Per message costs of email (was: Re: anybody else been spammed by "no-ip.com" yet?)

Jared Mauch jared at puck.Nether.net
Sat May 4 22:34:13 UTC 2002


On Sat, May 04, 2002 at 04:36:40PM -0400, Scott A Crosby wrote:
> 
> *blink*
> 
> So far, other than Jared Mauch <jared at puck.Nether.net>'s calculation where
> he neither confirmed nor disputed $.02/email, I've yet to see *one*
> quantified per-message price bandied about..
> 
> Are you also unsure of the per-message costs of email? I'd thought I'd
> find *someone* who could quantify a cost.... I certainly don't know and
> I want enlightenment too!
> 
> I'm surprised you can't quantify the per-message costs either.

	Well, my system (puck.nether.net) only gets upgraded
about once every 1.5-2 years.  It's not a pure e-mail server,
but let me give you an idea of my costs based on current replacement
cost:

	Operating System: $0
	machine:
		case $65
		disks $150 (aproximite cost today)
			(1x4g ide, 1x45g ide, 1x40g ide[7200rpm])
		cpu $100
		moterboard $100
		memory $250
		video, ethernet, etc.. $150
	--- cost: $815

	Now, i'll take a stab in the dark and say that I could
(if i wanted to) support 5000 users (mail) on this system
with no problems.  I've processed 6153 e-mail messages today
with an average size of 5k for my current 186 users.  (this
includes sending out mail for lists i host as well as delivery
to mailboxes of nanog and lists they are on).
	
	The system costs me (over 2 years) ~$1.116/day
	To instantly deliver one message would take 40Kb/s at an
estimated 8204 messages/day that doesn't count that much for overall
cost, but it is there.  That's .09 msgs/sec average.  Be sure to
scale this for your local site of course.

	So, lets scale this up:

	(assume Using 40g of disk for mail)

	So multiply by ~26.88 to take my 186 users up to 5k.

	That takes me to 2.55 messages/sec and 104Kb/s
constant bit rate.  104Kb/s @ $100/Mb= $10.4 in bandwidth costs.
(or ~$0.0000004/msg)

	If on average, my users fetch their mail once every two days
and fetch it all at that point, I'm looking at needing to store
~440k messages on disk, and on backup media also.  (I'll leave
that cost out right now).

	Storage cost/msg = ~$0.000005 (system cost per day / 864000)/2.55

	Taking these two into account, it has a low cost/day/msg.

	I don't have any support staff.  Add in a few (3) people @ 
65k/yr to answer phones for my 5k users that increases costs by $534/day
or $0.00242135/msg

	You can see where the real costs come in.  By needing someone
to go in and clear out their mailboxes when there are problems, it
increases costs drastically.  There are system costs, mine is very
cheap.  If I were to price it more realistically with some scsi
disk, controller, etc.. the cost/msg goes up based on that.

	There are some obvious user-number bars that are
created when you require more than one system to handle mail.  Assume
1 mail server per 10-15k users (depending on mail load) at
$1500/system (about double my costs, but serving 3x users) and you do
easily see where these costs/msg of the spammer start to hurt you.


> On 4 May 2002, Paul Vixie wrote:
> >
> > > I do find it amusing that nobody responded to my more relevant and
> > > intended thrust, about how putting a 'sender pays receiver for email'
> > > could cause a variety of new abuses of the email system.
> >
> > on the one hand, you're right that any micropayment system would have
> > to be very carefully thought out and even more carefully implemented,
> > lest it open the door to many and varied forms of microabuse.
> 
> Leading to more subtle abuses...... or benefits?
> 
> Will we have arbitrage trading by sending email through an intermediary
> for a lower fee.
> 
> Say, A charges $.03 email from B..
>      A charges $.01 email from C..
> 
> C charges B $.025 per email and forwards it A. C spends $.005 in overhead
> and keeps the $.010 difference as profit.

	The billing PIC in your new backbone router?  It'll tell the
difference between streaming video udp packets and voice to make
sure the various people get charged the correct long-distance rate
too?

> > on the other hand, that doesn't disprove the case, since even in your
> > example it would merely cause people to become a LOT more careful about
> > they mail they sent.  that CAN'T be a bad thing.
> 
> Apart from the balkanization of the lifeblood of the internet's
> communication systes, there's things like viruses, worms, zombie
> computers, etc.
> 
> There's also the transactional cost.. If a computer gets infected by a
> worm, who pays for the email it sends out? Who pays for the argument of
> who's responsible for the costs? Who's responsible for the tech support?
> 
> What if the user can't pay... Will ISP's have to insure themselves against
> email worms? Will people with insecure email clients be subsidized by
> those with more secure clients? Is that theft too?

	Communisim :)

> Its a cost of doing business. Its like restrooms in restraunts, a
> necessary evil. You can try to minimize the costs and stop abuse.  But the
> only way to avoid it is to leave the business entirely. Will you, like
> Donald Knuth, be giving up your email address?
> 
> I'm still hoping somone out there can quantify those per-message costs of
> email and spam.
> 
> Scott

-- 
Jared Mauch  | pgp key available via finger from jared at puck.nether.net
clue++;      | http://puck.nether.net/~jared/  My statements are only mine.



More information about the NANOG mailing list