Feature: mainframes – Be afraid … the mainframe is back

If true, it raises some interesting questions for the strategic consultancy business. Could mainframes become an architectural option once more? Allied to the whole question is the concept of “reverse downsizing”, where applications that migrated off Big Iron 10 years ago may find themselves back up there once again, courtesy of Linux, which may find its true role in the enterprise being a consolidation engine to replace Unix servers.

All heady stuff – but there are enough suggestive signs coming out of IBM, some of its customers, and the Linux world to warrant attention being paid to the proposition.

Rich Lechner, worldwide vice president of marketing for zSeries in IBM’s eServers range says: “The mainframe is the Manet in the attic.” The logic goes like this: organisations diversified off the host in the client/server and open systems arena to promote greater flexibility and use of PCs, which were sitting expensively around unconnected to back end systems. But that original expansiveness has grown to promote what he calls a “serial buying habit”, where servers, LANs and desktops proliferate endlessly through organisations. But “as you distribute computing, you need a stronger and stronger layer of management to control it”. E-business – especially as it manifests itself – calls for a unified image of the customer, and there is no better engine for doing so than the mainframe.

Lechner also contends “many organisations never migrated their core applications off the mainframe anyway”. This is certainly true of UPS, which manages to combine an immense IT infrastructure with innovative e-business applications like its electronic tracking service. UPS’s vice president of information systems, Jerry Skaggs, recently told journalists that his firm’s $16bn investment in IT included two data centres running 13 IBM CMOS mainframes and two of the newer zSeries high end boxes, plus a mere 901 midrange machines, including 217 AS/400s (now the iSeries), and 310,000 PCs connected by 3,500 LANs. The tracking system available from any web browser connects straight into the mainframe, which also runs the world’s largest DB2 database.

At the other end of the scale, JD Williams, the UK’s largest home-shopping women’s wear catalogue firm, also has the mainframe in a front-line position in its e-business strategy. Tom Fothergill, its head of e-business, explains that the company put its 16 major brands online using IBM technology after trialling the system on an NT box. “You don’t ever get downtime messages if you’re an IT manager about the mainframe. We had the existing skill set working with the 390 (mainframe), it could scale, it had a robust infrastructure, the systems applications were already there. We also didn’t want to introduce a sub-standard service through this new (Internet) channel.”

It could be argued that these companies are dyed in the wool IBM stalwarts. Surely it’s not surprising that they are finding new uses for their stable technologies? But non-IBM heartland companies are buying mainframes for the first time. Last month Venezuelan bank Banco Mercantil announced it was replacing 30 NT servers with one mainframe running two dedicated cryptographic processors. In December Danish ISP Telia revealed it was scrapping 70 Sun and DEC Unix servers and replacing them with a single S/390.

The idea is to use what even Lechner admits is “30 year old virtualisation technology” in an innovative way. Using the massive memory scope of a mainframe, it’s possible to create a partition, or processing state, where many applications can be run virtually, without need of a dedicated server of their own, thus centralising the application while fooling it into thinking it’s still running around free on the Unix prairie.

Note the word “many”. Telia is trying to run 1,500 applications this way. IBM says it is in serious discussions with “a large US telco” to put 4,000 Linux images on a single mainframe as a way to replace a Unix server farm. Peter Norris, IBM’s eServer Consultant for Europe, says he knows of another unnamed site looking at running 97,000 copies of Linux on a single machine.

But why? Cost is the surprising answer – especially since the prohibitive expense of running mainframe cycles was held to be the mainspring for the downsizing wave of the ’90s in the first place. Lechner quotes a report from specialist US research group Sine Nomine Associates into that mysterious US telco project giving staggering figures: the possible cost savings from running one mainframe versus 750 Unix boxes is $28m over three years ($7m versus $35m).

Alleged cost cuts come from a variety of sources: less people to serve the machines for a start (less than five full time support staff versus up to 25 for a server farm); less rent (important for many firms based in large cities); less maverick purchasing, or in Lechner’s perhaps suspect phrase, the “Sun box being used to support the coffee pot in the branch office”; and last but not least the juice to run it.

That claim would be easily dismissed – were it not for the fact that independent analysts groups broadly agree. In March, technology analysts Hurwitz Group and Matterhorn Group actually recommended the z900 mainframe running Linux as an effective energy saving device. While a typical configuration of 750 Sun servers costs approximately $620 a day in electricity to run, a single z900 running the same workload costs only $32 a day, a power saving ratio of nearly 20-1. The average server farm also requires some 10,000 square feet of floor space compared with only 400 square feet for a single IBM z900. At an average of 100 watts per square foot, the savings can be significant, Hurwitz concluded, while Matterhorn contends mainframes are inherently more cost-efficient than server equivalents.

The cost of electricity is obviously a more contentious issue in California than in Copenhagen, say. But plainly if there is economic and logistic sense in the idea that the mainframe could be a “green” alternative to many Unix boxes, customers will start taking note – which could be very good news for IBM. Colin Tenwick, vice president and general manager EMEA for its Linux partner Red Hat, says he knows of customers going out today with their chequebooks buying new mainframes.

Which might be even better news for IBM Global Services, given that after Amdahl and Hitachi Data Systems’ effective withdrawal from the market last year, it is effectively the only mainframe game left in town. It says at least three large US brokerage firms are investigating similar moves to Telia’s and its secret telco; UPS has confirmed it is at the experimental stage with Linux on a partition: given that UPS has conducted some 160 projects with IBM Global Services, is this bad news for anyone else trying to break into this new market?

Yes and no. Analysts are broadly sceptical. Phil Payne, managing director of Isham Research, says there’s a lot of interest in Linux on S/390, but any real take off will wait until much cheaper mainframes become available from IBM for this kind of use, which he thinks may be in mid-2002.

Mike Thompson, director of research for Butler Group, is harsher: “This is important, but it’s a far more complex issue than first appears and I can’t think we’ll see any real evidence of this kind of consolidation role for Linux for at least three years.”

Phil Dawson, programme director for Meta Group, adds: “I see IBM pushing this, but I don’t see any pull from customers. This is a case of customers not wanting to switch off their mainframes and looking for new things to run there. Linux is cool, but it’s for SMEs.”

To which IBM retorts: watch this space. “I have more customer pull than I know what to do with,” claims Doug Nielson, eServer technical consultant for IBM UK, pointing to the fact that IBM gave Warwick University its very own free S/390 in January as further evidence of grass roots demand for Big Tin skills and resources.

What is the message for the consultant in all this? Peter Slavid, business strategy manager for ICL, which used to sell mainframes but which has remade itself into a services play, says there are great opportunities. “Customers would rather have one box than separate boxes if possible, as it’s easier to manage, if not necessarily always cheaper. But there are consequences. They’d have to treat that box with an appropriate mainframe mentality, building in disaster recovery and continuity. There are change management issues to do with the cultural consequences of rolling departmental solutions back into the middle. Customers may look at IBM and say they’re not the right people to manage that kind of change, as they’re biased.”

Adds Payne: “Linux is going to be the landing strip on the mainframe for 64-bit Unix when it gets here. There’s a broad mass out there happy with Microsoft and Intel, and 1% having phenomenal growth beyond NT Microsoft can’t and doesn’t want to handle. There are major opportunities for management consultants working alongside IBM hardware partners to provide independent platform choice for that class of customers.”

Mainframes have always had the last word in real top-end power – IBM’s Power 4 Chip Module, due in September, will offer power and throughput through its 170 million transistors that will still leave Sun’s Sparc-V and Intel’s just-released Itanium 64-bit chip in the dust. The question is will IBM come true on this suggestion of offering that kind of power as a cheap Linux platform? If so, the market could undergo immense changes.

Smart consultants, as ever, will work on offering the best technical solution for their clients – even if it sounds as if they thought the League of Nations never really got the support it deserved, or that Popeye Juice is a safe route for the bronzage totale.

Related reading

aidan-brennan kpmg