Mainframes return to the business landscape

The much debated question of whether the mainframe is dead hit fever pitch in the early 1990s. Downsizing became the new buzzword and big iron was written off as passé.

The only way for any progressive company to go was client/server which, everyone said, was cheaper, more flexible and 'empowered' staff in a way that had not been possible before.

But at around the same time, analyst Xephon wrote a report that flew in the face of such accepted wisdom. Called The Dinosaur Myth, it claimed, against the odds, that the mainframe was far from obsolete and was in no danger of becoming extinct any time soon.

In fact, Xephon suggested that it provided much more cost-effective computer power than its distributed rivals, for all but the smallest organisations and, to pump up the controversy further, argued that it would continue to do so for the foreseeable future.

Ten years later, the same truths still apply, according to the analyst. For those who missed it first time round, The Dinosaur Myth mark two has now hit the streets, reaffirming the message that, in terms of total cost of ownership, you still can't beat the good old mainframe.

"We looked at the cost of running a mainframe compared with Unix and PC servers over five years, and mainframes were definitely cheaper for large and medium-sized companies," explained Trevor Eddolls, a senior analyst at Xephon.

The more applications you run, the cheaper the mainframe gets. "With Unix and PC servers, on the other hand, you may think you're getting a cheaper machine because you pay less up front but, by the time you've networked it, you're going to discover a lot of hidden costs."

Cost analysis
According to Xephon, the total cost of a mainframe's hardware, software and maintenance for five years is £4,750 per user. For Unix minicomputers it is £5,750, and for Intel-based PC servers, £7,500. Each of the machines in question supports several thousand users.

If applications software is added to the mix, the figures rise to £5,500 for the mainframe, £6,500 for Unix and £8,250 for PC servers. Support costs further increase charges to £7,000 for mainframes, £11,000 for Unix and £17,250 for PCs.

Finally, personnel costs up the total to £7,000 for the mainframe, £20,000 for Unix, and a staggering £26,250 for PC servers, making them nearly four times as expensive as their big iron brothers.

Josh Krischer, vice president and research director of enterprise systems and storage at Gartner Europe, believes that mainframes are better value for large companies because the pricing model of mainframe suppliers, such as IBM, has traditionally been structured to their benefit.

As a result, smaller customers pay proportionately more for each unit of work they undertake and, incidentally, often find it harder to justify the up-front costs to the board, no matter what the total cost of ownership might be like in the future.

In Krischer's opinion, mainframes are best suited to large companies, such as banks, insurers and airlines, where business continuity is the highest priority and downtime can cost thousands of pounds per minute.

Other advantages include the ability to manage the mainframe centrally, powerful security applications, and an architecture designed to handle many different types of application and to allocate resources as necessary.

Influence of the Unix vendors
And it is just these characteristics that have led to the gradual rehabilitation of the mainframe and, ironically given the witch hunt of 10 years ago, its appropriation by Unix vendors such as Sun Microsystems and Hewlett Packard.

"A lot of concepts surrounding the mainframe have not so much died as been overtaken by events," said Mark Lillycrop, managing director of independent consultancy Arcati.

"The mainframe platform is still there, underpinning huge amounts of corporate data and applications, but it's had to reposition itself away from the old concept of a very hierarchical environment.

"Instead I think we're now seeing a tendency to rediscover its strengths of architecture and approach, while managing it in a different way."

And one of these strengths is its very high utilisation rates of about 90 per cent. This compares with 20 to 30 per cent for Unix machines and as little as 10 per cent for PCs, which means that the latter are standing idle for most of the time. As a result, they are not as efficient as they could be which, in the end, costs money.

"If you look at the cost per user of running mainframe applications and compare utilisation rates, the mainframe is the most cost-effective by far," explained Lillycrop. "But in negotiations, when you're comparing one box with another, it's been difficult to make them attractive to the mid-range market because superficially they look more expensive."

This applies in particular to software licence costs which Meta Group vice president Rakesh Kumar suggested "are the single biggest issue for users".

"It's surprising that mainframe licences are growing at all, considering the pain here," he said. "People are trying to come up with better asset management tools, but it's still proving to be one of the most difficult hills to get over."

Kumar believes that the situation has not been helped by the exit of Amdahl and Hitachi Data Systems from the IBM mainframe plug-compatible market a couple of years ago, when they found that they could no longer sell enough machines to justify the huge development costs.

IBM's domination
While there are still independent mainframe manufacturers around, such as Fujitsu Siemens and Unisys with a couple of thousand users each, they are small fry compared with IBM, which has more than 80 per cent of the market.

As a result, Kumar said that some users have been complaining of price hikes of between 15 and 20 per cent, compared with increases elsewhere in the Unix market of between two and three per cent. It is a situation that IBM vehemently denies.

But pricing is undoubtedly an issue with mainframes, and one that has not been made any less contentious by the traditionally complex pricing model employed by the manufacturers.

For example, although IBM has attempted to simplify things, especially at the low end, by replacing Mips (millions of instructions per second) or capacity-based pricing with workload charging, where users pay for what they use, it is still a relatively complicated proposition.

Furthermore, many mid-range users are also wary of the fluid pricing and discounting model employed by mainframe vendors, even if they can manage to find any relevant applications.

"If you look at the applications market, the smaller vendors need to pick one operating system to write to, and Windows or Unix make more sense for them than the mainframe because they have much larger installed bases," explained Kumar.

"The bigger application players may support the mainframe, but very few people are using these packages because of technical and performance problems, and most have gone to Unix as the de facto choice."

Although IBM has now introduced a third-party channel programme to rectify the lack of new mainframe applications, Kumar pointed out that the majority of market growth has come from existing customers going after faster and bigger machines.

As a result, although the mainframe sector is at present a mature and fairly stable one, it is likely to shrink in size over time.

Shift of power
According to Meta, the capacity of the average data centre today comprises about 30 per cent mainframe, 50 per cent Unix and 20 per cent Windows. By 2010, the balance will have shifted to about 45 per cent Unix and Windows respectively, with only about 10 per cent of capacity being provided by the mainframe.

This situation could well be exacerbated by a projected skills crisis, particularly relating to mainframe systems engineers, although that trend is much more pronounced in the US than in Europe.

"Typical mainframe personnel are baby boomers, but they're now starting to retire and we're likely to see a skills crisis in the next five years," said Krischer.

"If the vendors don't start doing more training, the mainframe market will dwindle, so it's in their interest to get involved. But user organisations also need to start putting pressure on their suppliers because it's often just too expensive for them to train people internally."

One of the problems, of course, is that mainframes aren't glamorous, which turns off younger staff members. But IBM, for one, has recognised the problem and is trying to address the issue.

After a hiatus of 10 years or so, it has started installing mainframes into academic institutions, such as Warwick University, for teaching purposes.

"We don't expect mainframe-trained people to come streaming out of universities as a result, but if they have some knowledge, it's better than a few years ago when people were asking: 'What is this mainframe thing?'" said Doug Neilson, a systems consultant in IBM's enterprise systems group.

"We have done as much as we can to make our middleware such as the DB2 database, Tivoli [systems management applications] and Lotus Domino common across all our servers, so that the same skills can be used across all platforms."

Big Blue is also keen to introduce various self-managing and self-healing technologies, courtesy of its eLiza project. These go into its mainframes to reduce the need for large numbers of trained personnel, which are also inevitably costly.

For the time being, IBM claims that the mainframe market is in rude health and that the machines are the fastest growing of its server lines. Mips capacity increased by 65 per cent last year, and revenues by 15 per cent. It is "something that staggered a lot of people, even inside IBM", according to Neilson.

New customers
Moreover, he indicated that sales were not made purely to the installed base, with the company shipping as many as 85 machines to new customers over the same period. These were mainly dotcom start-ups which had opted to run the Linux open source operating system rather than the traditional z/OS.

Neilson explained the appeal. "You can run multiple copies of Linux on the same machine, which gives you a virtual server farm. This makes workload sharing and balancing easier, and makes it easier to manage everything in the same frame," he said.

"So Linux on the mainframe is mainly about server consolidation and having better service quality at a lower cost. While z/OS is aimed at running commercial business applications, Linux is great for newer front-end applications such as web serving and infrastructure stuff such as mail servers and firewalls."

But with or without Linux, Lillycrop believes that old-style mainframes will live on because the transactional data they handle is essential to the smooth running of many organisations.

He believes that the whole issue of whether or not the mainframe is dead will become increasingly irrelevant.

"The next stage in the life of the mainframe will be architectural benefits such as logical partitioning as a way of handling multiple workloads to other platforms, so that services become more consistent," he explained. "This means that the issue of whether the mainframe will live or die becomes less relevant as the cracks are papered over."

A HISTORY OF z/OS

1964:
IBM launches OS/360. It supports the System/360 mainframe's new architecture, which runs commercial and scientific applications.

1972:
The System/370 mainframe with dynamic address translation and virtual storage capabilities is announced. OS/VS1 was introduced for users of the MFT option (multiprogramming with a fixed number of tasks) and OS/VS2 for users of the multiprogramming with a variable number of tasks option. They had a capacity of 16Mb.

1974:
MVS Release 2 becomes available. MVS stands for multiple virtual storage (otherwise known as man versus system).

1979:
IBM starts charging separately for its mainframe hardware and system code. The latter is renamed MVS/370 and is Big Blue's last 24bit 370-mode operating system. It is withdrawn from marketing in December 1991 and from support a year later.

1984:
MVS/XA (extended architecture) starts shipping. This is a 31-bit operating system that can store up to 2Gb of address space. It is withdrawn from marketing in December 1992.

CASE STUDY: TNT "The mainframe is still very much alive. When you're talking about a networked business such as ours, where reliability is a key issue, it's an extremely cost-effective, resilient and secure platform to run your applications on." So says Dennis Beard, director of IT services at courier TNT. The company provides express business-to-business delivery services, handling documents, parcels and freight across 200 countries. As a result, its IT systems have to be operational 24 hours a day, seven days a week.The business is owned by Dutch company TPG NV, which employs more than 143,000 staff worldwide and generated sales of €11.2bn (£7bn) last year.TNT has been using IBM mainframes since 1984 and currently has three: the z87, RB6 and R36. All run the z/OS operating system and a mixture of DB2 and IMS databases. They are clustered using parallel sysplex technology and have a combined capacity of 1,732 Mips.The machines are located in the company's worldwide data centre in the UK and run its mission-critical operational applications. They are accessed by up to 12,500 active users every day by means of PCs and green screens.TNT's bespoke mainframe applications include sales and marketing packages, a track-and-trace system for parcels, and a worldwide rating and invoicing system. In-house apps
All the applications were written in-house using the PROIV fourth-generation language environment, which is sold by Northgate Information Solution's PROIV Technology unit."TNT's mission is to be the fastest and most reliable provider of express delivery services, so ensuring that our mainframe continues to be as secure and reliable as ever is an integral part of IT's role in supporting the business," said Beard.And it is this factor that made the company decide to stick with the mainframe despite widespread claims in the 1990s that client/server was the way of the future. But the company also uses Unix and Microsoft Windows NT in its country offices and depots where, for example, connections to Customs & Excise systems are demanded by local regulations."There was a lot of hype around distributed systems in the 1990s, but we've always made fact-based decisions through calculating the cost and the risk involved, which is why we've chosen the right platforms for each type of activity," explained Beard."The key is to be leading edge to gain competitive advantage, but mission-critical systems demand a well-proven, mature technology. It would also certainly be nice to see the maturity of the mainframe on other platforms."Tried and tested
In Beard's view, one of the main advantages of the mainframe is that it is a tried and tested environment which, unlike other platforms, boasts compatible and interoperable systems and application software, meaning that different versions of the same software work together.Moreover, he stressed that mainframes are easy to manage because well-documented processes and proven systems management applications make it easy to see where failures have occurred, and to plan ahead when undertaking tasks such as upgrades and maintenance."Availability is 99.99 per cent, measured every quarter. Failures are very rare and recoverability is very good. The mainframes have never gone down, which is critical for us," he said.The only downside is sometimes making a steeper capacity upgrade than Beard would have liked. TNT generally looks at upgrading its mainframes every two years or so, based on the growth in the volume of business that it handles."Because we support so many users, the up-front purchase price of software can be frightening, but the cost per user is better than other platforms and overall cost of ownership is very good," said Beard.TNT undertakes an internal audit every year to assess and review its business processes and how IT should underpin them. This is because of the changing nature of the global distribution market, where customers are constantly demanding quicker delivery, a faster response time from their suppliers, and "a faultless, friendly service".As a result, the firm is examining how to web-enable some of its mainframe packages, while also evaluating the potential benefits of Linux on the mainframe.TNT has also been taking steps to tackle the mainframe-related skills crisis by training staff internally. It has taken two computer science graduates from its graduate training scheme and assigned them to a team dedicated to looking after the mainframe. This comprises 15 personnel from a total data centre staff of 80.But when it has peaks in workload, TNT also brings in staff from zSeries business partner, Thesaurus Computer Services, to provide it with hardware and software support as required."Although we are concerned about the skills crisis, the key issue is about getting a balance between older, experienced staff, and those who are young and willing to learn," concluded Beard.

Like this story? Spread the news by clicking below:

 del.icio.us     Digg this     reddit!