Converged and virtualised IT installations can bring different power demands for customers says Stefan Buerki, Eaton Electrical marketing manager – Power Quality Australia and New Zealand.
IT environments today are typically characterised by unpredictable data growth accompanied by equally hard to predict performance demands from users.
New Zealand’s data centre operators are faced with responding rapidly, yet cost-effectively, to these challenges.
They are finding that legacy silo-based infrastructure no longer provides them with the flexibility that they need at a price that allows profitable operation.
As a result, resellers are increasingly turning to designing a converged infrastructure. Converged infrastructure is essentially a modular approach to data centre design and construction.
According to one of the major proponents of the converged approach, converged infrastructure 'is a systematic approach that brings all server, storage and networking resources together into pools of resources.
It brings together management tools, policies and processes so resources are managed in a holistic integrated manner'.
The basic building block of converged infrastructure is, therefore, likely to comprise CPU and memory, storage, network connection devices and, crucially for today’s systems, provision for virtualisation.
Physically, this basic building block can be realised in several different ways. It could be, for example, a bundled pod, a pre-populated rack or even a container.
In many cases, the choices of server hardware, storage array, networking and hypervisor will be fixed by the vendor, which means that the assembly conforms to a reference design that has been fully tested and proved in typical operation scenarios.
Whatever the choice, however, decisions have to be made about the provision of power for the modules and, as always, these decisions will be influenced by the overall size and power consumption of the installation, and by availability/ resilience requirements.
To take full advantage of converged architecture, however, an intelligent UPS system used in conjunction with an intelligent power distribution system and comprehensive power management software will be needed.
Irrespective of the approach adopted in their implementation, resellers need to understand that power protection systems for converged and virtualised IT installations must take into account fluctuations in the application load.
In some cases, the load can change from zero to almost 100% over a short time period, especially where strategies such as 'following the moon' are in place.
These changes in application load are reflected by similar changes in the power demand of the installation and, unless the power protection system has been designed with this in mind, the result is almost certain to be poor overall energy efficiency.
The second point is that effective power protection and management provisions are absolutely vital – even when resiliency is built into layers above the hardware layer – to guard against possible hardware damage by power transients, to eliminate the stability problems associated with zombie servers that are neither fully operational nor totally off-line, and to keep the higher layers informed, in real time, about power status.
Finally, resellers should design power systems from the outset to deal not only with fluctuating demand, but also future changes in demand, and with the clear understanding that failures can never be totally eliminated.
When all of these factors have been fully taken into account, the ultimate goal of combining maximum application availability with optimum cost will be achieved.