Five years ago, when IBM’s mainframe sales force was struggling to persuade customers that big computers still had a future, Hewlett-Packard Co launched a program called ‘The Mainframe Alternative’. The message was this: Swap your big, expensive, proprietary mainframe for open, client/server HP 9000 Unix computers. It will, they said, be cheaper, easier, and give you more software options than before. In the intervening years, HP has installed 3,500 HP 9000s in what it calls ‘mainframe alternative’ cases. Almost every data center in the world’s top 1000 organizations, they point out, has at least one HP 9000 computer. For HP, that is the positive interpretation of its mainframe alternative program. But IBM disagrees. Throughout that time, it points out, and in spite of all the mainframe’s difficulties, almost nobody took out an IBM or compatible mainframe. In fact, in the past two years, as organizations have increasingly consolidated mission critical operations around a few data centers, the mainframe has been enjoying the strongest period of growth in its 30-year history. Even HP will admit that, for most of those 3,500 ‘mainframe alternative’ sales, IBM itself probably did not even bid a mainframe. But recently, HP has been at it again. Bill Russell, the company’s general manager for the Enterprise Systems Group, hammered out this message: HP is taking charge of the data center. The mainframe is no longer an appropriate solution. One size doesn’t fit all any more. This time, he was armed with a new set of arguments, software partners, and services and financing options.
Implicit failure
But implicit in the message is that HP as well as the other Unix vendors were not nearly as ready as they claimed when they first launched their attack on the mainframe. It is only recently that people like HP have provided the RAS [reliability, availability, security] to allow these customers to look at alternatives to the mainframe, admitted Bruno Castejon, HP’s European systems programming manager. HP only recently launched its ‘five nines’ initiative, promising 99.999% uptime, intended to make HP’s high- end servers as reliable as IBM mainframes. But at present, even the systems in HP’s own data center are only at 99.2% uptime, which translates into 71 hours downtime a year. Systems management tools were also inferior, lacking a knowledge of data center management procedure and knowledgeable implementation partners. Neither could HP offer the wide variety of financing options that came from the IBM Credit Corporation and its rivals. This time, HP has partnered with Hitachi and NEC, both companies with years of experience selling into the data center. Meanwhile, it is attempting to correct the now widely held assumption that client/server computing is more difficult and expensive than mainframe computing, not cheaper and easier. This, says Russell, is not so: There has been a lot written about the cost of client/server. But there are savings if it is well managed. The resurgence [of the mainframe] is less to do with the mainframe and more to do with the disciplines associated with it. But why, if HP is faring so well without displacing mainframes, is it so intent on renewing its assault on the high-end? There are three reasons: First, HP expects that when the Intel IA64 processor is introduced next year, almost every systems vendor will have the capability to produce powerful, standalone Unix or NT servers. HP needs to aim for the high ground, aimed at the data center. Second, Unix systems still account for a large and profitable share of HP’s revenues. But as the popularity of NT picks up, competition will increase and margins will be affected. Again, by emphasizing its high-end capabilities, HP will be playing to the strengths of Unix. Today, NT plays only a peripheral role in HP’s data center strategy.
System consolidation
But the third reason is that HP has identified in the Unix systems market one of the same trends that helped drive up mainframe sales consolidation. During the last ten years, companies have bought hundreds of Unix systems from a variety of manufacturers. Now they are trying to cut costs by consolidating them into fewer systems and managing them with one set of system management tools. HP’s new mainframe alternative strategy will ride on what it believes will be a wave of consolidation as companies build Unix based data centers to manage their systems and control costs. And this time it has a key advantage: The hundreds or even thousands of Unix programs being used will not easily transfer onto an IBM MVS mainframe. This is not an attack on the mainframe. We are just saying it is not the right architecture for the consolidation of distributed servers. IBM itself is pushing its RS/6000 SP Unix server as the best system for consolidation for similar reasons. Once companies begin to consolidate around the HP 9000, HP then hopes that it will begin to take control. But can it realistically expect big corporations to reconsider their strategy of sticking with the mainframe? Most projections show that new applications are increasingly going on either Unix or NT systems, but mission critical mainframe applications not only remain, but are being modernized and extended. This is partly because the Unix technology, in spite of its huge advances, is still lagging behind the mainframe in key areas. Of the world’s biggest companies, almost all continue to run their companies on mainframes with the main exceptions being those, such as HP and Microsoft, which clearly have a reason for not using mainframes. HP’s own IT budget is slightly, but not markedly, less than than the average for the manufacturing sector in which it is classified.
Computer Business Review.
