Virtualised infrastructures in the data centre

An industry expert discusses the key to a successful management strategy within a virtual infrastructure, including automation and holistic adoption.

Virtualised infrastructures are gaining more ground in today's data centres because they help use expensive resources more efficiently. But with virtualisation comes a number of challenges related to management.

First, the nature of virtualisation itself. On one hand, it helps to reduce the number of physical resources that need to be managed (and server virtualisation is a clear example of this, where many virtual servers are hosted in a single physical server). But on the other hand, it allows for a nearly unlimited number of virtual resources to be created.

Server virtualisation is a good example; the number of virtual servers in a virtualised data centre greatly exceeds the number of physical servers before virtualisation. And every virtual server requires other virtual resources like LUNs, for example.

Virtualise the whole data centre, not just part of it.

Roberto Salucci, SNIA Europe Italian Committee member,

In a virtualised infrastructure, the total number of resources (physical and logical) to be managed increases dramatically and is a major contributor to management headaches. Business requirements add to these issues. One way of addressing them is through real-time (or near real-time) provisioning, one of the benefits offered by virtualisation.

Beyond increasing infrastructure efficiency, virtualisation enables flexible provisioning, an equally important benefit. In the current economic climate, IT departments must react in near real-time when business conditions change, because markets do no wait. In fact, in a virtualised environment, the word "flexible" is often coupled with "fast."

Another benefit of virtualisation is system availability, as this enables thousands of virtual resources and hundreds of physical resources to be monitored and managed quickly and without errors.

A Gartner study shows that 40% of unplanned outages are caused by operation errors, and these are more likely to increase with the complexity of the environment and the need to make decisions quickly.

Keys to successful management in a virtualised environment
Virtualise the whole data centre, not just part of it. Doing part will cause many more issues than virtualising everything. Strategies like, "I'm going to virtualise servers today, leaving storage for later when it will be needed" will not work.

Physical resources cannot be provisioned as fast as virtual ones, and are not as flexible. Matching virtual servers with physical storage or virtual storage with physical servers can be hard, but provisioning an application with virtual and real resources in a short time could be nearly impossible. To be effective, virtualisation must be adopted holistically.

Automation is the key
One of the characteristics of virtualisation is that virtual resources are perceived as inexpensive (a virtual server or a thin-provisioned gigabyte do not cost the same as the real ones) and hence tend to multiply. At the current pace of growth, the number of virtual resources that have to be managed in a data centre will overcome, in a few years, the capacity a human being can control. In other words, someone else will have to do the job in the future.

Options include system administrators ensuring efficient and effective management of the resources in a virtualised data centre, and employing management tools that enable policy-based automation.

These tools, triggered by specific events detected by the monitoring tools, will manage the virtual environment and take the appropriate actions, driven by a rules database. The policy-based automation systems can be instructed, for example, to monitor storage usage, predict possible shortages and prevent these shortages by dynamically provisioning thinly provisioned volumes to applications from appropriate pools of storage, or to move a virtual server from one physical server to another in case of performance problems or some fault that has been detected.

Automation environments ensure faster reaction times and dramatically reduce the possibility of operational errors inherent in repetitive tasks in a complex environment, and can consequently increase overall system availability.

The costs associated in the realisation of policy-based automation can be high, but a comprehensive cost/benefit analysis should show that the costs of not automating would be higher, and a return on investment analysis can show that the investment could be repaid typically in 12 to 18 months.

Orchestration is the future
The next challenge is related to the fact that current management tools are resource specific: there are storage resource management tools, network management tools, system management tools, application management tools and so on.

Each tool manages, and possibly automates, a specific set of resources with its own management console, its own GUI and its own Command Language Interface, and interfaces between them can be poor. Today, building a single policy-automated environment encompassing the whole virtual data centre could turn out to be a rather complex task.

As a consequence, a new generation of tools, called orchestrators, is appearing in the market. Their main benefit is the coordination, or orchestration, of all management tasks for the various types of virtual or physical resources that are intended to simplify the life of data centre administrators.

Roberto Salucci is a SNIA Europe Italian Committee member and a contributor to

Read more on Managing servers and operating systems