Building better boxes and pipes

Finance or engineering, or whoever will have decided to buy a server and some storage, and then they need to add to it, and after some months and years, suddenly there's a room full of servers. And they're still doing it." So says Martin Cooper, global operations manager at engineering, construction, IT and comms consultancy Arup – and it's a picture of the real world that will strike a chord in IT departments up and down the land. "Then you've got the cost of all those disparate servers, applications, the storage – and the management of that whole lump, with all the back-up, business continuity, disaster recovery plans... It's a huge and expensive pain." Absolutely. It's been said so many times: while companies' IT and network infrastructures are regarded by those that control the purse strings as vital, they're simultaneously dismissed as 'just the plumbing'. And more often than not, without a voice on the board or any senior position, nothing's going to change. The result, as Cooper says: "IT people are increasingly stretched, because although their responsibility has mushroomed, IT departments typically haven't grown. If anything, they've shrunk. And the bottom line is we don't have the time to put the TLC [tender loving care] into the departmental systems that they really need." Which in fact at best puts that 'vital' infrastructure at risk, and certainly means that it's costing potentially far more than it should. So what are we going to do about it? Worth getting some ideas from those that have not only recognised the problems but had the clout to do something about them and been remarkably successful. Needless to say, Arup's Cooper is one. His view, which is shared by many: "The thing is to centralise and consolidate." He concedes that that in itself is nothing like as trivial as it sounds, but insists that it's what his firm is doing, and says: "The key is planning. Once you know what you've got to do, you can do it. The hard part is to get a handle on your systems, the storage, the whole fabric of what you've got." And he emphasises the importance of taking a holistic view right from the outset, not simply focusing on individual problems: otherwise you simply perpetuate the difficult and costly IT issues. That done, however, he's adamant that storage will be central to most improvement projects. "Look at any server on the network and you won't find one running anywhere near 100%. But it's very common to see storage running at 98%. Processing is relatively easy: you can buy Blade servers to increase processing power, and managing all that is less critical anyway. If a server stops processing, the worst that can happen is you lose that capacity for a while. But if you get your storage or a back-up wrong, you can lose days, weeks or months of work." Consolidation focus Before consolidation, Arup's London IT infrastructure, which covers the HQ and 12 offices across the city, relied on around 100 servers using DAS (direct attached storage). There were several issues beyond the obvious. Arup relies on staff working flexibly and sharing information between teams, but with departmental proliferation that was difficult. Also, looking at the Building Engineering division – where the firm piloted its consolidation project 18 months ago – it's split into eight groups, each with dedicated e-mail and data servers, in three offices more than a mile apart. That meant 16 servers to manage individually. "Our staff needed storage for a range of data, including office automation documents, CAD/CAM engineering files and large spreadsheets," Cooper explains. "The existing systems required 16 separate back-ups and a high level of support. We wanted a solution which would simplify system maintenance, reduce costs, decrease support time and also reduce downtime on data recovery." He refers in particular to Microsoft Exchange Server: failures, he says, were around every six to nine months, and recovery times were more than 24 hours. "We had massive problems with Exchange going flaky – the database corrupting itself, for example." Arup's Advanced Technology Group went for Network Appliance solutions and its NetApp NAS (network attached storage) systems with a remote system mirroring the Exchange site. "We considered other companies, like EMC, StorageWorks and Hitachi, but none met all of our requirements [and] the NetApp Data OnTap operating system and Snapshot technology won us over," says Cooper. "We took 16 servers down to one. Our solution provides a robust central storage solution with integrated software that ensures rapid recovery times for Exchange." Significant savings And he adds: "We went through a meticulous migration phase to consolidate eight mail sites into one without a single outage, and the project was completed on time and 15% under budget." The result: storage consolidation reduced back-ups by 90%, with Snapshot also enabling, for example, system administrators to perform rapid online back-ups with minimal disk space. "That was Phase One: we implemented 1.6Tb of storage under the one system. We've just signed for Phase Two for the rest of the London offices, which will take that up to 136Tb, with 130 servers going down to three, and remove tape back-up altogether. Everything will go to disk, using Network Appliance and its quirky RAID4 technology, which is faster than standard and provides double disk parity in RAID sets." For him that's another big bonus, virtually eliminating disk failure risk – on top of what's already achieving everything he's asking of it. Before we move on, it's just worth catching some useful observations. Cooper says staff are now able to work across divisions, with CAD/CAM users in particular "finding file access significantly quicker." He also says: that users can now restore accidental deletions and the like directly from Snapshot copies without resorting to IT; that the last time Exchange had to be recovered it took just 25 minutes; and that the remote system makes management for disaster recovery easy. "The SnapMirror software on the NetApp server took about five minutes to set up and hasn't required any further attention." So there's food for thought. Moving on to networks, consolidation is again a significant driver, but the smart money is also on modern systems and in particular managed services. Peter Clarke, group IT manager for £500m Yule Catto, the world-wide speciality chemical group, for example, took the entire corporation onto a Vanco-hosted IPsec WAN (wide area network) almost five years ago – despite everything else, including a spread of ERP systems, being managed separately at the company level. "We wouldn't say 'here's the standard ERP system for the group; use it.' We believe that would just impact them negatively. But networking is something that can be the same throughout," says Clarke. He indicates that the companies, as they were acquired, already had WANs, some quite extensive, linking, for example, multiple sites across Europe and Malaysia. "We looked at putting in one network for them all and rolling it out for the whole group … but you mustn't underestimate the effort involved in international plumbing. We've tried dealing with telecom companies, and it's a nightmare. We couldn't do that ourselves." And that's one point: you can't do it. Another is you probably shouldn't. Clarke again: "The Vanco network handles all of that, and you have the security of a VPN [virtual private network] and resilience against web attacks and viruses. We couldn't develop anything like that with individual systems, with the level of security and availability and cost that they all need. And without it, you'd have the trauma of getting through 30 configured firewalls." Paul Rumble, senior vice-president at Electrolux IT Solutions, Electrolux's technology arm, makes another simple but key point in favour of the service option: "If the network fails, the business very quickly fails. But the technologies are changing very rapidly so it's very difficult to keep up with it all. By outsourcing, keeping your system up-to-date and secure is done for you." Cleaning up Electrolux is one year into a renewed five-year, $29 million contract with networks firm Equant, which provides its world-wide MPLS IP VPN (multi-protocol label switching – like a VPN but running over privately controlled circuits and offering 'classes of service') connecting 160 sites in 33 countries, from the US, to South America, Asia-Pacific and Europe. "We do know the technologies used, but it's not core to understand them. The issue is that in some respects we don't care because we no longer have to; we've outsourced the technical details. What we care about now has shifted to managing the supplier. We're purchasing a service." In fact, by moving to the global MPLS VPN, instead of multiple networks meshed together, Electrolux not only gets more secure, high performance, any-to-any connectivity in a private network environment, but the ability to prioritise traffic to maximise bandwidth utilisation and optimise business performance, and opportunities to go for integrated services, like voice and video for IP VPN. Rumble adds that by moving to a virtual IP VPN "you're just connecting into the cloud. You're just buying bandwidth at each location. It's as simple as that, and much more efficient than trying to deal with all the networks and telcos yourself." Incidentally, for Electrolux the Equant solution includes business continuity and disaster recovery. Just as important, Rumble also reckons it saves money. "It's difficult to say: nobody owns their own networks today anyway," he observes. "But moving from frame relay to IP VPN with a single supplier saves around 14% TCO [total cost of ownership]." And he agrees that managing the supplier, as opposed to the network, is far easier, so there are savings on top. How far should you go with outsourcing? What about on-demand computing, for example? Rumble sees advantages all the way: "We want to move to a position where we purchase more computing as a service. From an IT perspective, we just need to own the strategy around our applications. There's no sign of releasing that." Service with a smile Returning to Arup, however, it's a different picture at the moment – one of engineering-led precision control. On the LAN side, Cooper refers to heavy CAD/CAM and engineering traffic running over a Gigabit fibre network between the London offices and 100Mb to the desktops. "That's our Metropolitan network. Beyond that we have our own private WAN using ClearChannel, with leased lines hopping out to New York and a hub and spoke arrangement from there to the rest of the world. There's also VPNs for home and small office remote users." But although he's happy with that arrangement, saying "its good and reliable", he too reckons the next move will be for MPLS IP VPN with layers of redundancy. "We're predicting better service and lower costs, although I won't believe it till I see it," he says. "But if all the numbers pan out we could be saving one third of the total cost." Impressive; but if you're thinking 'I'll have some of that', hear Cooper out. "You have to understand the collection of all the parts. We constantly find people looking at just the storage or just the network. They need to look at the system as a whole. What about the server room, for example? Is there redundancy there? What about your WAN connections? Are there separate routes? Some companies haven't even got their power supplies covered. "Our advice is to make it straightforward so that managing it becomes easier and more obvious, and there's less to go wrong. The trick is first understanding how your equipment works, why it works the way it does, and the implications. You can have the best SAN [storage area network] in the world, but if it's in a server room and there's only one connection, you've got an instant point of failure." Time to get cracking.