Conference 7.286::digital

Would you please explain P.x?

Thanks,

Bob

    re .0
    
    i gave up trying to stop dec from destroying dec sometime ago. it's
    easier that way.
    
    seriously, we do some pretty dumb things with our networks that make
    bandwidth requirements almost mandatory. 
    
    this simply smacks at what appears to be corporate methodology of cost
    savings that says "take something away and see how bad it hurts". i
    know of NO customers of mine that would even consider that kind of
    approach. seems we're sort of admitting we haven't a clue about what's
    needed to run the business so let's start eliminating things and see if
    the bitchin' merits restoring it.
    
    i for one am sick of that kind of approach.

    
       Asinine.
    
       Aren't we supposed to be a networking company, telling other
    companies about how good networking is and our products are? 
    Presumably the reason that they're so good is that they help the
    customer increase profits/"success" in some way.
    
       Digital - we'll tell you how to organize your enterprise, but we
    don't follow the advice we sell.
    
    					-mjg

Sooo...who's the author/sponsor of this pilot plan?

								- paul

    Many of the comments in 2749.* apply here, imho.
    
    I would like to suggest that it is virtually impossible for expense-
    reduction "pilot projects" of this sort to be planned, much less exist,
    without their becoming general knowledge.  Therefore, it would seem
    prudent for managers who request such projects to take this into
    account, and to communicate openly about the reason(s) for the proposed
    action.  Beforehand.  Especially if the action might be construed to be
    counterintuitive or counterproductive.
    
    See .0 for the latest example of this not, apparently, being done.  
    

    Sorry folks but my immediate reaction is 'why not?'.If we have a higher
    level of service than is really needed for the job then it is
    justifiable to reduce it,especially if it saves money.An analogy for
    those who don't understand is ordering a seven seat limo when a taxi
    would do.Another point could be that given that we have a higher level
    of service than any PTT can offer in Europe(at a reasonable price),it
    seems a little brash to complain about reducing it-maybe we are a
    little spoilt?
    I am not convinced that our service levels are in tune with needs.We
    have amazing phone services and some slow lines for data-which seems
    odd.So I think it does need review.I am less convinced about our
    ability to do a tidy pilot but this is probably due to the complexity
    of big companies.
    SO,encourage them to do it,support it and debate the results.
    Don't prejudge the result-p20,p30 etc-I doubt that in reality many
    people will notice.
    
    IMHO of course

Hmmm, that's interesting. I have just been reviewing a couple of tools over the
last months that can help network planners/designers in such a process. One of
the tools that I reviewed allows you to put in your existing network and traffic
requirements, and shows you which lines are underutilized. you can than analyze 
alternatives with lower speed lines or other (cheaper) cariers. Actually, the
vendor of the tool told me last week, that they had done a similar exercise at
the BANQUE DE FRANCE. They had reorganized, layed off etc. and suspected that they
could do with less bandwidth. Instead of just lower the speed on some lines,
and see what happens, they signed up for a traffic analysis service,
put the data in the tool, saw that some lines were only used for 20% of their
capacity, redesigned it, and cut 25% of their telecom costs, without degrading
response times. Now if we are such a great networking company, why cannot we 
do something similar. I agree that we maybe can save some costs on our telecom
bill, but it has to be done with some thought. 

Oh yes, the tool costs ~80k$, but they recovered that largely during the first 
year.


Timo

    re:.7
    
    We have in house the expertise and tools to optimize our networks. 
    That is not the point.  The point of the proposal in .0 is to pessimize
    them:  Make them run badly, so that people begin to think that networks
    are a costly, rare resource, and depend upon them less.  The goal is to
    reduce total telecom spending by 20%, regardless of the impact on the
    user businesses, corporate bottom line, etc.  (This comes from
    Finance.)  This is a plan to micromanage cost and reduce demand by
    making the service less convenient to use.
    
    This proposal and the one in 2749 are related.  See 2449.23 for the
    source within Telecom.
    
    Definitions:  P-grade:  Given a certain amount of traffic carried on
    the network, and a Poissonian demand, this specifies what fraction of
    calls are blocked due to congestion.  Thus P.02 means 2% of calls are
    blocked, while P.20 means 20% are blocked.  
    
    For many  years, traditional Bell System practice was to engineer P.04
    or better for long distance.  Nowadays AT&T engineers for P.001 or
    better on its backbone, but by using more clever routing algorithms,
    they probably do it cheaper. (Of course Mothers' Day is a special
    case.)
    
    The proposal in .0 is NOT aimed at saving money via fewer lines.  Take
    HLO, for instance.  It is tied into the rest of the company via the
    Digital Lightwave Network, which provides three T1 circuits (24 voice
    channels apiece).  Based on traffic engineering principles, 72 channels
    can carry 55 Erlangs (average simultaneous calls) of traffic with a 2%
    blocking probability.  Those same 55 Erlangs can be carried with a 20%
    blocking probability using around 60 channels.  That's still three
    T1's, so nothing goes away -- channels are gratuitously turned off just
    to effect fast busy signals!  And since we don't pay by the T1 anyway,
    the savings from going from 3 to 2 T1's would be $0.00.  Any savings
    that result would come from demand reduction, which is simply assumed.

Hey, has anyone thought about telling Scott Adams (author of the
"Dilbert" comic strip) about some of these hare-brained ideas? He could
probably work them into some strips, and it would be a scream. In fact,
this whole notesfile is a gold mine of material for Dilbert strips.
Maybe we could even convince him to part with a small royalty :-). I
think his email address is [email protected].

	RE: .9 NO! Let's keep our dirty laundry in house.

			Alfred

    If you were running the Pony Express, would it make sense to have your
    "animal feed manager" decide to cut the rations of the horses by 20%?
    
    His little part of the world would show a cost improvement.  However,
    the horses would be weaker and run slower (degrading customer
    satisfaction with the service), and they would have to be replaced more
    often.
    
    This approach is called "sub-optimization" and it is, to say it mildly,
    not a recommended management technique. 

> The point of the proposal in .0 is to pessimize
>    them:  Make them run badly, so that people begin to think that networks
>    are a costly, rare resource, and depend upon them less.

 When are the people responsible for these egregiously stupid ideas going
to be held responsible for the detrimental effects they have on the way we 
do our business? In a properly run company, the people responsible for
these "ideas" would be terminated. Is anybody else tired of riding this 
rudderless boat, where fools make decisions negatively impacting thousands
yet they not only retain their positions, but they are rewarded for it?
As a worker, I am utterly disgusted. As a stockholder, I am outraged.
Digital stands only to lose by this proposition; our networks are already
woefully inadequate (at least for EFFICIENCY and TIMELINESS.) We should be
upgrading the networks, not downgrading them. And we should be sending
the people who are degrading our ability to compete with other companies
packing!

 The whole morale issue's guiding principle seems to be "how low can you go?"

 The answer is "only so low," and many of our company's most talented people
are voting with their feet.

 Since when do beancounters cause a company to make a profit? Maybe we should
make them deliver product.

 

    Where to begin....
    
    There has been a paradigm shift :^) in the provisioning of telecoms
    services over the past few months.  Traditionally, the job of the
    Telecoms function has been to support the needs of the business as
    expressed by the demands placed on the networks by the individual
    users.  Telecoms' biggest job was to reduce the unit cost of services
    by aggregating traffic loads and making use of cheaper (per unit)
    services offered by vendors.  This was done through a large number of
    clever technology tricks, and was not just demanding lower prices in
    the marketplace (although that, too, had its place).  It was Telecoms'
    job to handle every call or data session in the cheapest manner
    possible (consistent with good service), NOT to be a network policeman
    and tell people which calls they could make or which data they could
    transfer or which mail they could send.
    
    Recently, however, the Finance function has tried very hard to place
    Telecoms in the net police category.  The order now is not "make the
    call cheaper"; it has become "stop the calling".  Because that is the
    ONLY way that we can ultimately reduce telecoms expense by the levels
    that Finance thinks are reasonable.  It is worth ones job to argue the
    contrary opinion; that has been done repeatedly.  Everything said so
    far in this string has been stated clearly to the highest levels to no
    avail.  Well, no one has actually said "egregiously stupid", but the
    points hav been made.
    
    None of this is *reasonable*.  Nobody in Telecoms has any authority or
    ability to keep an individual user from making any call or any file
    transfer or any mailing that that user's management has determined to
    be in the interest of the business.  I've said this here before, the
    great bulk of telecoms expense is the result of individual users'
    actions, over which telecoms can exert little or no influence.  I am
    personally comfortable with this.
    
    All efforts, such as that outlined in .0, are aimed at (hopefully)
    changing (influencing) user behavior by decreasing the availability of
    the resource, so that users will not use so much of it.  I happen to
    think that this is tilting at windmills, but it is Telecoms only hope
    to impact behavior.  And remember that a fairly large number of
    telecoms people are being held responsible to reduce the telecoms
    spend, even though they do not control it!  This violates every
    management principle that I'm aware of, but this *IS* Digital....
    
    Unfortunately, the businesses who must use the telecoms infrastructure
    to run their business have not stepped in to intervene in the on-going
    "discussion" between Telecoms and Finance.  They just go ahead ordering
    more and more capacity, as if nothing had ever changed.  I have no idea
    of whether this is "good" or "bad", just that it is occurring.
    
    So, put yourself in Telecoms' shoes, Finance is saying "cut, cut, cut",
    and the businesses are saying "more, more, more" through increased
    demands, and Telecoms is squarely in the middle.  Except Finance is
    speaking from the SLT level and the businesses are (so far) only
    speaking from the end user level.  What would YOU do with your job on
    the line?
    
    Oh, by the way, over the past year Telecoms has cut close to 40% (or
    has plans to do so in the near future) of the costs they do control. 
    Unfortunately, they only control (directly) only about 20% of the
    spend, the rest is controlled by the end user.
    
    Being between a rock and a hard place is not comfortable.

>    If you were running the Pony Express, would it make sense to have your
>    "animal feed manager" decide to cut the rations of the horses by 20%?
>    
>    His little part of the world would show a cost improvement.  However,
>    the horses would be weaker and run slower (degrading customer
>    satisfaction with the service), and they would have to be replaced 
>    more often.

But how much food are the horses getting?  Just enough to sustain them in 
their daily duties or are they only eating half or 3/4ths of what you give 
them?  You didn't ask that question.

If you had 13 horses, but lost 4 for whatever reason, would you keep 
feeding the remaining 9 horses the same amount you gave the 13?

Dave

    Sounds like people will have to suck the Information Aquifer through a
    thinner straw ...	;')
    
    

    
    Well if Telecom is responsible for mail messages sent, I would
    recommend we teach people to use VTX for organizational announcements,
    etc.  What you read in VTX today, comes in electronic mail form
    a week or so later after each layer of management has forwarded it
    on.  That is a place where efficiency could be achieved.
    
    My 2 cents,
    Linda

    Name any other company that can afford 5 T1/E1 over Atlantic ?

    Our net is like any other part of our beloved DEC , no controls and
    no feedback from reality.

    I've seen a recent memo saying that DEC's IS costs more and produces
    less when compared with industry. Eaznet is a part of it.

    The logic behind Eaznet blown out of proportion is that it "increases
    the value of the company." Would 5 telephones on each desk have the
    same effect ? 

    
    re .14
    
>If you had 13 horses, but lost 4 for whatever reason, would you keep 
>feeding the remaining 9 horses the same amount you gave the 13?

    
    	If you worked for Digital, you'd feed them less even though
    	you'd be working them that much harder. Then when production
    	would suffer, as it ultimately would, you'd reduce your costs
    	by shooting 4 riders. Then you'd pat yourself on the back because
    	you showed a paper profit. This would entitle you to adjust....
    	you get the picture.
    
    	John

	re: Wlodek

>Name any other company that can afford 5 T1/E1 over Atlantic ?
>
>    Our net is like any other part of our beloved DEC , no controls and
>    no feedback from reality.

	Do you have real information on how other computer wendors
	have created their international net, or are you just speculating?


	I fail to believe that any company in our industry has implemented
	a network where the costs are controlled and distributed to the
	users (i.e. you have to pay for each byte you transfer across
	the pond etc...)


		pekka

    >If we have a higher
    >level of service than is really needed for the job then it is
    >justifiable to reduce it,especially if it saves money.
    
    The problem is that we don't have a higher level of data service than
    we actually need.
    
    I should be able to log into other sites without a several second 
    delay on my keystrokes.  I should be able to put an X window up on
    someone's screen at another site and give a demo without things hanging
    for 30 second or more.  I should be able to do shared development with
    someone at another site without having to worry about packing things
    up and moving them at night when there's adequate bandwidth.
    
    Furthermore, there are applications that only become possible with 
    increased network bandwidth.  The World Wide Web (distributed hypertext
    system) is one example that's barely possible with today's network.
    We've got teleconferencing (audio, video, and shared whiteboard)
    software running in Corporate Research that's unusable anywhere else in
    the company because the network infastructure isn't there.
    
    The situation is even more frustrating if you are technically familiar
    with networking practice and Digital's network implementation.
    
    Today, telling a computer company that you are cutting network
    infastructure is like telling a telemarketing company you are going to cut
    out some phone lines.
    
    -andy

    There are two problems here that both need to be addressed.
    
    One of them, as reflected in .0, is the technical weakness of the
    People Making Decisions.  (Actually, that's an in-joke.  In the Telecom
    function, nobody ever makes a decision.  So maybe it gets left to
    Finance by default.)  Telecom is tilting at windmills doing stupid
    things because they haven't convinced management of the fundamental
    nature of the business, which Bill explained a few replies ago.
    
    The second is that the users really DON'T know what it costs.  Now to
    be sure, we don't tell them.  And we don't "know" either.  Since I am
    working on the problem this week, I do know that we do not have valid
    cost-based chargebacks for our voice and data networks.  We have
    arbitrary (not cost-based) chargebacks per call for voice, so at least
    a cost center manager knows something is being spent, even if it is
    viewed as "funny money".  (It's not that funny, since it nets to zero.
    It's only funny because some calls are underpriced, some overpriced.)
    But for DECnet and IP traffic, there is no billing of any sort.  It's
    all perceived as "free".  And it ain't.
    
    Large-scale long-distance file transfer is very costly.  Copying
    big netkits across the puddle costs more than FedExing tapes.  It would
    be cheaper to copy once to a site and have users make copies across a
    LAN (free), but we chargeback $0.00 for the COPY and it takes a capital
    req. to buy a small file server.  This is just plain stupid.
    
    Video across wide-area data nets is very costly.  A single WAN (NOT
    across metropolitan fiber, which is moderatly cheap)  DECspin session
    costs many times as much per minute as a phone call, and much more than
    a PictureTel video conference (112 kbps).  But to the end user it's
    "free".  An X Window wide-area session can cost more per minute than a
    long-distance phone call or three.  But it's "free".
    
    The economics of telecommunications are very tricky.  Crass service
    cuts don't save money.  In fact I can mathematically prove (with a
    4-function calculator, the fanciest math I generally do) that it's
    cheaper to provide "P.01" service than "P.20" service.  This is utterly
    counter-intuitive and frankly nobody now in management understands or
    believes it, but it's valid using the latest tariffs and voice network
    technology.  But it can't produce instant paper "savings" for Finance.
    Until all of the SLT understands that Telecom isn't a simple "plug it
    in and there's dial tone, what's so tricky?" business, we'll keep
    fighting these stupid battles.  And the stockholders will suffer.

    Although I think Fred hit the fundemental problems right on the nail,
    the really galling thing in all this is the approach.
    
    I'm not a telecom expert but I *know* that we do an amazing poor job of
    designing networks in some locations.  Perhaps as a whole we do better,
    but in certain parts of the US Field, we'd get much better payback by
    designing networks properly than trying to control demand.
    
    As an example, in an area (region?) I used to work, we had 3 major
    sites that were each DTN hubs.  They also were data center hubs, i.e.
    housed much of the computing equipment for the area.  T1 lines were
    installed between the 3 sites to carry voice and data.  Portions of the
    T1's were held in "reserve" to allow the T1 switches to provide fail
    over capability.  This is completely insane as the PBXs, bridges, and
    routers terminating those circuits are completely capable of routing
    traffic in the presence of a failed circuit.  So it turned out that
    only about 1/3 - 1/2 of the actual capacity of the T1 circuits was ever
    able to be used.  Not a great use of our telecom money.
    
    Another example touched upon by Fred is the "I'll copy it myself"
    attitude.  Telecom *could* fix this problem if they'd offer the
    services.  Since it's their budget that needs to be reduced, why don't
    they put the file servers in place?  The same thing could be done with
    many of our distributed services.  As an example, someone could
    (should?) hack together a version of notes that allows shadowing of
    notes conferences.  I suspect a great deal of network traffic could be
    eliminated if the proper services were put in place.  I'd certainly
    rather copy a kit from a local server with immediate response time,
    than wait hours for it to copy across the WAN.
    
    I just wish I'd see indications that telecom is trying to work smarter
    as opposed to making its users work harder.
    
    -tl

    It seems that someone, somewhere has decreed "Cut costs!" and I suspect
    it was at, or near, SLT level in Finance.
    
    IMHO, that's a laudable aim.
    
    Then what's happened?
    
    It's cascaded down Finance into the various areas and plans, proposals
    and even edicts have been made without consideration of their
    interrelationship.
    
    So we have:
    
    "Save travel costs, by using economy fares"  which overlooks the
    difficulty of working on the flight, and overlooks the longer time to
    recover after a long haul flight.
    
    "Save travel costs, by using E-mail, phone calls, con calls and
    video-conferencing"  which overlooks the impact on circuits and PTT
    expenses.
    
    "Save PTT expenses by using the lines less"  which overlooks the demand
    that's just been created by cutting travel.
    
    I agree with the proposal that we should do a traffic study,  rather
    than explore the network over/under capacity by near-arbitrary
    reduction in service provision.

    
    >many of our distributed services.  As an example, someone could
    >(should?) hack together a version of notes that allows shadowing of
    >notes conferences.  I suspect a great deal of network traffic could be
    
    There already is a tool that allows (read only) shadows of conferences.
    In fact *this* conference has a shadow at MORTAL::DIGITAL. The tool is
    called PAN and it's conference is at ROCKS::PAN.
    
    			Alfred
    
    		

re: .21 by CARAFE::GOLDSTEIN

>    Large-scale long-distance file transfer is very costly.  Copying
>    big netkits across the puddle costs more than FedExing tapes.  It would

Can you put some numbers behind this statement?  (Not starting an argument
here, just interested...)

Mickey.

    re:.17
    
    Yes, we do have 5 E-1 (2.048M) across the Atlantic.  They are not there
    to be a "toy" (we do not build networks just for the heck of it!). 
    Each of them is totally justified based on the alternate costs of
    carrying the traffic offered.  If we remove any of them, DEC's cost of
    telecoms rises (if volumes do not go *WAY* down).  They are there
    because the business has arranged itself so that large amounts of
    communication are *REQUIRED* to run the business.  We could organize
    the company in a different way that would not require so much
    communications.  But as long as major business groups are split with
    some of the people in Europe and some in the US and some in APA there
    will be strong demands on the networks as these groups work to keep
    their businesses running.
    
    I have no solution to the problem because the company keeps doing
    things to increase telecoms usage, even in the face of all of the
    reductions in workforce that we've seen over the past two years.  Every
    group seems to want to use telecoms to reduce other expenses, and this
    may well be a proper investment.  But you can't put more demand on
    telecoms to reduce other expense AND still reduce telecoms.  In one
    region the arguement was made that we could reduce the TOTAL
    infrastructure cost by the objective amounts, but in order to do that
    we would have to use telecoms as a tool to leverage these other
    savings, and therefore telecoms costs would rise.  The response was
    that ALL infrastructure costs *INDIVIDUALLY* had to be reduced!  This
    is not rational behavior we are dealing with!
    
    Regards,
    
    Bill

    re:.25
    
    I don't have time to show it here, but I did some back-of-the-
    envelope calculations a little while ago which showed the true cost of
    file transfer as around $10./megabyte for international, $1-2/MB for
    non-local US domestic.  Very short hauls and LANs are cheaper.
    
    The  "demand reduction" folks in Telecom have created a single "$5/MB"
    number which of course averages it all together.  If you go with that
    number, then it's $5 if the server is down the hall or across the
    puddle, so there's no incentive to provide new ones.  I hope to get
    this view fixed; some of my econometric work may help.
    
    re:.22
    That reply pretty much sums it up.  Corporate Telecom (and Bill Mohn in
    particular) have redesigned their own T1 backbone (US IDN) to have very
    little slack at all.  BUt we don't own the "field".  Those guys put in
    their own circuits, often parallel to ours, under-utilized and higher
    in cost.  Due to improper cost allocations, they show a savings to
    their own cost centers vis a vis sharing Corporate IDN.  Again some of
    us are trying to put a halt to this but it is a higher-level management
    responsibility.

    It's my understanding that VTX is to be replaced by the Information  
    Aquifer Repository (i.e. no more VTX).
    
    +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
    
           <<< HUMANE::DISK$DIGITAL:[NOTES$LIBRARY]DIGITAL.NOTE;1 >>>
                        -< The Digital way of working >-
================================================================================
Note 2804.16          PLAN TO DEGRADE DEC VOICE/DATA LINKS              16 of 27
VFOVAX::BRAMBLETT                                     9 lines  30-NOV-1993 16:20
                         -< Distribution is an issue! >-
--------------------------------------------------------------------------------
    
    Well if Telecom is responsible for mail messages sent, I would
    recommend we teach people to use VTX for organizational announcements,
    etc.  What you read in VTX today, comes in electronic mail form
    a week or so later after each layer of management has forwarded it
    on.  That is a place where efficiency could be achieved.
    
    My 2 cents,
    Linda

At the same time we're being forced to use degraded network links. All sorts
of Corporate groups are mandating expensive client server point solutions to
all sorts of business "opportunities". I figure in a year or two they can 
preserve their jobs a while longer by consolidating all these lovely point 
solutions into some sort of "suite", which they can  then replace with 
point solutions etc...

It makes perfect sense to me...

dave

>        <<< Note 2804.27 by CARAFE::GOLDSTEIN "Global Village Idiot" >>>
>                            -< bucks per megabyte >-
>   
>    I don't have time to show it here, but I did some back-of-the-
>    envelope calculations a little while ago which showed the true cost of
>    file transfer as around $10./megabyte for international, $1-2/MB for
>    non-local US domestic.  Very short hauls and LANs are cheaper.

Is this really the *marginal* cost?  In other words, if I avoid downloading
a 1MB file have I actually saved the company $1-2?  It sounds more like
an average (mean) cost than a marginal cost.

Stating average costs as though they were marginal can lead to false
efficiencies and can even cause negative efficiencies if people
switch to other options that have a lower average cost but a higher
marginal cost (like FedEx'ing something instead of EMailing it).

It is my understanding that at any moment in time, most of our EMAIL (network)
services have already been bought and will simply expire unused if we do not
make full use of them.  On the other hand, each time we do or don't send a
FedEx, we do or do not incur an actual expense.

If we could identify where the marginal costs are, my guess is that we would
be able to reduce them without affecting service at all.  For example, creating
a selected number of shadowed notes files might reduce marginal network costs
an order of magnitude more than poking at a few random megabytes being
EMAILed around.

-steve-

    	I use this same philosophy with my lawn at home. If I don't water
    and fertilize it during the summer months, then I don't have to cut it.
    Sure, it gets all brown and looks nearly dead...and insects invade and
    destroy portions of it. But surely this analogy wouldn't apply to our
    corporate strategies!

.31> ...But surely this analogy wouldn't apply to our
.31>    corporate strategies!

Your analogy would be complete if you were to advertise yourself as
"Total expert in lawncare solutions!".

Lee

    re:.30
    You do bring up a good point on the difference between average
    and marginal cost.  The bucks-per-megabyte numbers were averages.
    Marginal is nonzero, but is lower than average.  It's not a linear
    function (bandwidth is procured in fixed increments) but there's
    definitely a declining slope.  I haven't done the analysis yet, but may
    in the future.
    
    Essentially no network facilities are "bought and paid for".  We lease
    bandwidth, and we own some bandwidth assets (which still have operating
    costs).  We own routers, but the Easynet has only begun to transition
    from ancient Microserver to DECnis boxes (which have orders of
    magnitude more capacity).  We own the phones, but they have operating
    expenses, including periodic required upgrades.
    
    Here's a quick back-of-the-envelope example of where some network
    activities may be costlier than you think.  Let's look at video.  If
    you ran DECspin at, say, an average of 250 kbps (long haul), then
    you're transferring about 1.25 MB per minute.  Let's say the domestic
    cost is on average $2.00/MB, and the marginal cost is only $0.75/MB. It
    still costs almost a dollar a minute.  Yet a PictureTel or similar
    compressor running over a Switched 56 network will cost under half
    that.  (Our _marginal_ cost of telephone bandwidth is on the order of
    under a dime a minute per domestic voice or Switched 56 call, between
    high-volume locations.  The average cost, is higher.)  The catch is
    that a circuit-switched call (DTN or MCI or whatever) generates a nice
    discreet manageable bill, while a DECspin session is just more data
    traffic, still basically unaccountable.
    
    As you can see, the issues get very complex.

>    As you can see, the issues get very complex.

For sure!  Especially when you try to assign a "value" to the benefits
we are getting from the various services.  This part of the equation is
virtually impossible to quantify.

-steve-

    By an accident I saw Reuter's network, these people are in selling
    information, no 5 E1s over pond. Some of our bandwidth is really
    justified with private voice that is cheaper .

    But data, no way. It's incredible luxury to sit at home in France and
    type it in in real time in US. Bandwidth costs a lot, a LOT !

    Is there a strategic direction for Easynet defined somewhere ?
    I mean in real sense of the word and not the generic BS we are used to.
    One of such statements could be "Easynet is a batch network" or
    "Easynet is a real time transaction processing network". 

    The "batch" one looks most real for me, so we should strive after high
    bandwidth utilisation and not bother about response times to much. The
    true real time users should pay premium price. 

    What we are seeing now is that owner came home and party is over.

    					

How can you bill a site which is forced by Corporate Mandate to install real
time WAN applications like CPCRS, CSMS, IRIS,  PIEC or RESUMIX to pay the 
piper? My site didn't originate those solutions, we're merely forced to 
implement them. It makes no sense to me to cut the telecomm bandwidth when
everyone wants to automate everything via these type solutions.

dave

    Hej, I'm just consultant giving a 0$ advice worth every cent of it !

    In reality, things are simple if you change perspective. Lets think you
    really own this circus and every $ out is from your own pocket. If you
    really try hard this approach, the answers are obvious. You make common
    sense business decisions and not try to hide behind pseudo science.

    Another thing struck me about this circus.  We *have* tools to model
    business problems of the kind being attacked here.  In fact we've even
    developed such tools and used them on smaller groups.  They were
    developed to help prevent "local optimizations" that have a negative
    global impact.
    
    But I guess we wouldn't stoop to using something we built and sell to
    our customers.
    
    -tl

    Is there any way to clean up some of the gross inefficiencies that
    exist?  Yesterday, I was in another local site to pick up a proposal
    from a tech writer.  The printer was in the next cube over from her
    workstation.  I asked why it was taking so long to print and she told
    me she had to copy the postscript file over to our A1 machine (in
    Chicago, we're in Detroit), then log into that machine to print the 
    file because there was no local queue definition for the printer.  
    It seemed pretty inefficient to copy megabytes of data hundreds of 
    miles so you could print it on a printer six feet away.
    
    By the way, she's been trying to get this changed for months and has
    just about given up.

    I believe that .39 illustrates the point that I've been trying to make
    here.  IM&T has been rapidly cutting their costs of operating and
    maintaining data centers by consolidating and centralizing data
    centers, so applications like ALL-IN-1 only have to be administered
    from one central site in a region rather than in all (or a large
    number) sites.  This saves the company a lot of money by making very
    efficient use of available (and declining) resources.
    
    However, in order to do this, substantial bandwidth must be installed
    over longer and longer distances.  This costs money, but it is (in most
    cases, anyway) cheaper than continuing to support a larger number of
    data centers.  This is what I mean by leveraging the network(s) to
    improve the overall economic efficiency of the company, and IMHO this
    is a "proper" use of telecoms (we sell these kinds of solutions to
    customers all of the time).  I cannot argue the merits of using these
    centralized applications to run the business; that is a business
    decision driven by the strategies of the business, and is a different
    "can of worms".
    
    The point is that Digital's overall expense picture is improved, but
    only at the expense of some added telecoms cost.  The problem we are
    trying to deal with is the edict that ALL expenses must come down,
    without regard to the interactions between different expense reduction
    activities and without regard for optimizing the TOTAL expense.  This
    kind of micromanagement misses out the "systemness" of the enterprise.
    
    We see this happening all over the company, with groups consolidating
    operations and generally acheiving greater efficiencies in manpower and
    other resource utilization as pressures to reduce headcount and other
    expense continue to be applied.  Unfortunately, these decisions almost
    always imply the need for more communications to make the actions work. 
    Until we totally re-engineer the way the company delivers products and
    services, we will continue to see increased reliance on telecoms to
    hold the company together.  I probably should have said "unless" rather
    than "until" in the last sentence.  Matrix organizations, worldwide
    account teams, worldwide development activities and all of the other
    ways that Digital works all require simply enormous amounts of telecoms
    as an enabler.  We could certainly do it with less, but not with the
    current organizations.
    
    I think there might be another whole topic in there somewhere....
    
    Regards,
    
    Bill

    >If you ran DECspin at, say, an average of 250 kbps (long haul), then
    >you're transferring about 1.25 MB per minute.  Let's say the
    >domestic cost is on average $2.00/MB, and the marginal cost is only
    >$0.75/MB.
    
    Is it really that expensive?  Here's my back-of-the-napkin:
    
    	T1, CRL<-->Palo Alto, $10K/month
    
    	1.5mbits/sec = 187500 bytes/sec = 502200 megabytes/month
    
    	$10,000 / 502200 = $ 0.02/megabyte
    
    (Another way of looking at this:  the T1 costs 22 cents/minute).
    
    Where is the additional TWO orders of mangnitude increase in cost
    coming from?  Packet overhead?  End equipment?  Salaries?
    
    The marginal cost of networking in the greater Massachusetts area
    should be very close to zero, thanks to our already installed fiber
    plant.
    
    -andy

    
    okay, a simple way to cut down on net traffic,
    
    write an extension to VMS BACKUP to do COMPRESSION. I bet most of the
    network traffic is large savesets being moved around, VTX and email
    don't move that much traffic. If we could compress files and then send
    them, think how much bandwidth we'd save! And it would make customers
    happy since they could use this too. Commercial compression tools like
    ZIP and so on can get a file to 30% of its original size, depending on
    contents.
    
    Plus at the CSC we could compress files to send down to customers,
    saving a fortune on modem charges. Transferring files to home systems
    wouldn't be the no-no it is now.
    
    Simon

	This is a good example. There seems to be a problem with our 
own internal groups MIS/IM&T to upgrade, install, use our latest
products. My manager has a new 1152 laser printer, but cannot use
it from Allin1 because the cluster is running old unsupported
versions of "Digital" software. We were told maybe in a year they
would consider an upgrade.....

	We in the field help our external customers see that it is
a good idea to upgrade, install our latest products. Some external
customers even run on "Beta Test" versions so that they can get a
jump on their competition. It looks like the message we Digital
present is:

	Do as I say, Not as I do......

FWIW
Renis

    re: .40
    
    I've heard this argument before about centralizing data centers to save
    costs and I've never bought it.  Please demonstrate the savings of
    consolidating data centers when many facilities need computer rooms
    anyway to house non-IM&T systems.
    
    I'll use an example again from the area I used to work.  We had IM&T
    run ALL-IN-1 mail systems and numerous software services systems. Most
    of this was housed in the three largest facilities in the area. Someone
    in IM&T decided they couldn't manage their systems in "so many"
    locations, so the moved all the systems to a facility that was nearly
    the most remote facility in the area.  A new data center had to be
    built, the rent wasn't substantially cheaper, the power was the worst
    in the area, and the comm lines were now mostly inter-LATA lines.
    
    Now a huge terminal network had to be built.  A separate data network
    was built.  And a separate telephone network was built.
    
    As was obvious, service quality got worse, costs must have sky
    rocketed, etc., all because someone decided centralized systems (ala
    IBM) were cheaper than distributed systems.
    
    We have the tools to run lights out data centers, remotely manage
    systems, support multi-protocol multi-purpose networks, combined
    data/voice networks, etc.  When are we going to use technology to solve
    our problems and not organizational gamesmanship?  We call ourselves a
    networking company, yet I think it's our customers that know
    networking, certainly not us.
    
    -tl

    re:.44
    
    I can't demonstrate any savings for data center consolidations because
    I was not involved in the decisions.  Or the studies.  All that you say
    may well be perfectly true; certainly the part about building terminal
    nets and parallel data nets, et al, is true.
    
    All that I am trying to say is that *someone* somewhere made this
    decision based on factors "well understood" by them, and telecoms had
    to support that decision.  Perhaps not all of the economic factors
    involved were taken into account; I don't know.  The sum total of
    hundreds of such decisions like this has the impact of requiring ever
    larger amounts of network to be built.
    
    As Fred has pointed out elsewhere in this string, the financial
    structures (internal) of this company creates a lot of "cost avoidance"
    through "cost shifting".  Often the costs avoided are lower than those
    shifted to another group.  This is nothing new; I've been here for
    nearly 15 years, and the practice was well-entrenched long before I
    ever appeared on the scene.  As long as people are incented to "save"
    money by shifting $100 out of their own cost center, even if it costs
    some other cost center $200 when they do this (or worse, ends up
    increasing the cash out the door by $200), this behavior will continue. 
    You can't blame people for doing what they are rewarded to do.
    
    Regards,
    
    Bill

    re:.42, Backup;
    
    I don't know of much backup going on over the wide-area network but
    with all the data center consolidations, it might well be major.
    There is a new version of FTSV (I think the Telecom folks are behind it
    -- the full-time telecom professionals, of whom there are many, do
    think of these things) which performs file compression automatically.
    Something like half the Easynet traffic is file copy.
    
    re:.41
    The difference in cost comes about because packet networks have lots
    and lots of overhead, bit-wise.  You can't look at bytes/month for
    interactive applications like video, becaue the line is only useful
    during the hours people are using it.  Thus we typically assume that
    voice networks run about 200 hours/month, with a coule of busy hours
    every day, and a few nearly-busy hours.  Batch file transfer, at least,
    keeps the networks humming 7x24.
    
    In a packet network, line occupancy affects average queue size; a
    75% line occupancy means the average queue is 3 packets.  This is a
    good starting point for IP.  Now you add packet overhead.  Add the cost
    of routers -- not cheap.  Add the substantial labor needed to run a
    router network -- it's higher, per bps, than telephone, by an order of
    magnitude or so.  (Telephone switches have bazillions of ports and just
    sort of hum along if maintained well and set up correctly.)  I think
    Digital spends about $5/MB on Easynet overall (WAN traffic), of which
    less than half actually goes to carrier bandwidth.
    
    Remember the comparison I'm making:  Video over packet vs. video over
    circuit.  Circuit has no per-packet overhead, and uses bandwidth at
    100% (via time-division multiplexing).  So we can run 24 x 64kbps on a
    T1 all at the same time.  Packet has bit overhead and other overhead
    that are higher than circuit, per steady-rate bps.  Where packet wins
    big is in burstiness:  Data traffic is not steady-state like voice. 
    Video is more steady than data, though slighly bursty.
    
    Within a metropolitan area (fiber optic or intra-LATA circuits), data
    bandwidth can be cheap -- at the margin.  Getting that first BPS to a
    site is costly.  Crossing LATA boundaries costs more, and international
    circuits are seriously expensive (bandwidth cost becomes dominant).

    
    re .46,
    
    Great, if newer versions of FTSV really does have a compression in it,
    then there should be a major effort by networks people to get all
    system managers to install the new version. 
    
    It seems we already have plenty of people looking for things to do,
    what with audits, policing of weather maps, and playing DECinspect
    games. Get these people to push out FTSV to every machine on EASYNET.
    Put some of these same sniffers out to find files not compressed by the
    new FTSV, and send both ends a nice email pointing out the new FTSV
    kit location.
    
    Simon

Continuing the FTSV rathole...

SO where is this new FTSV? I just installed V2.4 on both my VAX and AXP systems,
and I saw no COPY/COMPRESS qualifier.  It's just got the usual thing for doing
record blocking, which it's had for years.

BTW: The datacenter here in LKG offers, as others do, remote backups.  The
present scheme uses ordinary VMS BACKUP to move Gbytes of files over the
Ethernet. If this same stuff is happening between WAN-connected sites, it's
probably no wonder our telecoms cost is so high...


Is there any plan afoot to put some real compression into VMS BACKUP?  I think
that the old argument of "why, if we do that, people will buy less hardware!" is
very tired....

John

    
    I just looked in the FTSV notesfile, and 2.4 for a while had
    compression, but it turned out the public algorithm wasn't as public
    had been thought, so compression was pulled. According to note 726,
    compression will be added back in next year.
    
    Lets hope system managers can make compression the default, so we'll
    all be using without changing procedures or anything.
    
    Simon

    >re:.41
    >The difference in cost comes about because packet networks have
    >lots and lots of overhead, bit-wise.
    
    Overhead accounts for a factor of 100??!
    
    >Thus we typically assume that voice networks run about 200 hours/month
    
    Ok, then our T1 only moves 135 Gigabytes/month.
    
    >In a packet network, line occupancy affects average queue size; a
    >75% line occupancy means the average queue is 3 packets.  This is a
    >good starting point for IP.  Now you add packet overhead.
    
    A 56kbit data link will move 6.8kbytes/sec with TCP/IP (I just tried
    it).  That's 54.4kbits/sec, for a link utilization of over 97%.  So
    we move 130 Gigabytes/month.
    
    >Add the cost of routers -- not cheap.
    
    Ok, let's buy two and write them off in one month.  Add $20,000 to the
    cost of our line.
    
    >Add the substantial labor needed to run a router network
    
    One full time person for the T1:  add $10,000 (I still haven't figured
    out why it should take more to run a packet network than a voice
    network).
    
    This all totals to 130 Gigabytes for $40,000, or $0.31/megabyte.
    
    >I think Digital spends about $5/MB on Easynet overall (WAN traffic), of
    >which less than half actually goes to carrier bandwidth.
    
    The point I'm getting at is that $5/MB seems very high.  How is this
    figure derived?
    
    If the cost really is this high, then Corporate Telecom should have no
    trouble meeting a 20% reduction by slashing its generous overhead.
    
    If the cost isn't that high, then all cost/benefit analyses using this
    figure should be considered bogus.
    
    -andy
     

Re:.51

There is probably a 'burden' figured in of the floorspace for the
equipment in a computer room, floorspace for the staff, managers for the
staff, etc.

I still agree with your points though, $5/Mb seems way too high to me.

Lee

    	My pay stub today carried a reminder to get "Netwise". Where can I
    copy it from? ;-)
    
    Phil

    They also refered to VTX BCP (I think it was BCP) - I looked, there's a
    menu entry for NETWISE.  Of course I was abusing the network by looking
    at the netwise info in VTX, rather than tracking down a copy of Digital
    Today - but what the heck, it was kinda fun!
    
    --Scott
    

    re: .45
    >
>    All that I am trying to say is that *someone* somewhere made this
>    decision based on factors "well understood" by them, and telecoms had
>    to support that decision.  Perhaps not all of the economic factors
>    involved were taken into account; I don't know.  The sum total of
    
    Two comments on this.  In the US Field, it used to be that telecom and
    IM&T were effectively the same organization, so telecom having to
    support the decision made by someone else doesn't really come into
    play.  I do agree that the rest of the corporation wasn't (isn't?) so
    aligned.
    
    The other thing was the basic reasons given were that they couldn't
    manage computer and telecom systems remotely effectively.  I really
    believe it was much more an issue of "control" as there are certainly
    few technical obstacles in managing and operating those sorts of
    systems remotely.  The telling point was that they weren't any more
    "effective" in managing the systems after consolidation than they were
    before.
    
    Dysfunctional incentives might be the right issue, although I think in
    this case a lack of accountability and scrutiny are more the issue. 
    When these folks made their claims about how centralizing was going to
    solve "their" problems and indeed it didn't, someone should have gotten
    fired.  Most of our customers in that area did a much better job
    running their data centers and telecom operations than we did.  Our
    organizations were more interested in building empires than providing a
    service.  We should have been looking at outsourcing proposals or
    benchmarking our service with our customers.  It would have become real
    obvious that our service was terrible.
    
    What concerns me is I don't see this changing much.  I still see the
    organizational politics and infighting, the lack of accountability, the
    inability to make decisions based upon anything other than at the
    microscopic management level.  
    
    A simple example is the current work at home stuff.  We could move a
    large number of people to work out of their homes, meet the federal
    mandates, and save the corporation some big bucks, but no one
    organization can afford to do it with the budgeting and metrics scheme
    we have in place.  These sorts of problems take leadership from above
    that appears to be sorely lacking.
    
    -tl

    
    Lots of threads through this one note, but one in particular
    bothers me the most:
    
    When I pick up the phone to call a customer, will I be less likely
    to get through because of this?  If the answer is yes, then I'd
    like to hear the justification for creating this obstacle
    to doing business, comparing savings against reduced revenue.
    
    If the argument is that I should "try again later", then that's bunk.
    I don't have time to do something twice that could have been done
    once.
    
    And most importantly, what happens if it is a customer calling _me_?  
    
    
    As to the network file traffic issues - if someone or some group
    has a problem with the load, then why don't they fix it?
    Quantify and profile the traffic load, propose some solutions,
    prioritize their implementation, and then do it.
    
    No staff or budget?  Come up with a business plan and run it by
    management.  Let's use our own products and employees to reduce the
    amount of money we're giving to AT&T etc.

There used to be a time when we (this company) faced a problem
like this by designing and implementing a technology solution.

We used the solution, we refined the solution, and then we sold
it to others.  I thought that was where industry leadership
products came from.

But then we gave up the networks business, didn't we...

re: .46 FTSV, .47, and .50

	FTSV is a product developed by Corporate Telecommunications to 
do reliable batch file transfers. In November, we released FTSV V2.4.  
This release was due to have compression capability, but at the last 
hour this functionality was removed due to a patent issue with the 
algorithm we were using.  CT did not feel it was in Digital's best 
interest to pay for this license when we had other options available.

	A new release with compression capability is being developed 
that will ship early next year.  This release is using a compression 
algorithm that has been in use in Digital for some time, and does not 
have any patent issues.

	If you have any questions, please give me a call.


		Cindy Gregoire
		FTSV Product Manager

		DTN: 227-3955   CTHQ::GREGOIRE

    Perhaps this plan has already gone into effect.  Today, ping times from
    Chicago to Spitbrook have been averaging 7-10 seconds.  Seems like they
    must have decided to add more than 300 msec per hop :-(
    
    re: .58
    
    How about some effort into figuring out ways to eliminate file copies
    instead of optimizing them?  Just putting file archive servers in major
    sites with Archie front-ends would be a huge improvement.  I have to
    believe with all the equipment being dumped into DIAL each year, that
    we ought to be able to find some that can serve as file servers.
    Purchase some new large capacity disks and put together some easy means
    for people to put their stuff on the archive server and you're set. 
    Right now, a great deal of stuff gets copied from gatekeeper or one of
    the few other systems on the net.
    
    -tl

re Note 2804.59 by NUTS2U::LITTLE:

>     How about some effort into figuring out ways to eliminate file copies
>     instead of optimizing them?  Just putting file archive servers in major
>     sites with Archie front-ends would be a huge improvement.  

        I agree.

        This would be a perfect opportunity to try out ideas for a
        corporate-wide "information utility".

        If somebody has a file (program, kit, data, whatever) to make
        available to the Digital-wide internal community, there
        should be some way in which they can check-in their file and
        announce it, with sufficient description, for cataloging.

        The utility would take care of replicating and/or caching the
        file and distributing the announcement to various forms of
        catalogs.

        Requests for copies would be directed to local or regional
        servers and fetched from the nearest copy.

        The group (in IM&T) I'm in is building such a utility,
        although its original target goal is the distribution of
        corporate MIS products, e.g., database extracts and reports. 
        I do hope we can build it in a general enough way to serve
        some of these more general purposes.  It may be necessary to
        get some involvement from non-IM&T groups if that is to
        happen.

        Bob