Server platforms Toolkit

Making the case for the mainframe

Tags:
Figures,
Application Server,
Consolidation,
Engines

Cath Everett ZDNet.co.uk

Published: 26 Aug 2008 15:41 BST

...and wasteful in energy terms is that the boxes tend to be over-provisioned and workloads are not mixed efficiently.

But such factors are not the only reason that organisations are starting to show renewed interest in the mainframe. One of the most significant drivers behind IBM's 34 percent increase in MIPS last year — the company measures the market by total processing capacity expressed in millions of instructions a second — and a 32 percent rise in mainframe revenues over the same period, year-on-year, was the recent introduction of various speciality engines to run Linux and Java. Another driver is the ability to run IBM's DB2 database on the same machine as enterprise applications such as SAP or Oracle sitting on its WebSphere application server.

According to Phelps, the move has led to between 60 to 70 percent of all net new MIPS shipped over the past few years being used for consolidated application workloads that did not exist on the mainframe 20 years ago, with Linux now accounting for 11 percent of all MIPS installed worldwide.

The market needs some medium-sized applications vendors to represent stability for CIOs

Roy Illsley, Butler Group

The overall market itself, however, has grown on average by 15 to 20 percent in net new MIPS terms over the same time period. This is because, although the speciality engines are separately billable and cost about $125,000 each, they are not included in regular capacity-based licensing fees, which would add to costs significantly.

"This has allowed IBM to move into new areas without causing pricing problems in relation to legacy systems. So the speciality engines enable organisations to enhance their capabilities or run new workloads at a good price, which is part of a wider total cost of ownership [TCO] argument," says Phelps. Factors to be taken into account here include low administration costs due to the better reliability, availability and security of mainframes in comparison with rival platforms. As a result, he advises: "It's important not just to look at acquisition costs, but also TCO over a three- to five-year timescale."

Market dominance
However, while large enterprises generally have no problems in this context, the same is not true of smaller organisations as even the relatively low-cost speciality engines still account for a large chunk of their IT budget. Another issue is that, to generate TCO savings, it would be necessary to consolidate on average between 40 and 50 Linux boxes onto the mainframe to make it worthwhile.

To make matters worse, software pricing is tiered and works out proportionately much more expensive at the lower end of the MIPS spectrum than at the high end. And a further challenge at the applications level is that not only is the mainframe far from the first port ISVs think of when building new packages, but there is also no significant partner ecosystem to write them in the first place.

This means the market is dominated by a handful of vendors — IBM itself, Computer Associates and BMC — although Phelps indicates that as many as 300 new applications, including middleware, utilities and vertical market packages, did appear last year, which was the first time such growth had occurred for some time.

Nevertheless, says Butler Group's Illsley, the mainframe sector is "very skewed" by IBM's almost total dominance. "The market needs some medium-sized applications vendors to represent stability for CIOs and I doubt that many of them would want to put all of their eggs in the mainframe basket," Illsley explains.

A further downside is that, while support for environments such as Linux, Java and IBM's WebSphere application server means that skills in these areas are now portable, organisations may...