Information Age: News, analysis & insight for IT & business leaders

Information Age (IA): You believe that data centre technology is about to change radically. Can you explain why?

Bill McColl (BM): We're at a rare point in time when we universally agree we're in transition from using proprietary, expensive, manually-intensive systems to commoditised, low-cost, automated, virtualised systems. HP, Sun, BMC, CA, IBM, Microsoft, Veritas... everyone has basically the same message. It's all about automation, cutting costs, commoditisation and virtualisation.

IA: This is something that operators have aspired to for a long time. Why is it possible now? What technology is involved?

BM: The major vendors each have pieces of what is a pretty complex puzzle.

On the monitoring side, it is all about gathering, aggregating and analysing real-time information about what is happening across the systems and network infrastructure. There is a lot of information on this side, and there are a lot of technologies, from suppliers such as CA (Unicenter), IBM (Tivoli), HP (OpenView), BMC (Patrol) and Microsoft that provide a real-time picture of individual servers and applications.

But what has been lacking is a global view of what is happening in the data centre as a whole. Without being aware of the global view, making manual or automated decisions about changes can be problematic.

On the control side, there have been a number of very important advances in recent years. The first really important control technology was load balancing.

The second technology wave, which has grown over the past three or four years, has been in provisioning tools and technologies. This is about getting the bare iron and the software stack on stream as quickly as possible. In the past, it could take weeks to set up a server infrastructure with the right software stack on each machine. But using tools from companies such as Altiris, Opsware, BladeLogic and Microsoft, the time required can be cut to hours.

The third and in some ways the most profound change on the control side is in virtualisation. This has existed for some time on high-end machines and mainframes, but VMware, Microsoft Virtual Server and Xen [virtualisation products that enable servers to be partitioned so they can run multiple operating systems] opened up virtualisation to users of commodity servers.

IA: So virtualisation is the most promising technology in all this?

BM: Using the latest virtualisation tools and technologies, one can do things that are very powerful, exciting and new. For example, one can migrate a live transaction in real time from one machine to another. One can have 10 or more virtualised PCs on a single two-processor server. These new kinds of real-time control capability really begin to expand the incredible potential for automation within the data centre.

As an aside, it is interesting to note that virtualisation is now starting to eat into the need for provisioning tools. Rather than thinking about the process of setting up a new machine as a provisioning task - that is, as a systems admin task - in a virtualised infrastructure, setting up a new machine often becomes just a process of copying a pre-established software stack from shared network storage. The IT automation game is now moving away from auto-provisioning. Virtualisation and load balancing are really the core of the control side.

IA: And then there are technologies that help manage this according to business needs. How is this achieved?

BM: As IT automation increases, it is becoming clear that one needs a bridge between the monitoring side and the control side. There has to be a policy mechanism at the core - so that infrastructure change is driven at all times and in all conditions by changing business needs and priorities. The core of IT automation is the policy engine. This is the capability that Sychron's OnDemand Policy Server provides.

One of the longer term problems that remains is the absence of a single unifying, dynamic model of the data centre that all products can interface with. Up until now, every infrastructure management product has had its own internal model, but what one ultimately wants is to have a single source of truth - what's where, what's its status, what is happening with the network. At the moment, this doesn't exist.

Today, if one tries to adjust resources within a large, complex data centre, it might be an eight-stage process. One or more of the adjustments may fail. In order to handle these sorts of complex issues automatically and in real-time, the data centre ideally needs to have a single, unifying model that spans the monitoring, policy and control technologies used.