Performance?

[Bret Schuhmacher]

Good, points, Brian.  I neglected to add the hardware factor.  My 
question would have been better asked as, "are there any hardware 
benchmarks?" and I could extrapolate what I need based on those results.

I guess the first step would be to create a  load test program that you 
can configure to do N DHCPD requests (is this what dhclient does?).  I 
was planning on pulling the plug on the CMTS and causing all 8500 modems 
to reregister, but a load test platform would be nice so I don't affect 
8500 users ;-).

Bret



Brian J. Murrell wrote:
> On Tue, 2006-05-30 at 08:46 -0400, Bret Schuhmacher wrote:
>   
>> Hi,
>>
>> I'm new here, but I couldn't find this in the archives.
>>
>> How many registrations can DHCPD handle per second?
>>     
>
> Isn't this _always_ a product of what hardware (including network of
> course) it's running on?
>
>   
>> I have a client 
>> that wants me to add Postgres support to DHCPD and then guarantee it'll 
>> handle 8,500 registrations in 10 minutes (14.16666/second).  I need to 
>> know the baseline performance before I go guaranteeing anything.
>>     
>
> Until you are at the biggest and best of your slowest component, you can
> always get more performance by throwing bigger and better hardware at
> it, no?
>
> Maybe I am missing the point of the question.
>
> Maybe a good question here would be, are there any "multiple server"
> scalability options in the ISC DHCP server?  Certainly, I can divide up
> my address space among n DHCP servers, but I am asking about something a
> bit more intelligent than that, like the ability to give all of the DHCP
> servers all of the address space and have them share the load
> themselves.  Having to redivide my address space when I need to add a
> server would be painful to say the least.
>
> Can failover handle this (currently or in the future) or is failover
> (always going to be) strictly active/passive?
>
> b.
>
>
>
>
>
>