Nexxar is a new financial services firm, created in 2003, through the merger of a number of acquisitions, led by VC firms. It works as a backend for financial services – money transfer, money orders, bill payment and so on. It also offers “private label” services, acting as a processor for banks, phone shops and so on.
The other day I was lucky enough to speak to Nexxar CIO Wim de Ridder (thanks Gerald). Wim has been on quite a media tour lately if google is anything to go by [plenty of links to news stories if not corporate homepage]. Why does Nexxar make such an interesting IBM mainframe case study though? Primarily because it chose to consolidate the infrastructures of its acquisitions on a mainframe- with a board level decision- even though neither the firm, nor the IT people, had any mainframe experience. If there is one lesson to emerge its the level of resources IBM is willing to dedicate to customers considering interesting mainframe projects.
Towards an M&A Consolidation Platform
When Wim began the company consolidation project it was clear he needed a core infrastructure consolidation platform. The target companies had a mix of Sybase, Microsoft SQLServer, .NET and the usual suspects. But says Wim: “None of the firms had a technology that would allow us to consolidate, and allow organic growth and further acquisitions.”
The challenge, he says, was to build a shared set of assets, that were extremely scalable, reliable and so on. The company decided to build a new engine for consolidation, and decided upon Java Enterprise Edition (JEE) as the development platform. At that point, Wim said, the firm had not initially decided on a production environment because it was comfortable portability would come from writing to JEE, using WebSphere.
“I had no experience of mainframe. We looked at Intel, z, p and so on. Not everyone was interested in us as a customer. We weren’t a bank but a small company. That is where IBM stepped up to helping us. Each different sales force made their own case for their own platforms.”
“From a cost perspective we realised System z was a good choice. The way the licenses are charged for software are per processor – on Intel or Risc platforms – and cost of maintenance was dramatically increased.”
System z9 BC and the three bears
With no mainframe experience Nexxar needed clear proof points for what the System z could do. IBM stepped forward with an offer that it could use its Montpellier center of excellence for 4 solid weeks of stress-testing.
“It proved z was up to the challenge,” said Wim, “It also meant we could rightsize [capacity plan] the equipment we needed.”
Nexxar needed one integrated facility for Linux (IFL), with 25% of the system capacity for DB2 and zOS. But “then we started working on pricing…”
Quite simply – “The z890 was too small and the z9 was too big,” said Wim. Lack of granularity has long been a problem for entry level mainframe purchasing – the mainframe was either too hot or too cold.
Nexxar received extra attention from IBM’s mainframe team in Poughkeepsie though, enough to be invited into the early access program for the IBM System z9 Business Class (BC), which turned out to be “just right”.
IBM didn’t just provide access to products but to experts, as well, which is just as well as given Nexxar had to get up to speed in no time.
Selling zSeries to the Board
The facts presented by Wim, the CIO, as he sold zSeries to the board, are very interesting.
1. Pitching z for business alignment. Align the idea of a shared asset center with mainframe thinking. Then push the notion of optimised resources for that environment- CPU, memory and so on. With the mainframe “we can do production and web serving and even email from the one box, with a high Quality of Service.
2. Cost Matters. The calculus for the board was made quite simple. Total Cost of Ownership for 5 years, including hardware, software, maintenance, disaster recovery, expert services and training for $1.9m.
3. The feature sale. Wim sold the mainframe as an on demand box, that would grow as Nexarr did. This is one area where IBM’s marketing materials are legion, to help “make the sale”. At Nexarr On Demand is particularly attractive because of the bursty nature of the business. Being in the payment processing business, Nexxar finds that Mothers day, the back to school rush and Christmas are peak times-accounting for 25% of the company’s (and mainframe’s) yearly volume.
4. Mainframe security as a comfort blanket. One of Nezzar’s business offerings is called “private label” – where it provides seamless back end services for other banks. These banks tend to be cautious and risk averse and like what they know. “This was the sweet spot for Private Label: banks are comfortable when you tell them you’re running on the same box they use. But they want dedicated resource. We can of course provide that using logical partitioning and role based security through RACF.”
And so to Implementation: How To Overcome the Skills Gap
In order to ensure that a lack of zOS skills was not a problem Nexxar made two service agreements with IBM – first was about setting up, tuning, installation and documentation – which took 12 weeks from nothing to production. Second was production.
Included within the service agreements were 60 days of training, and a “resource bundle”-to have expertise from IBM on certain areas.
Think about that – 60 days of training across two employees, as part of a total cost of ownership pitch to the board. How many organisations are willing to properly train their employees-on modern or legacy technologies [don’t tell me Windows or Linux don’t require training to do right]
In the 12 week set up period the company contacted a head-hunter, to hire one DBA and one systems engineer. With HR legislation in Belgium it meant 16 weeks for the new hires to join. Both hires were in their late 30s, early 40s – which means experience and not too many grey hairs…
Nexxar is now in full production, with 800 employees, across 105 countries, servicing 100k locations. The business is growing year over year 35% organically, even without acquisitions.
In future Nexxar may consolidate its Lotus Notes mail to the mainframe.
Problems Solved
“When we went into production, IBM was here for a week, to monitor it. We had one issue-around memory allocation for WebSphere, and IBM quickly fixed it.
Disclaimers: IBM’s System z software and systems groups are both RedMonk clients
mainframe, System z, IBM, Nexxar, TCO, WebSphere, DB2
James says:
September 13, 2006 at 3:10 am
A very smart architect runs around corporate America yelling at the top of his lungs, the mainframe is not evil but COBOL is…
Bob Hoey says:
September 13, 2006 at 11:16 am
Jim, although I’m biased as the IBM Worldwide VP of Sales for System z, watch the low end space (< 1,500 MIPs) as we expect to see new footprint growth and increased use of speciality engines by customers because of the new economics of the mainframe. We expect the highest growth rates in China, India and Russia. We’ll see many more “Nexxar-like” wins in 2006 and beyond. Our Business Partners will drive 75% of this low end growth.
Peter Norris says:
September 13, 2006 at 12:13 pm
Bob – I first met this guy back in the mid 90s and I have always known him as James. But ignoring this minor point …..
I agree with your comment about more smaller users starting to understand the TCO and SLA benefits of running their business on one or two mainframes – rather than many little boxes. The momentum is building as more people understand the TCO/SLA issues of the alternatives and what is becoming an overwhelming case for the ‘new mainframe’.
Paul Anderson says:
October 23, 2006 at 7:46 pm
James(Jim/Jay/JJ), Bob and Peter, How interesting it is to read this. I was the IBMer that did the Java performance tuning for Nexxar at Montpellier, which was a fulcrum of the deal. I also have done some on site work on request a little while ago and to close the circle bumped into one of their chaps in Brussels last week where he asked me to back on site for a bit of a tune-up. So I’m fairly close to the account.
The thing I remember about Nexxar is that they had gotten the Java just about right to begin with, they were pragmatic and always kept in mind the job at hand and the potential transaction rate. This allowed them to reap the broader rewards of the platform outlined in this article.
OK we made a lot of improvements above the raw code, as you’d expect, but they were so willing to go through the process.
All in all a very good example of a complex but well engineered Java transaction at a realistic cost allowing the zSeries soar. 1 CPU running a business rather than >100 blades.
Paul Anderson.
Alex says:
April 25, 2007 at 1:53 pm
Thank You