Having sufficient power to run your Data Centre is essential, but it is an expensive commodity. Geoff Day, Technical Manager at CMS plc, provides a top level overview of the issues surrounding the management and monitoring of power in a Data Centre environment.
Power is essential to the efficient and reliable running of a Data Centre, but if it isn’t managed well it can prove costly to Data Centre managers and owners. As power costs are rising and its availability is limited, Data Centre managers are working within tight constraints. This means that they need to capacity plan, manage and monitor their power usage effectively in order to provide the service levels required cost effectively. Whilst this is a tightrope to be navigated, there are a number of design and installation best practices and an array of products that can be used to optimise power usage in a Data Centre.
Estimating Power Requirements
Estimating the power requirements for a new Data Centre is by no means straightforward. In the case of a co-location environment, it is difficult to accurately predict the power usage of each client, so assumptions would probably need to be made on a “per rack” basis. But if these assumptions are incorrect, whether under or overstated, there is an ongoing cost and impact. In the case of a dedicated Data Centre owned by one organisation with a good view of their near term requirements, the rapidly shortening hardware replacement cycles, which are now typically less than three years, makes the job of predicting future power requirements increasingly challenging. The obvious answer to both of these scenarios is to build in a buffer to provision for the difficulty in accurately forecasting power requirements, but again this has an associated cost.
Data Centre Design
The design of the Data Centre will also have a significant impact on its power usage. In addition to the power required to support the IT devices, additional reserves are required to support the operation of the Data Centre in the form of cooling equipment, UPS’s, PDU’s, lighting and switchgear. Industry estimates suggest that typically less than 50% of the power supplied to a Data Centre is used to support the IT devices. The Power Usage Effectiveness (PUE) metric has been developed to indicate the efficiency of a Data Centre and this is expressed as follows:-
PUE = Power to the Data Centre
Power used to support IT Devices
The smaller the number for PUE the better, but in the real world, a figure of 1, whilst perfect, is not realistic as there will always need to be additional power to support the Data Centre operation.
This is where the design of the Data Centre comes into play. The design and layout of a Data Centre has a significant impact on the heat generated and consequently the cooling equipment required. Cooling and power are inextricably linked so it is important that the design of the Data Centre optimises airflow as much as possible so as to minimise the power required for cooling equipment. The use of cold aisle containment in the Data Centre can assist the flow of hot and cold air to best effect, minimising the energy used by computer room air conditioning (CRAC). There are many additional products that can also be used to assist airflow such as blanking panels, floor grommets and roof tiles and these are relatively inexpensive and easy to install and retrofit.
Monitoring Power consumption is equally critical to the efficacy of a Data Centre. If we assume that there is sufficient incoming power, based on accurate planning and the design of the Data Centre has been optimised to manage airflow in a way that minimises the use of cooling equipment, how do we manage power consumption on an ongoing basis to prevent outages, or downtime which can be extremely costly?
The good news here is that there are many products and tools available on the market that can assist with power monitoring, and these vary in levels of sophistication and the associated price tag.
Inline meters can be installed in racks to measure and report the power usage within a rack, but as they are rack mounted they take up valuable real estate.
There are a wide range of PDU’’s available on the market place referred to as “Intelligent PDU’s” and these have varying levels of functionality including: reporting power usage, switching outlets on and off, sending alerts to software packages if triggered by an event outside pre-set parameters. These intelligent PDU’s are also configured to be used to bill clients for power consumption. They can easily be custom built to meet a specific rack requirement and can be vertically mounted so as not to use up horizontal rack space.
There are a number of software products available that enable the power monitoring and management at rack or cabinet level. These are useful mid-level products and ideal for use where racks are dedicated to specific clients in a co-location application.
At the upper end of the spectrum, Data Centre Infrastructure Management (DCIM) Software is available, which enables Data Centre Managers to manage pretty much every element of a Data Centre on a real-time basis. Typically these DCIM solutions enable the management of assets, provide network connectivity maps, power paths, floor plans and cabinet layouts. These software solutions assist with capacity planning and change management within a Data Centre, as well as reporting on trends and assisting with future capacity planning and provisioning of IT devices. A number of these DCIM solutions also integrate information from third party Intelligent PDU’s so that information can be managed from one dashboard. DCIM solutions are feature rich and obviously have a cost associated with them that can price them out of reach in some applications.
The good news for Data Centre Managers is that there are products available at all levels to assist with the efficient management and monitoring of power in a Data Centre. From the simplest to the most sophisticated application, there are products available that are fit for purpose.
The process of planning, managing and monitoring power in a Data Centre is not for the faint hearted, and there are some general best practice principles I would recommend.
1. Planning – detailed planning is essential
2. Estimate Power requirements at rack level
3. Optimise airflow to minimise energy spent on cooling
4. Deploy monitoring tools that are appropriate to the application
5. Ensure you are able to monitor power consumption at rack level