Lights Out On Lights Out Computing
by Robin Bloor & Carol Baroudi, Partners
When did it happen? We’re not sure we can put a date on it, but we believe it was sometime in the 1990s when ?lights out computing? died. If you are a recent joiner to the IT industry, then the words ?lights out computing? may not mean much to you. We certainly don’t hear them as much as we used to.
Lights out computing meant completely automated data centers that ran without any need for ?operators? to do anything other than monitor what was going on. And thus it meant automation of data back-ups and fully automated job scheduling and, actually, the automation of nearly all operational activity. ?Lights out? meant that the lights were no longer needed in ?the computer room? – or the server farm as it has become. And that’s the point.
Lights out computing was a mainframe/data center idea that was actually realized in some sites until the proliferation of server computers switched the lights back on again in a big 24×7 way. With tens and then hundreds and in some organizations, even thousands of server computers, the idea of a fully automated IT infrastructure is no more than a dream and the concern is how to keep the number of operational staff from increasing.
How big is the problem?
Well naturally it varies, but a survey of data centers, which an HP executive quoted to me, suggested that the number of data center staff was increasing at a rate of about 2 percent per year, and the number of operational support staff was increasing at about 10 percent per year. And, by the way, finding skilled staff to do these jobs is no trivial task.
What are they all doing?
Well, many things. They are managing data storage (SANs, archiving and back-ups), managing operational problems (application failures through to hardware failures), managing databases, scheduling workloads, manning the help desk, managing the network, managing security, managing the desktop, managing email, testing and implementing patches, configuring and installing new devices, new software, new machines and more. Where there used to be a handful of operational staff running the IT infrastructure, there?s now a gradually growing army.
So why isn?t this multitude of tasks automated?
Well to some extent it is. If you look at each specific area, you discover that for just about every kind of operational task there is some software product and usually several, that can reduce the labor involved, sometimes even quite dramatically. Many things that are not automated right now in many sites could be automated. And, as it happens, most sites I?ve talked to are implementing some of these solutions in an attempt to keep the lid on the staff numbers. But they simply do not have the bandwidth to solve every problem at once. And the overall problem is like the Hydra that Hercules fought; cut off one head and another two grow in its place.
The Underlying Issues
There is also a deeper problem. The system management frameworks that were marketed so effectively about a decade ago, did not deliver. The idea was that they would act as the glue to provide a coherent and integrated infrastructure management capability. I never came across any site that implemented these frameworks with any success. Underlying all the operational management problems is the fact that the software products that deliver automation do not integrate well.
Furthermore most IT Departments are currently organized on a silo basis in the operational area. It seems to be a general rule that where you have silo systems you have a silo management structure, and that tends to be what has happened in IT Operations.
Such a structure is effective at dealing with localized problems, but performs poorly when it has to deal with problems that span the silos. And among the problems that span the silos are service level and support problems, scheduling problems, change management problems and forward planning problems. These are also the problems, which, if they are poorly handled, give IT an unwelcome visibility.
In my opinion, this is also the area where the highest payoff can be achieved by intelligent automation.
The Usual Suspects, Plus One
In the long run the large system management vendors; BMC, CA, IBM/Tivoli, HP and Sun, will deliver – because they have to. But right now they themselves are also going through the pain of having to transform silo product portfolios into end-to-end capabilities with high levels of interoperability. There is however a way of attacking the problem directly that I have come across recently ? which involves providing interfaces between many of the popular ?silo solutions? and also providing a workflow design and implementation capability between these tools.
The company that provides this capability, Opalis, is getting traction with its software in large IT Departments, where operational management is starting to spin out of control. There may be other products like this, by the way, that I?ve not yet come across. New products in the system management space are like London buses. They rarely turn up in ones.
In any event, it?s an interesting idea when you think about it: Automation for the automators. IT Operations managed using workflow technology. The cobbler?s children wearing quality shoes.