Data center efficiency is Critical to business needs as computer requirements grow, density increases, and power and cooling demands climb. There is a changing face of the data center against the backdrop of IT budgets being slashed and the rising cost of power and cooling.
A recent AFCOM [a leading association for Data Center professionals] survey suggested that 38% of data centre managers have been asked to cut their operations expenses by 15% .
I have created this blog based out of my learnings and experience in handling the Unified Communications Products in the Data center which gave me a good exposure in energy consuming including the power, cooling and virtualization techniques. There are some day-day task to be performed on the inventory, energy consumption, server utilization which will help you manager your Data centre with high availablity and reduced operational costs. To summarize just prepare a simple check list on the said task and monitor the utilization.
How do you respond to this when many critical components such as energy bills and maintenance expenses are skyrocketing?
Moreover, the conventional capacity planning methodologies could fail as the meaning of scalability does not always mean how to scale up—the infrastructure should also be able to scale back if required and this is not something for which many DC managers are prepared. Hence it is important that DC managers first acknowledge the problem, and adopt a tactical approach towards future proofing, or in the current context recession proofing, their data centers.
Improving efficiency is one of the key ‘mantras’ that all professionals cite when it comes to constraints in resources. For this you need to measure and benchmark; how do we measure DC efficiency?
- Data center efficiency: Green Grid’s PUE (Power Usage Efficiency), McKinsey’s CADE (Corporate Average Data Center Efficiency), Uptime Institute’s SAVE (Site Analytics (to) Validate Efficiency) are some of the metrics that have emerged for measuring data center efficiency. Tracking PUE (Power Usage Efficiency) which gives a measurement of power usage for IT load compared to the supporting infrastructure load is a popular metric in the industry. While none of us want to comment on Google’s claim on running world’s most efficient data centers at a PUE of less than 1.21, PUE measurements give us a yardstick (in the absence of a better one) to measure and compare even though the number of variable factors at play are numerous including climate, age, utilization, resiliency and redundancy requirements.
Once you identify the causes of power usage, then there should be a mechanism to allocate the cost to various user communities. In many organizations, this exercise ends with the classical IT vis-à-vis facility team battle. The price book, provisioning and billing systems should also be able to respond to these changes.
- Obsolescence policy: Another important aspect is the retirement and obsolescence policy. A recent survey suggested that only one-third of the organizations surveyed have a documented and approved equipment retirement and obsolescence policy. If DC managers take a close look at the hardware inventory in the DC, this often reveals that many pieces of equipment are still plugged in and using DC resources without a clear production application running on these machines.
- Design capacity considerations: One should not forget the resource requirements for current applications and their scalability requirements while deciding upon the configuration of the computing fabric. We often find many architects/IT managers provisioning overcapacity in most components, be it processor, memory or storage. The increased awareness and availability of various upgrade options allow many architects to handle the processor and storage, but mostly insufficient attention is given to memory. It is a close second to processors in leading the energy drain in the data centers and needs special attention.
- Virtualization trends: From the computing and storage fabric, virtualization solutions are available in switching and routing platforms as well. This also gives an opportunity for Dynamic Power Management (DPM), an emerging technology that monitors the power usage of servers and turns them off when they are not in use. In a virtualized environment, computing load can be moved and concentrated on specific servers and idle ones can be powered off.
The study of alternate sources of energy and their evolution is another important consideration. Solar energy and wind power are possible contenders when we consider ‘other’ sources of energy. The current high cost of photovoltaic panels which constitutes solar energy infrastructure makes this unattractive at the moment and the payback takes years together. Wind power and its possible application in the DC environment is still being discussed.
- Green or eco-DC: Green initiatives which have gained significant momentum in the recent past can compliment these initiatives. Gartner has done significant work in this area and suggests a short-term, mid-term and a long-term strategy for Green IT and identified several important areas to focus on, such as advanced cooling technologies, virtualization and consolidation, use of modeling and monitoring tools, process design and server efficiency and integrated energy management etc. In the next two to three years, many green technologies will mature and become important and sometimes mandatory from a regulatory perspective in the DC.
- Data center automation: There is also an increased need to automate, have ‘Lights Out Operations’ and standardize the process framework followed in these facilities. ITSM offers a good methodology and an excellent governance practice and DC managers are urged to take these seriously to bring in efficiency in their delivery and support processes. It is important to evolve a customer-centric service culture especially during a downturn.
At the end of the day, how many DCs would sacrifice computing performance for infrastructure efficiency remains to be seen. To cope successfully, it would be the role of DC managers to create a balanced approach. This is a paradigm shift in their profiles and I wish them luck.
I would like you to share your Ideas and Suggestions based out of your Experience in this blog to help users learn more. Thanks for your time.