Contents'
1! Introduction ... 1!
1.1! Background ... 1!
1.2! Retrofitting Existing Data Centres ... 1!
2! Air Management ... 2! 2.1! Containment ... 2! 2.1.1! Blanking Panel ... 2! 2.1.2! Containment Curtains ... 3! 2.1.3! Resilience ... 4! 2.2! Improved Airflow ... 4! 2.2.1! Floor Grilles ... 4! 2.2.2! Foam ... 4! 3! Cooling Infrastructure ... 5! 3.1! Airside Economisation ... 5!
3.1.1! Example of Airside Economisation ... 6!
3.2! Wetside Economisation ... 7!
3.2.1! Example of Chilled Water System Economisation ... 8!
3.2.2! Example of DX System Economisation ... 9!
3.2.3! Example of Evaporative Wetside Economisation ... 9!
3.3! Other Retrofit Options ... 10!
3.3.1! Underfloor Fans ... 10!
3.3.2! Variable Speed Drives (VSDs) ... 12!
4! IT Infrastructure ... 13!
4.1! Initial Analysis ... 13!
4.1.1! Characterisation of IT Applications ... 13!
4.1.2! Examination of CPU Usage ... 13!
4.2! IT Power Management ... 14!
4.3! Consolidation ... 14!
4.4! Virtualisation ... 14!
4.5! Server Refresh ... 15!
4.6! Storage Optimisation ... 15!
4.6.1! Virtualisation and Data Management ... 15!
4.6.2! Capacity Management ... 15!
5! Other Initiatives ... 16!
5.1! Metering ... 16!
5.1.1! Handheld Multi-Core Amp Meter ... 16!
5.2! Fire Suppression ... 16!
5.3! Lighting ... 17!
1
1 Introduction
This guide outlines a selection of practical retrofit options to increase the energy efficiency of a data centre. The emphasis is on proven measures based on modern cooling techniques and the latest cooling infrastructure.
As well as the retrofit options outlined, based on the practical experience of the Data Centre Special Working Group (SWG) members, the report also examines ways of improving the energy efficiency of IT infrastructure being managed in the centre. Too often, energy efficiency initiatives concentrate on the cooling infrastructure without examining how the IT infrastructure is being managed from an energy efficiency perspective.
1.1 Background
The Data Centre SWG is part of SEAI’s Energy Agreements Programme (EAP), which supports large industrial energy users in implementing energy efficiency projects. EAP participants agree to implement the EN16001 Energy Management Standard and to pursue an aggressive programme of energy efficiency action and investment, in return for relationship support, advice, networking and financial supports.
SEAI developed the SWG annual initiative as an initiative within the EAP. An SWG is formed by SEAI with members drawn from either the EAP or the Large Industry Energy Network (LIEN). SWGs may focus on a particular area of technology, an area of special interest of members, or the Energy Management System (EnMS). In this instance, the SWG is made up of participants from the data-centre industry.
This report is aimed at data-centre managers looking for practical ideas on how to improve the energy efficiency of a data centre. It attempts to provide a holistic approach to reducing energy consumption by encouraging managers to improve air management, reduce the cooling load, optimise the IT load, and examine other potential initiatives.
1.2 Retrofitting Existing Data Centres
Retrofitting existing data centres for increased energy efficiency is inherently complex, from two perspectives:
• Existing Operations: The data centre being retrofitted is operational and must remain so. All renovation activities must take place while maintaining continuity of service.
• Building Structure: Many older data centres are hosted in buildings which, through their original design, present obstacles to incorporating the most up-to-date energy-efficient methods and infrastructure.
Consequently, not all options described in this document may be relevant or practical for a given data centre.
The guide is broken down into the following four sections, each dealing with an aspect of energy efficiency in a data-centre environment:
1. Air Management – managing air distribution through better containment and improved airflow
2. Cooling Infrastructure – optimising the cooling load through more efficient use of cooling infrastructure
3. IT Infrastructure – reducing the energy consumption of the IT infrastructure through greater use of IT power management, increased consolidation, the use of virtualisation technologies, server refresh, and storage optimisation
2
2 Air Management
A proven best practice for better energy efficiency in data centres is the implementation of better airflow management. Airflow in a data centre operates at three levels: in the server room (perimeter), at row level, and at rack level. For the purposes of this document, containment is applicable at the first two levels, i.e. at the room and row levels.
The most common approach is to (1) separate hot and cold airflows through practical initiatives such as hot aisle/cold aisle containment, and (2) improve airflow through the optimal positioning of floor grilles and better extraction of hot air.
When the mixing of hot and cold air has been reduced or eliminated, the cold air supply temperature can be gradually increased.
2.1 Containment
The two main methods of containment are hot aisle containment and cold aisle containment. Each method provides significant energy savings compared to a non-contained arrangement. Hot aisle containment is preferred when building a new data centre, as it is better suited to take advantage of airside economisers and can accommodate higher rack densities. However, in a retrofit situation, where there is a raised floor and low headroom (or no dropped ceiling plenum), cold aisle containment may be the more practical option.
Containment is a logical extension of the policy of organising data centres in a hot/cold aisle configuration. The primary purpose of hot/cold configuration is to guarantee the delivery of air to the inlets of the IT racks within the recommended environmental limits. Hot/cold aisle configuration helps in this by directing the supply air to the point of need and by ensuring that exhaust air is not directed towards rack inlets. However, even with hot and cold aisle configuration, it remains difficult to guarantee the correct environmental conditions to the inlet to every single piece of IT equipment. Optimising hot/cold aisle configuration, installing blanking walls and panels and generally managing airflow effectively can allow a much higher proportion of equipment to be running within the recommended operating conditions.
Without containment, the temperature profile from the bottom of a rack to the top of a rack can be quite large, due to the short-circuiting and mixing of hot air into the cold aisles. Consequently, air needs to be supplied at a much lower temperature to guarantee recommended inlet conditions are met at all points – particularly, for example, at the top of the rack. When containment is installed throughout a data suite, the supply air temperature to the cold aisles can be raised to a much higher level since the temperature profile of the air in contained aisles is highly consistent.
This guide does not intend to concentrate on specific hot aisle/cold aisle containment systems as there is a wide array of manufacturers in the market. Instead, the focus will be on inexpensive but effective containment methods.
2.1.1 Blanking Panel
Unused slots in racks need to be blanked off in order to guide the flow of air and prevent the mixing of hot and cold air. A practical and inexpensive method of blanking off slots is to use polycarbonate sheets, cut to the shape of the gap being filled.
3
Figure 2.1: Polycarbonate Blanking Place (source: Vodafone)
2.1.2 Containment Curtains
A proven and relatively inexpensive method of partitioning a data centre into cold aisles and hot aisles is by means of PVC curtains. The advantages of curtains are that they:
• Can be cut to take account of the varying heights of equipment in the data centre • Can be easily removed and reconfigured when adding or removing equipment • Permit freer movement of personnel in the lab
• Are inexpensive to install and maintain
The photograph below shows a curtain arrangement for partitioning aisles; in this instance, for hot aisle separation.
4
2.1.3 Resilience
Data-centre resilience should not in any way be affected by installation of hot/cold aisle containment. There are a couple of relevant points. Air volume to the cold aisles should be somewhat greater than required (e.g. N+25%) to ensure continued environmental control in the event of failure of a number of Computer Room Air Conditioner (CRAC) units. As part of the standard data-centre airflow management regime, air volumes should be monitored to maintain required levels as IT equipment load evolves. Also, control practices should be put in place to ensure cold aisle integrity is maintained and thus avoid the occurrence of hot spots due to a breach of the containment (e.g. a ceiling tile becoming dislodged).
In the event of a mains fail and temporary shutdown of CRACs pending generator activation, although the initial temperature in the cold aisle may be higher than in a non-cold aisle scenario, the rate of increase of temperature will be less due to the server fans pulling air from the reservoir of cooler air from the floor void via the cold aisle.
2.2 Improved Airflow
In a cold aisle containment scenario, airflow into the racks from a raised floor plenum can be optimised by means of well-positioned floor grilles, and through the use of foam or grommets to prevent air leakage through the floor.
2.2.1 Floor Grilles
There are many different floor grille types being marketed by a range of manufacturers. The ideal floor grille maximises the direction and power of the air flowing through it. An effective grille reduces short cycling and increases the stratification level of the air, concentrating as much cold air as possible on the front of the racks.
2.2.2 Foam
The application of polyurethane foam is a good means of assuring containment. It prevents bypass air, stops recirculation, and makes gap filling quick and inexpensive. It is ideal for raised floors as it can be applied from above.
The photograph below shows foam placed around cabling to prevent a potential loss of cold air from under a raised floor plenum.
5
3 Cooling Infrastructure
Better air management allows data-centre managers to examine next the behaviour of the cooling load. As highlighted in the previous chapter, when hot aisle or cold aisle containment is installed throughout a data suite, the supply air temperature to the equipment can be raised to a higher level since the temperature profile of the air in the contained aisles is highly consistent.
When the mixing of hot and cold air has been reduced or eliminated, the supply air temperature can be gradually increased in line with American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) recommendations, leading to better Co-efficiency of Performance (COP) of the cooling infrastructure. By raising supply air temperature, energy can be saved in a number of ways. For example:
• Raising chilled water temperatures leads to more efficient chiller operation and reduced transmission losses.
• There is a greater free cooling window for both direct and indirect free cooling.
• There is less wasteful latent cooling (condensation on the cooling coil in the CRAC) and also, as a consequence, reduced rehumidification (energy-intensive when steam humidifiers are used).
• Even though the heat load in the room remains the same, the cooling load may be reduced somewhat at higher supply temperatures due to greater room heat losses as a result of raised ambient room temperatures.
Furthermore, there is the reduced volume of air required from the CRAC units. Normally, excess volume is needed to minimise mixing from hot aisle to cold aisle. With hot/cold aisle containment, only the required volume of air, plus some extra for redundancy, is produced, which also allows for a slightly positive pressure in the cold aisle. The net effect is that fan speeds can be reduced, which presents an additional energy saving. (It should be noted that the energy consumption of a fan motor decreases exponentially with a decrease in fan speed.)
The rest of this chapter outlines the economisation options available in a retrofit situation, in order to reduce the costs of cooling a data centre.
3.1 Airside Economisation
Airside Economisation uses air as the heat transfer medium by bringing cool outside air directly into the data centre. Airside economisation is also called Fresh Air Cooling or Direct Free Air Cooling. Typically, Air Handling Units (AHUs), based on the roof of the data centre (see the figure below), draw in cool air that is then ducted into the data centre. The airflows through the servers exiting in a hot aisle and the heated air is then vented through a plenum and returned to the AHUs. The advantage of airside economisation is that it provides more hours of economisation than waterside economisation. Depending on the climate expensive pumping of water is not needed, there is less use of water resources and there is no refrigeration cycle, during normal operation.
6
Figure 3.1: Diagram of Air Economisation (source: www.datacenterknowledge.com)
3.1.1 Example of Airside Economisation
In 2010 EMC Ireland implemented an airside economisation project in its data centre in Cork. AHUs (see example in Figure 3.2 below) were installed to admit outside fresh air, with the fresh air being filtered to prevent ingress of particulates. If the outside temperature is too low, the fresh air is mixed with return air from the server room. If the outside temperature is too high, the fresh air is cooled by a cooling coil. The air is then directed into a plenum down the side walls and into a raised floor plenum, where it is directed up through the racks. The heated air is returned through the AHU.
7
Fresh air mix sections were added where existing recirculated AHUs were in place, along with high efficiency motors and variable-speed drives.
The figure below shows air mix sections being added to the AHUs.
Figure 3.3: Addition of Air Mix Sections to Air Handling Units (source: EMC)
3.2 Wetside Economisation
Wetside Economisation uses a cooling tower, evaporative cooler or dry cooler to cool water or a fluid that is then piped into the data centre where it acts as a heat transfer medium. Wetside economisation is also referred to as Free Cooling or Indirect Free Cooling.
'
Wetside economisation is primarily used in combination with Chilled Water systems and Direct Expansion (DX) systems:
8
• Economisation with DX Systems: In a DX system, the refrigerant cycle is contained in the CRAC unit. When implementing wetside economisation, a separate economiser coil is installed in the CRAC unit. When ambient conditions permit, the economiser coil allows for the data-centre heat load to be cooled by external air, removing the need for the CRAC’s normal refrigeration cycle.
3.2.1 Example of Chilled Water System Economisation
EMC installed a Dry Cooler in sequence with the existing chiller. The water returning from the CRAH is first cooled by the dry cooler, when the outside air is sufficiently cool, before being forwarded to the chiller to be cooled further (see Figure 3.4 below). The chiller refrigeration compressors are power-consuming and expensive to run. Installing a free cooler removes heat from the fluid before it reaches the chiller and reduces the amount of heat to be extracted by the chiller compressors. Typically, in low ambient conditions, when the free cooler is fully used, less than 1 kW of power input can achieve 30 kW of cooling.
Figure 3.4: Diagram of Dry Cooler and Chiller (source: EMC)
9
Figure 3.5: Installation of dry coolers (source: EMC)
The CRAHs are now equipped to always free-cool first. When the data-centre setpoint temperature is not being achieved, the CRAHs increase the speed of their fans. If at this stage the data-centre setpoint temperature is still not reached, the floating setpoint on the chillers decreases until cooling is achieved.
The CRAH controls were upgraded and Modbus linked to the Building Management System (BMS), to ensure that the floating chiller setpoints are operational.
3.2.2 Example of DX System Economisation
For a CRAC unit containing a Direct Expansion (DX) type compressor, a new free cooling coil was added and connected to an outdoor dry cooling unit. The compressor on the CRAC was also retrofitted with electronic expansion valves.
These works involved removing the CRAC units individually for retrofitting offsite, as the installation of a pre-cooling coil was deemed to be too labour-intensive to retrofit within a live data centre. It’s important that there is sufficient redundancy of CRAC units to ensure continuity of cooling while the retrofit is taking place. If not, a form of temporary cooling will be required.
3.2.3 Example of Evaporative Wetside Economisation
As we have seen, wetside economisation typically uses one of two means of dissipating heat to the atmosphere – either a dry-cooler, as in the EMC case above, or by using evaporative cooling. The case that follows describes how Intel Ireland Ltd improved the energy efficiency of its largest data centre by using evaporative cooling.
Using the thermodynamic effect of evaporative cooling, wetside economisers typically use condenser water produced by cooling towers to cool the warmer water returning from the data-centre air handlers. Similarly to the EMC case, this ‘free’ or rather highly efficient cooling reduces the load on the chillers, thus lowering power consumption.
10
evaporation by drawing air through the fill. The cooled water is collected at the bottom of the tower and redistributed to the chiller for further cooling or directly to the data centre.
The Intel decision to use an evaporative economiser was primarily based on return on investment. Dry coolers are typically a more efficient option, having greater operating hours relative to a tower. However, Intel already had an underused tower onsite. The capital cost of purchasing the dry coolers and the civil cost involved in installing them drove the decision to use evaporative cooling.
The basis for the installation was the reuse of the existing tower. The option involved the installation of a heat exchanger, controls, valves and a primary pump (see Figure 3.6 below).
Figure 3.6: Evaporative cooling (source: Intel Corporation)
Return water passes through the secondary side of the heat exchanger and, when environmental conditions permit, this return water is cooled by water from the tower, which is pumped into the primary side. If conditions are not favourable, the water continues on as per the original design, to be cooled by the chillers.
With changes made in the data entre to segregate the cold and hot aisles, supply water temperature was increased from 6°C to 10°C. The evaporative project saved approx. 50% on cooling costs relative to the base scenario.
3.3 Other Retrofit Options
3.3.1 Underfloor Fans
11
• There is increased heat exchanger area, increased filter area and reduced internal losses due to deflection of air, contributing to a reduction in input power requirement
• Underfloor fans are a more efficient use of data-centre space The figure below shows where the underfloor fan is located.
Figure 3.7: Diagram of an underfloor fan (source: Weiss Klimatechnik GmbH)
There is a new EU rating system for AC Motor Efficiency. Comparing the new system to the old system:
New EU system (also used by China, USA, etc) OLD EU system
IE4 (called ‘super premium motors’)
No equivalent
IE3 (called ‘premium-efficiency motors’)
No equivalent
IE2 (old best, called ‘high-efficiency motors’)
EFF1
IE1 (old system, called ‘standard efficiency’)
EFF2
No equivalent
EFF3
IE4, which is a permanent magnet (synchronous) motor technology, is the best of the new system. There is also a DC technology, commonly referred to as EC (Electronically Commutated) Plug Fans, that is being used for small fan motors in data-centre CRAC units. These are very efficient and are rapidly gaining in popularity for use in data-centre air-conditioning.
12
Figure 3.8: Installation of EC fans under a CRAC unit (source: EMC)
It is advisable to monitor the acoustics of the fans before and after installation.
3.3.2 Variable Speed Drives (VSDs)
The installation of VSDs to CRAC units is highly recommended. When modifying the cooling in a data centre for greater energy efficiency, the two main control variables are the chilled water/refrigerant temperature and the volume of air being cycled in the room. Circulated airflow can be reduced either by reducing fan speeds or switching off CRAC units.
13
4 IT Infrastructure
Better energy management of the IT infrastructure affects the energy consumption of the data centre as a whole. A decrease in IT-related energy consumption not only reduces the IT load and by extension the total energy used by the data centre, it should also reduce the energy consumption of the data centre’s supporting cooling infrastructure. There is therefore a double benefit to making the energy use of the IT infrastructure as efficient as possible.
This section of the guide concentrates on five areas relating to improving the energy efficiency of IT infrastructure:
1. Initial analysis to classify the infrastructure and applications in terms of availability and use, so that an effective optimisation strategy can be developed and managed
2. IT power management applications and techniques available to data centre management 3. The role and impact of consolidation and virtualisation in the drive to increase energy
improvements
4. Refreshing server equipment to reduce energy consumption
5. Optimisation of storage equipment through virtualisation, data management and capacity management
A useful reference is the EU Code of Conduct for Data Centres, which addresses IT equipment and services in its Best Practices guide,1 advising on:
• Selection and Deployment of New IT Equipment – section 4.1 • Deployment of New IT Services – section 4.2
• Management of Existing IT Equipment and Services – section 4.3 • Data Management – section 4.4
4.1 Initial Analysis
It’s important for data-centre management to gain a definitive understanding of the current usage of IT resources, before it can plan the necessary steps to greater energy efficiency. To know where to go, you have to know where you are!
4.1.1 Characterisation of IT Applications
Enterprises and organisations host many different applications, and each application has its own profile in terms of availability, use, user base, expected life-cycle and so on. Each application needs to be characterised to understand the organisation’s requirements for that application.
For example, Intel uses the acronym DOME (i.e. Design, Office, Manufacturing and Enterprise) to define the four main computing environments in the company. Different environments require different availability of service. An office environment will have applications that will not be needed outside office hours, thereby offering the possibility of shutting down these applications at night-time or at weekends. Furthermore, some applications may not be mission-critical and may require less redundancy, thereby consuming less energy.
It’s important that data-centre management understand the applications being run in its facility in order to better manage the overall use of IT resources, from an energy consumption perspective.
4.1.2 Examination of CPU Usage
Historically, IT equipment has been run very inefficiently, often with CPU usage at less than 10%. However CPU use on its own is a blunt measurement – a server may have high usage but be running redundant applications. Often over time applications are no longer used, but are still left running on
1 http://re.jrc.ec.europa.eu/energyefficiency/pdf/CoC DC new rep form and guidelines/Best Practices
14
servers, to no practical purpose. Efficiencies may be gained through consolidating legacy services, and powering off unused or redundant servers and applications.
Also, with the advent of virtualisation technologies and the resultant capability to run multiple ‘virtual’ machines on a real machine, CPU usage can be greatly increased, often by a factor of 10. Consolidation and virtualisation have direct benefits in terms of energy consumption. An idle server can use between 30% and 60% of its maximum power. As a simple example, using virtualisation technologies to consolidate eight low-usage machines onto one busy machine, though increasing the power used by that machine to its maximum, eliminates the power consumed by the other seven machines, thereby reducing overall power consumption by up to 70%.
Before embarking on a consolidation or virtualisation exercise, it’s important to characterise the applications being used in terms of required availability, user base, expected life-cycle and so on.
4.2 IT Power Management
Equipment vendors provide a range of specialised IT power management applications, some examples of which are listed below:
• Intelligent Platform Management Interface (IPMI) is an industry standard that defines a set of common interfaces to a computer system, enabling administrators to monitor and manage the system’s behaviour. The standard is supported by over 200 equipment vendors, including Dell, HP and NEC.
• Intel® Intelligent Power Node Manager is available with Intel Xeon processor 5500 series. This enables power management of individual servers through modulation of the
processor’s performance (P) and throttle (T) states. It measures the power consumption of a server and can limit the power to a targeted power budget. It also reports the server power consumption, sending alerts if the targeted power budget is being exceeded.
• Intel® Data Centre Manager (DCM) is a software development kit and reference user interface that provides power and thermal monitoring and management for servers, racks and groups of servers in data centres. DCM can aggregate data from systems enabled with Intel® Intelligent Power Node Manager, allowing administrators to monitor and limit the power consumption of banks of servers.
4.3 Consolidation
The objective of server consolidation is to ensure that each server in a data centre is used to its maximum utilisation, thus reducing the number of servers being used overall. Consolidation centralises applications on fewer servers, thus avoiding the situation where servers are running at low utilisation or running applications that are no longer being used.
Effective consolidation is based on a comprehensive understanding of the applications being run on the servers, in terms of each application’s criticality and its pattern of usage. For example, applications that are not needed outside office hours can be consolidated on servers which can then be powered down at night-time and weekends.
The term consolidation is often used in conjunction with the term virtualisation. However, they are different activities. Some server consolidation can be achieved without virtualisation, whereas virtualisation is a form of consolidation.
4.4 Virtualisation
Virtualisation enables the installation of multiple virtual machines on a real machine. This increases the use of each real machine (ideally a usage target of 80%), enabling the consolidation of IT equipment in the data centre, with follow-on potential reductions in power usage and overall cooling demand.
15
• Application Profiles. Some applications do not lend themselves to virtualisation as they may be based on specialised legacy equipment that cannot be virtualised. It’s also important that applications sharing a common platform have the same availability profile, e.g. 24-hour operations, office hours, night-time batch applications, and so on.
• Higher-Density Equipment. Virtualisation leads to increased rack power density, with power and cooling requirements similar to those of blade servers.
• Increased Server Criticality. Each physical server becomes more important in a virtualised environment, with more applications affected in the event of a server failure.
• Hot Spots. Cooling requirements become more localised as the rack density increases, increasing the need for better air management (as already documented in section 3 of this report). Cooling requirements can also become time-dependent as servers are dynamically stopped and started in response to, for example, office hours.
• Reduction in IT Load. The reduction in power used by the IT infrastructure, arising from virtualisation, presents data-centre management with the opportunity to gain energy efficiencies from reducing the cooling load in response.
Perversely, virtualisation increases the Power Usage Effectiveness (PUE), because a reduction in the IT load means that there is often an excess of fixed cooling capacity. It is only when the cooling load is decreased correspondingly that the PUE can be brought back to its original levels.
4.5 Server Refresh
New multi-core servers have greater capacity and are more energy-efficient than older servers. By continuously refreshing the server stock (e.g. every four years), a data centre benefits from greater consolidation onto blade servers while substantially reducing energy consumption.
Non-energy benefits of server refresh include construction cost avoidance (reduced need to expand facilities), and lower network costs (reduced switch port requirements).
Therefore, data-centre management should consider a strategy of accelerated server refresh combined with other initiatives addressed in this chapter, such as consolidation and virtualisation.
4.6 Storage Optimisation
The approach to the energy optimisation of storage infrastructure is similar to that for server infrastructure – i.e. analyse the current situation in detail before implementing a well-planned consolidation of infrastructure.
4.6.1 Virtualisation and Data Management
Storage virtualisation is an important energy optimisation enabler, allowing multiple systems to share a common physical storage device. Methods for optimising data that can be used in conjunction with virtualisation include:
• Storage Tiering. Applications and data can be mapped and aligned on storage tiers based on the data value, eliminating the storage of low-grade data on expensive devices.
• De-duplication. De-duplication methods examine data to find and erase redundant copies. A phased overall approach to virtualisation and consolidation is required: beginning with assessment, followed by a pre-production deployment, then moving on to non-critical workloads, and finishing with the remaining workloads.
4.6.2 Capacity Management
Capacity management technologies offer opportunities to improve use of storage devices:
16
• Thin Provisioning. Thin provisioning improves use by assigning physical disk space only when data is written, rather than when logical volumes are created.
5 Other Initiatives
5.1 Metering
What isn’t measured can’t be managed. It’s important, therefore, to determine and monitor the different types of energy consumption, such as data-centre electrical demand, IT load, cooling load, transmission losses, UPS losses, lighting load, and so on.
Often in data centres, energy usage is monitored at room level, with little visibility of usage at row and rack level. There are practical means of building a more granular picture of energy consumption, with for example, localised readings being taken in the data centre using a handheld multi-core meter.
5.1.1 Handheld Multi-Core Amp Meter
The electrical load at rack and server level can be measured using a non-invasive multi-core amp meter. The measurements provide an electrical KW value and corresponding cubic-feet-per-minute value for every work cell in the data centre. Consequently, areas of overcooling and hotspots can be identified.
The figure below shows an example of a handheld multi-core clamp meter.
Figure 5.1: Handheld multi-core clamp meter (source: Megger)
5.2 Fire Suppression
17
construct, such as light polycarbonate, which will either collapse in the event of sprinkler activation or, more than likely, have melted well in advance of such an activation.
Different considerations may apply for rooms protected by fog or gas systems. The general point, though, is that it should be possible to agree a system that is acceptable to a loss insurer while minimising retrofit costs.
5.3 Lighting
Lighting usually accounts for less than 1% of a data centre’s energy consumption. Nevertheless, this consumption can be optimised through the use of Passive Infra-Red controls (PIRs) and high-efficiency light fittings.
A PIR ensures that lights only come on when movement is detected within a range, which can be up to 20 metres. This is more energy-efficient than the more traditional method of using switches to turn on and off the lights when entering or leaving a space. It is important to choose the right PIR in order to avoid the possibility of heat dissipation from server racks causing control issues on lighting. Energy-efficient light fittings, such as the compact fluorescent light bulb, are more energy-efficient, produce more light per watt, and have a longer lifespan than the traditional incandescent bulb. However, they do come at a greater cost, so there is a trade-off in terms of energy efficiency versus capital cost.
18
Glossary of Terms
AHU Air Handling Unit
COP Co-efficiency of Performance
CRAC Computer Room Air Conditioner. In a CRAC unit the refrigeration occurs either partially or completely with the unit.
CRAH Computer Room Air Handler. In a CRAH unit the refrigeration occurs outside the unit, i.e. by means of a chiller.
DX Direct Expansion. A type of refrigeration compressor in a CRAC unit. EnMS Energy Management System
