The manager of a large data centre was once asked what the ideal state of automation would look like. He thought for a while, and said: “It would be managed by a man and a dog.” What would they do? “Well, the man would be there to feed the dog,” he replied, “and the dog would be there to bite the man in case he touches something.”
Humour aside, the allegory contains more than just a bit of truth. As more and more data centres turn to the concept of virtualisation in order to pool and share IT resources, thus hoping to squeeze more performance out of existing infrastructures and increasing the ability to respond quickly and flexibly to changing business needs, human intervention is becoming more and more of a liability. Therefore, if you do virtualisation, you will almost certainly have to think long and hard about automation.
However, virtualisation can mean many things to many people, virtually anything from simple server partitioning
– putting more than one application onto a single machine – all the way up to creating a sophisticated fullyautomated, 24x7 “lights-out” data centre where a man and a dog sit warily eyeing each other.
“The reality is that you actually can have virtualisation without automation. Most customers who have already implemented a form of virtualisation have done so without automation,” says Malcolm Garstang, HP EMEA programme manager for Virtual Server Environment Solutions, “but in terms of agility, speed of deployment, transfer of resources, cost savings, and resource optimisation, the minute you start to automate, either by taking steps or people out of the process through automation, that’s when you start to see true or greater or maximum business value.”
Say you have a consolidated server, running three or four applications simultaneously, and suddenly there is a surge in demand for one of the applications. “There are two ways to solve the problem,” Garstang maintains. “You can either take resources from somewhere else and assign them to that application, either by bringing in more CPU capacity or bringing in more servers, or you can take that workload and move it virtually to another more capable server. “Unfortunately, most IT organisations are not mature enough to do that automatically. Virtualisation is about using the tools that are available today from companies like HP to automate that process and to make it happen transparently to the users and to the IT department.”
The problem, he believes, is that many IT professionals today don’t feel comfortable with the concept of automation, and have processes established to ensure effective control and management of their resources. “They want to have a human being saying, “Yes, I want to do that, I want to move that application to another server, yes, I want to add more processing capability to that server.” Some of that is due to a lack of understanding in the capabilities of the tools and the level of trust in that ability to maintain service levels, and some of it is down to some other things like managing cost and licensing. So while many of the tools are there today, general opinion is that automation is really still the promise of the future.
So, what are the tools that can help data centre managers achieve that vision? “It depends on the kind of automation you’re looking for,” Garstang says. One of them, HP’s Global Workload Manager (see article Virtual baskets, real savings), he believes can be a key enabler for automation because it uses predefined policies to enable the automated movement of resources between servers. Others tools, such as ProLiant Essentials Virtual Machine Management Pack, Virtual Connect for HP BladeSystem, and HP’s Virtualization Manager for the Virtual Server Environment products, allow seamless, dynamic movement of virtual machines from one server to another or from a virtual to a physical machine, or from blade to blade. All of this effectively managed through a single management tool of Systems Insight Manager.
While some data centre managers may eschew the idea of handing over control to a piece of software, they may actually welcome tools that serve in more of an advisory capacity, Garstang feels.
“In a virtualised world, if you’re going to dynamically move applications in virtual machines around, then it’s very important to understand the potential impact on the other applications when you bring a new application into that shared resource pool. HP has a tool we call the Capacity Advisor, which is linked with some of our other management tools such as Systems Insight Manager, and basically captures real performance data on the server workloads. It then gives you the ability to take workloads from, say, virtual machine A and, without losing the resource, move it logically to virtual machine B, to see what the impact would be on the total system. This is not theoretical, like, ‘We did this before so this is what it’s got to do in the future’ – it’s using real data. The net is that it will tell you beforehand whether you will have enough capacity in the new server to cope with the additional application. You can then either automate or continue to move the process manually. It’s up to you.”
In many cases, unless managers have that information, they may find themselves running out of resources literally from one second to the next. The danger lies in sudden over-utilisation, as two or more applications on the same server experience usage spikes at the same time. “You may have 70 percent free capacity one moment and none the next, especially if the spikes don’t allow the dynamic matching of those applications. Having a tool that gives you advisory information to enable you to effectively move resources around dynamically – and, should you choose to, to automate that process – is definitely a cool thing to have,” Garstang believes.
It’s either that or a dog.
|
Printable version
