Re: "The Kernel is the Problem, Not the Solution"

[Amir Chaudhry <amc79@xxxxxxxxx>]

On 14 May 2013, at 09:07, Anil Madhavapeddy <anil@xxxxxxxxxx> wrote:

> Yeah, it's all the other stuff -- managing the iperf instances, making sure 
> they're not all hogging CPU, splitting it across aliases, interfaces, bridges 
> and VMs, etc.
> 
> All a bit tedious. Isn't there some magic load generator that already helps 
> out with this sort of thing?

I think that would be the same magical load generator that Euan was asking for 
a couple of weeks ago :)

ac

> 
> On 14 May 2013, at 00:00, Stephen Dolan <stephen.dolan@xxxxxxxxxxxx> wrote:
> 
>> You've probably seen this before, but it's really easy to multihome
>> interfaces on linux:
>> 
>>    for i in $(seq 1 254); do ifconfig eth0:$i 192.168.42.$i; done
>> 
>> gives you 2^24 distinct (ip, port) pairs on one box.
>> 
>> Stephen
>> 
>> On Mon, May 13, 2013 at 10:50 PM, Anil Madhavapeddy <anil@xxxxxxxxxx> wrote:
>>> On 13 May 2013, at 22:36, Dave Scott <Dave.Scott@xxxxxxxxxxxxx> wrote:
>>> 
>>>> 
>>>> http://highscalability.com/blog/2013/5/13/the-secret-to-10-million-concurrent-connections-the-kernel-i.html
>>>> 
>>>> Some quite interesting stuff in there.
>>> 
>>> I definitely want to try out our existing TCP stack with a few million 
>>> connections and see how we fare.  Perhaps we should give this a shot at the 
>>> Xen hackathon on Thursday?  The hard bit is finding enough interfaces on 
>>> the Linux test clients so they don't run out of ports :-P
>>> 
>>> -anil
>> 
> 
>