Load times for Large Subnets

Larry Apolonio isc-dhcp at rh73.com
Tue Oct 9 17:51:12 UTC 2007


Simon Hobson wrote:

> Larry Apolonio wrote:
>
>> I was wondering if there is a way I could speed up the startup time of
>> our DHCP server.
>>
>> I've compiled 3.1.0 on a 64 bit RHEL 5 box.  I then created a dhcpd.conf
>> file which had 144 subnets with 410,000 available IP addresses for 
>> lease.
>
>
> That's nearly 3,000 addresses per subnet, is that required ?
>
> I believe that reducing the number of available addresses can make a 
> significant impact on startup times - so if you only needed (say) 300 
> addresses, cutting each subnet down to whatever is needed plus a 
> reasonable margin should deal with the problem.

Yes, they are fairly huge subnets, some are /19, /20, and a few /21.  
All non rfc1918.

But I did find an answer to my question, in the source code there is a 
folder called includes/dhcpd.h

There is this section
#if !defined (LEASE_HASH_SIZE)
# define LEASE_HASH_SIZE        100003
#endif

Basically I up'd that number by an order of magnitude + some (I think it 
was 20000003) and my load times dropped from 20 minutes to 6 seconds.

For those wondering, I believe 20000003 is a prime number.

This section was in dhcp-3.1.0 and dhcp-4.0.0.a3 but it was not in the 
RHEL SRPMS dhcp-3.0.5.

There certainly was a huge cost in memory, I think it took nearly a GB 
of RAM, but dhcperf showed similar results in 4 way handshakes. 

The leases file was also in tmpfs, also switching to an AMD box did 
improve performance a little with its Hypertransport technology.

Does anyone know if there are any other problems that will arise when I 
increased this number?  What other tweaks are there that I can do to 
improve transaction performance.  Would be nice to be able to do 100,000 
four way handshakes per hour.

Thanks,
Larry Apolonio



More information about the dhcp-users mailing list