Richard Seltzer's
home page Publishing home
Articles
about DEC
mgmt memo
.
Volume
7, #5 July,
1988
On May 5, 1988, over 600 senior managers
attended Digital’s State of the Company Meeting in Merrimack, N.H.
The theme for the day was "One Company, One Strategy, One Message
- Leading the Way to Enterprise-Wide Computing." Many of the
presentations focused on "The Desktop and Beyond" and "Digital’s
Transaction Processing Strategy, Systems and Uses." The following
are summaries of the speeches.
The
One
Company Theme And Digital’s Evolving Strategy by Ken Olsen, president
Financial
Update
by Jim Osterhoff, vice president, Finance
Decwindows
Solution
by Roger Heinen, Corporate Consulting Engineer
Our
Transaction
Processing Strategy by Bob Glorioso, vice president, High
Performance Systems
When we started Digital, we picked areas in
which we could be different and unique. They were simple ideas.
Wall Street didn’t understand them.
We had the unique idea that we wanted to make a
profit. In those days you were just supposed to grow, not make a
profit, and then sell your company. We said growth wasn’t a goal.
We wanted to do a good job and let growth come automatically.
That’s still our goal.
At MIT, we got the idea of interactive
computing, enabling people to do the things we see so commonly now
with personal computers. That was a hard idea to explain. Finally
it sneaked up on people, (we like to think that it was because of
us), and it mushroomed 25 years later.
Next came the idea of timesharing, which was
first done on a computer we gave to MIT. We were ridiculed for
this idea. We stuck with it; and, in time, it became the most
important part of our company and probably still is.
When interactive computing became popular, we
realized that making small computers was going to be too easy. It
was clear that this was not our place to make a contribution. Our
contribution was to do the bigger job — the job of integrating a
whole organization around the world, taking advantage of our
strengths in VAX, VMS, DECNET and Ethernet products. The press
never understood that. We are still criticized because we are
still not making personal computers for the home.
Rather than just be one of hundreds of
companies making personal computers, or one of dozens of companies
making workstations and changing their architecture each time a
new chip comes along, I’d rather be different and take all the
arrows that come our way.
When my friend, the head of Sony, came up with
the idea of the Walkman, he didn’t ask the marketers for advice.
Marketers are taught to ask customers what they want, but
customers only want what they have seen. That’s not where new
ideas come from.
There’s a very important lesson here: market
surveys are a serious danger. If you did a market survey a year
ago, you’d have found that most personal computers were bought in
stores by individual users from companies, that they included all
kinds of hardware options, and everybody’s personal computer was
different. Such a market survey would guarantee that that was the
way it would go forever. But you know very well that can’t be
true.
My friend from Sony knew that his job and
position were dependent on being right. That’s true of us, too. We
have to be right and then be brave.
Today, the strategy for which we are being so
roundly criticized is something we have planned very carefully and
developed very well. There are a number of details which we have
to improve. There are always ways we can do things better. But
with regard to our strategy, I couldn’t think of a better position
to be in today than the one we’re in now.
A corporation is a collection of people, and
the leaders can only do a very limited amount of the creativity
and the work. The work has to be done by the people at all levels
and at all jobs. And they have to have the freedom, the knowledge,
the motivation, the acceptance to do creative things. The job of
management is not to do the inventing. Our job is to make sure
there are goals, to make sure everybody knows where we are going
and is in a position to be creative.
From the start of the company we have taken the
part of a scientist — to be rational, analytical, calculating, to
do the smart thing. A few years ago we developed the strategy that
says — one protocol, one architecture, one software system, one
networking system. I had little or nothing to do with that except
complain a little bit. But it was clearly my job to say that if
this is the strategy, we’re going to follow it. That is the part
of the manager - not to do the inventing, but to make sure there
is a goal. That is an example of what I mean of being analytical,
scientific and rational.
We are academic in that good sense of being
very critical of our ideas and always driving for the truth. It’s
quite common that people fall in love with their own ideas and
defend them like they were the only ideas they ever had. But we
want to optimize and improve, change, modify, but always drive for
the truth. As long as we do that, we have a good future.
When we said four or five years ago "We are one
company with one strategy and one message," we were changing the
company from being many product lines competing with each other,
to putting all our resources together. We pulled it off better
than I ever could have dreamed. The idea of 120,000 people working
together was inconceivable. We’ve done a miraculous job. We
succeeded because of the quality of the people we have. We should
all be very proud of it. We had a simple message, and it worked
well.
Now we are being beaten up by the press again —
for inconsistent reasons. They criticize us for not doing
everything, and then criticize us for not doing one thing only.
You can't answer both of those charges at once.
A few years ago, we said a major part of our
emphasis was on one architecture and one software system. We then
paused and said we would also do Unix*. Now we’ve modified our
statement to say the "one message" and the "one strategy" is that
we have two software systems. Obviously, our strategy evolves as
we strive to improve.
The first VAX-11/780 computer was a completely
different machine than the last, when we stopped building that
model eight years later. One by one, over the years, we had
changed every module. Our VMS operating system has changed
drastically over the years and will change much more drastically
over the next ten years.
Yes, we change. But we do have one strategy as
a company, and we have to be sure we follow it.
We sell Unix and VMS software with equal
enthusiasm. That does not mean that they are equal or that they do
the same thing. There is a place for one and a place for the
other.
You may hear that VMS is an old, obsolete
software system. Just the opposite is true. Unix, which we now
love, is 19 years old; by the way, over those years, we’ve been
the largest seller of software and hardware and service for Unix.
The VMS software system — the only other major software system —
is just 12 years old. So in years, it’s the newest.
When Unix started, it was never really planned.
It was made to be a small, informal, austere system for one user,
done on PDP-7, and later on PDP-11 computers. It was never planned
to do anything else. That was the delight of it — it was easy to
use. In the early 1980s, Unix got to be very popular because it
was free and it was easy to learn. We came out with our ULTRIX
software to add formal discipline and quality.
From the start, VMS software was laid out to do
the whole job and to last almost forever. It is modular so parts
can be improved and the rest of it doesn’t change. It is
disciplined, planned and organized. That's been the reason for
its success. Therefore, I claim the VMS system is the more modem,
both in age and in organization.
You may hear that VMS software is not
"transportable" and that Unix is. There are three ways in which
software can be transportable — over time, over a range of
equipment from very small to very large, and from the equipment of
one manufacturer to another.
In terms of time, VMS software that was written
ten years ago will play today on modem, better, faster, bigger and
less expensive systems; and it will play ten years from now on
even better systems.
VMS software is also transportable from very
small to very large VAX systems. If you have a branch office that
has two people or a branch office that has two thousand people,
you pick the appropriate size machine, and the same software is
transportable. If your company grows or your taste for computers
grows, the software is transportable to bigger computers.
Our software is not readily transportable to
other manufacturers’ equipment. This is one of the arguments for
Unix. But this argument has problems. Our goal is to make the best
Unix; IBM’s goal is the same; so is Hewlett Packard’s, etc. That
means everybody’s Unix is different.
But for those customers who want simplicity,
who want to take the responsibility themselves, who want to be
able to buy different computer systems and put them together, we
offer Unix. Those are our traditional customers, our friends. We
offer systems, we take care of those customers, and we’re going to
be the best supplier.
That is not in conflict with our main strategy.
We can do two things. We’ll satisfy that market and also the bulk
of the computer market that wants one manufacturer to do the whole
job. We have to sell both those ideas.
Our contribution to the major corporations of
the world is to be trusted and competent to do the whole job,
worldwide. We have the equipment for tying everything together. We
have the line of computers that can do almost everything. We have
the sales people who know how to sell to banks and insurance
companies. They do it with competence, and they are trustworthy.
We have the design people for designing the systems. We have the
people for installing them. We have the people for servicing them.
And we have by far the best software services organization in the
world.
We are doing the whole job, the complete system
- all the computers and the software systems, the networking. Our
combination of computers will do it all.
We did poorer than planned last year, but in
any measure of management we did very well. The way we were able
to adjust and cutback from the volumes that we had budgeted was a
real accomplishment.
Budget time is when we have to decide what we
want to do in the coming year. Of course, we all want to do an
infinite number of things; but they don’t add up, and you can’t
sell them all. The customers couldn’t swallow them all, and the
sales people couldn’t remember them all. So at budget time we have
to make decisions and face what we really want to do.
So, just like every other year, when we add up
the budget this time around, there’s more that we want to do than
we can possibly sell or pay for. Even though in my heart 1 know we
did great last year, we did make a mistake in budgeting higher
than we should have. We have to be absolutely perfect this year.
We’ve got to work hard to make sure we don’t budget one more thing
than we can sell. There’s no point in having two computers do the
same thing, or in having one more computer than the sales people
can remember. And, after the budget is done, we know and are in
agreement about where we’re going.
We tend to get all wrapped up in the business,
which is important. But remember the other things, too: your
family, your wife or husband. You owe it to the people you’re
responsible for to work hard, but business can’t be everything.
So as we push and drive hard for business — which is the
excitement, the fun and the responsibility — remember there are
other things, too.
Ken recently asked the provocative question.
"If we’re doing so well why do I feel so bad?" By most measures we
really are doing well: 20% growth, 13% pre-tax profit margins and
$2 billion cash in the bank aren’t all that bad. But that doesn’t
change the fact that we feel bad.
To describe our financial situation clearly, we
have to look at both the static and the dynamic conditions of the
situation. Consider an illustration of an automobile straddling a
railroad track just a few feet in front of a freight train. That’s
a snapshot of a moment in time. It doesn’t show us what’s
happening. We can’t tell whether the car is moving, and if it is
moving, how fast. We can’t tell whether the train is stopped or
moving or in which direction it is going. Likewise, a snapshot of
the present condition of the company — how much we owe, how much
we own, how much cash we have in the bank — gives an incomplete
picture. We can gain a reasonable understanding of the situation
only by looking at where we’ve been and where and how fast we
might be going.
In 1985 our static condition was okay. Our
asset management as reflected in the balance sheet was no better
or worse than it had been historically. Our capital structure was
sound, but we had just added $750 million cash from additional
borrowings. Our dynamic situation was generally poor. Profit
margins were in the single digits, and the volume gains we were
achieving weren’t returning much to the bottom line. Good asset
controls were being implemented, but we hadn’t seen the results of
those yet. We were not living up to our potential, and the low
return we were providing to our investors was depressing the
market value of our company.
In 1986 our static condition was very good.
Results of inventory and receivables management actions had
become visible. Cash balances were up to $1.9 billion from $1.1
billion the year before, and we had reduced our debt. Inventory
turns had improved by a factor of 50% from historical Digital
levels. Our dynamic situation was improving, but the distance
between revenue and cost was still too narrow, given the rate at
which we were growing.
In 1987, just a year ago, our static condition
was triple A, with an exclamation mark; and our dynamic condition
was very good. We had made outstanding progress. Our return on
investment was the highest in recent history; our stockholders
were happy; and the investment community was expecting all of
that to continue.
Today, our static condition is still triple A.
Our dynamic performance on the other hand isn't what we planned it
to be, or what the outside world expected it to be. I would assess
it as disappointing.
This might provide a way of answering Ken’s
question. A snapshot of the company would indicate that we are
indeed doing very well, but we feel bad because our dynamic
performance recently has been disappointing.
Another reason why we feel bad is clear from a
graph of the price of our stock. We’re the same company we were a
year ago, only 20% larger. We’ve moved up on the Fortune 500
listing from 44th to 38th. We’re still triple A. But a year ago
the stock was selling at about $170 a share and was going up. So
you wonder what we’ve done to deserve all this rough treatment
that we’re getting on Wall Street these days.
A year ago, when we were making plans for 1988
we were in the midst of a 24% growth year - a strong rebound from
14% growth the year before. The computer market was still not
particularly strong, but we were out-performing virtually all of
the competition. Our operating profit margin was 17% - a giant
leap from 11% in FY86, and well in excess of the budget. Inventory
turns not only surpassed the record of the prior year, but were
among the best in the industry. And our cash balance was a strong
$2.1 billion after spending about $800 million to repurchase our
stock. Our product leadership position was gaining wide
recognition among computer buyers. We could demonstrate our
uniqueness and its value to the customer, and we could charge for
it. In the financial community as well as in the computer industry
our triple A image was gaining recognition. In brief, we had
strong, positive momentum.
In that environment we evaluated alternative
planning strategies for FY88 — determining the appropriate balance
between minimizing cost growth and maximizing marketshare. We
evaluated such factors as product strengths, competitive
advantage, customer satisfaction and loyalty, image, financial and
human resources, and profit momentum, (i.e., whether our profit
margins are rising or falling). An evaluation of our condition
clearly indicated an aggressive strategy for FY88, tilting
strongly toward market growth. The objective was a combination of
revenue growth and budgeted profit margin that would provide an
excellent profit result for FY88 and advance our market position
for the benefit of future years.
Our plans for FY88 were roughly comparable to
the results of FY87, with slightly less margin, but higher growth.
Our plan provided for substantial investments, building on
momentum that began in FY87. This meant investments in new
accounts and a higher level of service for existing accounts. It
meant investment in U.S. Industry Marketing, DECWORLD the event,
and Application Centers for Technology, as well as Computer
Integrated Manufacturing and process technology improvements in
our plants. It meant new high, mid-range and networking products,
and special efforts to penetrate the financial services industry.
Many of these projects were building on investments that were
begun in 1987 or even before. In other words, instead of just
sitting back and reaping the rewards of prior year investments, we
planned to continue to make even heavier investments for the
future — paid for. of course, by higher revenue growth.
For the first three quarters of FY88, our
revenue is up 21%. That isn’t bad. (It’s close to our average
since 1981). But it’s not as good as our plan.
Our expenses in the first three quarters have
been less than the budget. The effects of lower-than-planned
product sales, higher-than-planned services volume, and changes in
the mix of products and services sold account for some of this
difference. Compared with assumptions in the budget, the dollar
has weakened, increasing the dollar equivalent of cost incurred
overseas. Aside from these factors, attention to costs and a
cautious attitude, initiated after October 19. have contributed to
significant savings from the levels budgeted. We achieved this
despite an unexpected increase in the cost of dynamic RAM memory
chips that we import from Japan.
Following the pattern for expected revenue
growth, we expected that costs would grow by increasing amounts
through the first three quarters. We have kept costs under budget
in each of the first three quarters, due in part to increased
attention to hiring and expense controls. We show the effects of
that to a minor degree in Q2, and much more visibly in Q3.
The combination of revenue and cost performance
tells the profit story. Despite running below our budgeted cost
levels, our rate of cost growth has exceeded revenue growth for Q2
and Q3. Even in Q3, when expense controls helped our cost growth
bend downward, we still couldn’t react as quickly as the change in
revenue.
Our pre-tax profit margins for QI of FY87 and
QI of FY88 were about the same — close to 13-1/2%. But in Q2 our
margins were below last year by about three percentage points, and
in Q3 they were below last year by about five points.
In absolute terms, Digital had higher profit
margins than such competitors as IBM and Hewlett Packard in the
latest quarter. We are still doing well, but don’t forget the
illustration of the car and the freight train. Direction and speed
of movement are the critical factors in looking beyond the present
state of affairs into the future. As an outsider looking at
Digital’s recent performance, you would be disappointed in the
decline from last year and perhaps somewhat concerned about the
future direction.
During our strong period in the market last
year, the price of our stock just about kept pace with the market
in general. But since October 19th we’ve fallen substantially more
than other stocks.
I believe there are two key elements to turning
this around, to converting "disappointment" back into optimism.
First, we have to tell our story more clearly — to convince the
public that what we have is on the leading edge of enterprise-wide
computing; that we are a major factor in workstation products;
that we are very serious about on-line transaction processing;
that we are not going to be stopped by the hot companies in the
low end. Second, we have to recover our profit momentum.
To really get costs under control, and still
accomplish our new program and our growth objectives, we have to
do a better job of re-directing resources to areas where
incremental needs are the greatest and manage headcount growth
more carefully.
We also have to pay careful attention to
capital expenditures. These are investments for the future, and
the costs associated with them are spread over future years. They
affect future profits much more than they do current profits. The
effects of capital expenditures are seen directly in the form of
depreciation. But they also drive other costs, such as project
expense, operating expense, new product or new plant start up,
material and inventory related costs and the like. This year our
capital expenditures took a sharp jump upward to a running rate of
about $1.5 billion from roughly $750 million a year ago.
A portion of these expenditures is for the
purchase of office space that was previously leased. The balance
represents an eventual net addition to our cost base.
There’s a warning signal here, that the cost
pressures that we are feeling today are likely to be with us for
awhile longer; that at this level of spending, we may be building
in too much cost ahead of revenue.
The challenge is to make sure each investment
is supported by a good business plan with a promise of a good
return; that each investment makes sense not only on a stand-alone
basis, but also in the context of the total investment portfolio
of the entire company; and that each investment is backed up by
the commitment of everyone whose participation is required to make
it successful.
Borrowing on a phrase from Abraham Lincoln,
it’s not enough that our capital investments be commitments by the
company. They must also be commitments of the company and
commitments for the company - that is, an integral part of the
company’s strategic direction.
Our past strategy has been very simple and
focused. It has produced remarkable products and remarkable market
success. It has resulted in: o a broad range of compatible
computer systems;
o a very high quality, highly functional
software system which covers all the major usage styles
(workstations, timesharing, production computing and real time);
o industry-leading networking with wide-area
networking capabilities and local-area networking products and
solutions; and
o a very rich base of applications from
Digital, third parties and customers.
Our strategy has been so well aligned with
customers’ need for distributed computing in the enterprise, that
its technical and market success has probably exceeded our
original expectations. Our customers want to push computing closer
to the business and closer to the end user to get responsiveness
in application development, application deployment and the actual
use of the applications themselves. They also want to increase
organizational effectiveness by connecting all the computing
resources of the business — including those of multiple suppliers
— together in a single enterprise-wide computer network.
The motivation behind our evolving strategy is
simple: to strengthen our position relative to IBM and assume the
leadership role in enterprise-wide information systems.
Recently, we’ve been focusing on six areas:
o the applications environment;
o transaction processing and databases;
o desktop systems;
o enterprise-wide networking;
o industry standard operating systems and
o the integration of non-Digital architectures
into the network.
Applications environment. Over the years,
Digital has had the lead in the distributed applications
environment based upon its well-architected DECnet, VMS, and
ALL-IN-1 components. However, since these components were
designed, technology and application requirements have changed.
These changes have motivated us to update the Digital application
environment, across the board. The program to provide this
updating is called the Applications Integration Architecture
(AlA). Our graphics-based, windowed-user interface and compound
documents are of central interest to AIA. (Compound documents can
simultaneously include multi-sized text, graphics, images,
spreadsheets and so on.)
Transaction processing and databases. Many
enterprise-wide applications require very large, high-performance
databases. Transaction processing is a style of computing that
facilitates secure and formal changes to a database. Digital has
initiated a major program to enhance our capabilities to
state-of-the-art levels in this area.
Desktop systems. Digital has a very strong
position on the desktop with our terminalbased desktop systems.
Sales of our terminals are rapidly increasing, and terminal-based
systems will remain important to Digital for the foreseeable
future. In addition, Digital has a major program to create
leadership 32-bit desktop systems, which build on our existing
VAX, VMS and ULTRIX architecture strengths.
Enterprise-wide networking. To support our
leadership role in enterprise-wide networking, we are evolving the
DECnet protocols to the international standard OSI protocols. At
the current time, IBM’s SNA* is the de facto standard for the
limited kind of enterprisenetworking being done by most
customers. However, SNA is a closed, complex, IBM-controlled
architecture, with an hierarchical structure that doesn’t meet the
real needs of enterprise-wide networking. In contrast, the OSI
networking standards are relatively simple, are open for
multi-vendor attachment to the network and are peer-to-peer rather
than hierarchical. In other words, the OSI standards support the
style of computing that we think is appropriate.
We also are developing wide-area networking
infrastructure and product set, so we can truly say that we offer
wide-area networking solutions. In addition, we are developing
network management tools, to help our customers manage large,
complex networks.
We should keep in mind that DECnet networks are
fundamentally easier to manage than SNA networks. We don’t need
the level of tools, the level of people, or the level of
processing to manage a DECnet network that the SNA network
requires. IBM has been quite effective in marketing this
liability of their network as if it were an advantage. We should
be very sensitive to that when we face IBM on the enterprise
networking question.
Industry-standard operating systems. In recent
years there has been considerable industry and customer attention
to computer hardware and software standards. One area that’s a
particular interest to Digital is industry-standard operating
systems. This has shown up as a customer demand for Unix. This
demand results, in part, from a customer requirement for a
standard, vendor-independent operating system interface. Customers
would like to write an application once and then run it on
different vendors’ equipment. Customers also want to have access
to the many interesting applications that have already been
written for the Unix interface.
Digital wants to meet this demand for the Unix
interface. But we want to insure that the interface is a true
industry standard — one that has an industry-wide, as opposed to a
single-vendor-dominated process for defining the standard for
establishing and evolving the standards.
Digital’s operating system strategy is to
develop aggressively both VMS and ULTRIX software within the
framework of a common, single-system architecture. Currently this
common architecture includes the VAX hardware, the DECnet
communication architecture and the Application Integration
Architecture (AIA). We also recognize that VMS and ULTRIX
software have different strengths and weaknesses in their
capabilities and their available applications. Therefore, we will
market and sell even-handedly both VMS and ULTRIX software,
trying to match real customer requirements with the basic
capabilities of the respective systems.
In addition, we are working with multiple
vendors to insure that Unix interfaces are true industry standard
and not a proprietary interface of a single vendor.
We expect to see a large number of common
applications across both VMS and ULTRIX software. The fact that
Digital has the two most important computing environments, VMS and
Unix, and that we have common applications that run on both, will
be a very important message that is unique to Digital. There, of
course, will also be some applications that are specific to the
VMS or ULTRIX environments.
Integration of non-Digital architectures into
the network. Customers have major investments in the computing
architectures of multiple vendors and need to integrate all these
architectures into their enterprise-wide networks. In the past,
Digital has had numerous separate architecture integration
efforts. We recently expanded these efforts and linked them
together into a more formal overall program.
This program has been variously called Network
Application Services or Network Application Support (NAS). There
are three classes of network services: applications access;
business communications; and information and resource sharing. We
want to provide the ability to:
o access applications anywhere on the network
from any desktop on the network;
o communicate with any other user on the
network, using any of the styles of business communication, mail,
conferencing, videotex, and electronic document exchange; and
o share printers, documents, files, databases
located anywhere on the network from any appropriate application.
The NAS program provides these services for the
Digital architectures and key non-Digital architecture targets.
The non-Digital architecture targets, selected on the basis of
their importance to our customers, currently include:
o MS-DOS* and OS/2* targets because of their
importance in traditional business personal computing;
o the IBM SNA* environment because of corporate
data stored in that environment;
o the Cray supercomputing environment used for
high-end technical and scientific computing; and
o the Apple*/Macintosh* environment, which is
increasingly attractive for personal computing with a modem user
interface and ease of use.
Overall, Digital’s mission is to be the leading
supplier of enterprise-wide information systems and to assume, as
appropriate, full responsibility for that system. That
responsibility extends all the way from the basic engineering of
the system through integration, testing, installing, maintaining
and evolving that system. Not all customers will want Digital to
assume all these responsibilities at all times. But Digital’s
unique position and competitive advantage will result from its
broad capabilities. In addition, we have to ensure that we have
the necessary product and support architectures, so that this
mission is profitable for Digital and cost-effective for our
customers.
Our strategy can be summed up in three points:
o a desktop-to-data center family of computer
system products based on a single system architecture which
includes both VMS and ULTRIX operating system components, and
meets all the functional needs for enterprise wide computing;
o complete local and wide area networking
products built to the emerging OSI standards; and
o a comprehensive set of network application
services to build network business applications that include both
the Digital and key non-Digital computing architectures.
Digital today has a huge opportunity in
computing from the desktop. It is estimated that revenue
associated with this style of computing — desktop devices,
servers, networking
Digital today has a huge opportunity in
computing from the desktop. It is estimated that revenue
associated with this style of computing — desktop devices,
servers, networking and services — will make up over 40% of
worldwide computer revenue in 1991, almost $60 billion.
This market is changing very rapidly, due in
large part to our competitors’ efforts to move their low-end
products, primarily personal computers, up the capabilities
ladder. They’re trying to make their low-end systems handle more
of the mainstream computing of the enterprise. This will create
problems for their customers and opportunities for Digital.
Our competitors’ low-end technologies are
powerful enough to be used for what we would call "real work," but
there’s a catch. They require that customers buy new applications,
new operating systems and new communication approaches. Customers
are locked into expensive and time-consuming conversions or they
can scrap what they have. Customers who choose to convert may be
in for an unpleasant surprise. For example, an article in the
March 7 issue of COMPUTERWORLD indicates that it may costs as much
as $8000 to convert an IBM* PC-AT* to run OS/2, when it’s
available.
Compare that approach with ours. The same
messages we’ve used with such success in the mid-range and the
high-end apply to the emerging desktop market: one architecture,
with compatibility from the desktop to the data center. With this
strategy, we’re already positioned to be the winner in the desktop
battle of the 1990s.
In the 1990s, we believe most people will need
what we call "enterprise-wide" computing. This is an environment
in which all of the computing resources and information of a
corporation can be easily used by anyone to achieve the goals of
the enterprise. Enterprise-wide computing enables people to use
large amounts of information and easily communicate and share
information with other people, creating large gains in
productivity. Enterprise-wide computing is a change in computing
style that will be as important and as wide-ranging as the shift
from batch to timesharing, and we are positioned to take
advantage of it. We’ll be ready with the world’s best networking,
the most useful software environment and the best desktop
products. We intend that Digital will be the leader in
enterprise-wide computing, and the primary vendor of these
computing systems during the 1990s.
How will we achieve this ambitious goal? Here
is the one-minute version of our desktop strategy:
The large majority of users want to do simple
word processing, some electronic mail and perhaps some spreadsheet
work. They don't want to be burdened with file backups, changing
software or system maintenance. For these customers, we offer a
simple terminal. For customers who require graphics capabilities
to do presentations or to plot data, we pro
vide graphics terminals. For those
professionals, engineers, scientists and architects who require
graphics and local computing, we offer the VAXstation family. For
those people who already use and need MS-DOS applications, we
offer the VAXmate computer. And, for those customers who have
already invested in a desktop device from another vendor, we offer
the highest quality award-winning integration products in the
industry. We integrate these devices into a VAX/VMS environment
so users can communicate and share information throughout their
organizations. We intend to be the number one integrator and
service supplier of the major desktop devices.
Our desktop strategy is simply providing
customers with the best computing tools to get their jobs done. We
will implement our desktop strategy in two ways.
The first and most important approach is all
Digital. We’ll offer premiere enterprisewide* computing systems —
a pioneering approach to desktop computing built on our
strengths. It will be faster, richer and more productive than
anything else on the market. And, of course, we'll continue to
offer leadership terminals and workstation products.
Our second approach is a recognition of the
fact that many of our customers have already made a large
investment in PCs and other desktop approaches. We’ll bring their
islands of PC computing into our enterprise-wide computing
environment by offering Digital added- value networking and
application integration, along with ongoing Digital services. This
will help our customers become more productive by communicating
and sharing information effectively.
We’ll offer the desktop user the same
architectural application and communication advantages we offer
today in the mid-range and the high-end. The key to this is the
phrase "Digital’s Added Value." Many vendors will be selling PCs
and local area networks and some sort of server/client
arrangement. But only Digital will make the desktop a full member
of the enterprise-wide computing environment.
Our Network Application Support (NAS) strategy
is the key element of our effort to bring other vendors’ PCs into
our enterprise-wide computing environment. We’ve had a very good
year in the integrated personal computing business. VMS Services
for MS-DOS is the leading integration product in the market. It
has won numerous awards.
Our relationship with Apple has generated a lot
of publicity. We will do joint development with Apple. In the
context of our strategy, we are simply bringing Apple desktop
systems into our enterprise-wide computing environment. We’ve
already done it with IBM PCs. and later we'll announce it with a
variety of other PC clones. When IBM gets OS/2 organized, we’ll
offer support for that as well. The Apple announcement was just
one part of our approach to enterprise-wide computing.
For a great many customers, the terminal is
still the preferred desktop device. For certain applications it
may continue that way forever. It was only several years ago that
industry analysts were predicting the death of the terminal
market; that terminals would be replaced by personal computers.
Well, that didn’t happen. The plain truth is that many people are
perfectly satisfied using terminals. Terminals give them all the
access to computing they need.
We’ve had a great year in the terminal
business, perhaps our best ever. VT320, VT330 and VT340 shipments
all exceeded forecasts. We are shipping more than 500,000
terminals this year. The market for our terminals is growing at a
phenomenal rate. Our customers want leadership terminals and
that’s what we’ll give them. To continue our leadership, we’ll
continue our efforts to reduce prices, and we’ll enhance the
capabilities of our terminals so they can participate fully in the
enterprise-wide computing environment.
We’ve also had a great year with our
workstation products. This fiscal year, we became the fastest
growing workstation vendor in the industry, surpassing the growth
rate of Sun Microsystems. We leaped over Apollo, to become number
two in the workstation industry; and we’re heading toward the
number one position. We’ve expanded our workstation offerings by
introducing the lowest-priced workstation in the market and, at
the high end, a powerful real-time three-D graphics workstation.
Our workstations have been successful with
technical users, but we face a serious challenge if we want to
continue our success in the workstation market. This year the vast
majority of all the workstations sold run Unix, while 90% of the
VAXstation systems we’ve sold run VMS software. We are working
very hard now to increase our share of the Unix workstation
market.
We face another challenge in the workstation
market. There are seven times more commercial computing users
than technical users. We’ve done well with the technical users;
and, of course, we want to continue our success in the technical
markets. Our challenge is to keep our loyal technical users, while
at the same time winning commercial desktop business. We will do
this by offering a computing environment that merges the power of
workstations with a new generation of personal product
applications and greatly improves ease of use in system
management. We will deliver the Digital Application Integration
Architecture to the desktop.
The new desktop computing environment
represents a fundamental change in the way we must think about
computers. Computer hardware used to be a scarce and expensive
resource. We had to allocate it carefully and share it. We had to
change the way people worked in order to get the most benefit from
our expensive computing resources. Now computing is relatively
inexpensive. It’s the human resource that’s scarce. We can put
processing power and huge amounts of storage on the desk at a very
low price, but you can only stretch that highly trained engineer
or commodities trader so far.
How do we solve this problem? We use the
availability of computing resources to help stretch the scarce
resource — the person. We create a computing environment that is
so powerful and so easy to use that the person becomes much much
more productive. We make it possible for the person to share and
interact with information easily and instantaneously. The
computing environment helps the person to work naturally and
efficiently almost without regard for the mechanics of computing.
It means that all the work of computing - translations, running
applications, exchanging files, loading operating systems, etc. —
is going on elsewhere. The users do their jobs, and the computing
environment helps them, acting as a silent partner. When you
dedicate more power to each user and share important applications
through a fast transparent network, the usual obstacles to
productive computing disappear.
In this new computing environment, users will
do word processing where on-screen fonts look exactly like the
printed page. They will work in a completely integrated
application environment. They will use applications under VMS,
Unix and MS-DOS — the three most popular operating systems.
They’ll do this simultaneously, and they won’t need to know which
operating system they’re using. Users will do real work. They
won’t need to spend any time thinking about operating systems,
interfaces or file transfers. Systems management will become a
non-issue as far as the user is concerned.
In this computing environment, resources out on
the network are exactly like resources at the desktop. The user
won’t even know what’s running locally and what’s running
remotely.
This computing style is a long range strategy.
We’ll continue to enhance the capabilities of the system. As
independent software vendors bring more of their applications into
this environment, as DECwindows becomes a de facto standard, and
as the cost of processing power continues to drop, the environment
will become that much more productive.
To conclude, I want to leave you with three key
points. First, we have a winning desktop strategy for the 1990s.
It builds on our strengths and expands our offerings in ways that
will help our customers succeed in their enterprise. Second, our
desktop products, most importantly the VAX computing environment,
will help customers save money and run their businesses more
efficiently. Third, our desktop computing products will be easy to
sell, easy to install and easy to use.
DECwindows is a company-wide program to enhance
all of our VAX/VMS and VAX/ULTRIX inter active applications, to
exploit VAX workstations and provide the same easy-to-use visually
sophisticated user interface. It is also a good example of the
work that the Applications Integration Architecture Program is
doing for us.
Today’s customers have to deal with a very
complicated network and need many computing systems to interact in
that network. Our answer to that customer problem is Digital’s
Network Application Support (NAS) — the highest quality
application environment in the industry. The Application
Integration Architecture (AIA) is a set of software standards that
help unify that environment.
The primary purpose of the AIA program is to
provide a management and technical forum that focuses our efforts
and insures the highest possible quality of our software
standards. A quality software architecture tracks and anticipates
changes in technology, while at the same time ensuring a level of
stability in the Digital product set that customers can depend on
as they make their computing system investments.
We want every VMS and ULTRIX workstation to
bear a strong resemblance to one other. They should look and feel
the same and have identical applications for all day-to-day needs.
To achieve this goal we organized an engineering and marketing
effort called the DECwindows program and outlined a four-part
workstation software strategy: o provide a common user interface,
o use a common programming interface, o use industry standards,
and o deliver it fast.
The result of this strategy will be a family of
DECwindows workstations.
The first and most important of these is our
common user interface, the DECwindows environment, which will
give our products the same look and feel. As a result, users will
be able to move from one application to another, or from one
workstation to another, without being retrained.
To be successful in the workstation market, we
need as many applications as possible; thus the second element of
our strategy is aimed at attracting application designers. We have
to make programming easy for them. Therefore, we provide a single
application programming interface for both VMS and ULTRIX
software. This programming interface provides all the software
tools and mechanisms that are necessary for designers to construct
applications which have our common look and feel.
Saying that the DECwindows programming
interface is the same for VMS and ULTRIX software is terrific, but
saying that it’s based on an acknowledged industry standard is
even better. We picked a standard called "X Windows," which
emerged from our collaborative work at MIT with other computer
vendors at Project Athena. Frankly, we picked the winner. X
Windows is clearly the standard in the industry for windowing and
graphics on workstations. Today the specifications for X Windows
are managed by an MIT-sponsored group called the X Windows
Consortium, of which we’re a member, along with 23 other prominent
computer vendors. As with the strategy for a common programming
interface, this element of the strategy — using industry standards
— is aimed directly at attracting application developers.
The fourth element of our strategy is quick
delivery. To make sure that our DECwindows plans work to our
advantage, we told our customers and application designers our
intentions so that they could prepare. Their initial reactions
were overwhelmingly positive, but they need quick delivery. We
immediately formed a DECwindows program team, with a coordinating
manager and a lead technical architect. This team works with all
the engineering groups, the marketing groups and Field teams to
coordinate the entire program.
Applications are the key to success in the
DECwindows program, so we set the highest priority on the needs of
the application developers who actually write the software. We
decided to make the first release a benchmark for functional
capability and performance, and we gave ourselves twelve months to
deliver the initial software to the developers. Once the
applications were ready, we could ship the product to the end
user. We also decided that if workstations were going to be
commonplace in Digital networks, then the DECwindows environment
had to be a standard, basic component of VMS and ULTRIX software.
The DECwindows programming interface provides a
set of software tools, which make it easy for an application to
display all the ingredients of our DECwindows look and feel. These
tools supply pop-up menus, boxes on the screen to capture user
dialogue, clip boards for moving data between applications, scroll
bars, text and so on. In addition, our programming interface
exploits our workstation hardware to deliver very high performance
graphics. An application can use and intermix a variety of
programming interfaces, ranging from industry standard graphics
libraries, all the way down to some very primitive functions for
high speed drawing, text display and color support. Each of these
capabilities represents something different in the DECwindows
system, which plays an important role in supporting our visual
user interface.
The most unique aspect of the DECwindows
architecture is that it works across a network, and the user’s
display need not be on the same computer as the application. The
DECwindows approach provides a way for users to separate
application execution from application display. A DECwindows
application can run on a VMS or ULTRIX application engine using a
variety of display devices spread throughout the DECnet OS I
network, including our own DECwindows workstation products or
industry standard PCs and workstations. Users most often will
choose to run their application on the same computer as their
display; and, in fact, the DECwindows environment is optimized for
this case. The user has the freedom to run an application on a
machine that has more computing resources than the office
workstation without giving up the benefits of the common user
interface. For example, a customer who has a graphics
application, which requires access to a large data base, can run
the application on the machine with the data base and run the
display on a local workstation.
DECwindows applications have to work for all
customers in our worldwide marketplace. We’ve designed a very
simple and easy way for DECwindows applications to come in
different natural languages. We’ve separated the program of an
application from the description of what it looks like on the
screen. Then we supply a special editor which allows you to
replicate that description in various languages, such as English,
French, German, Italian, Spanish, Dutch, Japanese and so forth. We
ship all of the replicates to the customer. When users log on,
they can select the language in which they’d like to see all their
applications appear. This is a very powerful and easy way to
insure that we have a completely international product.
When you sit down at a DECwindows workstation,
the first thing you’ll notice is that the user interface allows
you to simultaneously see the display of many applications. Each
window on the screen represents an application at work somewhere
in the network. The windows can be positioned to overlap like
pieces of paper on your desk, or they can be positioned like tiles
in a mosaic. You set up the screen, and you direct the action.
Many people are already using DECwindows
software today. We have over 50 internal software projects
implementing DECwindows applications and hundreds of active users.
Many people are working to convert our workstation and interactive
applications, including software development tools, ALL-IN-1,
network management tools and electronic publishing software to the
DECwindows environment. We also have a whole new set of DECwindows
applications.
In addition, we have an extensive program to
support outside application developers. We’ve already trained
hundreds of people to use the DECwindows interface and develop
programs for it, and we’re training more each month. These
designers are now converting their existing applications and
developing new visually sophisticated applications to exploit
DECwindows workstations. These developers are interested because
the DECwindows approach is new and wonderful. But, they’re also
interested because it is based on the X Windows industry standard.
This means that they can leverage their DECwindows development
investment on other platforms that supply the X Windows standard.
In fact, we’ve even had interest from competitors in adopting our
extensions to the X Windows standards as their software standard.
DECwindows is a product we can all be proud of.
It’s an important new Digital software architecture. It’s a design
center for world-class interactive applications. It’s a superior
implementation of the X Windows standard. In summary, it’s a key
opportunity and competitive advantage for Digital.
To describe what a transaction processing
system is and how it works, I’ll use the analogy of a metropolitan
library system, with a central library and a number of branches.
The total system includes the libraries themselves, the books, the
roads that connect the libraries, vans that go back and forth
among the libraries, and people. People work in the various
libraries, and people borrow and return books. In addition to
those things and people, the relationships among them are also
important. For instance, a specific road connects a branch library
with the central library or a specific person is now trying to
borrow a specific book at a specific branch. Those objects and
their relationships constitute the data base, which is an integral
part of any TP system.
In addition to the "things" and their
relationships, thousands of changes and actions are occuring all
the time. For instance, vans are transporting books, and people
are coming to various branches to borrow and return books. All
those "actions" can be viewed as transactions in a transaction
processing system. Hundreds and thousands of those actions are
occurring all the time. We should note that most of the actions
are isolated from each other.
Let’s zoom in on one of the specific actions: a
customer wants to borrow a specific book. The customer in this
case is the TP application. The librarian wants to service that
request and can do so in any one of different ways. If the book is
in this library, the librarian simply comes up with the book. If
the book is not in this library, the librarian will issue an
inter-library request, send a van to get the book from another
library, and give it to the customer. Once the librarian has the
book and is about to hand it to the customer, something important
happens: it is documented that the book has been borrowed on the
library card. Think of that documentation as a contractual
commitment, as the completion of this particular action. In a
crude way, this scenario illustrates what constitutes a specific
transaction.
We can consider the collection of librarians at
each branch as analogous to the transaction monitor or
transaction server in computerized transaction processing. The
whole picture - the data base, the transaction server and the
applications - constitutes the transaction processing system.
Most transaction processing systems in the past
and most in existence today use "flat files" rather than data
bases for storing information. The reason for this is primarily
one of performance. Data base systems have been around for a long
time, but they weren't able to fill the performance requirements
of transaction processing systems. We are changing that.
Flat files will remain faster; and, in some
situations where the performance needs are extra high, we may
still want to offer flat file systems. But there is definitely a
trend in the industry toward using data bases for storing the
information in transaction processing systems.
We use TP systems in our real-world, day-to-day
activities. They become critical components of our lives. For
example, transaction processing systems often take care of the
reservations and the cargo systems for airlines. They are key
parts of the airlines’ operations. If the TP system stops, the
planes are grounded. The same thing happens with banks. There have
been examples where the TP system failed and the consequences were
disastrous. Several years ago, the TP system of a bank in New York
was erroneuosly handing out billions of dollars; and for a period
of three days the bank had to borrow money to cover the shortage.
Since these TP systems become part of our
real-world activities, they can’t be slow. They have to work as
fast as the real world works. There’s heavy demand for fast
response times. Typically, about 90-95% of the actions have to be
finished in one to two seconds. That means that data bases are in
a total state of change. They’re updated in near real time. At any
one time, they represent that particular part of the business or
part of the real world that they’re implementing. Because they can
be a critical part of a company, we refer to them as "bet your
business" systems.
For example, look at money. Back in history,
gold and silver coins represented value. Later, we abstracted that
to paper money. Now, money has become electrical impulses on the
data base. It’s very easy to see that if the data base loses its
data, you lose money.
Consider your personal checking account. When
it comes to your bank balance, you don’t accept 99% accuracy. It
has to be 100% accurate all the time.
If there is one common theme in the transaction
processing market, it is that mistakes are not accepted. Period!
In computer language the term is "data integrity."
Now we know that computers are computers, and
software is software. We blow bits. We have bugs. Mistakes do
happen. So, what do we do about that?
We translate that into: errors happen, but they
can’t go undetected. We’re building better hardware and better
software that help us to find the errors when they do occur. And
we have a lot of computer solutions to help solve the data
integrity problem. You hear about recovery, shadowing, concurrency
control, etc. I’m not going to go into those in detail. But I do
want to give you a visual image of a TP system at work.
Imagine a data base that has good, accurate
information. A transaction comes in at time TI, and some kind of a
change happens to the data base. You could say that the data base
has changed its state; all we did was to update some information
in the data base. Another transaction comes in later at time T2,
and the data base gets updated again. Then something bad happens —
a big error of some kind. It might be an overflow of a variable in
the program of it could be that lightning struck the system. The
question is, what do you do when the error happens?
We can’t let mistakes go through. This is where
the transaction processing system has an advantage over real-world
situations. We can stop the transaction processing system, rewind
time and replay back to the point just prior to where the mistake
occurred. So, we go back to a place and time where we knew the
data was accurate. We also have put these transactions on
something like a log, to remember them. Once we go back, we are
able to replay the transactions against the data base and make
sure we don’t have that error happen again. Then we can continue
on our merry way. Now, obviously to have these errors occur, to
rewind time and to replay it, is bad news, because the real world
is halted while you’re doing this. The planes are grounded. The
money is stuck on the data base. This replay can take anywhere
from seconds to minutes to hours to days, depending on the
complexity of the transaction processing system and the error that
occurred. This is the basis for the very strong requirement that
transaction processing systems must be highly available — with
down times on the order of minutes per month or even year, on
systems that are running 24 hours a day, seven days a week.
Basically, the picture of a transaction
processing system is relatively simple and easy to imagine. The
complexity comes in making sure we have the integrity, that we
have the performance and the capacity to capture a major share of
that market, which, it is estimated, should be worth about $60
billion in 1991.
Transaction processing differs from timesharing
in style and in metrics. In a timesharing system, there are often
many users, but each user typically has his or her own data base
and own application. In a transaction processing system there are
also many users, but the data are not owned by an individual. The
data are common to everyone, and the applications are typically
very few — for example, buying and selling stock.
The metrics are different too. We’re familiar
with MIPS (millions of instructions per second), VUPS (VAX unit of
processing — referenced to a VAX-11/780), cost/performance,
dollars per MIPS and the number of users that a system can have in
timesharing. In transaction processing the key metrics are
transactions per second, dollars per transaction per second and
response time.
Transaction processing implies
"bet-your-business" applications. When you realize that people are
betting their livelihood on your system, you take a different
approach to the way you design it.
The marketplace is big - about $26 billion -
and it’s growing at about 20 to 30% per year. By 1991, we expect
it to be about $60 billion.
We approach transaction processing, as we do
other complex problems, with an architecture. An architecture
defines interfaces, which are what makes things compatible, but
not the underlying implementation. Therefore, it forms a basis for
product development strategy and for the way groups interact to
implement the strategy.
The Digital Distributed Transaction
Architecture complements and extends our VAX, VMS, DECnet and
other current architectures. It consists of two parts — a front
end and a back end. The front end provides the interface to the
user. Often, in transaction processing (TP) systems, the
interface to the user consists of the forms that users fill out to
enter data. It also handles interfacing the forms to the rest of
the system. The rest of the system consists of a transaction
server and the resource manager. The transaction server (sometimes
known as a "transaction monitor") deals with the information in
the application itself. It queues requests from multiple users so
that they will be dealt with in order. It also interfaces to the
resource manager — the database system.
We’ve defined the TP system architecturally and
have defined interfaces between the key pieces. That allows us to
split those pieces off and implement them wherever we like. For
example, we might split the front end of the two pieces and put
them on separate processors, such as MicroVAX computers. Or, in an
application that involves heavy interaction with the forms
system, we can put the forms processing next to the user. Now that
does a couple of things for us. That approach
off-loads the main system and makes the communication between the
front end and the main system fairly compact. It also cuts down
the communication costs.
I’m sure that you have all heard horror stories
in the industry about the cost of downtime. The only way we’re
going to get a system to be reliable and available is through use
of redundancy.
There are two ways of using redundancy. First,
there is redundancy in time. In that case, you roll back to the
time before the error and correct things. We do that in VAX-
cluster systems. If a node goes out, we configure the cluster
around the failed node. That solution is hardware and software
based, and it takes time. There is also redundancy in space. In
that case, you use more space for the computer. You use more
parts; so if one of them fails, the others can carry on with the
task.
Digital’s Distributed Transaction Architecture
provides a number of advantages. It’s fully distributed in the
network. It puts processing power where it is needed. If your
application requires multiple database activity per transaction,
you put your power at the database side. If you have complicated
forms processing, you put your power in the forms at the front
end. It also minimizes communications overhead.
Let’s take a look at the way TP systems are
implemented today and give you a glimpse of where we’re going in
the future. Today, they’re typically layered on top of the
hardware in the operating system in the database. The pieces
required to satisfy the most stringent requirements of a TP
system typically reside in the TP monitor.
Because transaction processing is a business
where people are betting their business it requires additional
focus on the kind of support that we provide our customers. There
are pieces in place to do that. We have a design consulting group
within the TP systems engineering organization to help design
large TP systems. We are disseminating information to the
Applications Centers for Technology in demos. We’re developing
competency centers. We have training programs, such as TP
University, under way right now, training people to be real
experts who can train other experts.
We have application development environments
that fit the customer style of developing applications. We have
the lowest application development time. We have the best
distributed transaction processing system. We can integrate
things that other people don’t even think about. We can integrate
transaction processing, office, decision support and timesharing
in a single environment. We have the widest range of systems
without changes in the software. And we have competitive cost of
ownership.
A recent survey done by DATAMATION asked
customers what they expect to do in transaction processing over
the next year. The survey says that today 1.3% of their systems
are from Digital, and by the end of next year they expect 10.4% of
their systems will be from Digital. They seem to want us.
In summary, the TP opportunity is now. We have
market demand. We have products. We have performance leadership.
We have price leadership. We got here with the hard work of a lot
of people. Multiple groups have put together what we have in
transaction processing in record-breaking time.
I’d like you to leave today with the following
message: Digital is a leader in transaction processing with
enterprise-wide networking and its family of VAXcluster systems.
To do that I’ll be covering four basic topics. These are:
o the importance of VAXcluster systems in
Digital’s Product Strategy,
o attributes of VAXcluster systems,
o the role of VAXcluster systems in our
transaction processing strategy and offerings, and
o the evolution of VAXcluster systems and their
impact on Digital’s Distributed Transaction Processing
Architecture.
VAXcluster systems provide flexibility,
availability and expandability. They give the customer the ability
to configure a system that has no single point of failure and,
hence, the ability to meet the most stringent requirements of
their business critical applications. For example, VAXclusters
allow the customer to design a system that has multiple copies of
a data base transparently, and updates are executed to all copies
simultaneously via shadowing. In a similar fashion, the customer
can design a system with n + 1 redundancy of processors, each
processor executing part of the overall task; and in the event
that there’s a failure in one of the processing elements the task
can be shifted to the spare processor.
With regard to expandability, a customer
purchases a VAX 8700 computer to solve today’s problems, and, as
the business grows, can add more computing resources, storage
resources, or hierarchical storage controller resources to meet
changing application needs. In effect, customers are not
restricted to the set of resources purchased on day one, but
rather can expand their computing environment, protecting their
investment over time. Our customers say that this is an important
competitive advantage.
VAXcluster systems are the vehicle that Digital
has used in the past and will continue to use in the future to
build highly available systems. Today, we have an installed base
of about 6600 VAXcluster systems worldwide. About 16% of all VAX
nodes are in clusters. About 64% of our top 200 accounts have
VAXcluster systems in their installations.
The key attributes that VAXcluster systems
provide that are important in transaction processing are high
availability and data integrity. In simple terms, high
availability means that when you have one opportunity to access a
critical application or a critical piece of data, you want that
application or that data to be operational. In order to accomplish
this, and given the fact that failures are an inevitable state of
nature (Murphy’s Law), a system can be designed to guarantee that
applications or data be operational when needed. Typically, such
systems are designed in such a way that the time that the system
is down is minimized as much as possible. This involves,
naturally, the use of redundancy, checking mechanisms to detect
and isolate failures, and protocols that allow for the re-start of
applications.
One such approach is known as "redundancy in
time," which involves the execution of different actions on
multiple components. Upon the detection of a failure the execution
of the action stops, the problem is isolated, and fail over to
another component that can do the same application. For example,
say you have an airline terminal reservation system, consisting of
two processors. One is handling flight 44 to San Francisco;
another is handling flight 41 to Miami. If there’s a failure in
securing a seat on flight 44 to San Francisco, you detect the
failure, isolate it, and fail over the handling of that
particular booking to the other computer. Then you restart the
application and complete the transaction.
To differentiate Digital’s offerings from those
of the competition, I focus on four key attributes: availability,
data integrity, service and price/performance.
We offer conventional systems, such as a VAX
8700 standalone computer. The downtime of that system can be
measured in tens of hours per year. If you have a failure of a
hardware component, you call Field Service, and it may be two
hours before you can replace that component.
Our high availability VAXcluster offering
reduces that time significantly. With specific configurations, we
can reduce the downtime in a VAXcluster system to just minutes per
year. We do that with redundant components, such as multiple
processors.
We build our high availability offering around
our conventional systems and guarantee that they provide higher
levels of data integrity. For example, we provide a reliable
interprocess communication mechanism.
In terms of cost, of course, the conventional
system is the least expensive. A high availability system, because
of the built-in redundancy, always increases the cost.
In the 1990s and beyond, with our Distributed
Transaction Processing Architecture, we will separate the basic
functions associated with the execution of every transaction, that
is, stimulus capture (front end), application execution and data
base access (back end), and associate as much computing resources
as possible with each element of the transaction. To understand
the differences between stimulus capture, application running, and
data base access I will use an example. Consider the typical
purchase in a department store of a videocassette recorder (VCR).
This simple event (transaction) invokes all of the elements of a
Transaction Processing system. First there is the stimulus
capture, this could be "forms'' manipulation, that is, the clerk
interacting with a video monitor on a prescribed form where the
number and price of the unit is entered. On the other hand, the
stimulus capture can be of a different nature on other
applications, like credit card verification, where the stimulus is
simply running the credit card through a credit verification
machine.
The second step in the transaction involves the
running of the application. In the example above, most likely the
application is a warehouse and inventory control application that
will make sure that the desired article (videocassette recorder of
the particular brand and model) is available.
Lastly the execution of the transaction will
involve a data base access, namely modifying the data base to
reflect the purchase that just occurred.
Every one of the steps in this transaction will
be distributed in Digital’s Transaction Processing architecture.
This distribution of functions to computing resources will
provide us with a competitive advantage.
In 1988, VAXcluster systems provide high
availability and data integrity for all functions. That is, the
VAXcluster system does the whole job: stimulus capture,
application execution and database access. In future years,
VAXcluster systems will execute the backend functions only;
therefore, supporting high availability; and data integrity for
the application execution and database access.
Our ability to distribute stimulus capture,
transaction servers and data base access will allow us to cover a
much wider range in performance going from the low end of 16
transactions per second to the hundreds of transactions per
second.
In summary, high availability systems are
important in the transaction processing environment for three
reasons:
o They meet the most stringent requirements of
our customers - downtime of less than three minutes per year.
o They insure data integrity.
o And they allow you to have a large range of
performance.
Our customers are trying to manage the same
kinds of environments that we are in our industry. They’re trying
to manage change, to become more competitive and to do a better
job at servicing their customers. They’re trying to improve their
quality and to reduce their costs. They’re trying to be leaders in
their technology. In some cases, they’re going through radical
shifts in technology, trying to get high quality products and
services to the market much faster. Many of them are dealing with
de-regulation or re-regulation.
When organizations recognize they need to
change, typically, their first inclination is to change the
organization or change the people. The products and the services
are more difficult to change. And the information systems that are
used to run the business are very difficult to change. And all of
these elements are interrelated.
The information systems approach that customers
are using today in an IBM environment can help set the stage for
describing our advantages.
Mission critical, transaction processing
production systems are "you bet your business" applications.
Examples include reservations systems, inventory control systems,
funds transfer, a lot of banking applications, order processing
and billing. These systems have to be efficient, flexible,
reliable, and highly available. Historically, batched production
systems were primarily focused on accounting. They were used to
measure the business after the fact. What our customers want now
is on-line transaction processing systems that still take care of
the accounting, but that also allow them to run their businesses
more effectively.
Some of us remember the IBM environment of the
1960s. We remember standing in line with our card decks trying to
get computations done, only to find out, a day or two later, that
we made one simple syntax error. We also went into the computer
room between 2 and 5 AM to get computer time to do testing and
debugging because the rest of the time the system was running
production, and there was no way that the MIS department was going
to upset that.
The mainframe computing environment hasn’t
changed much since then. The punch cards have been replaced by
terminals, but the act of writing programs and compiling is still
very much a batch process. There are batched compiler queues that
always get last priority on a mainframe, after production. The
Finance Department, which usually runs the MIS Department, is
focused on making overhead people more productive. Hence, the
focus is on accounting for the business after the fact.
We know our customers want distributed
responsive systems and want to put computing out on the front
lines of their organizations, but this 1960s environment is where
they’ve come from, and change comes hard.
Their present environment is very costly. It’s
a model that worked well in the 1960s, when computers were very
expensive and networking was relatively inexpensive. But these
mainframes are very complicated to operate, and require large
staffs of operators and system programmers to keep all of the
software working. The communication costs are very expensive and
increasing. Growth is very expensive and disruptive. When you want
to get more computing power, you have to swap a CPU or add more
CPUs; and the cost of adding them is very high because they have
to be added in large increments. It really gets messy when you add
the number of data centers that a large company typically has:
multiple mainframes, hundreds and even thousands of terminals,
and all the standalone PCs that employees have purchased because
they got tired of working with that difficult computing
environment.
Because the mainframes aren’t easily
accessible, users bought PCs or terminals connected to
minicomputers. IBM’s TP software doesn’t work on PCs or on
workstations. It only works on the mainframe. Therefore, people
who want to do end-user computing and transaction processing need
two sets of user devices and two networks, at the minimum.
Now, the economics of computing are also
changing. The cost of putting the computing resources closer to
end users is becoming lower and lower. The costs of networking are
becoming higher and higher. So the mainframe terminal-network
model is becoming economically as well as technologically
obsolete. That’s why the mainframe business is not growing.
For example, consider a manufacturing company
that has separate data bases, multiple networks, isolated local
area networks, PCs and workstations, limited-function terminals,
and huge investment in old software and training. By separating
its computing resources, that company is not tied together; rather
it is split. Imagine what happens when such a company wants to
make dramatic changes in the way it does business. Their
information systems often limit how quickly they can change.
Contrast that picture to Digital’s internal
Easynet network, and the way we conduct our business worldwide. We
have one network around the globe that supports 445 locations in
32 countries. We have over 28,000 computers and 80,000 users on
that network. The same network is used for end-user computing,
production computing and software development. We do our order
processing, inventory management, accounting, and mail on the same
network. We have integrated data bases, so users can get the
information they need about orders, shipments, revenues, etc. And
it's all one uniform network. The system doesn’t care whether
we’re doing TP or end-user computing or software development.
When our customers see what we have, they say
that’s the direction they want.
Customers really want to concentrate on their
businesses and not on their computers and networks. They’d like
systems that change easily as their business conditions change.
They’d like highly productive systems that allow them to focus on
and change their business environment. They want systems and
networking capabilities that will help them tie
their organizations together. They want
investment protection. They want to know that we’re going stick
with them for the long term. They want a wide range of
applications, but they have limited resources for development. So
they want to acquire applications that have already developed.
They want to know that service and support is available when they
need it.
Other computer companies such as IBM, Hewlett
Packard and Tandem also claim to have these capabilities.
IBM has the major marketshare in on-line and
batch production systems. Their "on-line" style means remote
terminals connected over telephone lines to mainframes. They have
fast, batch TP and build large systems that can handle large data
bases. They also have the ability to handle very complex systems
projects, both inside their company and for their customers. They
can provide the customer support that’s required, when it’s
required and where it’s required.
Their problem is that they have multiple
operating systems, multiple TP systems, multiple data base
systems, multiple networks and poor software development tools.
The cost of ownership — from the standpoint of people, equipment,
software or facilities management — is higher than anybody else in
the industry. And IBM does not have any distributed TP products
today.
Hewlett Packard has a very loyal customer base
that has stuck with them through the years. They do a good job of
customer support. They have an image of supporting industry
standards. They recently brought out a new line of RISC
architecture computers with good price/performance. They are
selling those machines into the laboratory and science markets
where they’ve always been a strong competitor of ours. And they’ve
told the world that they’re going to be a factor in TP as well.
But Hewlett Packard has four operating systems
that they are trying to support. They emphasize Unix, but there
are no TP applications on Unix today. They have not developed a
range of enterprise-wide solutions, and their networking
capability is limited. They lack expandability, and they also lack
the equivalent of clusters or symmetric multiprocessing.
Tandem more than any other company has been
recognized as a leader in TP. They’ve done a magnificent job of
marketing. They offer low-cost, high-performance, expandable TP
systems, and have a good distributed relational database
technology. They also do well at IBM interconnect: so their
systems are often used as front-end systems to IBM mainframes.
But Tandem has a proprietary development
language and operating system, and is narrowly focused on TP.
Except for connections to IBM, they have a poor multi-vendor
network approach, and application development tools are also
lacking.
How do we win? We will win with our single
well-integrated architecture, our networks and network management
capability, our distributed TP and data base products, and our
efficient application development tools; we will win with the
interoperability that we offer among desktop, mid-range,
departmental and data center computing. We have the broadest range
of applications in the industry. We have a commitment to industry
standards and open systems. We’re the only company that can
deliver all of this today.
We have a simple goal: to be number one in
distributed production systems and to tie those systems together
well with our end-user computing environment and our software
development environment; to work well with the mainframes and
central computing resources customers have today; and, eventually,
to replace those mainframes as our systems become better at
handling the large jobs.
* Unix is a trademark of
AT&T Bell Laboratories.
* IBM, PC-AT, OS/2 and SNA
are trademarks of International Business Machines Corporation.
* MS-DOS is a trademark of
Microsoft Corporation.
* Apple and Macintosh are
trademarks of Apple Computer, Inc.
A summary of Gary Eichhorn’s State of the
Company presentation on "Research as a Market — a Window of
Opportunity" will appear in the next issue of MGMT MEMO.