Keeping cool and saving money Keeping the datacentre cool is like chasing a constantly moving target. As newer equipment is introduced into the datacentre, the temperature balance and the energy consumption change. IT and Facilities Managers need to work together to keep things under control. Striking the right balance between the cost and energy required to cool the datacentre and run hardware can be difficult. The measure of how efficiently you are using energy in the datacentre is defined as PUE. An efficient datacentre should have a value of 2, but what does this mean? To arrive at the PUE for your datacentre you take the amount of power you are consuming and divide it by the power used to run equipment. If you are consuming 15MW of power but your IT equipment only consumes around 5MW then your PUE rating is 3. This is relatively inefficient and shows that too much power is being wasted on cooling the environment. As you reduce the power used to cool the datacentre, more power is released for hardware. But this increases the amount of heat that needs to be removed and thus the cooling bill goes up. So what can you do to strike the right long-term balance? The first step is to consider what you are buying. Processors and high performance memory consume a lot of power and require plenty of cooling. There are low power versions of both processors and memory, but they are often more expensive and have a lower compute capability. There is no simple answer as to whether it is more efficient to buy a powerful server with higher power and cooling requirements, or two lower power but more efficient servers. The application you are running will often dictate what is required of the server, with applications such as databases and financial or scientific calculations requiring the maximum processing capability. In comparison, servers running multiple virtual machines are able to move their workloads around to make the best use of resources.
Another deciding factor is cost. The new low power processors and memory are more expensive as vendors look to recoup the costs of development and production. So the financial decision doesn’t just include the cost of power and cooling, but also the cost of acquisition. This is less than a quarter of the cost of a server over a three-year period, but the energy savings must be carefully calculated. It is not just about buying new hardware. Over time, equipment gets upgraded and even when a server is replaced that does not mean it is no longer used. Often a server replacement program brings in new hardware for the most demanding applications and reassigns older hardware to less difficult tasks. The days of upgrading processors and memory are long gone for most organisations. This is because the cost involved, especially in downtime and engineer time, usually outweighs the cost of buying a new server. Yet there are things that can be changed over time and which will have a significant impact on power and cooling costs. Power supplies and fans are becoming more efficient. There are also products that vendors standardise across their entire hardware range. It is worth looking at the replacement costs for some of these units as they age. Cooling technology with fans has also taken a huge leap forward. The more efficient a fan is, the less power is required to cool processors and memory. This does not alleviate the problem of cooling the air once extracted, but it does create an immediate power saving. Management solutions like HP Insight Power Manager allow you to set limits on the amount of power being used by server and storage blades. This allows IT and Facility Managers to agree on how to match power usage to cooling capabilities, especially where there is a limit on the amount of cooling available in the datacentre. Hitting any moving target is hard. And when that target falls across two departments, IT and Facilities Management (FM), it is especially challenging. Using lower power hardware and replacing less efficient components will help IT meet the restrictions placed upon it by FM in terms of power and heat. Buying energy efficient hardware may be more expensive
from an initial Capital perspective, but if IT is to help reduce Operational Expense, it must be given the necessary budget increases to buy the right hardware. In the long term, it will pay off.