"The UNIX Value Proposition"
Keynote address by Lew Platt, Hewlett-Packard chairman, CEO and president to Uniforum '96, February 14, 1996
(Visual #1) Title slide
Good morning, fellow UNIX lovers.
I don't use that term just because it's Valentine's Day, but because UNIX is near and dear to my heart -- and to that of the Hewlett-Packard Company.
It's impossible not to feel strong allegiance to a technology that has spearheaded the open systems movement.
Despite my long love affair with UNIX, I must admit to feeling somewhat inadequate as I address you this morning.
Because, you see, when Richard Jaross called a few weeks ago to discuss this speech, he led me to believe he had some implicit expectations.
He described with great nostalgia the uproar Andy Grove created at Uniforum two years ago, when he challenged the UNIX community to get its act together -- and the equally big uproar Ed McCracken caused last year over the issue of changing Uniforum's name.
As Richard described these episodes with an obvious fondness, I got the sense that he was hoping I'd say something equally controversial.
I was worried I'd let you down, until my salvation arrived ... in the form of Scott McNealy, when he agreed to do tomorrow's keynote.
So the pressure's off me. If you want outrageous, wait until tomorrow. Scott's sure to oblige.
Seriously, I really like Scott. I think he runs a great company -- as does Ed McCracken . And the fact that we can respect and tease one another -- and both collaborate and compete -- is part of the UNIX community's unique character.
Today I'd like to discuss that character -- in the form of four value propositions UNIX has represented to the many customers who have embraced it.
I don't intend to talk a lot about HP and any of the whiz-bang technologies under development in our labs. I'm not here to promote HP, but to promote UNIX. I want to speak as a member of the UNIX community.
To the customers here today who have pioneered open systems computing by moving to UNIX, I want to thank you for the faith you have demonstrated in this technology.
We at HP know we must earn that faith on a daily basis by helping you use open systems technology to achieve your business results.
To the UNIX suppliers and developers here today, I'd like to present a challenge -- which is as close to outrageous as I'll get.
To my UNIX colleagues, I'd like to suggest that if we want the future to be as exciting as the past decade, we must strengthen our focus on customer needs. Our challenge is to find new ways of delivering on the value propositions that have made UNIX as successful as it is today.
Host/Server Revenue by OS
But first let's recognize how far we've come. I had cause to appreciate our progress when, the other day, I looked at the first Uniforum keynote address that I delivered back in 1988.
The title of my speech was: "Maintaining the Momentum: Can UNIX Make It?" Well, in the intervening eight years, the answer has become unequivocally clear -- "Yes! UNIX can make it. Indeed, it has made it -- reaching a degree of success few of us would have imagined not so very long ago."
In fact, looking back at my 1988 speech in preparation for this one, I was struck by how modest our aspirations were. Because in those remarks, I cited predictions that UNIX systems sales would represent 13 percent of the total computer market by 1991.
Well, we badly underestimated the appeal of UNIX. By 1991, UNIX already represented 21 percent of system sales, according to IDC.
And here you see the growth rate of UNIX since then, compared to other leading operating systems. It's a record to be proud of.
Quite simply, UNIX has succeeded because it has offered real value to customers. Let's take a moment and remind ourselves of the value propositions UNIX has offered.
First, UNIX has stood for technical excellence ... the rich combination of multitasking and functionality needed to perform complex technical tasks.
UNIX can handle applications that consume multi-terabytes of data ... such as the weather forecasting application shown here.
Its ability to scale upwards provides the performance, security and high availability that enable UNIX systems to rival mainframe computers in mission-critical roles.
In fact, UNIX has not only rivaled mainframe performance. It has done so while offering a better price point. Here you see a graph of the cost ratio of various mainframe technologies versus high-end UNIX.
A ratio of one represents price parity. And you can see that, even in 1998, it looks like the mainframe will be 50 percent more expensive for a given performance level.
Back in 1988, we thought UNIX would find its niche in technical computing. It has gone much further than that; UNIX is running key business processes in enterprises around the world -- ranging from Boeing, to Federal Express to the Singapore Stock Exchange.
In short, UNIX took advanced technology out of the research environment and brought it to the manufacturing floor ... to the data center ... and to the retail outlet.
And along the way, you -- the UNIX community -- spun off a number of significant technologies that have furthered the information revolution.
For example, without UNIX and its advocates, there would have been no internet, ... no World Wide Web, ... no glimpses of the evolving information superhighway and all that it will make possible.
UNIX has also stood for choice.
It runs on more microprocessor architectures than any other operating system.
Its scalability is unparalleled -- allowing it to run on anything from a laptop computer to a supercomputer.
UNIX has provided portability and compatibility. Today, for example, applications written to the UNIX '95 specification need only be recompiled to run unchanged on either a DEC or HP system.
And the interoperability UNIX provides has made it possible for customers to select a variety of systems and still have that heterogeneous environment operate effectively.
The degree of choice UNIX has provided has ushered in new rules of competition for computer vendors.
Customers gained new power, since they were no longer "locked in" by proprietary architectures.
And, of course, with that increased customer power came lower prices, which, in term, spurred more demand.
Adaptability has also characterized UNIX. Its stable set of published software interfaces has made it possible for customers and ISVs to tune the implementation of the operating system to meet their specific needs, without affecting the application interface.
This concept -- that standardized interfaces encourage innovative implementations -- has been one of UNIX's most important contributions to the world of computing.
The easiest way to represent this truth is an analogy you may have heard before -- the electric plug.
It defines the connection between the electric utility and the appliances that plug into it. That standard interface has made possible the development of electric razors, yogurt makers, VCRs -- you name it. And those appliances don't care if the electricity was generated from nuclear fusion or gerbils on a treadmill.
If we tried to draw the computing equivalent of this concept, the role of the plug would be played by portability interfaces like UNIX '95 and interoperability interfaces such as X.25, 802.3 or TCP/IP. And, as with the plug, these standards have made it possible to innovate on either side of the interface.
The inherent adaptability of UNIX allows customers to partition the operating system, application, database management and user interface in a way that best supports their businesses.
Features within the OS have made it possible to develop applications such as HP OpenView to manage large, distributed, heterogeneous networks.
And technologies built on top of UNIX, such as NFS, have made it much easier for users to get files over a network -- and for network administrators to manage the files on that network.
In short, UNIX has made possible distributed computing.
Last, but by no means least, UNIX has attracted the applications needed to solve most business problems.
There are literally thousands of them ... twelve thousand on HP-UX systems, for example.
Many of the critical mainframe applications have moved to UNIX servers.
And, of course, UNIX supports applications not available on other operating systems, such as high-end graphics.
(Visual # 17 )
The value propositions UNIX represents to customers -- technical excellence, choice, adaptability and applications -- have made this technology more successful than any of us would have imagined just a decade ago.
UNIX has changed forever the face of computing. We see its impact in two related phenomena.
The Open Enterprise
The first is what we at HP call the open enterprise.
Perhaps the most fundamental contribution UNIX has made to customers is that it has helped them evolve their proprietary, legacy systems into an IT architecture that is more flexible.
Such a flexible architecture allows customers to address today's business challenges, while simultaneously preparing their organizations for an uncertain future.
They're not sure who their new competitors will be in the future -- nor where they'll come from. They cannot foresee geopolitical events or the acquisitions or reorganizations their company may undertake.
The only certainty is the need for a robust network infrastructure, ... and the ability to connect teams of people from throughout their organization ... and provide them with the information they need to do the best job possible.
Most importantly, they need speed -- the ability to respond quickly to changes in technology, customer needs and the competitive landscape.
The rapid rise of commerce on the public internet is a case in point. Many companies are scrambling to understand and realize the potential it presents.
On some level, so are we at HP. But we were well positioned to respond to the opportunity of the internet because we had the flexibility that derives from openness.
When we began developing our own internal network back in 1986, we chose to base it on open systems technology -- namely, TCP/IP.
We now probably have one of the largest private TCP/IP networks in existence. It includes 140,000 hosts and transmits 5 terabytes of information every month.
And because we built our infrastructure with open systems technology, we had the flexibility to connect our network to the public internet when it reached critical mass.
And so, we're doing commerce on the public internet. Access HP has 9,000 pages of information on the web. Between 30,000 to 40,000 customers look at these pages every day -- with an estimated 700,000 pages viewed daily . And these numbers are increasing every week.
We've used the internet to reengineer some of our most important business processes. For example, we now distribute software updates via the internet. We put them up on an HP server, along with information about the update, and allow customers to pull the code when they're ready to deal with it -- in other words, when it fits their schedule, not ours.
So that important idea of organizational flexibility -- the open enterprise -- has been the first new computing paradigm that owes much to UNIX.
Evolution of Computing
The second phenomenon is really a magnification of what's happening at the enterprise level. Actually it's the context within which a business operates -- or the information infrastructure, as some people call it.
This image summarizes the evolution of computing ... with a decade-by-decade snapshot of where we've been and where we're headed. You may have seen it before, because we at HP use it often to characterize computing's history and future.
We're now in the fourth wave -- characterized by cooperative computing made possible by open systems of clients and servers.
UNIX and the UNIX community -- with their emphasis on open interfaces and distributed processing -- deserve a lot of credit for making it possible to reach this stage.
We now stand at the threshold of truly pervasive computing -- a stage when we so take it for granted that we only notice its absence, rather than its presence. Vice President Al Gore and his friends call this phenomenon the information superhighway.
At HP we prefer to call it an information utility ... because the term denotes something as ubiquitous as electricity ... and the development of specialized information appliances that can plug into it.
Actually, even the electric utility metaphor has its limitations. Because the information utility won't require that users find a plug, thanks to wireless technologies.
Having led the evolution of computing over this past decade, what can UNIX do as an encore? Certainly, the UNIX story doesn't conclude with this sense of satisfaction. Comfort with the status quo simply isn't an attitude that characterizes the UNIX community.
Instead, we are seekers. We're a restless lot, uncomfortable when constrained by everyday assumptions and limitations.
So I suggest that we commit ourselves to taking computing to that next phase -- the stage when people take information technology for granted, just as they do electricity. What does that take? (Visual #21) I'd like to address the four technical prerequisites shown here, ... but not in their architectural order.
First, it will require scalable, high-volume hardware components.
Most of the computing power in this revolution is not going to come from proprietary architectures. Instead, it will come from enormously powerful merchant-technology microprocessors, used as building blocks of complex systems.
Through our joint venture with Intel we expect to leapfrog at least one generation of computer power.
We're delighted with the progress we're making, and here's Andy Grove to provide a short progress report.
(video to Andy)
(Visual #22 again)
Thanks for the update, Andy. Thanks, too, for demonstrating an example of how the information utility may be used. Although I wonder if the people who were at Uniforum '94 might think the red Miata was an even classier way to arrive.
As you can conclude from Andy's remarks, scalable, high-volume hardware components are very close on the horizon.
A second prerequisite for pervasive computing is network services. Some already exist, such as distributed file and print services. And Novell's NetWare Directory Services -- which has been available in PC environments for some time now -- will be on SCO UNIX this month and other UNIX systems later this year.
So services are emerging ... (Visual #24) as are the third prerequisite -- namely, network-aware applications that take advantage of those services.
The fourth requirement for the development of an information utility is probably the least advanced . It's the challenge I want this community to embrace -- namely, the creation of a robust, open, network-centric operating environment.
UNIX is ideally positioned to meet this need.
The challenge is to take the ideals and contributions UNIX has stood for -- and to refresh and reinterpret them in the light of today's realities and the needs our customers have expressed.
So let me take the same four value propositions I described earlier and suggest an agenda for the next generation of UNIX.
Moving forward, technical excellence means developing an industry-standard UNIX architecture to take advantage of new features that are emerging in today's and tomorrow's processor designs.
We now have an excellent process in place -- the 64-bit initiative -- to move this architecture forward and to achieve broad support for its specifications and software programming conventions.
Actually, the 64-bit title for that initiative is somewhat misleading. Next-generation UNIX is probably a more accurate description, because of the range of advances contemplated.
The work will include common software management -- of great interest to ISVs -- and agreement on a set of distributed system-management APIs, among other things. I pledge HP's commitment to work with this group.
We also are fortunate to have an organization to provide the required leadership this effort entails -- The Open Group, whose creation was announced just this morning.
I'm personally delighted to see OSF and X/Open combine their respective strengths in this new worldwide organization. It will strengthen and streamline the entire open systems process.
A second agenda for technical excellence is scalability. While this is an historic strength for UNIX, customers are looking for more.
We need to develop a standard version of it that can easily and efficiently scale up to 32 or 64 processors in a single system for those workloads where symmetrical multiprocessing is the best answer.
Other customers have workloads most efficiently addressed by computing clusters. And we all recognize the need to make it easier to configure those clusters.
In other words, we need the software equivalent of the high-volume, scalable hardware that is emerging.
We also need to look beyond increased performance and scalability of the core operating system. If we're going to help build the information utility, UNIX will require a significant leap forward in reliability, connectivity and security.
We should challenge ourselves with questions like these:
Can we make UNIX more self-managing?
What if capacity problems and system failures could be predicted?
Could we create systems that are self-diagnosing and self-healing -- like the ones shown in the "Synergies" video some of you watched earlier this morning?
How can we make internetworking pervasive and truly secure?
Let me say just a few words about security. I'm proud to say that HP -- in conjunction with Informix and Gemplus -- are developing a new generation of secure "smart cards."
These will allow multiple applications to talk to UNIX servers running an International Cryptography Framework that HP has developed.
We believe this approach satisfies U.S. government security requirements, and we're optimistic about its chances of approval. The framework simplifies the task of procuring an export license, and demonstrates quite compellingly that open systems can be secure systems.
So -- as this brief listing of unmet needs should lead us to conclude -- there's plenty of room left for UNIX developers to exercise their creativity to achieve even greater technical excellence.
Let me make just one comment about choice -- because I want it to stick with you.
Earlier, I suggested that the path to UNIX's future success lay in taking the principles that have made it great ... and refreshing them in the light of changes in technology and customer needs.
Customers are choosing PCs to help run their enterprises. That means we'll see Microsoft and Novell software at both the client and the server level, as this three-tier architecture would suggest.
And that, in turn, makes it imperative that UNIX embrace interoperability with the PC world. I noted with great pleasure that this conference agenda has several sessions on integrating UNIX and PC environments. I believe that's wise.
Instead of fortifying fortress UNIX, we must build bridges. Customers have made it very clear to us that they want help in integrating their enterprise and desktop computing environments.
At HP, we've been building bridges for years, because we operate in both spheres. And our recent reorganization -- in which we brought together our UNIX /enterprise servers and our PC/desktop activities under the same management -- is a sign of our commitment to redouble our bridging efforts.
OK, let's move on to adaptability. Anyone who's been around UNIX for any length of time may remember the days when it was a slimmer, more compact operating system that was easier to work with.
We should start by removing the redundancies and deadwood that have made UNIX bulkier than it need be. SCO has already started on this, and that's good news.
We then need to create a modular architecture like the one shown here which enables system vendors to highly tune the operating system for advanced processor capabilities and new system designs -- without sacrificing application compatibility.
In order to retain simplicity, today's architectures sacrifice design flexibility when incorporating new concepts like NUMA or clustering.
This increased adaptability will result in the introduction of myriad advanced systems that are fully compatible.
UNIX continues to attract applications at a healthy pace, but we can take steps to increase its attractiveness to software developers.
First, we can continue to advance the capabilities of UNIX. I've already suggested a few areas that are ripe for innovation. We should also provide innovative services -- such as the transmission and storage of video -- that can serve as a foundation for new kinds of applications.
Secondly, we can offer a version of UNIX that provides a high-volume, scalable UNIX porting target. This would present a number of benefits to ISVs:
... lower application porting and support costs;
... a broad target market; and
... easier software distribution.
This is a key goal of the HP-SCO work that is currently underway. Actually, our collaboration embodies all the UNIX challenges I've discussed today.
We are working on a three-dimensional architecture with unprecedented scalability and parallelism. The networked design will enable UNIX to play a pivotal role in the information utility.
Here you see a road map of where we're headed. I've removed some of the details from this slide because the print would be too small for you to read.
We're eager to share those details with you, and I encourage you to attend a briefing session immediately after my speech. It's in room 120, across the hall. We've got space for a crowd, so feel free to drop by.
For now, I just want to say that this arrangement is significant because together, SCO and HP and our development partners have the motivation, ... momentum, ... market presence, ... and resources required to move the UNIX implementation forward.
We intend to do so in a way that implements the emerging consensus on a 64-bit architecture interface and that works well with leading 32-bit and 64-bit microprocessor architectures.
Competition and Collaboration
Looking forward, I believe that we in the UNIX community can continue to balance two very legitimate forces -- competition and collaboration.
At times, we compete, and the customer wins.
Competition breeds innovation aimed at emerging customer needs. And innovation, by definition, isn't standard.
At times, we collaborate, and the customer wins.
Collaboration breeds consensus on a set of standard software interfaces that can be implemented differently -- but in ways transparent to the customer. Customers reap the benefits, in terms of connectivity and interoperability.
We go through those steps, over and over again. Competition... innovation ... collaboration and ... consensus on standards.
Every time we do it, we raise the bar. The technology moves forward. The consensus on standards covers a broader and richer set of interfaces.
And the customer gains the advantages of both competition and collaboration.
This is a paradigm that's unique to the UNIX community, and it's precious.
My pledge to you today is that the Hewlett-Packard Company will honor this precious balance. We want to help move UNIX forward in the spirit of competition and collaboration that have made it -- and will continue to make it -- a real success story.