Companies still running IBM or Unisys mainframes and minicomputers or systems from Informix or Wang are living evidence of the lock-in power of proprietary technology. Once you bet your business on these platforms you are in for the long haul.
The most important challenge for any IS department today is keeping systems in line with business needs on a restricted budget. Associate Professor Mike Dewe at Canterbury University sometimes finds himself lying awake at night wondering what might happen if he were to lose the IT specialists who know the legacy student system inside out. “Having a system like this has become too critical. We’re still dependent on individuals who aren’t getting any younger. The code is not well-documented and I really do worry that the university would have serious problems keeping the system going if we were to lose one or more of these key people.”
Life extension efforts
The present system has been evolving for around 15 to 20 years. It was originally based on a Digital Vax platform running VMS and now runs on an Alpha series minicomputer with code written in a mixture of Cobol and Forte. More recent life extension efforts are based around Microsoft’s .Net tools.
Dewe, who wears the dual hats of IT director and MIS projects director, says there are long-term support issues handling just about every kind of database you care to dream of. “It is getting untidy, a problem typical of systems that evolve in a university environment. There aren’t the disciplines one might expect from a professional software company.”
The university has been considering a change for three years, but commercial systems only recently became widely available in New Zealand. In 1999, the university planned to revamp its Oracle financial systems and had no HR system apart from rudimentary spreadsheets and filing cabinets. It was aware of the need to replace its outdated student system, which was unable to meet the needs of the changing educational market.
That’s when things got tricky. Students protesting over increased fees occupied the registry. Ironically the amount of the fee rises happened to coincide with the estimated $11.5 million budget for the proposed new system. “The timing was atrocious and everyone got cold feet.”
After much consultation, an upgrade of financials to Oracle 11i was decided upon, along with a PeopleSoft HR/payroll system, leaving the IT department to try and extend the life of the legacy student system for another two to three years.
Buy or build dilemma
That time frame has well and truly passed, and while a preliminary request for proposal has been conducted, the decision to buy or build is still being weighed up in great detail. “To build would mean a total re-engineering of an in-house system. To buy would inevitably result in major customising to suit the Canterbury University environment. Any replacement exercise would require comprehensive planning and 18 months to two years to implement,” says Dewe.
“All the good books tell you ‘if you can buy it, don’t build it’. But there’s a lot of evidence that student administration systems are complex, that different institutions have different requirements and always require a lot of customising, and that many institutions have stayed with their in-house build options as a result.”
Then there’s the need for synergy with existing MIS systems and interaction with other universities. If you buy, do you buy from Australia, the US or locally, and what are the long-term support issues?
To get answers to these and other questions, the university had to go back to the various providers. “Quite frankly, the money is not in the actual software, it’s in the implementation side. It costs significantly more putting it in, no matter whose system it is. Then there are the migration issues.”
Dewe reckons the ultimate system would need to provide enrolment, track students’ performance and establish a student’s eligibility to graduate. “This would require a rules engine to define how you earn a degree to get a qualification with web front-end, and the ability to migrate existing information.”
A shortlist should be ready by March or April 2004. “It’s not something you rush into. It’s a major investment that must be done properly.” Meanwhile, a web enrolment front-end is being plugged into the system as an interim measure to make it user-friendlier.
Functional but inefficient
Few legacy environments can be simply left to run on endlessly. As businesses require greater performance and functionality, limitations become increasingly apparent.
More complex coding is needed to get data out in acceptable formats and more maintenance and tweaking is required to keep software and hardware ticking over.
Even the most miserly of managers will be aware, without careful change management, that business-critical systems of the bygone era may well cause the crisis in the next era.
IAG New Zealand’s chief information officer Catherine Rusby says moving from legacy systems requires constant vigilance, risk management and an eye for the unexpected. “You need to be resilient and prepared for changes you don’t expect. We were optimistic we could move the systems of State Insurance across to IAG in a short time, but then we acquired NZI and had to double our scope.”
The core legacy systems are different and require manual intervention between accounts and financial reporting to aggregate key reporting information. Each business had begun to develop strategies to move from ‘green screen’ technology to a more open Windows-based approach but with different platforms in mind.
State was heavy involved in direct insurance and NZI’s experience was as a commercial broker. The businesses were different but complimentary and IAG had to determine whether having single applications supporting all channels was a valid strategic direction.
“We felt that a single platform would mean less hardware and software to maintain, less complexity, easier management and lower cost of ownership over time. We wanted to leverage as much of the back-end as possible and decided to continue down the path IAG had settled on.”
To that end IAG is in the midst of Project Endeavour, a major business re-engineering exercise, which will deliver a full suite of back-end applications and business processes. That includes replacing its own Oracle and Unix-based legacy systems with software from UK-based Serious running on SQL server. This will bring a single view of management and customer information across the businesses.
Getting a common infrastructure in place was an important pre-requisite. More than 85 per cent of customer transactions are completed over the phone, so the phone network and the integration of computer and telephony applications is pivotal.
In fact, the entire voice and data infrastructure will also be placed as part of a $40 million Project Converge with TelstraClear creating a single IP network for voice, data and videoconferencing. The integration for IP telephony and applications for call centres and other voice-based applications are being supplied by Agile, which is installing Avaya switches and media gateways.
The new integrated approach to technology is believed to be one of the largest ever deployed in New Zealand bringing new generation technologies and significant business efficiencies across the State, NZI and IAG insurance brands.
IAG now has about 1800 staff across the country including those at 10 call centres and 42 sales centres and branches. The new IP-based technology will mean operators can work seamlessly from anywhere in the country and it can also operate a virtual head office between the main centres.
Desktop and server management has been outsourced to Gen-i. The shift has required a number of bridging efforts to ensure continuity and minimise risk. Citrix thin client technology used by State and IAG meant the companies could communicate with each other fairly quickly.
A storage area network (SAN) to enable cross company data sharing and a read-only data warehouse for frontline staff engaged with customers to look at historical data with a browser look-up were the next steps. “We’re going to convert current data, but historic data will stay in the warehouse as long as is required,” says Rusby.
Substantial scripting and coding is necessary to convert current customer and policy information to the new system from several different sources. “We know what the target looks like but we have to work out the translation scripts.”
The end result will be a common infrastructure for everyone, including operating systems, LANs, databases and computer telephony integration (CTI). Older systems requiring paper shuffling will be displaced with automated workflow and project management. “We’re going to be retiring three legacy systems on to one new platform,” says Rusby. The go live date for the first phase is early 2004.
While essential data needs to be moved off systems with an uncertain future, some may happily grind away for years in splendid isolation. Allan Green, network systems manager with pharmaceuticals company Bristol Myers Squibb (BMS), says his company has derived great benefit from continuing to support legacy applications.
He was forced, however, to re-evaluate his options through the regional consolidation of financial shared services data centres, which required everything from the Asia-Pacific region to be relocated to Melbourne.
That meant Green would be responsible for managing a plethora of different servers running different applications in a warehouse already becoming overcrowded. Some rationalising would be unavoidable. BMS’ legacy environment, while rich in operational functionality was silo-based, complex and inefficient.
In the past the marketing department for example, would bring in an application to run on its own server. Under the consolidated model, Green wanted to preserve the richness of the legacy environment, but expand capacity with the best storage area network (SAN) he could find; which in turn required an upgrade of its AS 400 servers. “We consolidated to 23 servers down from 100, which provided more space and allowed us to implement SAN infrastructure to maximise disk space more effectively,” says Green.
Some of the legacy systems had to go, as IBMs eServer option could not support BMS’ existing Nortel server environment. He shut down some of the oldest, least functional legacy systems, supporting a gradual consolidation that would leave the silo-based mentality behind. “We’ve brought more legacy applications across to the new environment smoothly and without the previous headaches of integration,” he says.
When you have always done things a certain way, your systems shape work practices. A lot of intellectual property is tied-up in those processes. Tossing out the old cannot occur overnight, it requires a careful exit strategy.
Fairfax Group, which acquired the publishing interests of INL in New Zealand, is relying on in-house innovation to extend the life of its Collier-Jackson (CJ) advertising system, while embarking on a replacement strategy to retire its ancient Atex editorial system.
“In the short term it’s a complicated process, because you’ve got work practices built around those older systems that are not easy to change,” says group operations officer Brian Ashton.
Fairfax realised there was some life left in the CJ system upon which it had come to rely. Replacing both editorial and advertising systems and the underlying infrastructure and workflow would have been too much change. “It may not have the reporting dynamics and flexibility of a modern psuedo financial system, but we can probably still do that with some innovative thinking and the use of modern reporting tools.”
CJ, which has been running on Vax minicomputers and is about to be moved up to Alpha servers, remains stable providing considerable functionality across Fairfax publications. It takes bookings for each edition across the chain and stores information about clients and how much they spend.
“We’re looking to make enhancements around it to give the business the information it needs to be more progressive. We’ve been using software called TM1 which is an OLAP tool to extract data that was not otherwise possible to obtain, and we’ll be building an easier web-based reporting on the front-end of that,” says Ashton.
Editorial content presents a more complex set of problems. Atex was first deployed around 1985 on old J11 mainframes, which ran classifieds and paging, taking INL from the hot metal days into the first stages of electronic publishing. While there haven’t been any updates for a while, the company did enable PC emulation so operators could access word processing and email alongside the text-based interface. Workflow and page flow management is done manually.
The major failings are that it’s text-based and uses dumb terminals; so definitely not WYSIWYG. “In a modern newspaper you need to see what the page looks like.”
Another major issue is support. “Our support contract is through Australia. There’s only one engineer in New Zealand and we probably have more expertise in our company than with our contractor. That’s not unusual when you have legacy systems,” says Ashton.
He agrees the old system should have been retired some time ago and the acquisition by Fairfax moved things on. “There was always a need to balance out the risk to the business against the cost of replacement. We haven’t lost a newspaper in the four years since I’ve been here, so I think we’ve done a good job of managing it.”
Fairfax has six Atex sites – the three largest are the Dominion Post, the Sunday News and Sunday Star Times and the Waikato Times. The Dominion Post will be first to get a makeover, and that is expected to take a year. The others will be replaced progressively. Ashton believes a complete swapout could take up to four years.
The group will take a page from Fairfax Australia’s books by taking on the Intel-based Genera System from Atex Media Command. The system will run on Microsoft Windows using an SQL server platform with PCs at the front end. “We really need to move on to give stability to our newspapers and provide some new functionality and workflow. We’ll replicate some of the functionality, but rebuild the rest,” he says.
The research director of the CSIRO’s ICT Centre, Dave Abel, says IT systems become outdated because operational needs change quicker than technology, and faster than companies can invest. He says most companies made their last IT investment in the 1990s but few operate in the same business environment as they did then.
The IT industry had a habit of selling applications as silver bullets. Some organisations locked themselves into using particular applications, thinking they would not need to be upgraded. “There’s massive investment tied-up in the hardware, software and the business processes that have crystallised around that. You have to be a brave organisation to let go of the value within that.” And he warns of the risk of moving to a brand new system. In the past, such systems have often sucked-up massive investment, failed to deliver or been terminated before the project was finished.
Rather than ‘forklift upgrades’, he suggests incremental change can remove some of the risk. “With incremental change rather than massive leaps, gradually you get to the point where you can shut one part of the legacy environment off altogether. It goes back to change by 1000 cuts rather than a nuclear bomb every 10 years.”
CSIRO has been able to re-purpose its legacy environment for new business requirements although web services add another level of possibilities. “Web services allow you to extricate from your existing legacy systems certain functions by writing some sort of router,” he explains.
For example, a web services-based approach to managing and integrating disparate legacy systems using a uniform language such as SOAP allows users to access a large number of discrete data sources from legacy servers and applications. This uses the connectivity of the internet to avoid the bottlenecks common to hub-and-spoke exchanges, where all data must pass through the same point.
Rather than treating a legacy system like a monolithic application, you treat it as a system that can perhaps carry out 50 to 100 discrete operations. You then build a new system that can aggregate those operations into a meaningful business process, says Abel.
Gartner research director Steve Bittinger agrees web services will be increasingly crucial to getting the most out of legacy systems. “It helps us glue together the old stuff and the new stuff.” However, he warns many legacy systems will struggle to survive as technology develops.
You can put a web services wrapper around any old system and use it to extract valuable information from legacy architecture, but the internet-based, always-on IT systems of a web services-enabled business will leave most legacy systems behind, says Bittinger.