I do not have any hard numbers, but I believe you are going to be
more-so bottlenecked in the disk I/O area.
If you do not find any answers here, I believe there are a few users in
#dspam that might be able to help (scooby2, I believe).
Happy Easter,
Kyle
Brandon Macmillan wrote:
Hello List!
First off I want to introduce myself. I work as a system admin at a
large virtual ISP provider. We are currently looking at replacing our in
house spam filtering system (a collection of commercial and free
blacklists, whitelists, SPF record checks, reverse DNS entries check,
etc) with a proper spam solution. We have evaluated commercial spam
filtering appliances such as Iron Port spam gateways but have found that
initial purchase costs and yearly maintenance costs are astronomical for
a deployment of our scale. As always I would like to turn to the open
source community in hopes of finding an effective solution. We already
use many open source projects on our systems, such as Apache, MySQL and
Linux, so I hope that my proposal to use open source for spam filtering
will be successful. I have previously experimented with spam assassin
but found that the system requirements do not scale well for larger
deployments. I learned about DSPAM from a friend who is running it as a
spam filter on his home machine.
I would like to setup DSPAM as a mail relay solution due to the fact
that our actual MTA is closely tied to our provisioning system and can
not be easily adapted to work with another delivery agent. I would
basically like to re-implement the Iron Port systems that we demoed
using DSPAM. That is, point the MX records at the DSPAM boxes, have them
filter the SPAM and then relay the (HAM) email to our real MTA.
I would like to build a test case, 1 box filtering spam for one of our
smaller client domains for testing, training and demonstration to my
manager and if all goes well we could very well scale up to 500,000+
users right away.
What I need to know beforehand is what sort of system requirements I
should expect for large deployments such as this. I see from the DSPAM
home page that someone has done a 300,000 user install before, does any
one know roughly how many boxes they use? How would the spam database be
distributed over multiple spam relays? I assume that for a large
deployment we would setup one machine to handle the DSPAM database and 2
or 3 or 4 machines to handle the load of processing mail. I have the
option of purchasing hardware to build out this spam processing system
so if there is anything in particular I should be aware of such as fast
SCSI disks, lots of RAM, etc. please let me know.
I realize that there is no simple answer to my questions but a response
such as "4 dual CPU Xeons with a gig of RAM each should do it" would be
a great help at this stage in my research.
Thank you in advance,
Brandon Macmillan
IT Systems Administrator
IP Applications
Office: (604) 630-5675 Cell: (604) 834-2912
Email: [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>
Web: www.ipapplications.com <http://www.ipapplications.com/>
This electronic mail transmission contains confidential information
intended only for the named person(s). Any use, distribution, copying or
disclosure by any other person is strictly prohibited. If you receive
this transmission in error, please notify the sender immediately so that
we may correct our internal records. Please then delete the original
message. Thank you.
