However, when running a resilient data centre with a high level of redundancy, measures need to be taken in order to ensure a good balance of load across multiple UPSes in order to ensure efficient operation of each.
With virtualisation taking preference in many organisations today, a new dimension of complexity is introduced in ensuring a good balance between energy efficiency and performance due to the dynamic nature of a virtualised IT system.
A common misunderstanding today is that metrics, like Power Usage Effectiveness (PUE), provide a unified method to understanding data centre performance. While PUE certainly has its advantages, the very definition makes it misleading when the primary goal is striking the right balance between energy efficiency and performance. For instance a 200kW data centre running at 40% load would have a fairly poor PUE by definition.
One way of improving the PUE would be to add more IT equipment to get the load up, which would automatically improve PUE, but which doesn't make the data centre more efficient if all that IT equipment just runs idle!
Also a virtualisation effort, where many servers are retired by consolidating multiple servers onto fewer virtualisation host servers, will automatically improve the efficiency of the IT load, but (illustrated by an increased PUE) unless the facility equipment is equally scaled at the same time, the bottom-line efficiency improvement that was scoped will not be reached!
Complexity like this is why a continuing focus on device-level optimisations will never enable today's data centres to reach their ultimate potential in terms of efficiency/performance balance. Rather a shift in focus to a combined device-level and holistic system-level efficiency optimisation is needed.
This type of complexity drives many organisations in the direction of deploying a smart toolset like Data Centre Infrastructure Management (DCIM). Only by having the full transparency provided by a DCIM tool can the data centre planner be enabled to understand the implications of any planned changes and take measures to mitigate any risk imposed.
By using a DCIM tool that allows documentation of the entire lifecycle (from planning, through commissioning, into operation, and ultimately decommission) can the data centre planner make fact-based decisions to avoid problematic configurations that do not support the journey towards the right efficiency or performance balance.
When doing system-level data centre efficiency optimisations it has proven valuable to divide the approach into 4 phases, as described below.
Step 1: Perform a discovery and categorise findings
The first phase focuses solely on establishing a clear picture of the current state of affairs in the data centre by examining the equipment deployed and its respective performance. In this phase, an understanding of the type of equipment utilised will be established as well as an indication of performance telemetric for each server.
Sign up for Computerworld eNewsletters.