----- Original Message -----
From: "Meng Weng Wong" <mengwong(_at_)dumbo(_dot_)pobox(_dot_)com>
To: <spf-discuss(_at_)v2(_dot_)listbox(_dot_)com>
Sent: Tuesday, February 17, 2004 8:54 PM
Subject: Re: [SPF-Discuss] Performance issues
On Tue, Feb 17, 2004 at 08:35:44PM -0500, Hector Santos wrote:
|
| In any case, we are doing some final performance testing with a beta
team.
| I tried to provide some info in this area in the near future.
|
I would be interested in hearing about the DNS burden of SPF compared to
DMP.
thanks
meng
Meng, good news and bad news.
How do you want to see this? I will give you my preliminary analysis. If
you want to the data files, I will make them available for your own
analysis.
Legend
cip - client ip
cdn - client domain name (helo/ehlo)
crp - client return path
First, I took a January 31 smtp trace log and created a data file with cip,
cdn and crp for each session. On this date, there were 3395 connection
attempts. Most either dropped or were rejected at HELO for syntax or format
error (domain literals spoofs are rejected). In total, there were 341
sessions reaching the RCPT TO: state.
Our wcSAP suite was set to run the data using two orders (two separate runs)
FLT (Internal Filter white/black list)
DMP
SMP
RBL
FLT (Internal Filter white/black list)
SMP
DMP
RBL
Here is what I found (ignoring RBL and FLT analysis):
First, Order played no role. The same results. I was thinking the dns
cache would help one or the other.
Good News:
When the DOMAIN is valid in DNS, SPF is on average 3-4 times faster than
DMP with ranges in the milliseconds (i.e, 30 ms to 120 ms)
Bad News:
When the DOMAIN is invalid in DNS (NXDOMAIN), SPF is on average 4-6 times
SLOWER than DMP with ranges in seconds (i.e, 6-4 secs to 1- 0.5 secs).
I don't have the time at the moment to further analyze this result but it
was pretty apparent.
---
Hector Santos, CTO
WINSERVER "Wildcat! Interactive Net Server"
support: http://www.winserver.com
sales: http://www.santronics.com