In 
As data centres strive to become highly available as well as efficient, an important aspect of the entire data centre infrastructure that needs special attention is their power distribution strategy. Traditional ways of distributing power to the racks are being constantly challenged with power densities as well as power consumption costs among data centres constantly rising. A recent study conducted by IDC found that the number one challenge facing data centres today is power/cooling. In this study, more than half of the respondents are taking a serious look at how to reduce data centre costs by controlling power consumption. In addition, an EPA study mentioned that the power consumption costs of a data centre can account for as much as 30 percent of a companys IT budget.
According to a study by the US Environmental Protection Agency, 20% of the operational cost of a data centre is for energy and 60 billion kilowatt hours are used annually to operate data centres within the United States alone. IDC research says by 2010 the expense to power and cool the worldwide installed base of servers is expected to grow to $44.5 billion, equivalent to 70% of new server spending. Growth of data centre energy consumption within APAC increased at a rate of 23% between 2003 and 2005, compared against 17% for Western Europe, 14% in the US, and 14% in Japan.
Clearly, understanding and controlling energy costs will help meet business objectives. Beyond controlling energy costs, IT administrators are always being asked to find ways to make IT operations more productive, while at the same time improving availability. Technologies that can effectively monitor and control what is going on within the IT infrastructure while at the same time reducing staff workload can help focus resources on whats really important to the business and impact on the environment.
Challenges faced
Increasing energy costs are becoming the nemesis of every IT organization, with these costs robbing money from more productive business purposes. Engineers are required to make many trips to the data centre site to recycle the power so that SLAs can be maintained, and MTTR reduced. It has become very difficult to accurately assess whether additional equipment can be safely added to an individual equipment rack or to a data centre. This further adds to difficulty indentifying power wasting devices and lack of granular power control. There is a need for environmental, real-time power consumption analysis, pre-emptive alerts monitoring that give complete understanding of hot and cold spots within the data centre, allowing for the optimization and tuning of air conditioning assets.
Major power distribution considerations
To overcome such challenges and support GREEN Power Management one must discuss how to choose the appropriate power distribution products at the rack level. Interestingly enough, several of these considerations also help data centre operators manage and control the largest portion of data centre power consumption, which are the actual IT loads.
Voltage to run all the IT equipment: Most modern day IT equipment typically requires 1-ph power to operate. However, most IT equipment can accept voltages within the range of 100V 250V. It is recommended that customers choose the higher voltage means a lower current draw for the equipment which translates into lower losses and better efficiency. Industry studies have revealed that running servers at higher voltages can bring as much as two percent efficiency advantage to the servers alone. Counting the added benefit of lower cooling, the real gains might be as high as three percent to four percent. In view of this benefit, several enterprise level UNIX, blade servers as well as large networking switches are being built with power supplies that can only accept 200 250V.
Electrical circuits within the rack: Most IT equipment today comes with dual redundant power hence data centre operators should plan on bringing a minimum of two sets of completely independent circuits (N + N) into each rack coming from different power sources handling anticipated load capacity of the rack.
Load densities >5KW / rack, customers should consider bringing 3-phase power to the rack level which provides several benefits. This helps on Lower cost of cabling and higher reliability of electrical infrastructure using rack-mount PDUs with phase level metering. Reducing the circuit can help in maximizing the air flow both under the raised floor and within the racks keeping IT equipment from overheating. Increased rack scalability can also help in plugging in more equipment in the future.
Physical characteristics of power distribution units: One must consider three physical characteristic for rack mount PDUs like Form, Type and Number of Outlets. The Form factor the IPDUs are either vertical or horizontal within 1U of rack space. Most modern day IT equipment are standardized with an IEC input power connector and modular input cords except for customers in North America have some equipment that has a hardwired input cord with a NEMA plug. The PDUs should have the appropriate number of outlets required to power all IT load within the racks and we should keep in mind the power supply redundancy.
Other Important features within power distribution units: In order to ensure the highest availability levels for all mission critical equipment, it is important that data centre operators consider power distribution units that have metering as well as switching capabilities.
PDUs having metering and current thresholds capability protect rack from overloading and also helps data-centre operator to gauge actual power consumed which helps in appropriately sizing the upstream electrical infrastructure when adding new racks.
Optimising the electrical infrastructure improves the efficiency of the entire power chain, thereby cutting down on the power consumption costs.
Remote recycle, turn ON/ OFF is required for non-responsive servers or network equipment when in hung or distress situation. This helps to minimize the downtime keeping operational costs low by reducing the number of expensive trips to the physical location of equipment.
Several modern day servers are equipped with embedded service processor technologies such as IPMI, Dells DRAC or HPs ILO. But still switched PDU solutions provide several other enhanced capabilities that are important to improve the availability of the loads. Some of these features are Sequential startup, shutdown and Electronic over-current protection:
Remote management of power distribution: In order to make the maximum utilization of monitoring and control features and to help realize the vision of a lights-out data centre, it is important that switched and metered PDUs can be managed remotely.
One must consider that Remote management should have centralized, secure and single window to manage console and Intelligent Power Device Units (iPDUs). PDUs should work in OUT-OF-BAND management and should be give global integrated view of all PDUs within the data centre. PDUs should have features of auto discovery and grouping capabilities for outlets, which ensures that the power to the entire server can be recycled.
_________________________________________________________________________________
The author is the Regional Manager - West and Industry Specialist (BFSI), Avocent India
Add new comment