Allianz consolidates from 60 servers to 1 mainframe in 48 hours

Insurance giant goes with a z10 mainframe and zLinux for its virtualisation and consolidation path

There aren't that many organisations which would consider consolidating 60 servers holding the most critical customer-facing applications down to one mainframe in the space of a weekend. But that's exactly what insurance giant, Allianzdid in May this year.

Allianz CIO Steven Coles said the planning for the migration of more than 60 Wintel boxes to one z10 mainframe was more than a year in planning.

"We have got two mainframes; one is a z890 which we have had for a while and manages our COBOL workload. More recently we have acquired a z10 which runs zLinux and we have migrated most of our web-facing applications onto that machine," Coles said.

"We have seen tremendous growth in our web front ends that support our business goal to be a leader in sales and distribution and they have become increasingly important. Traditionally they have run on over 60 separate pieces of kit out of data centers where capacity is becoming more expensive and scarcer."

In short, the driver for Allianz was to see how it could rationalise the infrastructure as much as possible, ideally onto one box. It did so through the deployment of a new mainframe and zLinux. A move which Coles says was pretty seamless for customers but one that provided the company with "significant benefits".

"The build up was around a year in terms of planning and at the time we were one of the few guys in Australia looking at using the technology in that way," Coles explained, adding the project was referred to as the 'Mainframe Virtualisation Project'. "So we spent a lot of time talking to IBM specialists globally and their clients to reassure us. We came out the other side with minimum disruption and from the end-user perspective it would be tough to realise the change had taken place."

For Coles the initial catalyst was challenges in gaining access to data centre space in metropolitan Sydney, which has been an issue faced by many corporations in recent times.

"That was causing us some concerns strategically as we were growing at pace and these applications were critical to that growth," Cole said.

While migrating to virtualised environments on x86 platforms has been a move made by many Australian companies, the take up of mainframes has received far less attention. While globally, IBM's mainframe revenue dropped 26 per cent in the July to September period, over the past two years Big Blue has had two solid years in Australia.

Across the country IBM has roughly 50 clients with 150 mainframe boxes between them (excluding IBM Global Services). And of this, six per cent are new customers over the last 24 months.

"It has really ticked all the boxes. It reduced the dependency on a data centre, it reduced the complexity from over 60 servers down to one box, it enabled us to put a lot more robustness around it in terms of DRP and scalability, and was environmentally friendly as well," Coles said. "It took our kVA power usage down from about 40 to 4 – so a significant environmental benefit as well." Plus it significantly reduced our IT running costs and paid for itself in just over a year.

In the next 18 months Allianz will also migrate its z890 COBOL workload over to the z10.

"I don't think we yet fully realise the potential benefits we are going to get as an organisation from having the majority of our applications on one physical box ," he said. "At the moment we have duplicate tables and code to support passing information to applications. If they sit on one box it will enable us to simplify the environment we have today. I have a sneaking suspicion that these longer term benefits may be greater than we think."

Sign up for Computerworld's newsletters to stay up to date.

Have you considered moving to a mainframe?Email Computerworld or follow @computerworldau on Twitter and let us know.

Tags mainframez10AllianzzLinuxServer consolidation

More about Allianz Australia InsuranceBig BlueIBM AustraliaIBM AustraliaLinux




Tell me more!! Is the world realizing that one mainframe is much better value than a multiplicity of smaller servers interms of throughput, costs and environmental impact?
Has Allianz consolidated environments in the process or just set up separate partitions for each box? Has throughput increased? What impact on future growth will the z10 provide?



John, you should join the System-z Linux group on linked-in.

Also, have a look at this -,289142,sid80_gci1367755,00.html

Allianz decided to ditch the Windows machines in favor two IBM z10 mainframes running Red Hat Enterprise Linux (RHEL), now in version 5.3, on the mainframe's Integrated Facility for Linux (IFL) processor. IFL allows users to run Linux workloads on big iron without the hefty price tag.



Not surprising. The admin staff costs for zOS have been far less than Wintel for decades, but zOS has struggled against the tertiary education marketing force of desktop wintel use by practically all non-technical education streams (who end up as business managers). Things are quick to develop on Wintel, if it fails you reboot, close and reopen - you cannot open enterprise apps to that risk without it costing big money. Most business types don't have enough transparency into the IT numbers, so cannot make the informed decisions they want to make. Mainframes are up 24 x 7 with much less corporate risk, less power costs, better storage speed and less admin costs. Add zLinux blinding I/O and virtualisation and you have a marketing winner in a language that business types can understand. So yes, not suprising at all, overdue though.



Nice idea, but why mainframe?

Sun big iron gives you everything that a mainframe has at a fraction of the cost, space and power consumption. What Sun doesn't have is the army of consultants and sales staff that are ultimately paid for by the high procurement ans sustainment costs of the mainframe.

The other question I have is how many mainframe customers did IBM lose in Australia over the last 24 months. They only gained 6, but I know they lost more than this. Please, if this is a real news article then do the homework rather than just reprinting an IBM press release.

Stan Murawski;


I would like to see the real numbers for the cost of that z-10 system that took on that 60 servers worrh of "web facing" workload. Then I'd like to comparte that to the cost of an HP or other mainframe alternative systems able to run that some workload. Putting 60 not too busy or older/slower server's workload onto a modern box is no big deal and I bet the HP alternatives is dramatically less bucks. The number of Linux images should be the same either way (60, 1 or in between) so that admin/support cost should be constant between the two alternatives.

Regarding the comment that " ... on Wintel, if it fails you reboot, close and reopen - you cannot open enterprise apps to that risk without it costing big money" it all depends on how you want to run your shop. It's about process and good procedures consistently followed. if you apply that discipline to Wintel you get goodness and, conversely, if you apply to your z-10 an zLinux box the lack of discipline your "if it fails yhou reboot" statement implies then you'll get a similar bad result.



I am guessing that the uneducated comments from Stan from Microsoft (Stan Murawski, Systems Technologies Evangelist, Bay Area Embassy of Microsoft Corporation) are coming from a good place and are not meant to really disparage or question the integrity of the CIO of a reputable company that spent a lot of time and effort making a well educated & informed decision. I am sure these numbers aren't secret & if you really want proof that Allianz made a very well educated & informed decision why don't Microsoft or HP approach them directly. Instead of adopting the usual practice of spreading FUD.



They won an award at the Redhat Summit in Chicago in the Carving Out Costs category for this project and so likely the comments are on the mark.



For those comments that if it doesn't work just reboot are very naiive indeed. That is not acceptable for sophisticated online businesses. Let's not have your personal bias detract from a CIO's business decision.



You have no idea of the mainframe mix of ins and outs in Australia, so don't make remarks without facts. Sun has been losing customers left and right this year after Oracle's bid - just ask the consultants.



With 5 to 10 fold core reductions from SUN or HP to zLinux and 40 to 80% reduction in application liscence costs and 60%+ reduction in environmentals we are going to see truck loads more of these stories.

Colin McCabe


This is on the mark. It's interesting to see just how much interest there is on System-z and Linux (read Red Hat!). Have a look at the system-z linux group ( on Linkedin, and you'll find out why it is the one of the fastest growing groups on Linkedin. With all the comments below, it's interesting to see some clear FUD that is disputing the financial and technical acumen of one of the most respected CIO's around. The main adopters in ANZ of System-z linux workload over the past 18 months is clearly financial services organisations (See BNZ Article). There are more that haven't gone public yet as well. Of all the types of organisations that will look closely at the ROI/NPV/TCO of any technology, at the moment surely banks and insurance orgs are the most likely to kill off any non sound financial decision (GFC anyone?).

Stan Murawski


Yes, I was a System Technologies Evangelist working for Microsoft. No I don't think it's a bad idea to consolidate 60 underutilized servers. I am saying only that I think there is an even lower TCO and therefroe and even better ROI when consolidating to System x (or HP or Sun or whatver). Regarding the claim that I my comments are uneducated, please explain. I am trying to learn what it really costs to put such a workload on zLinux. If I learn that I am wrong I will change my position, which I would hope you would do too, becuase I am trying to ahve a fact based discussion and ROI and TCO descions can't be based on dogma.



Let me see, already have a Z system that run COBOL application + wintel servers.

option 1) rewrite the COBOL application to SUN/HP then consolidate to those platforms
option 2) consolidate everything on z Series platform (old system+ all wintel).

Let's be honnest here... regardless of your affinity with IBM, Linux .... You are not going to move mainframe applications to a different platform, on the other hand moving wintel is relatively easy. If what you are looking for is consolidating with what you already have, they made the right choice for their need.... introducing sun/HP in the equation would simply not make sense or would cost a lot more. Now would that be the right answer for everyone? probably not.... but in this case, it just makes sense, so stop bashing IBM, Linux - and give them that one, will you?

Mike Schwab


If you want to try z/Linux, download a mainframe emulator and install z/Linux on it. You get about 50 MIPS per thread on at 1GHz+ processor, about 1 MIPS on a Pentium 90.

Colin McCabe


More on this, including some $'s at

I think Stan Murawski is not going to like the numbers, nor the moving from Microsoft line...



Phil wrote "Sun's big iron gives you everything that a mainframe has at a fraction of the cost, space and power consumption".

"... gives you everything that a mainframe has" is the comment of someone who knows Sun hardware and who thinks they know IBM mainframe. You cannot even begin to compare the two.

Check out for instance the power consumption of a z10 ELS and you will be very surprised at how few KwH it requires. And how many power cords do you need for a Sun M9000 with expansion cabinet?

Scott Mc


It's always interesting to read comments coming from Techies in relation to TCO. I am one of those business types that end up in management positions but also posessing an education in Technology.
20 years of distributed computing has led us to the mess we have today. The drivers behind the uprising against mainframe are as valid today as they were 20 years ago. The problem is that the self interest approach to IT management (increment by increment) has delivered and unmanageable disaster that cannot be understood let alone managed. Sad for the little boxes (that have grown to become very fat) that the price performance when TCO is actually measured (including labour networks, data centre etc etc and most importantly lifecycle management) is still firmly in the favour of mainframe. Simple question - how can expensive equipment running at very low utilisation levels be more cost effective especially when the management of that environment includes a very complex and infinitely variable software stack? No non mainframe environment in my experience has ever been rationalised. Virtualisation of X86 is not the answer - it's just a better answer than non-virtualisation and maybe a step on the way to the answer. There is no doubt that for small/medium enterprises that mainframe is probably not appropriate but for most large enterprises with signigicant centralised computing requirements the best answer is still mainframe and probably will be for some time. Final point the driver of the move away from mainframe was not the business types - it was the new age techies that were educated in an anti-establishment world - the exact environment that led to the creation of companies like SUN. It is the current generation of architects that need to stop and learn from the past.

Comments are now closed

Symantec identifies sophisticated, stealthy 'Regin' malware