Why is automation such a big topic in IT management all of a sudden? After all, it isn’t as if it were exactly new: automation batch processing has been around since the halcyon days of mainframe computing back in the 1970s.
However, automation is back on centre stage again, as IT managers struggle with rising costs and increasing risks. Data centres are actually spending anywhere up to 65 percent of their budgets today on routine maintenance tasks, says IDC, with only about ten percent left over for innovation and development.
At the same time, availability issues keep putting additional pressure on IT managers already hard pressed to keep their straining systems up and running. Here, the human factor plays an important role. Gartner* estimates that 40 percent of unplanned downtime is attributable to operational errors – someone pressing the wrong button, activating the wrong script, deleting the wrong database or erasing the wrong file.
(* “Up and running”, The CTO Forum, January 2007)
Combine both, and you have the ultimate IT manager nightmare. Enter automation: by reducing operation errors and routine tasks, automating key parts of your data centre management can win back valuable technical, financial and human resources, freeing them up for those important strategic and innovative projects. All in all, this sounds like a pretty good way to please your friends from the business side of the company.
But where to start? At the very beginning, says Phillipe Roux, Automation Programme Manager at HP. “Automation is a journey, not a project,” he believes, so planning the trip down the automation highway can be crucial to a company’s actually reaching its desired destination.
One good embarkation point, depending on a company’s individual requirements, can be configuration management: managing features and assurances through control and testing of changes made to hardware, software, firmware and documentation. The object here is to gain comprehensive insight into what systems are actually available and what they can do at any given moment. This in turn enables IT managers to reallocate resources, for instance shifting them towards development of new applications.
Another place to start can be software distribution and patch management, those tedious and repetitive tasks that are the bane of every IT manager since they freeze up expensive manpower. Besides, they tend to be highly prone to human error, as every computer expert since Murphy can attest.
Once automation has gained a firm footing inside the data centre, managers can tackle some of the more demanding tasks such as server and storage provisioning, automated backup recovery and archiving, and resource allocation designed to better meet IT’s obligations from service level agreements.
However, IT managers should avoid the mistake of relying too heavily on technology as a save-all, says Roux. “Technology is only about 30 percent of the game,” he maintains. The rest, he feels, is about establishing the right processes and making sure that organisations adapt in a timely and efficient fashion.
“You need to write foolproof processes that will do the job,” he says. This means ensuring that appropriate actions are triggered automatically whenever certain things happen. If demand on your e-commerce servers becomes too great, response time will go down, causing more and more customers to break off a transaction or to surf over to the competition’s website.
By automatically provisioning additional CPUs, storage or servers, systems like HP’s Global Workload Management software can in effect inject oxygen into your system, making sure that customers keep coming back for more.
Processes like these must be written into model-driven automation schemes, in which each anticipated event is carefully vetted to ensure that the model, and hence the system, know exactly what to do in a given case. “These events need not only be technical,” says Roux, “they can be business scenarios.” So here again, the need to know how business processes should work is essential.
IT automation can therefore have significant impact on the organisation itself. After all, one of the reasons top management usually signs off on automation projects is because they expect IT to be able to make do with less – less money, less manpower, less risk, less fuss and bother.
Automation can allow for remote control of businesscritical systems, with managers, for instance, sitting at a dashboard many miles or even continents away, monitoring performance and workloads, while less qualified personnel onsite can be relied on to insert the right cartridge into the right server when told to do so. “This doesn’t mean you actually have a lights-out data centre,” Roux insists, “but it‘s getting pretty close.”
Naturally, some IT people may have some qualms about losing their jobs to automation, but such worries are usually groundless, at least in Roux’s experience. “Yes, you may have to re-skill some of your staff,” he believes, ”but in the end what we are talking about here is freeing them up for tasks that are more demanding and hence more rewarding.”
One question IT managers frequently ask is, “Do I have to automate 100 percent from day one?” The answer, Roux says, is no. In fact, he feels that by starting out with smaller projects and by concentrating on low-hanging fruit, automation can create “instant success stories” for IT professionals.
Take identity management, for instance. Instead of doing the hard things first, like asset provisioning or identity federation, IT managers can opt for the relatively simple task of automating password reset. Or they might choose to clean up their directory systems first through a process called “deprovisioning”: getting rid of dead accounts and unused software licences, thus in effect paying for the project out of direct savings. Other “quick wins” may include automating the backbreaking work of backup and recovery.
Another good place to start might be storage provisioning. Today, many data centres use what experts sometimes refer to as “thin provisioning” of storage resources, namely allowing space on large-scale centralised computer disk storage systems, SANs, clusters, and storage virtualization systems to be quickly allocated to servers on a “justenough and just-in-time” basis. Roux calls this method “light overbooking” and compares it to airlines accepting more reservations than the plane can hold on the theory that a certain percentage won’t show up anyway. However, as he cautions, “Passengers and IT customers really hate it when they get bumped – so don’t!”
Of course, most data centres today use different systems from different vendors, which can cause some major compatibility problems. One way to effectively automate and manage a multi-vendor storage system, is through assigning one storage system the role of a central hub, Roux suggests. HP’s ProLiant storage solutions, for instance, can act as a kind of traffic controller within a multi-vendor storage environment, ensuring that each system is put to best use. In addition, HP has recently acquired Opsware, the manufacturer of powerful tools specifically designed for automating multi-vendor capabilities (see article Factoring out the human factor).
Like automation, virtualisation is itself another hot-button topic in IT management today, and there is often some confusion about the two. No, they are not synonymous, says Roux. However, you can’t have one without the other. “If you do virtualisation, you almost certainly will have to think long and hard about automation.”
That is exactly what people at HP Labs are doing. Described as “the new Nirvana of data centre computing,” IT automation has been placed at the very top of the agenda there. By applying well-known techniques used in physical systems such as control theory, statistics and formal mathematics, the scientists at HP are working to develop IT systems that are more automated, less prone to errors, and more agile and predictable.
"Our vision," says Kumar Goswami, who leads much of HP Labs' automation research, "is one in which the user expresses the 'what', and our tools do the ‘how’." He also feels there is much to be learned from looking at systems like automatic transmissions in cars or the autopilot feature of an airplane.
One result may soon be systems that can automatically diagnose businesses’ problems. "This is a very difficult problem for our customers,” Goswami says. “They get tremendous amounts of data from applications and from the systems, and they’re under pressure when the system is not working properly to diagnose what the issue is and fix it in a timely fashion."
By using statistical and machine-learning techniques, HP researchers may soon be able to help customers shift more quickly through data and determine what will be most useful in solving a problem. In addition, they are developing technologies to take a digital "snapshot" of the system when it is not functioning properly and compare it to previous snapshots to determine if a particular problem occurred in the past and how it was resolved.
"By automating tasks like these, you're able to create an agile repeatable, and more importantly, a very predictable IT environment," Goswami says. "That's what CIOs today are really looking for."
|
Printable version
