[comp.protocols.tcp-ip] milnet billing

stanonik@NPRDC.NAVY.MIL (Ron Stanonik) (06/27/89)

We've heard rumors that milnet sites might be charged for network
access (something like $100k per year for a psn connection, plus
packet charges).  Currently end-users (ie, milnet hosts) are not
being billed; apparently dca is picking up the tab, though we understand
that "informational bills" are being sent to overseeing organizations.

Well, this raises many questions.

1) Is any of it true?  Will overseeing organizations be receiving
real bills, and should end-users expect to pay those bills; ie,
need to budget for them?

2) What charging rates should be expected?  In particular, how will
packet charges be figured (ip packets or psn packets)?

3) If ip packets are the measure, will a gateway be billed for packets
routed through it, or will the orginating network be billed?  The
information all seems available for the latter; ie, the address is
in the packet, and the network is presumably registered with NIC.

4) Is there a better forum for these questions?

Thanks,

Ron Stanonik
stanonik@nprdc.navy.mil

barns@GATEWAY.MITRE.ORG (Bill Barns) (06/28/89)

[I've had to post this separately to the list because this mail
handler I use has some warped ideas about header processing.  A
copy DID go to Ron Stanonik, notwithstanding the header above.
I thought others might care to see it also.]

Hello Ron,

I'll bite, since you may not find other takers.  In fact, I'll even
break my usually unbreakable rule about NEVER reciting the questions
before responding.
 
> We've heard rumors that milnet sites might be charged for network
> access (something like $100k per year for a psn connection, plus
> packet charges).  Currently end-users (ie, milnet hosts) are not
> being billed; apparently dca is picking up the tab, though we understand
> that "informational bills" are being sent to overseeing organizations.

It's disappointing that the word doesn't get out better than it does.
Personally, I'm not surprised; in theory, this has all been a matter of
general knowledge throughout DoD for quite a while.  In case there are
others in the same boat, I'll try to cover a lot of territory with this
answer.  Of course, this message is unofficial.

It was made a matter of policy in 1986 by JCS memo MJCS-137-86 that
there would be "usage sensitive billing" on the DDN.  This is somewhat
a consequence of OMB Circular A-130 (December 1985), which generally
requires policies of this sort for most big information technology
facilities in the Government.  A-130 has extensive fine print defining
its scope, too much to include here.  JCS MOP 195, the main policy
directive governing DDN, has been adjusted to include billing-related
responsibilities.

DCA is not picking up the tab for MILNET hosts in general at present,
and never entirely did in the past either.  There is currently an
administratively determined cost split among DoD organizations.  For FY
1989, the Army, Navy, and Air Force are each paying $2,320,000 per
month.  Smaller agencies are also paying negotiated fees of various
amounts, loosely based on how many hosts they have on the network.  A
few users pay $6,300 per month per port.

DCA gets budget money for certain costs of managing DDN, and they get
budget money for their own use of DDN.  The usage part goes into the CSIF
(Communications Service Industrial Fund) where it is commingled with
the contributions from the rest of DoD, not only for DDN but for DSN,
ARPANET, and whatever the other 15 or so common user systems are.  There
is only one CSIF and its revenues balance its expenditures every year.
The individual systems don't have to balance.  The operational costs of
DDN (and DSN, etc.) are all paid out of the CSIF.  Certain management
and "other" costs come out of the DCA budget.  Certain overhead costs
such as the Monitoring Centers are paid out of the CSIF and thus are
distributed over the users.  The CSIF chunk of overall DDN cost is large
compared to the other part.  So, basically, the users are paying, but
at a high organizational level.

The "informational bills" are being sent to give managers a handle
on their usage levels.  I believe that those sent so far are all from
a sort of pre-prototype billing system.  The gears of the real phase-1
("Interim") billing system are just starting to grind now.  The ultimate
billing is embedded in an existing DECCO billing vehicle called the
Carrier Cost & Obligation (CC&O) report, which is where your billing
would normally show up if (for example) you leased a telco line.  Since
there is not much room in the CC&O for useful detail, there will be a
new report called the DDN Supplemental Billing Report.  The "bills"
sent thus far contained information that will be used in the supplemental
report, but the format and some other details differ.

> 1) Is any of it true?  Will overseeing organizations be receiving
> real bills, and should end-users expect to pay those bills; ie,
> need to budget for them?

The generic answer is "yes" but the details will depend on management
decisions in your chain of command.  The intent of the system is to
create an effective feedback loop which makes the consequences of
user behavior fall upon the users.  Just how far down the org chart the
budgeting will descend, and how soon, is somewhat at the organization's
option, but the JCS directives refer to the "lowest practical level."
 
> 2) What charging rates should be expected?  In particular, how will
> packet charges be figured (ip packets or psn packets)?
 
The charging is per 128 octet (or less) subnet packet sent.  If an AHIP
(1822) host sends a 512-octet (excluding leader) regular message to the
PSN, or if an X.25 (Basic or Standard) host sends an X.25 packet with
512 octets of data, it counts for 4 packets because it generates 4
subnet packets.  If the host sends 513 data octets, it counts for 5
packets.  Yes, this means that IP and TCP headers are billable data.

In the earlier planning stages we seemed to have a "tariff du jour" but
this seems to be settling down.  DCA issues a formal rate letter every
few months with projected forward pricing for all the common user
services, including DDN.  For hosts, cost components include access
line charge(s) if circuits must be leased between your host/gateway and
the PSN, monthly access charges which depend on access line speed and
single/dual homing, and kilopacket charges.  All CSIF charges have an
overhead factor of about 1% applied to cover some of the CSIF cost of
doing business.

There is a complication here in that all of the Services are not cutting
over to packet charge allocation simultaneously.  Therefore, there are
two rate schedules for MILNET, one with packet charges separate, and one
with no packet charges but a considerably higher monthly charge.  The
classified networks have their own price schedule, with no packet charges
at this time.  There's another complication, too: to try to ensure some
stability of real expenditure vs. projections as this scheme begins, a
number of special temporary rules have been worked out in various
organizations to confine the budgetary hits between various ceilings
and floors.

Here is a sampling of the planning rates for FY 1990.  The table of
access rates represents dollars per month per host.  Note that
special rates exist for dual homed hosts.  Below that are the
kilopacket rates, which are added onto the "MILNET (Plus Pkt Chg)"
column for those to whom it applies.  Terminal rates exist, too,
but I don't feel like typing in the entire chart.  I believe that
the Navy will be on packet charging during FY 1990 and the Army and
Air Force will be using the composite rate with no packet charges.

Access Single/     MILNET       CLASSIFIED   MILNET COMPOSITE
Rate   Dual    (Plus Pkt Chg)                  (No Pkt Chg)
------ ------  --------------   ----------   ----------------
50/56K Single       2,100          5,600           4,150
50/56K Dual         2,700          7,300           5,650
19.2K  Single       1,600          4,800           2,250
19.2K  Dual         2,000          6,250           3,050
9.6K   Single       1,000          4,000           1,350
9.6K   Dual         1,300          5,200           1,800

Traffic Precedence    Kilopacket Charge
------------------    -----------------
Level 1 (Routine):
   - Peak                   $1.30
   - Off-Peak               $1.00
Level 2                     $3.00
Level 3                     $4.00
Level 4 (Flash)             $5.00

> 3) If ip packets are the measure, will a gateway be billed for packets
> routed through it, or will the orginating network be billed?  The
> information all seems available for the latter; ie, the address is
> in the packet, and the network is presumably registered with NIC.

Loosely speaking, bills go to PSN ports.  If you have a gateway
connected to a PSN, the TSR that got you the port (or a subsequent
change) specifies the Program Descriptor Code (PDC) to be charged for
the packets that port sends, as well as for its monthly charge, and
generally for its access line costs if any.  There are also billing
mechanisms for TAC usage; details are a touch byzantine, but the
philosophy is analogous.  My report (referred to below) tells about
this.

The PSN doesn't look at IP headers.  Also, there is quite a bit of
non-IP traffic in DDN - somewhere between 5% and 25%, probably toward
the middle of that range, but since nothing looks, deriving these
numbers is a difficult proposition.  Then there is GOSIP...

There has been a serious discussion going on for almost a year about
upgrades for the second phase of the DDN billing capability.  MITRE (in
the person of me) did a study of technical issues and alternatives.
Peeking at IP headers was among the possibilities considered.  The
study report was published a couple of months ago for authorized
distribution.  It is pending public release authorization - "due" in a
couple of weeks, but it will get here when it gets here.  DoD people
can probably get copies now; quite a few were apparently distributed to
various "key players" in the services.  There is also a somewhat shorter
DCA document ("Analysis of Potential Future DDN Billing System Features")
covering some of the same territory; for maximum coverage, you probably
need both.  If you think it's appropriate for you to receive these
documents and you have difficulty obtaining them, drop me a note and I'll
try (as my time and patience permit) to get it taken care of through an
appropriate channel.

The sticking point with all enhancements is that the services (users)
will have to pay for developing and fielding enhancements, and I gather
that they are not sure that an elaborate centralized system is worth
the cost.  Military organizations with views on this should be working
with their service or agency's TCO or other appropriate focal point to
get their positions considered.

> 4) Is there a better forum for these questions?

Among net mailing lists, probably not (sigh).  Most of the key players
in this drama don't inhabit mailing lists.  You can talk to me
informally about technical issues, for what that's worth.  DoD
Services/Agencies all have TCO's and/or other kinds of focal points for
DDN, and that is the channel to input any official opinions.  There
have been a number of miscellaneous documents, briefing slides, etc.,
distributed to those people at various meetings on the subject, and
some of that information might be useful to you.

Bill Barns / MITRE-Washington / barns@gateway.mitre.org / (703) 883-6832