Testing the water
IBM's revamped mainframe software pricing is an attempt to stem the tide of NT and Unix at the high end. As usual when Big Blue dives in, it's worth being aware of the implications.
The latest bombshell from IBM - a radical streamlining of itsem the tide of NT and Unix at the high end. As usual when Big Blue dives in, it's worth being aware of the implications. mainframe software licensing - is an attempt to wrong-foot its competitors in the NT and Unix camps and protect its installed base. It is the latest strategy in a four-year campaign to bring the vendor back to the forefront of the industry it once dominated.
IBM has made a series of tactical and strategic moves - technological, commercial and organisational - to strengthen its hand against its rivals.
The latest pricing move amounts to moving the cavalry into place to counter-attack the new model army of plug-compatible processor supplier Hitachi Data Systems (HDS), plus storage vendor EMC, ISVs such as Computer Associates and competitors such as Sun Microsystems, Hewlett Packard and Microsoft.
Although the changes apply to the mainframe world, they are likely to have a knock-on effect throughout the business. What IBM is doing is phasing out the old processor-based licence charges for its System/390 (S/390) mainframes in favour of usage-based licences. Unlike the PC sector, where software is purchased outright, or the mid-range and Unix world, where it is licensed on a per seat or usage basis. Mainframe software has always been licensed according to the size of the machine. The bigger the mainframe, the higher the monthly licence fee, irrespective of the processor power consumed or the number of users logged on. Therefore, a piece of software that uses five million instructions per second (mips) and has six users would cost the same on a 600mips machine as software that used 500mips and involved 1,000 users.
IBM has 10 methods of charging for mainframe software, all of them complex and little understood even by aficionados. The vendor is planning to phase these out over the next two years and bring in two separate means of charging.
The first, parallel sysplex licence charge (PSLC), already exists and is aimed at high-end users. The revised pricing structure is called usage licence charge (ULC) and is aimed at existing mainframe users who want to add applications to their systems.
Up until now, users have been reluctant to pilot applications on an S/390 because of the high licence fee. A new application that uses only 15mips would be charged at the same rate as an existing one using all the mips in the system. Many users choose a Unix or NT box to pilot their applications, causing IBM to fear they may stay with those operating systems rather than port the tested application to the S/390. Under ULC, users who consume less than 25 per cent of processor power on the mainframe will be charged according to the power they use.
ULC has huge implications for hardware, software and OS suppliers. Robin Bloor, chief executive of consultancy firm Bloor Research, believes the lessons will not be lost on other vendors, particularly Microsoft, whose rapacious attitude to software licensing he blames for many of the iniquities in the industry.
'Since 1994, Microsoft has been doing the industry a favour by behaving like a pirate in licensing, such as charging $400 per seat for terminal servers,' he says.
Bloor adds that the impact of IBM's moves will force other suppliers, particularly those in the mainframe and mid-range sectors, to introduce similar policies: 'Everyone is going to have to fall in line behind this.
The competition has no option but to drop its prices. The interesting thing is that in some areas, such as OS/390 and MVS, IBM is dominant, but in others it has no presence at all. Vendors can be highly competitive.'
While IBM insists its latest moves are aimed at the existing user base, there is no reason why it should not attract customers previously wedded to other suppliers. Tim Jobson, regional director for high-end systems at IBM, says: 'The primary objective is to make the S/390 platform more cost-effective than others such as NT and Unix.'
There are already versions of DB2 for Sun and Hewlett Packard implementations of Unix, which offer an alternative to Oracle and other relational databases.
The software licensing move is just the latest in a series designed to salvage IBM's position in the market. For instance, the vendor's shift from bi-polar technology to complimentary metal oxide silicon (CMOS) to power the S/390 was a huge break with tradition. Bi-polar or emitter coupled logic (ECL) technology is fast, but expensive to produce and run. Because the chips generate a great deal of heat, the machines have to be water-cooled, which requires separate computer rooms and specialised plumbing.
When IBM moved over to CMOS in 1994, analysts were critical, arguing that the vendor's large mainframe customers needed the older ECL. This proved to be the case and IBM lost its market lead to HDS and its hybrid ECL/CMOS Skyline system.
The latest generation of IBM machines, based around generation 5 (G5) of the CMOS chips, has narrowed the gap, although not sufficiently for HDS to lose its leadership at the top end of the market. But according to David Hammond, marketing manager for high-end systems at IBM, the vendor now has three G5 S/390 systems that out-perform Skyline at the top level.
The other key move on the hardware front was the introduction of the Multiprise 2000 in 1996, the first mainframe IBM sold through the channel.
The Multiprise is a low-cost, entry level mainframe that appears to be designed to attract newcomers to the market, although IBM denies this.
Jobson says: 'Is our focus on getting new customers who don't have an S/390 on board? I would say not. But customers are willing to consider Unix as a standalone development system, so why not look at the Multiprise and software from our business partners?'
To further encourage users to move into the market, IBM has introduced another option - application growth environment pricing (AGEP). Under this scheme, a customer who wants to buy a machine for application development will be entitled to a usage-based licence that, according to IBM, 'could provide customers with a reduction of up to 75 per cent in their IBM software costs versus standard capacity-based pricing when running one or more specific applications on a dedicated S/390 server'.
Although IBM is playing down the competitive potential of the licensing changes, there is no doubt in Hammond's mind that it has the ability to attract customers from the likes of Hewlett Packard. It was HP that was in the best position to take advantage of IBM's financial difficulties at the beginning of the decade. It concentrated on the development of enterprise Unix servers and later turned its attention to NT. While IBM dithered and looked inward to stave off further financial disasters, HP was busy eating into the IBM mainframe accounts, not least because it was able to offer cheaper software licences. But the boot is now on the other foot and HP has hit a sales downturn. As UK consultant Xephon pointed out in its July bulletin, Insight IS: 'The palpable deceleration in Hewlett Packard's revenue and profit growth coincides with two other clearly discernible events - a resurgence in IBM's fortunes and an attendant stabilisation of HP's previously growing influence within long-time IBM shops.'
Insight IS also contrasts the re-growth of mainframe sales with growing scepticism about the suitability of Unix and NT as enterprise servers in the current age. 'Despite reduced revenues, IBM's mainframe shipments continue to increase, raising doubts about the feasibility of a wide-scale migration to either Unix or NT,' it says.
Naturally enough, HP does not share these doubts, nor is it worried that IBM's licensing changes will slow down its own sales into the S/390 environment.
Indeed, far from accepting that the changes are either a result of IBM's desire to placate and keep its existing customers, or an attempt to win customers from the competition, HP believes it is pressure from the Unix vendors that has led IBM to revise its pricing.
EJ Bodnar, European competitive programme manager at HP, says: 'From our perspective, this is purely a tactical defensive move. The vendor's customers and ISVs are still writing applications in Unix and NT. We have 3,500 mainframe conversion stories to tell. The open applications on the mainframe are not there yet.'
Bodnar also questions whether ISVs will be willing to go along with IBM's licensing strategy. 'There are things out of their control in the mainframe market. You have to question whether the ISVs will go along with this,' he says.
The attitude of the ISVs is crucial to IBM's success, particularly where it does not have a competitive product. If the ISVs continue to charge capacity-based licences for their software, users will be deterred from taking up IBM's offer of usage-based licences. But on the other hand, IBM has been putting pressure on the ISVs to fall into line.
The problem is that ISVs and other suppliers are no longer overawed by IBM. Microsoft in particular, knows it is in the ascendancy while Big Blue is coming out of decline. Computer Associates has made a practice of buying up ailing companies and converting them into CA products. Even Compaq - once an upstart - has acquired Digital and Tandem, both of which play in the larger systems field. But IBM is fighting back and its moves in the mainframe software licensing market are not just designed to hit HDS or other mainframe plug-compatible suppliers.
HP, Microsoft, Sun and Compaq are also targets. IBM is tired of seeing its user base eroded and has begun the fight back. Ten years ago, Open Systems was anathema to IBM. TCP/IP, the standard communication protocol for Unix systems, was derided by IBM. Companies such as HP and Digital, while snapping at IBM's heels, were still second-rate players while Compaq was only a PC supplier.
Like every other supplier, IBM is selling more systems and software through the channel and will continue to do so. S/390 third-party sales aren't large either in terms of money or processing power, but they represent the direction in which IBM is heading.
Resellers, IBM partners, competitors and independent vendors would be well advised to keep an eye on what IBM is doing on the big systems side.
It offers more than one clue to the direction in which IBM chief executive Gerstner and his cohorts are taking the company. As in the past, IBM may fail to market its message effectively, but that message eventually tends to filter through.