Re: [PATCH] Mass rDNS performance

[jah <jah () zadkiel plus com>]

Thanks for testing Brandon!

My initial thought is that your djbdns box can handle a greater number
of outstanding requests than the code allows it to reach.  With a
hostgroup of 4096 the patch would allow a maximum capacity of about 91
and it would need 4096 requests to get 4096 responses to reach this
level (i.e. no timed-out requests and all responses either OK or NX). 
The mean capacity would be about 60 in this case and total time would be
around 3 times longer than if the mean capacity were nearly CAPACITY_MAX
- which is roughly what you're seeing.

To test this, you might specify --min-hostgroup 16384 which should allow
capacity to reach around 180 (again, with no timed-out requests and all
responses either OK or NX) and a mean of about 120 - so resolution time
for the patched version should be about half (and about 50% slower than
current nmap).

I'll look at the capacity increase formula again - I'm thinking that
capacity += 8/floor(capacity) should increase fast enough to suit your
box, but this would likely be way too fast for many people so it looks
like I might need to revisit the idea of "slow-start".

If you still have them, I'd be grateful if you'd post the final stats
output from each hostgroup in each of the of the two versions of nmap
(grep "DNS resolution of 4096 IPs took").

Again, thanks for the testing.

jah

_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://SecLists.Org