi.
>
> At tuenti, we have found a performance degradation on UDP that can be
> solved using a range of ports instead of a single one.
>
> More info on the problem and the solution can be found here:
> http://blog.tuenti.com/dev/using
On Feb 28, 2011 5:04 PM, "Trond Norbye" wrote:
> I don't have any huge objections to this patch, but I'd rather change the
logic so instead of specifying the number of ports and have them in a range
I'd prefer if we used multiple -h hostname:port and -U hostname:port. That
would allow you to bind
On 1. mars 2011, at 00.00, Adam Lee wrote:
> On Feb 28, 2011 5:04 PM, "Trond Norbye" wrote:
> > I don't have any huge objections to this patch, but I'd rather change the
> > logic so instead of specifying the number of ports and have them in a range
> > I'd prefer if we used multiple -h hostna
On 1 March 2011 10:04, Trond Norbye wrote:
> I forgot that -h was used to get the help text and not the host name ;-)
> Anyway. I pushed a path that gives you the functionality through the -l
> parameter:
> -l interface to listen on (default: INADDR_ANY, all addresses)
> may be
On 1. mars 2011, at 10.42, Jaime Medrano Navarro wrote:
>
> That syntax is ok for me, but you're missing the most important part
> of my patch: change the mapping between udp sockets and threads.
>
Sorry, I forgot that part ;) I'll add it.
Cheers,
Trond
> If only one port is provided, we s
As Jaime pointed out, the main purpose of the patch is not make more
flexible how memcache listen on ports, but when using udp be able to
use 32 consecutive ports with one thread per port listening to get
huge improvements in performance, less system and contention.
For this you usually need all p
On 1. mars 2011, at 15.30, Guille -bisho- wrote:
> As Jaime pointed out, the main purpose of the patch is not make more
> flexible how memcache listen on ports, but when using udp be able to
> use 32 consecutive ports with one thread per port listening to get
> huge improvements in performance, l