tuning for maximum dhcp performance

[David W. Hankins]

On Mon, May 12, 2008 at 01:19:03PM +1000, Andy Hood wrote:
> > to be?  But if the coding to implement such options requires as much
> > relocating of code/function
> > as allocating for the hash table after the leases are counted, then
> > this idea would hardly
> > be worth the trouble.  Might as well just count leases.

yeah, at least my plan is to go straight for the automagic solution,
with config over-rides for tweaking.

> Calculating primes is pretty quick. At least for numbers up to 2^32.

could do.  i'm a bit wary of heavy sounding calculations (not so much
for CPU use as maintability).

> Did you intend using the same hash algorithm for IPv6 addresses? There are
> 82025 primes below 2^20 which lets my algorithm work up to 2^40.

for v6 we do things a little differently.  so far we haven't needed a
way to lookup an ipv6 address directly.  we always seek into that
lookup from an IAID/DUID pair, which approximately is the number of
clients you have, not the number of v6 addresses in dynamic ranges.

so we use our current (somewhat questionable) "opaque string of octets
hash algorithm" to do hash lookups by IAID/DUID concatenated.

-- 
David W. Hankins	"If you don't do it right the first time,
Software Engineer		you'll just have to do it again."
Internet Systems Consortium, Inc.	-- Jack T. Hankins