C P: The issues here are mainframe or enterprise-class server computing. We're going to assume that the terms "mainframe" and "enterprise-class server computing" are interchangeable and try to be as agnostic as possible. We conducted a study about people's attitudes with respect to mainframe and mainframe-class computing. We want you to question the data. So we'll just kick it off here with asking the group: Upon what type of system does your company run the majority of its enterprise mission-critical systems. This is how it turned out [in our study]. [Some 75% used mainframe systems, see chart ] Any surprises here?

MP: Well, I guess I'm a little astounded by the light showing for Unix vs. the mainframes.

CP: Clark, were you surprised at the large tilt towards the mainframe side?

CM: No, not particularly. I think the key word here is mission- critical. I think that denotes non-stop, 7-by-24 availability, all those sorts of things. And I think there are a few companies that have been leaders in pointing Unix systems that way. But I think it's just starting to explode.

GR: Let's take an $8 billion company. I think Star has already demonstrated the fact that you can run an $8 billion company without a CMOS machine. And I feel very comfortable in the statement that you can run it cheaper than you can with a CMOS machine. That's one dimension. The other dimension is that the user population is not interested in that world anymore for the most part, and if they had their druthers, they would go a different direction. Whether that means Java or another suite--I won't get into that. But either way you go, it takes you away from those top two lines. And my advice is that they would prefer NT as a choice. They will take Unix when NT doesn't work. Longer term, that's going to have to shake out as to which one of those two is going to win the battle. And Unix may stay for quite some time as a database server. But at the very low end, where the masses are, I think I've heard Bill Gates or someone from Microsoft say they don't really want the whole world, just the bottom 80% or 85%. And I really believe that's what this thing is going to shake out to be--the world of intranets and office tools and all that is going to dominate the low end of the operating environment and yes, there's going to be big servers in the world and they may well be Unix servers for some period of time. And there will be some CMOS servers if IBM can make that work properly.

CF: Users do love NT. Because they love GUI, they love point-and-click. That's something that in the eyes of IBM for years, we didn't do for a long time. We had green screens, character-based. ... But I think what we're going to see is a new breed of applications; a whole thing that's come about in the last couple of years [is] that the industry has monitored enterprise applications. And [with] a logical application set, some of them will run on S/390, some of them will run on Unix, some of them will run on NT. And the user really is not going to know where it is. He's got an NT workstation but DB2, CICS, whatever has got the data. I think that is as important a trend as centralization and consolidation, because that's almost like there is this one application but where does it run?

CP: This question is meant for Kevin or Claire. What are your mission-critical systems and where do they run?

CC: At Pepsico, two major divisions do things two totally different ways. Frito-Lay is mainframe-centric; it stores vast quantities of data on its mainframes, runs all its mission-critical applications on mainframes. Pepsi-Cola, on the other hand, runs its business in a distributed fashion, it uses Hewlett-Packard HP 9000s running Unix. What it has done is to centralize all those servers in a very large data center that's probably equal in size to the Frito-Lay data center. I don't think as an organization we've taken a look at what the cost structure is in both to see which approach is more cost-effective. But they both se em to work pretty well.

GW: From an IBM perspective, in the mid-90s--'92, '93, '94--there was a significant decline ... and now we see a tremendous resurgence. Things like E-business and E-commerce and Java--all those things run on a mainframe and they basically have allowed a great opportunity because the green screen has gone away. The other end is a browser that makes, from an application perspective, System 390 in the mainframe world a very, very good world to move ahead.

CP: In the next [survey] question, we're asking what type of system you predict that your business will run its mission-critical applications in the next two years? You can see here what some of the changes are. Obviously, the growth of CMOS over water-cooled; the growth in Unix servers and fairly modest growth in perception of smaller platform servers as running mission-critical business systems.

PK: We've been looking at tools that give access to legacy data through Web-browser type interfaces, so that a lot of the data and information could stay in the same place but would become usable in very different forms and it might not show up on the chart like that. One of the numbers floating around is that 70% of all the business information is on mainframes. And as Claire states, it's extremely expensive to move that data anywhere. So when people are saying: "Well, maybe there are cheaper tools we can put in between on our intranets," that might not show up there. Data and applications may look as if it was a PC.

CF: That's the whole new application, that's the SAP application paradigm: Two years from now, I might have a CMOS mainframe feeding that Web or Internet server. Our job as vendors is to make sure that it's solid. If you go to a browser-based application set and the application logic may be running back on X system, but if you can't get to it because your Web server is not running, that mission-critical function--maybe not mission-critical to your application, but it's a mission-critical infrastructure.

CP: But I think [respondents] were saying: In two years, [when] I think mission-critical systems, I don't yet think Web server.

MP: The numbers say something else to me, too. I mean, one, it says the mainframe never went away. The second thing is, looking at what people say about the present and the future, is that most of the critical systems, 90% I think, both present and future, are going to be on centralized systems. And centralized computing is really where it's at for enterprise mission-critical systems. I think, to a great extent, that speaks to the failure of the client-server approach in spite of all the press, all the hype, all the marketing, all the investment companies have made--very few people either have today or are thinking about having in the future, mission-critical enterprise systems up in a fat client-server environment. And the reasons are it's very d ifficult to deploy, it's very difficult to support.
The other thing, just to contradict Gary, Unix right now works better than NT and most of the enterprise things that have been done on different platforms, either today or in the future, as indicated by the study, will be Unix. And yet in the press, it's like an overwhelming crescendo that speaks to the inevitability of the demise of Unix and the rise of NT. NT is an important operating system. It certainly gave Unix a run for the money. But I don't see the trend in your numbers. Your numbers are that Unix is a very key part of the move to certain large-scale server systems. And [my] last ...point is that the users like GUI, not necessarily NT, they like the Web, as well, and that provides us an alternative. So I will go on the record to say that we believe in the Web, we believe in Unix servers, and we think that's where the future is going.

CP: I can see Frances sh aking her head on everything you're saying.

FW: Well, it would be interesting. I'll go back to the point on trends--it would be interesting if you had done this two years ago or four years ago to see. I think it would have been far more dramatic, that people would have said in '95 or '93 that 50% of their applications would have run on client-server. There's been a huge dose of reality that says that's not feasible for mission-critical workload at this time. So I think results are tempered by four or five years of trying to move applications over to network, over to that platform.

DM: [With] most of the companies that I've talked to, profitability is still their No. 1 mission. Those mainframe legacy systems are just real hard to convince a group of people that you need to redevelop all of your systems in light of the client-server difficulties with management and all that. I mean, you get a real good secu rity blanket with the MVS world and the manageability of networks and mainframes. So you go to the Web servers that don't run and you take a system that runs 99.9% and run it on the system that might get you 80% availability. After all that redevelopment, people have asked: What did I get?

CP: We're going to look now at this manageability question. We wonder how many data-center folks will operate that they would define as mainframe-class computing. It's interesting: It's about 60% for one site and then the rest is divided between two, three, four or more. So then we asked: Do you plan to centralize your distributed systems? Kind of open-ended question and we didn't give them a time frame: 61% said yes, they do; 19% said no and 20% said they already have. So 80% said that they have or they plan to centralize their distributed systems. Why centralize your distributed systems? This is the order of reasons why people said that they wanted to centralize the management of their distributed systems--they range from easier management down to simpler storage strategies.

Continue on to page 2

Back to News in Review

Send Us Your Feedback

Top of the Page