Powering the Data Centre

Maintaining business continuity and optimising energy efficiency through effective uninterruptable power supply (UPS) systems.
November 2012

Data centres are the basic platform that organisations have become more dependent upon. The failure of any aspect of the data centre can lead to a loss of business capability that is expensive at many levels for an organisation. However, the loss of power to the data centre will bring all IT systems within or dependent on it to a halt, and recovery from such an occurrence would be difficult for an organisation of any size. Alongside existing issues of accidental loss of power to the data centre, organisations must start to look to the future, where the possibility of forced power cuts looms as power companies struggle to maintain constant supplies. With the increasing cost of energy, the most cost-effective approach to ensuring that an IT platform can continue to operate, even when power from the grid system is no longer available, has to be a high priority for any business.
Clive Longbottom Quocirca Ltd Tel : +44 1189 483360 Email: Clive.Longbottom@Quocirca.com Bob Tarzey Quocirca Ltd Tel: +44 1753 855794 Email: Bob.Tarzey@Quocirca.com

Copyright Quocirca 2012

Powering the Data Centre

Powering the Data Centre

Maintaining business continuity and optimising energy efficiency through effective uninterruptable power supply (UPS) systems.
Data centres are power hogs, and the increasing price of energy is increasing the business focus on the facility Current energy strategies are likely to introduce more pressing issues in the future A UPS is not just a standby power system Modularity is a key factor in providing the flexibility and scalability a modern data centre requires Modularity is not just about single systems
With data centres now being a major part of any organisations energy usage, businesses are beginning to focus on how to bring the energy costs, run up by their IT departments, under greater control. Alongside virtualisation and cloud computing, the role of the UPS should be regarded as a prime area where energy savings can be made. The inertia of governments caught between the rock and the hard place of fossil-based energy generation versus renewables energy, alongside the move against nuclear after the tsunami in Japan, means that many countries are moving towards an energy deficit, which will result in rolling brown outs and even to planned power outages. Data centre managers must start to plan for such a scenario now. A modern UPS must be capable of ensuring that power feeds are fully smoothed, removing artefacts from the feed and protecting against over and under voltages, current and voltage spikes and surges and other disturbances such as those caused by radio frequency interference. Monolithic UPSs are no longer cost effective, as incremental needs to grow or shrink capacity are difficult to manage. Modern UPS systems should be modular, enabling better fractional growth, fractional systems redundancy and flexibility of physical installation, enabling the UPS system to be built around the rest of the needs of a data centre facility. Increasingly, organisations are looking to use a more distributed approach to UPSs, integrating them into racks and rows. Any UPS supplier must be able to support centralised and decentralised operations, with management systems that can regard all the UPSs as a single operational item. Data centres continue to be energy inefficient. With many servers running at below 10% utilisation, much of the energy used within a data centre is wasted. However, improving the efficiency of a UPS by just 1% means that a fully-supported 2.5MW facility will be saving 25kW of energy, without doing anything else to the IT equipment. In the case study within this report, Iliad Datacenter managed a 6.4% energy efficiency improvement in its facility. High efficiency modes and enhanced management systems not only ensure that a UPS runs efficiently, but can also extend the life of the various components, ensuring that the overall system provides a better return on investment and lower cost of ownership. Intelligent usage of such modes can provide higher energy efficiencies a key consideration in an age of uncertain, but rising, energy costs. Intelligent and reliable load sharing across modules within and across UPSs avoids any single point of failure while intelligent failover systems, combined with battery management capabilities, will ensure optimised life for critical components at minimum lifecycle cost.

Overall energy efficiency should be high on a list of a buyers priorities Special operation modes can add value without impacting performance and optimise energy consumption Load sharing and intelligent failover management are key requirements

Conclusions Modern UPSs are far more than just a black box providing power support for a few minutes while auxiliary generators take over. UPSs are complex systems that need to interact seamlessly with the different components of the data centre. A UPS can be the backbone of the critical infrastructure that supports the aim of minimising cost while maximising reliability and availability of power supply. Increasingly, UPSs are the main support when a power failure occurs, with generators only being powered up where the power failure is perceived as being for a significant period. A modern UPS needs to be capable of reliably addressing what-if scenarios: providing high-efficiency modes that enable greater energy efficiency and lower running and maintenance costs; allow for flexibility and scalability in implementation and utilisation; and minimise up-front capital expenditure costs, as well as providing the core capabilities of ensuring adequate coverage during power outages and in smoothing feeds from the energy grid.

Quocirca 2012

-2-

Powering the Data Centre

Executive Summary
Todays data centres are big users of energy. Jonathan Koomeys 2011 report for The New York Times estimated that global energy usage by data centres was running at around 1.3% of total demand in 2010, down from an expected 24% in his earlier 2007 paper. However, figures from DCD Intelligence estimate that energy usage grew from 19GW to 24GW from 2010 to 2011 and will grow to 38GW in 2012. So, data centre energy usage is growing fast but as a percentage of world energy usage it is shrinking, as emerging economies ramp up their overall energy usage. Although new generating capacity is coming on line in these emerging economies, the developed world has to deal with ageing generation plant, along with arguments around what the mix of fossil, renewable and nuclear generation capacity should be. These discussions have stalled decisions and an energy deficit is predicted by the regulator OFGEM for the UK as soon as 2015. Against this background, the dependence any given organisation has on its IT systems has been inexorably increasing just as the continuous availability of grid-based power is under threat. Historically, the focus has been on dealing with accidental power outages, where a main grid line fractures or a distribution line is brought down by bad weather, or whatever. When data centres were relatively small and the amount of power required to keep them running was manageable, large monolithic uninterruptable power supplies (UPSs) were used as part of the data centre fabric to provide enough power to bridge the gap between the failure of the grid feed and the starting up of auxiliary generators. Now, power requirements have often outstripped the capabilities of the monolithic UPS and generators, leaving data centre managers with a need to prioritise what is protected and what should be left to fail during a power outage or to see everything cease working when grid power fails. The move from fossil-fuelled central power generation to nuclear has stalled, affected by issues such as the cost of provisioning and de-provisioning nuclear plants, along with the concerns raised by the 2011 earthquake and tsunami in Japan. At the same time in the EU, existing fossil- and nuclear-fuelled generating capability is scaling down as power stations reach their end of life and fall foul of EU-driven pollution laws. Huge investments are being made into green energy initiatives yet, in many cases, these will fail to come on line before existing capabilities are decommissioned. Also, due to the inconsistent generation capability of the main renewables (solar and wind), they still need to be backed up with other, more predictable generating capability bringing back a need for new fossil or nuclear generation capability. This is what is leading to the energy deficit, which could well result in a need for rolling black outs, particularly for large businesses. For a data centre owner, this leads to the need for a complete review of their energy use strategy firstly, how to minimise energy usage; and secondly, how to ensure that the data centre can cope with switching over from a grid energy supply to a backup power supply without any loss of business capability, or, at worst, with a graceful and controlled loss of low-priority systems. This strategy has to be able to cope with any planned data outages caused by rolling blackouts, with poor energy quality caused by the grid having to modify its feeds through lowering voltages (brown outs) or with accidental outages caused through natural disaster or man-made error. The period of time between a power outage and auxiliary generators taking over is the key area where a UPS provides its worth but is not the only one. This period of time should be relatively short and at most be measured in the low minutes. However, even such a short time, for a moderate-sized data centre, will involve large amounts of power needing to be made available to keep everything running. Also, if a single large UPS system is used, there will be a lack of granularity with the UPS system simply acting in the same way as the grid; all systems continuing to run, draining the UPS itself until auxiliary generators take over or grid power resumes and takes over. If there is only one UPS supporting the data centre, there is also the problem of what if it fails? Therefore, redundancy is needed, which, at a standard 2N or even a 2N+1 level, will require two or more UPSs that can both support the whole data centre individually. This is not just wasteful at a capital level, but also at an energy level with one or more UPSs being kept topped up just in case.

Quocirca 2012

-3-

Powering the Data Centre

However, a UPS strategy should also provide other value added capabilities. It should be able to continuously cleanse the power as it is fed from the grid or auxiliary generators so that data centre equipment receive the right voltages without spikes, surges or other artefacts. It should also be energy efficient and provide advanced functions and capabilities such as high efficiency modes that provide a greater degree of flexibility, reliability and futureproofing than has been available previously while minimising energy usage. Finally, it should also provide the greatest possible degree of scalability to deal with future needs of expansion and contraction. Economics are putting a lot of strain in the data centre business models and scalable systems are one way to maximise the use of capital. The aim of a modern UPS has to be to provide high availability through reliability while ensuring a clean and effective power feed at the highest possible energy efficiency.

The role of the UPS

A UPSs main role is to keep systems running through a power outage. No matter how well a data centre is configured, how good its power usage effectiveness (PUE) score is, how much virtualisation and cloud computing has been built in if there is no power to the facility, it is just full of dead equipment. However, keeping the power feed live should not be seen as the only function that a UPS provides. A UPS should be seen as equally important while grid or auxiliary generator power is on online, as it should act as a means of smoothing the power feed before it is distributed around a data centre. This capability may not appear to be important at first glance, as many assume that a grid supply is a grid supply. Being able to smooth supply fluctuations out to a constant 230V supply (for the UK) puts less strain on the lower quality transformers found in IT equipment, and gives rise to higher overall systems availability. Even at this level, most IT equipment would manage with a reasonable power feed that fell within the legal limits of 230V +10% -6%. Where a UPS really proves its worth is when dealing with other artefacts and problems within the power feed. Such issues with power quality can take the form of spikes or surges of over-voltage, or large drops of under-voltage power. A spike is a very short duration over-voltage artefact, generally caused by a fault in the grid system or a direct consequence of a lightning strike. Often, such spikes can be measured in thousands of volts way above the limits of the power supplies of the equipment being fed, and a massive spike may spill over from the power supply into the various internal electronics of the equipment itself. A sustained over-voltage is a surge and can have similarly catastrophic impact on IT equipment, which will often have low limits of over-voltage protection. A low voltage could cause the equipment to drop out, restarting as the voltage comes back up to the norm but this will mean that the software applications and the functions dependent on the hardware will become unavailable and transactional data will be lost, impacting hard on the business. Also, many pieces of IT equipment (for example, spinning hard disk drives) do not like being shut down unexpectedly and may fail on restart. A suitable UPS will smooth the spike or under-voltage out, again ensuring that the quality of power is maintained. In dirty environments, radio frequency interference (RFI), for example from switches in equipment on factory production lines, can affect the stability of the grid power supply, introducing artefacts into the sinusoidal waveform of an AC feed or impacting the voltage or current of the feed, which could impact the IT equipment using it. Again, the transformers in much of the IT equipment are not built to deal with such problems, and could fail or go to a safe mode of switching off if they sense that the incoming power feed is not right. A good UPS should be able to deal with variation in the mains frequency, in harmonic distortion and in line noise, providing a constant, clean power feed. However, the UPS should also to be as energy efficient as possible. Losing 5% of power fed into a data centre through UPS losses is now unacceptable, yet many UPSs are still running at such losses, or more. The choice of UPS operating mode, and the manner in which value-add capabilities are implemented in order to minimise energy losses while maintaining power quality and core capabilities to the UPS functions, is the new battleground for UPS manufacturers.

Quocirca 2012

-4-

Powering the Data Centre

It is therefore critical to ensure that any UPS chosen for a data centre is capable of dealing not only with power outages, but also with the problems that can occur through the variability of the power feed coming into the data centre, while ensuring that energy efficiency is maintained throughout the different working modes of the UPS. The UPS cannot be chosen just because of the time it can support a given workload on a power outage any decision has to include how reliable and flexible the system is, the quality of the UPS continuous output power feed and how energy efficient the system is. Whereas, in the past, improving one of these may have meant a negative impact on the others, modern UPS systems can ensure that all three areas are optimised.

Types of UPS
There are different technology approaches to the workings of a UPS, which mainly come into one of three groupings and then there are hybrid systems of the future: Offline Also known as standby UPSs, these tend to be systems with basic functionality such as battery power backup capabilities and basic surge protection. Also known as standby power supplies (SPSs), such devices are not generally aimed at the data centre, being used more for providing support for PCs and other office equipment. SPSs suffer from a lag between power going down and the SPS taking over although this is generally only of the order of a few tens of milliseconds and is good enough to prevent a PC from going down, but it is a poor approach to critical equipment within data centre equipment. Line interactive Line interactive UPSs are based on the same approach as an SPS, but add more capabilities. The use of an autotransformer means that over- or under-voltage situations are dealt with automatically without the need to use the batteries, so improving energy efficiencies and increasing battery life. The time lag between power failing and the UPS taking over is generally much less than a standard SPS, but overall energy efficiencies tend to be lower than on-line systems. On-line On-line UPSs introduce all the capabilities that a data centre or facilities manager should be looking for in a system. These UPSs use a double conversion system, where the rectifier and inverter are directly coupled. The batteries are also directly connected to the inverter so that, on any loss of power, the rectifier drops out of the circuit and the batteries take over immediately, with zero lag between the two supply systems. These systems provide the cleanest possible supply to the downstream equipment. However, the systems tend to be expensive and are difficult to modularise in a manner that provides the flexibility and scalability required in a modern data centre. Future technologies Hybrid Between pure line interactive and on-line UPSs lie a range of systems that bring together various aspects from both sides. The idea here is to provide all the advantages of an on-line system, while pushing the energy and cost advantages of a line interactive system. Ensuring that the trade-offs between the two approaches do not impact the main capabilities of the UPS is where each vendor differentiates itself or fails in ensuring that their UPS meets the real requirements of the environment it is installed in. Modular systems need to be integrated together in a manner that still allows the UPS to take over immediately on a power outage. Workload management is required to ensure that individual modules are not overstressed, causing a domino effect should a single module fail. Energy efficiency needs to be managed effectively through additional capabilities, such as special high-efficiency modes and high quality components in the transformer and battery charge management subsystems.

Quocirca 2012

-5-

Powering the Data Centre

Prioritising for business support

By taking a more strategic view of how UPSs are used in the data centre, a far better environment can be created that better supports the business while lowering costs. The first thing to look at is ensuring that the data centre itself is fit for purpose. Here, virtualisation helps in driving up server and storage utilisation rates, resulting in fewer servers and storage systems being required. More intelligent cooling strategies ranging from hot and cold aisles to evaporative or free air cooling further cut down on the energy needs. Bear in mind that such a data centre is only fit for purpose if it is working. Therefore, a UPS has to be seen as a core part of the underlying foundation of the data centre. It provides the power needed to ensure that the data centre runs not just when a grid power outage happens, but through all the different aspects of problems that can be introduced into the main power feed. A UPS isnt just there to keep the servers running, it has to be there for the storage, for the network and for the other equipment that is needed to ensure that the data centre keeps on supporting the business, twenty four hours a day, 365 days a year. The current use of UPSs should be reviewed. Quocirca has found that in certain cases where Facilities look after the UPS and auxiliary generators, but IT look after the IT equipment, the increasing amount and density of IT equipment in place outstrips the capacity of the UPSs. With many older UPSs being incapable of intelligent load shedding, when a grid failure occurs the UPSs cannot provide the amount of power that is required, and they then fail as well. As the auxiliary generators will not have kicked in, the whole data centre therefore fails, with catastrophic consequences. Even if Facilities and IT work together to try and synchronise total data centre power requirements with total UPS capacity, for many organisations the capital costs for regular UPS and generator incremental addition or replacement does not fit in with budgets. Therefore, the first decision that needs to be made is whether all IT equipment within the data centre needs to be kept running through any power outage. This has to be decided upon at a business level: the business needs to define which business processes are key, and IT can then define which technical services are required to maintain those processes. If the business decision is that the data centre as a whole has to be kept running, then a multi-module UPS approach that allows for incremental growth and for in-built redundancy will be the best approach. However, in many cases, it is far better to look at the needs of the different workloads within the data centre and how these should be prioritised to ensure that each workload is provided with the support that it needs. For example, a mission-critical SAP installation will need to be prioritised to run continuously. A less critical business service, such as payroll, may need to be kept running at certain times, but could be allowed to be shut down gracefully at others. A low priority system, such as a near-line storage system, could be shut down at any time. Virtualisation can help here. For physical systems, the only level of granularity for prioritisation is to close down the servers that underpin a specific application. With virtualisation and particularly with a private cloud architecture where resources are shared it is possible to take a far more granular approach. For example, simply shedding workloads by shutting down virtual machines will lower the energy requirements for modern physical servers, as under-utilised CPU resources will result in a lower energy draw. However, it is also possible to throttle workloads, so enabling better prioritisation high priority workloads can be given as much resource as they need; lower priority workloads can be shed completely or provided with less resource but still given a capability to run, even if this is at a lower response or throughput rate. Physical servers can then be shut down as the overall resource usage is lowered, enabling the data centre to run for a longer time directly from the UPS, often negating the need for the auxiliary generators to be switched in at all. Such prioritisation then allows the right UPS strategy to be chosen on a more granular level.

Quocirca 2012

-6-

Powering the Data Centre

Granularity of UPS
Todays UPSs come in many forms, as well as covering different technological approaches. The use of monolithic UPSs is changing as more modular designs, consisting of multiple components that can operate in a high availability manner within a single system, come into play, enabling different scalability models and flexibility in how energy can be managed. Also, smaller UPS systems installed as in-rack or in-row units, but operating as a single managed UPS system, have become more common and are being used in order to provide much greater levels of control and support. With a modular UPS, a large system can be built up that provides a capability for incremental increases in energy management without the need for replacing the whole system. It can also enable built-in high availability, with the various modules sharing workload automatically and re-allocating the workload across remaining systems should there be a failure in any single module. Such modular systems also tend to come with management systems that enable greater control of what the UPS does and how it will deal with the various areas of the data centre should shut down of systems be required, or the auxiliary power be insufficient to deal with overall power draw of the IT equipment. Even where a distributed set of small UPSs are used at the rack and row level, it is important to ensure that these can all be seen and managed at an aggregate level. More, smaller UPSs can present more issues around areas such as battery management, and it is necessary to ensure that each system is monitored individually, and also, at a higher level, to ensure that issues can be dealt with quickly and effectively.

Buyers Guide to UPSs

There are 4 main areas that a UPS buyer should look out for: Energy efficiency A modern UPS should be capable of running at above 98% energy efficiency. Some systems are capable of exceeding this, running at an overall efficiency of 99% or higher. Not only does this provide on-going savings against incoming energy prices, but also helps in meeting externally driven regulation requirements, such as the carbon reduction commitment (CRC). Such systems can also provide positive data for an organisations corporate social responsibility (CSR) statement. For example, multi-module UPS systems tend to run at very low levels (usually between 15% and 30%) of their full capacity due to redundancy needs (e.g. use of 2N or 2N+1 approaches). This calls for an operational mode where the UPS can concentrate the power in the least amount of power modules and thus increase the efficiency levels of the overall system (like many pieces of equipment, UPSs have lower efficiency levels when they operate at low levels). These modes must not sacrifice reliability; therefore the unused power modules should be kept in such a state that if needed they will seamlessly start up and protect the load. Many UPSs now have enough built in intelligence that allows the systems to select the operational mode based on a set of input parameters. This means that if the quality of electrical mains is good enough, the UPS will maximise the systems efficiency up to 99% by allowing the power to flow straight through its static switch while providing a certain degree of filtering and surge protection. Should the electrical mains go beyond the desired safety window, the UPS will automatically go to double conversion (full protection mode) so fast that the load will not be aware of the switch over. In looking for the best energy efficiency from a UPS, a buyer should look for the specific high-efficiency modes that best suit their requirements, while ensuring that these modes do not then introduce shortcomings in the overall

Quocirca 2012

-7-

Powering the Data Centre

capabilities of the UPS, such as lower load support times or a lowering of systems reliability. Also, such capabilities should not come at too high a cost those vendors who can provide good energy efficiencies built in to the cost of their average systems are the ones to look for. Reliability It is no use having a UPS system in place if it fails when it is most needed. Taking a large, monolithic UPS with no inbuilt redundancy and no intelligence in load sharing or in how to use effective failover means that systems redundancy can only be provided through having extra UPSs in place. This leads to a 2N or 2N+1 approach, where the main UPS is backed up by one or two UPSs that are exactly the same, so that on the failure of one complete system, another one can take over. This results in there being two or three UPSs where the extra ones will either be running at low utilisation levels (so impacting overall energy efficiency a 2N system will only ever run at an overall 50% utilisation, a 2N+1 generally at considerably less than 50%; for example, a simple 2N+1 system where a single UPS is required would result in 33% utilisation), or will only be run in the rare case of failure of the primary UPS. Even in this case, the backup systems still need to be kept fully charged and maintained, so impacting overall energy efficiency, capital costs, on-going maintenance costs, as well as space requirements. A more modular approach means that a fractional N+1 approach can be used. Here, as each module is a fraction of the overall total power support, its failure can be covered through a similar module. Taking an example of a 2.5MW data centre, using five 500kW modules means that high availability can then be provided through the addition of one or two extra 500kW modules not more 2.5MW systems. This then requires the capability for effective load sharing not only on failure (UPS systems must be able to load share even in the event of a communication failure between UPSs), but also during normal running to ensure that all modules are equally utilised, so extending the overall life of the total system. This is also important where different modules feed different equipment, but where the failure of one module can lead to another module it shares loads with taking over, so ensuring that support is continued across the overall system. It is also important to ensure that the chosen UPS system is fit for purpose in the environment it is implemented in. For example, a single-system modular UPS may be most useful where the whole data centre has to be kept operating on power failure, whereas a more granular system based on discrete UPSs at row or rack level may be more suited to a prioritised data centre where different servers may need to be kept running for longer periods than others. Therefore, a UPS buyer must make sure that the systems chosen are capable of running as a single managed system in their chosen environment either as a single facility UPS or as a distributed, granular system. It is also important to ensure that the UPSs support different configurations to achieve the desired reliability at the site. For example, parallel UPS systems can be installed in either a centralised or distributed configuration. Centralised systems utilise one single static bypass switch for all the UPSs, while distributed systems utilise one static bypass switch per single UPS. Both configurations have their advantages; a scalable centralised system will provide full fault current capability on day one, something extremely desirable in scalable systems. On the other hand, a distributed system may provide greater availability at the UPS system level by providing several static bypass switches instead of only one. The bottom line is that it is important to understand the implications of each configuration and make the most out of them according to the users particular needs. There are other areas to consider when it comes to reliability. The lifecycle of the batteries in the UPS is a critical area, and many systems have low battery life,caused by relatively unmanaged charging circuits and a lack of granular management of the individual batteries and cells. More advanced systems will monitor each battery for its health and will also monitor areas such as battery temperature to ensure that over temperature conditions are avoided. Special high-efficiency modes can ensure that temperature envelopes are maintained even under heavy load and rapid charging. Battery life can also be extended through the use of advanced battery management capabilities, checking the status of batteries on an on-going basis and carrying out actions that are required to optimise battery life and inform administrators in a timely manner where problems are identified. Flexibility

Quocirca 2012

-8-

Powering the Data Centre

UPS systems also need to be flexible. As discussed above under reliability, the age of large, monolithic UPSs is coming to an end, and the age of highly modular systems is now upon us. Using modules means that systems can be added to as required rather than having to decommission an existing 2.5MW UPS and replace it with a new 5MW system, an existing 2.5MW UPS can be extended through adding additional modules as required to become a 3MW, 4MW or 5MW as required. This pay as you grow approach means that capital costs are better controlled, and existing systems are cost optimised over a full lifecycle. Additionally, a well-engineered modular UPS should allow for flexibility in how the modules are physically configured. With space in many data centres being at a premium, it may be that a long, thin configuration down a data centre wall may be the best way to deploy, or it may be that a square configuration in the centre of the data centre is a more optimal. Indeed, as more data centres move toward a modular design in themselves, and the use of scale out block computers, such as VCEs V-Blocks, Ciscos UCS and IBM PureFlex systems, the flexibility of the UPS implementation becomes critical. For those who require containerised data centres or use containers as a part of their data centre configuration such physical flexibility is not just a nice-to-have, it is an imperative. Being able to choose how modules are physically configured should be high on the list of priorities when looking at buying a new UPS system. A modularised system that is flexible in how it can be physically configured also provides other capabilities, such as enhanced capacity to deal with prevailing environmental conditions. For example, in data centres running closer to the American Society of Heating, Refrigeration and Air conditioning Engineers (ASHRAE) upper temperature envelope, a flexible physical implementation can allow for more targeted heating while maintaining optimum running conditions. Serviceability A UPS requires regular maintenance and, due to the nature of how it works, will require some components to be replaced on an occasional basis. For example, a failed battery or even a failed full module must not result in the need to replace a total system. Many UPS systems require access to both front and back to be able to service parts. When combining serviceability with flexibility, a system that is space constrained may only be able to provide real access from one side and, as such, buyers should look to those UPS systems that enable all serviceable parts to be accessed from one side. Even so, buyers must ensure that this does not then introduce further problems. For example, some systems may provide single-side access, but this then makes extracting failed parts more difficult. The mean time to recovery (MTTR) for any component should be available from the vendor, and this provides a representation of how easy it is for each item to be accessed, removed and replaced. Another factor is how well a vendor and its supply channel can manage spare parts it is all very well being able to gain easy access to the parts in the system, but if you have to wait weeks or months for spares to come through, then the UPS is not providing the support that it should do. A buyer should ensure that its chosen provider maintains a good stock of spares with suitable logistics to get these parts to their site within agreed service levels.

Quocirca 2012

-9-

Powering the Data Centre

Case Study - Iliad Datacenter

Iliad Datacenter is one of Frances largest data centre operators. It had decided to modernise its DC2 data centre, a 5,000m facility in Vitry. Part of the modernisation was to move to a 2N+1 redundancy architecture, leading it to a full review of its existing UPS strategy. A key driver in choosing a new UPS was the need for overall energy efficiency. With the DC2 facility needing power backup totalling 2.5MW, any gains in energy efficiency could lead to appreciable savings. As a data centre operator, Iliad needed full power protection to the facility. Energy costs amount to 70% of Iliad Datacenters operating costs, and it estimates that the cost of energy is rising at an average of 7% per year. The system that Iliad Datacenter decided upon was based on a series of Eaton 9395 UPSs. A chain of five 9395s was implemented, providing a capability for covering the whole of the data centres 2.5MW against power outage. Three areas of functionality within the 9395s were seen as being of major importance to Iliad. The first is the in-built Energy Saver System (ESS), which enables a straight-through power mode when incoming grid power is within determined limits, so increasing the overall efficiency of the UPS. Secondly, the Variable Module Management System (VMMS) enables the UPSs to make the most of their modularity, running individual modules at optimum load levels as required while keeping others unloaded. Should the overall load require another module to be switched in, this is done within 2 milliseconds. These two key technologies allow not only reduced power consumption and the minimising of any environment impact, but also lowered operational costs significantly. Additionally, Eatons Hot Sync technology allows the UPSs to be run in parallel without the need for a common logic bus, which can be a single point of failure in other modular UPS systems. Hot Sync also enables any problems within any single UPS module to be detected at an early stage, and for that module to be removed from the chain in a manner that allows the other UPSs to remain running with the extra load being shared between them. This feature greatly enhances the reliability of the system and, when used in conjunction with enhanced technologies such as VMMS, allows the system to gain both efficiency and reliability simultaneously. Through implementing the 9395 UPSs, Iliad Datacenter has found that the overall energy efficiency of the DC2 facility has risen from 92% to 97.9%. Based on the success of this implementation, Iliad Datacenter is now implementing more 9395s in its 11,800m DC3 facility. This Tier 4 facility is much larger than the DC2 facility, and will require 24 UPS modules. Iliad Datacenter will be making full use of ESS, VMMS and Hot Sync to maximise the energy efficiency of the facility. For Iliad Datacenter, the main aspect is the impact on its bottom line. As a major data centre operator, its total energy requirement runs at 24MVA. Through using Eatons products, the real-term improvement in energy efficiency is 6.4% a possible saving of over 1.5MVA in energy. Combining this with better levels of energy availability and a tested capability to push the systems beyond their stated capabilities, Iliad Datacenter believes that Eaton has helped it optimise its energy usage in its data centres.

Quocirca 2012

- 10 -

Powering the Data Centre

Conclusions
A modern UPS is far more than a battery back-up for the time between a power outage and auxiliary generators kicking in. There is an increased likelihood that grid power feeds will be of an uncertain quality, with brown outs and even planned outages becoming more of a possibility. Therefore, any UPS system must be able to act as a primary power feed cleanser, ensuring that the feed into and around the data centre is managed within acceptable levels. The keys to acquiring the right UPS system are in ensuring that the right levels of efficiency, reliability, scalability and serviceability are attained. Special operational modes, such as high-efficiency modes and means of load sharing to provide high overall efficiency, are where buyers will see the biggest differentiation between vendors. With the highest energy efficiency in UPSs now being around 99% and the lowest being around 94%, a 5% saving on energy in a 2.5MW facility will be 125kW. Furthermore, as any power losses are shed as heat from the equipment, improved efficiency will bring secondary benefits as less power will be needed for cooling purposes as well as reduced stress in certain components, extending the devices lifetime as well as increasing reliability of the overall system. With special modes requiring up to 20% less cooling and ensuring longer working lifecycles for batteries, the return on investment of a modern UPS is rapid, and the cost of ownership over an extended period will be lower. For reliability, one of the key features is bullet-proof load sharing. Most UPS systems are modular and therefore load sharing must be smart and reliable enough to avoid single points of failure (e.g. the UPSs must continue to load share even in the case of loss of communication between the individual systems). Additionally, a chosen UPS system has to be able to provide rapid switchover from grid power to backup power on any grid failure; should ensure that the power feed is maintained between strict limits, smoothing out any artefacts in the feed that could damage the equipment dependent upon it; and must be able to deal with any component failure itself without the whole UPS system failing. UPS systems also need to be far more flexible than in the past, being able to grow and shrink incrementally as the data centres needs vary, and also to be installed in a physically flexibly manner to meet the needs of increasingly valuable space within the facility. All this should be achieved without compromising the reliability of the overall installation. Modularity is becoming the norm, from being able to build a single UPS from sub-system modules to using smaller UPS systems in rows and racks. Such modularity should also improve serviceability, with the various subcomponents being more easily accessible for servicing or replacement. Whatever approach is chosen, the overall UPS system must be able to be managed as a single entity, and must also be able to operate and provide support across the facility should any single module fail. As the usage of IT changes, with co-location and cloud computing becoming more important in the mix of facilities being used by organisations, total flexibility around a UPS installation will be more of a requirement. When combined with the uncertainties around grid power availability and the quality of the feed, now is the time for data centre managers to review their existing approach to their energy usage and to ensure that whatever systems are chosen can be fully explained to the business to gain its support.

Quocirca 2012

- 11 -

About Eaton Corporation

Eaton Corporation is a diversified power management company with more than 100 years of experience providing energy-efficient solutions that help our customers effectively manage electrical, hydraulic and mechanical power. With 2011 sales of $16.0 billion, Eaton is a global technology leader in electrical components, systems and services for power quality, distribution and control; hydraulics components, systems and services for industrial and mobile equipment; aerospace fuel, hydraulics and pneumatic systems for commercial and military use; and truck and automotive drivetrain and powertrain systems for performance, fuel economy and safety. Eaton has approximately 73,000 employees and sells products to customers in more than 150 countries. For more information, visit www.eaton.com.

REPORT NOTE: This report has been written independently by Quocirca Ltd to provide an overview of the issues facing organisations seeking to maximise the effectiveness of todays dynamic workforce. The report draws on Quocircas extensive knowledge of the technology and business arenas, and provides advice on the approach that organisations should take to create a more effective and efficient environment for future growth.

About Quocirca
Quocirca is a primary research and analysis company specialising in the business impact of information technology and communications (ITC). With world-wide, native language reach, Quocirca provides in-depth insights into the views of buyers and influencers in large, mid-sized and small organisations. Its analyst team is made up of real-world practitioners with first-hand experience of ITC delivery who continuously research and track the industry and its real usage in the markets. Through researching perceptions, Quocirca uncovers the real hurdles to technology adoption the personal and political aspects of an organisations environment and the pressures of the need for demonstrable business value in any implementation. This capability to uncover and report back on the end-user perceptions in the market enables Quocirca to provide advice on the realities of technology adoption, not the promises. Quocirca research is always pragmatic, business orientated and conducted in the context of the bigger picture. ITC has the ability to transform businesses and the processes that drive them, but often fails to do so. Quocircas mission is to help organisations improve their success rate in process enablement through better levels of understanding and the adoption of the correct technologies at the correct time.

Quocirca has a pro-active primary research programme, regularly surveying users, purchasers and resellers of ITC products and services on emerging, evolving and maturing technologies. Over time, Quocirca has built a picture of long term investment trends, providing invaluable information for the whole of the ITC community. Quocirca works with global and local providers of ITC products and services to help them deliver on the promise that ITC holds for business. Quocircas clients include Oracle, Microsoft, IBM, O2, T-Mobile, HP, Xerox, EMC, Symantec and Cisco, along with other large and medium-sized vendors, service providers and more specialist firms. Details of Quocircas work and the services it offers can be found at http://www.quocirca.com Disclaimer: This report has been written independently by Quocirca Ltd. During the preparation of this report, Quocirca has used a number of sources for the information and views provided. Although Quocirca has attempted wherever possible to validate the information received from each vendor, Quocirca cannot be held responsible for any errors in information received in this manner. Although Quocirca has taken what steps it can to ensure that the information provided in this report is true and reflects real market conditions, Quocirca cannot take any responsibility for the ultimate reliability of the details presented. Therefore, Quocirca expressly disclaims all warranties and claims as to the validity of the data presented here, including any and all consequential losses incurred by any organisation or individual taking any action based on such data and advice. All brand and product names are recognised and acknowledged as trademarks or service marks of their respective holders.