Is your infrastructure right for your business?

Putting IT costs through server and application consolidation projects should be high on the agenda for hard-pressed manufacturers today, but so should maintaining a secure, resilient network infrastructure. And so also should building an integrated system able to respond flexibly and fast to inevitably changing business requirements – whether they're about producing more with less, sustaining growth with minimal investment, or changing direction. Agreed so far? Either way, having invested considerable sums on enterprise IT at all of the levels, probably in piecemeal fashion and with 'varying' success in terms of return, you're probably loathe to spend more, whatever the need. At the very least, you'll want a steer on prioritising further investment and, in light of what's always fast-changing technology with new cheaper, better possibilities, reasonably precise ideas about cost/benefits and what's going to deliver most where – and the implications. That being the case, we asked the 'great and the good' to do that job for us all – to provide advice and guidance on the plethora of potential initiatives around that all-encompassing word, infrastructure. Fact is there's a lot to go at here – hardware, software, services, networks – and a lot of dependent choices. Also, what might appear to be top line priorities for some will be anything but for others, while others again think of it all disparagingly as simply the plumbing. However, consolidation should be right up there. As Mike Lucas, technology manager at services and software tools provider Compuware, says: "There's a big wave of server consolidation now aimed at reducing complexity, reducing licensing and maintenance costs of the platforms. Where we were all on mainframes and moved to many smaller machines, now we're going back the other way." Hardware and software Note straight away, this isn't just about hardware: application consolidation is arguably where much bigger money can be saved. Moving from multiple bespoke suites to ERP, or reducing the number of instances of an ERP system and going for a single provider in a multi-site, multi country operation, can, over time, save a fortune – although it's still not a trivial exercise. Lucas again: "You reduce the complexity and the cost o f management, you cut the number of staff and the skill sets you need: you're cutting the costs that sit behind it all. In a federated control environment, you have a lot more people with shadowed responsibilities so there are a lot of hidden costs in terms of people with other functions who should be doing something else." There is, however, a trade-off. The more you consolidate, the more network traffic, bandwidth and capacity you need (although you can mitigate that through thin and browser-based clients) and the more assistance you'll require from network support, deployment and analysis tools. All of that is handlable, and it's all a good way to go anyway – there are efficiencies and economies of scale to be had from today's better management tools, single user interfaces and network, server, application and desktop management systems. But you need to be aware of the importance of capacity planning, trending, performance analysis, getting into the detail of usage patterns, providing appropriate tools, defining roles and ensuring proper stress testing. This is business-critical. Lucas makes another point that's open to some debate. Looking at the hardware, the application, database, web, desktop and file and print servers, he notes that whereas, in the past, applications were built for specific platforms, now virtually everything is on flavours of Unix or Windows. "So the obscure platforms can be progressively chopped out, and users can go mainstream with HP, Sun and IBM." Commodity systems And there's open source. The point: it's the application that's strategic, not the platform. "All servers look the same, and commodity pricing is now with us for the high end servers too." And hence even mission-critical applications can run on large, but low cost server clusters – and the same is happening in CAD/CAM, with the promise of Intel boxes running Unix for compute-hungry analysis and simulation. "Hardware prices are just becoming less and less of an issue." Sounding worth it? Clearly, but it's not necessarily as straightforward as it might seem. Although, on the face of it, you can cut costs, for example by using more open source (at least at the server level), manufacturers need to look at the total cost of ownership and resilience of alternative solutions. Messing with mission critical systems away from the tried and tested platforms is a risky business. Either way, it's worth noting a thought from Steve Craggs, European chairman of the EAI Industry Consortium. He reckons it's likely that chopping down the number and type of boxes to just a few is going to deliver the easiest bang for your buck, because, apart from the networking issues, it's a relatively transparent exercise. But it's not going to deliver the biggest saving. Moving to the same set of packages and tools will save you more, but the return will be slower. Craggs: "If you can reduce your infrastructure from 50 applications running on 20 platforms to three on one, that's far cheaper to run and there will be quality of service improvements." You've also just upgraded your entire system, so there should be great opportunities for innovation and competitive advantage. You could, for example, get into web-based supply chain interaction, or profit from enterprise-wide business intelligence, or harness web technologies for rapid order entry, product configuration, scheduling… There's another twist here though. Steve Ashurst, European sales and marketing director for legacy applications transformation tools firm SEEC, reckons that investigation and capture of companies' legacy application environment, followed by simplification and migration to Java, is another important option to consider. He also suggests that, once migration has been done, we should think about outsourcing management of modernised composite applications to an offshore company – again to save money. Hidden value Moving on to that other hot topic, enterprise application integration (EAI), one of the points that everyone in this area raises almost immediately is that it's not just about connecting applications. Another is that costs, complexity and what can be done have changed dramatically in the last two or three years alone. And a third is that while it delivers little immediately apparent value in terms, for example, of competitive advantage, the key is that it provides a foundation from which companies' aspirations can be met in the lowest cost, most agile manner. Why? Because you've created the potential for a 'real time business' with immediate information sharing, automation and the rest; you've re-used your existing application infrastructure, rather than ripping and replacing; and, given appropriate technology choices, your systems and people are open to any and all modern IT-enabled business initiatives. So there are several questions. What should you integrate? What are the issues? And which technologies should you consider? John Dillon, European marketing director at WebMethods, brings us up to date. "The 'A' in EAI was necessarily for applications, and EAI used to be essentially about just connecting them. But industry has recognised that businesses don't only have seamless, straight-through processes; there's also the need for human interaction, so the concepts of business processes and workflow become relevant too." And hence the range of business process management (BPM) tools vendors today. "All that history of solving integration problems has got us to where we are now – thinking about enabling applications and people to work together more effectively using the sum of their systems." It's fair to say that ERP systems captured manufacturers' imagination early on by providing business integration, albeit in a monolithic 'single application' format. But Dillon makes the point that companies still have to communicate digitally with people and departmental and/or legacy systems not touched by their ERP, as well as with partners, suppliers and customers. "Now, the EAI arena is evolving primarily because of Web services and in particular the standards underneath them," he says. And he adds: "Web services have put the 'service orientated architecture' [SOA] back on the agenda." Back? Yes, SOA isn't a new idea – it's the latest acronym for a concept from the '90s around object-orientated software, when the name of the game was code re-use. Now it's more about exposing functions in existing applications as services, making them re-usable and callable, linking business data and workflows through them – and providing a messaging backbone to make all of that work. That's where the commercial advantage starts to come in – you get away from traditionally isolated departmental systems and data structures. The difficulty, of course, is understanding the inventory of code you have, and getting the granularity right to produce useful mapping – although most believe it's not that hard: for typically 60% of requirements it's accessible and can be wrappered. Meanwhile, with standards – notably J2EE versus Microsoft .Net, and Web services embracing both – creating a robust SOA is more feasible for more folk. "This is the next wave for the IT industry," says Dillon. "And it also means the re-emergence of plug-and-play and best-of-breed. As standards are proven to be working and embraced by everyone, there will be significant business benefits." Although Wolfgang Gebhard, who runs product marketing in Europe for EAI software developer Tibco, warns that there's still plenty of grunt work: "Standardisation, like that in Web services, is very helpful for integration, but it's not sufficient." However, Mike Gilbert, director of product strategy at IT consolidation specialist Micro Focus, goes as far as to suggest that "SOA is the second age of computing." Meaning? "The first age was about automating functional departments; the next age is about automating entire businesses and enabling them to become flexible and responsive to events and change. Functional silos are now saturated: the job of the next five to 10 years is going to be bringing them together with a simple, standards-based but powerful infrastructure architecture." Integration on tap Rather than getting too bogged down in the detail, the key point here is to understand that in cost/benefit/difficulty terms there's been a step change in the last couple of years. Dr Ian Howells, vice president of marketing for integration big-boy SeeBeyond, notes that whereas five years ago integration was very expensive and a big company undertaking, today's choices are "very much cheaper, and the time to roll them out is very much less." Indeed he suggests a five-to-one ratio of productivity with modern systems and tools against older systems. There's also more providers: total infrastructure providers like IBM and HP, users' own enterprise application vendors, and the specialist integration system vendors, like SeeBeyond itself. Which means you have a choice. As for the technologies, there's also a choice, and which you need depends on what you're trying to do. Everyone I spoke to was pretty consistent here, mentioning: portals for 'loose connection' – presentation really – internally and externally to databases and applications; data warehouse technologies for non real-time; OnRamps and adapters for data conversion; and messaging and brokering systems and SOA-based enterprise service buses for navigation and transmission. Gill Corfield, technology support manager at software and services firm Compuware, is good value here, structuring the options according to the three main layers – the data, business logic and presentation tiers – and making two main points (see panel, page 39). The level you attack the issues from dictates the outcome: in other words, if you're thinking is stuck in point-to-point integration at the data layer, you'll achieve an objective, but no more for the business. And businesses need to think of integration as a dynamic and ongoing process, always moving to match changing requirements. If that's sounding too heavy, Tibco's Gebhard, reckons users "shouldn't care about the different technologies; what matters is … solving the business problems." For him that's where the definition process starts, and it leads to the information assets to connect, how tightly, and thus to the adapters, Web services, 'wrappering' technologies (page 44), messaging technology backbone and the rest. Compuware's Lucas sums it up nicely: "Go for business enhancement projects first. Look for cost reductions as ongoing housekeeping projects – they're part of running the business responsibly. But watch for resources: that needs periodic review, and companies are generally well advised to consider outside help at least for the 'discovery' and solution roadmap phases of any big project involving business improvement."