Data Centre Infrastructure: Power Efficiency and Protection

Abstract

The rapid expansion of e-commerce necessitates expanding the capacity and dependability of data centres in order to provide services with the proper level of quality. A green technology that has a lot of potentials to reduce CO2 emissions is optimization data centre design. However, a large data centre required a large amount of electricity because the capacity of the racks is higher, which required more potent cooling systems, power supplies, protection and security. These will increase the cost of the data centre and render it unusable for services. In this chapter, we provide a design for a tire-four data centre that will be situated in Cyberjaya, one of Malaysia’s best locations. This design’s primary goal is to offer highly functional and high-quality e-commerce services, particularly food delivery. Each component of the data centre has been carefully developed to deliver a range of services, including the administration of IT infrastructure, co-location, cooling system and protection. Additionally, advice and support have been given to guarantee that the suggested design outperforms competing data centres in terms of dependability, power efficiency and storage capacity. The analysis, synthesis, and evaluation of each element of the proposed data centre will be considered in this chapter.

1. Introduction

Meza is one of the native data centre firms that offer a range of services, including the administration of IT infrastructure and top-notch security and data management. Meza is anticipated to be constructed in a number of data centres throughout Malaysia. More than five million people in Malaysia use a food delivery application company’s services, and that number keeps growing. Due to the current infrastructure’s inability to handle the massive quantity of data processing, users may have a bad user experience as a result of the lengthy response times from servers and delayed processes. In order to accommodate the company’s continuous growth, Meza has been chosen by the corporation to build a data centre. Customer relationship management, which unifies all client communication into one inbox, will be needed by the data centre to conduct online food orders and payments.

For this chapter, a data centre design with the necessary components will be provided, and each component of the proposed data centre will be examined, synthesised, and evaluated. These data centre components including power usage effectiveness and efficiency, cooling system and protection. Other data centre components, such as storage infrastructure, networking and environment have been discussed in another chapter Data Centre Infrastructure Design and Performance.

2. Power system

2.1 Electrical power system

Power is an important element that gives life to a data centre and maintains the IT infrastructure even when an interruption takes place. The value of the power system in the data centre cannot be emphasised enough. Power is one of the significant factors when it comes to cost estimation of colocation services as well. Additionally, the type of currents used to power the servers, switches, routers and associated IT infrastructures are of two different types. One of them is DC (Direct Current) and the other is AC (Alternating Current) [1].

Meza has decided to use 277/480 (277 for Single phase or 480 for three-phase power) Volt AC power supply for the food delivery data centre because it removes the PDU (Power Distribution Unit) and passes directly the power to the server cabinet at a higher voltage. It is also energy efficient and provides a decreased load for the cooling systems. Furthermore, it has increased consistency [2]. Moreover, in [3] reinforcement learning method is applied for automating energy efficiency.

Since Tier 4 has been chosen for this data centre, there are a few components that will be discussed which will allow the data centre to be up and running 24/7. The components that Meza has decided to use are:

• Automatic Transfer Switch

• Backup Power Supply

• RPDUs (Rack Power Distribution Units)

• Backup Generator

2.1.1 Automatic transfer switch

When there is an interruption of power supply, an Automatic Transfer Switch (ATS) is used. It is an electrical switch that moves the source of the power supply from the main source to a backup source. If the primary power source detects a power loss, the ATS triggers the substitute power sources which will provide a continuous power supply [4]. There are four types of ATS, and after a copious amount of research, Meza has decided to use Closed Transition ATS. This switch functions in a way where a little pause is not accepted in the power source supply which allows it to detect a power blackout and power fluctuation earlier and allows the data centre to have a smooth switchover while running the main power supply and the backup supply simultaneously without any interruption of services in the data centre [4] (Figure 1).

Figure 1.

A comparison between traditional ATS and closed transition ATS [5].

2.1.2 Backup power sources

The need for backup power resources is very crucial to provide with uninterruptable power supply for the data centre in the event that the primary source of power supply fails. Two of the backup power sources chosen are:

• Uninterruptible Power Supply (UPS): UPS backup power is a critical component. Without it, fluctuations in power and outages can push workloads down, damage equipment and result in hefty payments. The main aim of it is to preserve the data centre infrastructure until stable power returns or long-term alternative power backup systems are up. There are mainly three types of UPS, they are:

  • Standby/offline UPS

  • Line Interactive

  • Online/Double Conversion

• The Online/Double Conversion UPS has been recommended by Meza. Online/Double Conversion UPS is the kind of UPS that, during normal operations, provides AC load power through a Rectifier and Inverter Combo and uses an AC power inverter during power failure. The output power supply therefore still stays ON and switching is not required [6]. This UPS has been chosen because as stated by [7] that it provides the highest degree of security for a facility by shielding IT equipment from the raw power utility. This system operates by power conversion from AC to DC and back into AC. Additionally, this UPS is said to be the only kind of UPS with zero battery transmission time and is suitable for sensitive applications. That is why this UPS is the safest to use for any mission-critical equipment for any facility such as the food application data centre.

• Backup Diesel Generator: When the main power source is disrupted, backup generators are used to provide electricity for the data centre. Power interruption due to grid failures can lead to a high risk of operational loss in data centres. Furthermore, the system and components of data centres should function constantly and require a reliable, continuous power supply 24 hours a day, 7 days a week. If the event of a power lapse, files can be lost or corrupted, mainframes can malfunction [8].

2.1.3 RPDUs (rack power distribution units)

The RPDU generates no power but rather delivers power from the available power supply. In the world of data centres, the RPDU is able to monitor, manage and regulate the power usage of many devices. It can supply vast quantities of electricity and can be accessed via the local or remote network. RPDUs can withstand high power density and are immune to greater temperatures in order to satisfy the ever-changing needs of the data centre [9]. Rack PDUs can be categorised into two categories which are non-intelligent PDUs and smart PDUs [10]. There are mainly four types of PDUs, they are:

• Basic Rack PDU

• Metered Rack PDU

• Monitored Rack PDU

• Switched Rack PDU

Since the aim of the proposed data centre is to be robust, highly reliable and highly effective, only a smart PDU will be discussed in this paper which will be the modern Switched Rack PDU.

Switched Rack PDU is a power management unit that can be installed on a typical industry rack and have the capability to power on and off remotely for individual outputs. Moreover, the Switched Rack PDU offers outlet control for rebooting locked devices as well as remote access capabilities to power and environment information. It also offers current, voltage, power (kW), apparent energy and cumulative energy measurements per outlet (Figure 2) [11].

Figure 2.

A switched rack PDU design [12].

2.1.4 IT space allocation

The above picture showcases an electrical power system supply arrangement in a data centre as well as how the cablings are handled (Figure 3).

Figure 3.

Proposed IT space for the electrical power supply [10].

2.1.5 Justification

As the data centre for the food delivery company has been decided to go with tier 4 data infrastructure, the electrical power supply has been tailored to those needs. Two main utility power grids have been arranged. The data centre has also been equipped with a Closed Transition ATS because this automatic transfer switch will make sure that there will not be any power outage when transferring the power from the main supply to the backup supply which will ensure a smooth service for the food delivery application. Furthermore, the data centre has been equipped with Online double conversion UPS because, with this, there is very less chance of electrical load loss. It is also efficient and has a good PUE (Power Usage Efficiency) as supported by [13]. Moreover, a backup generator has been placed in case both the power grid somehow blackouts.

Additionally, for the rack power distribution unit, Switched Rack PDU has been placed because it will allow the higher officials of the data centre to remotely monitor all the data load and electricity consumption and as well as remotely change the voltages for the racks/servers which will make the energy consumption of the data centre even less. This statement is also approved by [14]. Lastly, the power utilisation of this data centre is highly efficient because of the necessary features that have been added which are also supported by [15] that by adding features such as power-saving “standby” modes, energy management software and efficient cooling systems, data centres can become more energy-intensive. Such improvement in efficiency will produce significant energy savings and reduce the electricity grid load.

2.2 Fire detection system

Fire detection systems are made for the early identification of fires while time for the safe evacuation of individuals is still available. In order to ensure the health of emergency response workers, early detection also plays an important role. Furthermore, fire damage and operational downtime can be minimised as monitoring measures begins while the fire remains low. Most alarm systems supply emergency responders with information about the location of the place where the fire has started which speeds up the process of controlling the fire [16].

There are mainly three levels in a data centre that needs to be protected, they are:

• Building Level: The primary goal is to avoid fire in the building and its workers. Fire sprinklers and portable extinguishers are the most common forms of fire safety.

• Room Level: This level focuses on specific rooms such as the data centre cabinet, surveillance room, etc.

• Rack Level: The last standard of fire safety at a data centre is on the rack level. Data servers, storage and network infrastructure must be protected using this fire protection to minimise fire damage (Figure 4) [17].

Figure 4.

The importance of rack-level protection by illustrating a monthly cost [17].

So, to minimise the downtime and loss of data for the food delivery data centre in case of a fire breakout, the recommended smoke detector, and the fire alarm system will be discussed in this chapter.

2.2.1 Smoke detector

A Smoke Detector is an electronic fire detection unit which senses the presence of smoke automatically. Smoke detectors are typically managed by a central fire alarm device, operated by building power with a battery backup in building infrastructures [18]. There are typically 3 types of smoke detectors. Air-aspirating/air-sampling smoke detector has been recommended by Meza.

2.2.1.1 Air-aspirating/air-sampling smoke detector (ASD)

This smoke detector has become very popular as it can detect fires at a very early point, even before smoking happens as well as before an open fire and before excessive smoke happens. ASD can locate fires considerably faster than point or beam detectors, meaning the first signs of smoke can be reacted to quickly. This early detection is important for sensitive and high-risk infrastructure. These detectors can also be set to a traditional point detector sensitivity level (Figures 5 and 6) [21].

Figure 5.

How an air-sampling smoke detector works [19].

Figure 6.

Performance levels for fire detection systems [20].

2.2.2 Fire alarm system

Once the smoke has been detected, the smoke detector will notify the fire alarm system. A Fire Alarm System is intended to alert people to an emergency in order to protect themselves. Whatever detection system it is, the sounders can be used to alert building staff about the risk of fire or evacuation if an alarm is activated. The Fire Alarm Control Panel is the ‘brain’ of the fire detector system. The central hub gives a status indicator to users for all the detector signals. This system can also be programmed to simulate an alarm for regular fire drilling and evacuation, so that all workers know what to do in case of a real fire [22].

Generally, fire detectors are of three types which are Conventional fire alarm system, Addressable fire alarm system and Wireless fire alarm system as stated by [23]. After plentiful amount of research, it is believed to go with Wireless Fire Alarm System for the food data centre infrastructure because the data centre already has many cablings with the IT equipment and resources. So, wireless fire alarm system has been chosen over addressable fire alarm system and regarding the cost efficiency, they both will cost roughly similar price because even though wireless alarm system is expensive but for an addressable fire system, it will take a high cost to arrange all the cablings as stated by [24].

Wireless Fire Alarm System: The solution of choice for many applications is wireless fire alarm device. The huge versatility and endless combination of wireless alarm devices make it a good choice for organisational sensitivity. Each unit of the range communicates with self-optimising amplitude and frequency through sophisticated bidirectional encrypted radio transmission. Additionally, multi-directional integrated antennas guarantee the virtual removal of signal corruption. Furthermore, wireless fire alarm systems have shown that they provide the best security in the premises in a reliable and cost-effective manner (Figure 7) [26].

Figure 7.

How a fire detection system along with a fire alarm system generally operates [25].

2.2.3 Recommendation

After critically analysing the different types of smoke/fire detectors, Meza has decided to use Air-Sampling smoke detector because since the data centre will handle a huge load and huge processing power for the food delivery infrastructure, this early smoke detection system should be installed in case there is an overload and overheating which may cause a fire. As supported by [20], it can help with relatively early fire warning detection in rooms containing IT and telecommunications equipment. Additionally, the very early warning of smoke for the staff running the infrastructure is a crucial aspect of air sampling smoke detectors. The early warning capability enables managers to evaluate a smoke long before it enters an emergency state and activates a fire suppression system.

Lastly, wireless fire alarm system has been suggested as well because the data centre needs to be protected explicitly from fires and needs to be cost-effective and more reliable. As stated by [27], wireless fire alarm system is easier to deploy with low downtimes and has easier maintenance. It also has higher reliability, is more cost-valuable in the long run and can be repositioned easily if necessary which are the key points that are being kept in mind by Meza to make the data centre more efficient, robust and safe; see Figure 6.

3. Fire suppression systems

For a data centre fire suppression system is mandatory. The food ordering application holds many data and computes many processes and it is mandatory to keep the data centre safe from a fire outbreak if happens. A fire suppression system is a collection of designed units installed with the application of a material to extinct flames. The fire protection device typically has a built-in component that identifies fires by flame, smoke and other alert signals at the beginning stages. These are connected to an alarm device to warn in the presence of the fire and to take action to avoid the fire. Most fire detectors activate an application of an exterior material upon identification and/or warning immediately to extinguish the fire. However, certain fire suppression devices are issued manually (Figure 8).

Figure 8.

Fire suppression system 1 [28].

3.1 Level of protection

There are three levels of protection for data centre. Which will ensure the safety of the data centre and the information stored in the system. First level of protection is building-level fire protection. The key goal is to defend the buildings and their workers from fire. Fire sprinklers and handheld extinguishers are the most widely used type of fire protection. The construction policy for handheld extinguishers requires that, if the class (A) combustible materials are in the workplace, there is a portable fire extinguisher for every 3000 square meters. A building may also use passive fire safety, including the construction of firewalls and floor mounts which dramatically delay the expansion of the fire in other areas of the building [29].

Second level of protection is room-level fire protection. The National Fire Protection Association (NFPA) sets the standards for room-level protection. The water still occurs in the piping of a wet piping network which will escape instantly after triggering of the warning. The downside to this device is that the pipe will leak and spill on the room facilities. The most commonly employed space fire safety is a pre-action device. The triggering of the sprinkler device needs at least two fire detection points. Other devices divide the space into sprinklers, so they just go off in the quadrant triggered. The best solution for data centre fire safety is fire sprinkler devices. Novec 1230 and FM-200 are the two rising safe agent gas systems. By raising the fire heat by absorption, they contain the fire. Such gases have zero ozone loss, rendering them biologically and humanly free. The physical footprint is smaller than inert gas systems as no agent is required to occupy a whole space. Electrically non-conductive, non-corrosive, sterile agent gases do not leave any traces upon evaporation. It renders them the best fire prevention contractor in data centres. As with fire sprinklers, the space is equipped with a tube network [29].

Third level of protection is rack-level fire protection. Relevant appliances and loss control need this fire protection. Although the mandatory fire sprinters protect the building and the room from fire, the appliance is not unprotected, which is worth 57 per cent of the cost in the room. To order to conserve money, the hardware has to be secured from a fire on a shelf. The implementation of a preconceived automated fire deletion mechanism safeguards the unit with the identification and deletion of the fire within seconds until it is triggered by the complete flute or sprinkler machine. It avoids the disruption to the facilities done by a water-based sprinkler and allows huge amounts of agents to be released into an expensive overall flood container [29].

To ensure the protection of the data stored in the data centre all three levels of protection should be taken. This will help the Meza team to protect the data centre of the food ordering application system. And reduce the cost of damage.

3.2 Types of protection

The data centre for food ordering applications can be protected from fire in a variety of ways. The first is the water-powered sprinkler device that manages the flames, stops them from spreading and avoids structural harm. The sprinkler device is considered as an inexpensive and simple approach utilising about 25 gallons per minute of water. This leads to certain risks, which may be greater than the fire loss if electrical conductivity of data centre appliances, is because the spot where the fire takes place would tremendously wash, which will take the business more.

The Water Neck System, a recent entrant to the water-based fire protection program, is another water-based method. The spreading of the water in the specified area needs strong pressure pumps, which is advised only for wide areas, as it resulted in poor fire output in order to prevent flooding the area entirely, particularly in the fire is blocked. In fact, after suppression, the mist systems leave residual vapour, and costs and a problem exist with the equipment.

The clean agent, which can extinguish the fire very easily and connect the fire damage to the data centre equipment in its location which does not need water, is also a way of protecting the data centre from fire accidents. The main purpose of this sort is to protect the resources that are important, dynamic and essential. It is distinct from other forms as it does not involve extinction washing so no corrosive or residue is left behind. In complex environments, it can extinguish fires in blocked or three-dimensional areas. In addition, two forms of cleaning agents are usable. Firstly, halocarbon agents including carbon, hydrogen and halogens including fluorine, thus creating dangerous effects for those in the fire because fuel is being polluted [30] and causes a breathing problem. Secondly, inert gas agents are prepared for gasses such as nitrogen, argon and carbon dioxide which are less risky to humans and less damaging to the resources than the first type. Both agents are regarded as electrically non-conductive and can be used in typically covered areas.

3.3 Recommendation

Kidde, fire system, a global pioneer in the development and manufacturing of fire detection and safety systems between clean agents-inert gas type and water sprinkler, was the focus of the previous debate and several real-world experiments, such as a study. As a Meza team, it suggests the clean agent-inert gas network that is installed for the food ordering application data centre, as well as the different benefits and functionality it provides, so that a fire will easily be extinguished and damage in a certain region limited, as well as no cleaning after the fire takes place. Therefore, the expense is inexpensive to introduce and eventually it can be built and managed very effectively and does not affect the safety or atmosphere of people. Therefore, beyond the core of the data centre, we do need to use the sprinkler network for assisting in certain circumstances by positioning several halocarbon agents.

3.4 Network infrastructure

See (Figure 9).

Figure 9.

Network infrastructure 1 [31].

3.4.1 Cabling

Despite too much emphasis on various forms of technology and how organisations really bring their networks to use, it’s simple to think about the physical framework that allows every data centre networking system feasible. Cabling is a massively critical element of data centre architecture. Weak cable implementation is not just unclear it can obstruct airflow, hinder the correct removal of hot air and stop cold air from entering. Cable air damming over time may cause overheating and failure of equipment, resulting in costly downtime. Data centre cabling was usually mounted beneath a high level. Designs have, though, improved in recent years to at least some flexibility for overhead cabling that also helps to lower electricity costs and minimise refrigeration needs. In order to maintain continuity of output and easy usage, well-managed facilities using organised cables procedure. Unstructured point-to-point cables cannot be mounted at all but are also correlated with higher operational expenses and severe maintenance issues. A successful first move to networking for the food ordering application is the proper cable control [32].

3.4.2 Connectivity

The abundance of ISP networking solutions is one of the key benefits of a carrier-neutral data centre. Basically, a data centre, like any other person, links to the internet: through a different service provider cable. Moreover, in comparison to a traditional house, data centres provide many links with different vendors, enabling the food ordering application a variety of choices. The various networking solutions often provide a lot of flexibility, and it is nearly always easy to reach the external internet. Blended networking solutions often have major protections against DDoS assaults (Figure 10) [32].

Figure 10.

Network infrastructure 2 [33].

3.4.3 Routers and switches

Cabling from data centres is difficult enough even without routers and improvements to guide network traffic flow through and across the facility will approach nightmarish rates of difficulty. These devices act as unified nodes that enable data to move as easily as possible from one location to another. Properly installed, they can handle vast volumes of traffic and form a vital part of the topology of the data centre without losing efficiency. Incoming public internet data packets first reach the edge-routers of the data centre, which evaluate from where any packet arrives and where it has to go. It transmits the packets to the core routers that form an additional layer at the deployment stage. Such appliances are more aptly defined as switches, as they handle traffic in the data centre networking infrastructure. The grouping stage is named a set of key switches as all traffic is guided inside the datacentres. When it requires data to move through not physically linked computers, it must be transferred through the main switches. Having a wide number of addresses for the core to handle and sacrificing pace becomes necessary for individual servers to contact each other, and the data centre networks prevent the issue by linking server batches to a third layer of switches. Often these classes are called pods which encrypt data packets to allow the core to recognise what which traffic should be guided to rather than handling individual server requests for the food ordering application [32].

3.4.4 Servers

Deployments with high-density servers appear to have higher cabling, cooling and power usage specifications. The food ordering application wants their equipment in racks with convenient access to direct connections and individual cross-connections that provide better efficiency, pace and reduced downtime effect [32].

3.4.5 Direct connection

Perhaps the internet of the data centre is not quick enough to meet the demands of a client. It cannot require the lag or downtime of a cloud service [34] provider link. Data centres may in these situations give them the benefit of directly linking the server with one cross-connection to the servers of the provider. Through the direct cable running between servers, customers may work better when reducing latency and downtime. Although data centre networks are complicated structures which must be carefully maintained to guarantee high-quality efficiency, in any facility the basic framework is focused on exactly the same principles. By improving these networks, data centres will also be an enticing venue for businesses wanting to position their IT systems with a third-party supplier while providing a selection of creative offerings to their clients [32].

4. Cooling system

In most data centres, the server room’s temperature is always lower than in other places in the same building. For every single data centre, when the data collected on servers are growing, and a continuous increase in processor performance, it will lead to the servers will produce more heat. When the server’s temperature reaches a critical point, it might cause the server not to be able to work properly, and the processor will slow down the performance to avoid overheating. When coming to the worst-case scenario, extremely high temperatures will burn the processor, eventually cause services interruption, and require replacing the hardware for resuming the services. For many reasons showing why a data centre needs a cooling system, and the cooling system is one of the essential components in the data centre. Hence, selecting a cooling system that can operate continuously, and the reliability is the top priority.

The cooling system operates in a variety of ways and has different levels of performance. Generally, there are two main methods for data centre cooling, which are an air-based cooling system and a liquid-based cooling system [35]. In liquid-based cooling system is a reduction of heat from the server by exploiting the properties of liquids [36]. For the air-based cooling system, commonly there are three types, transitional cooling, hot aisle containment, and cold aisle containment. These cooling systems cool down the server room temperature via cold air.

According to the research, the air-based cooling system method has commonly been used for years, and it is simple to compare to the liquid-based cooling system. For a liquid-based cooling system has the risk of leakage across the server rooms, it could damage the components if not appropriately used [37]. Therefore, in this project, we will focus on the air-based cooling system.

4.1 Hot aisle containment system (HACS)

Hot aisle containment system (HACS) composes the hot aisle to collect the hot exhaust air from the server at the back of the racks, and cold supply air is brought in by the equipment at the front. This is similar to a traditional hot aisle/cold aisle arrangement where data centre racks are arranged in alternate rows, cold air intakes facing one way and hot air exhausts facing the other, it keeps hot air in one aisle and cold in the other. While the traditional hot aisle/cold aisle cooling system, which only uses rack placement, worked well in a low-density environment but did not completely isolate the aisles and prevent hot and cold air from mixing [38]. The only method to stop the hot and cold air from mixing up is to form a physical barrier. This is where the containment systems are coming in. Typically, the containment forms a physical barrier from the top of the server racks to the drop ceiling, and it contains the hot aisle, and the exhausted hot air directly back to the cooling units [39]. The cold air will come from the CRAC unit, and the environment outside of the hot aisle becomes a large cold air plenum [40]. This is ensuring the hot and cold air are isolated (Figure 11).

Figure 11.

Hot aisle containment system [35].

Pros of HACS

• Overall the server room is in a cold environment.

• Any leakage of conditioned air will not affect it, as it goes into the cold aisle.

• More effective in cooling.

• Able to use standard fire detection system without having any obstruct.

Cons of HACS

• The cost of the construct is higher.

• It needs a contained path to return the exhaust air to the cooling unit.

• Hot aisle’s temperature easily becomes higher [41].

4.2 Dynamic cooling management and optimisation

Cooling management and optimises system continuously optimise airflow in the data centre equipment, performing improved reliability and availability. The system uses a dense array of temperature sensors to discover precisely where the hot spot is in the data centre. It helps identify potential equipment risks and automatically eliminates up to 95% of hot spots. As the IT load changes, integrated machine learning automatically adapts cooling to varying IT loads to balance the dynamic data centre environment. The system adjusts the cooling need with the lowest possible energy consumption, achieving the immediate cost savings and a suitable amount of cooling in the data centre (Figure 12) [42].

Figure 12.

Dynamic cooling control [42].

4.3 Justification

In order to achieve a tier 4 data centre, all components in the data centre must be fully redundant, including the cooling system. A cooling system design for a tier 4-rated data centre should fulfil the requirements below:

Redundant components—a backup of equipment for a cooling system such as:

• CRAC Units

• Chillers

• Fans

After comparing the air-based cooling system, the hot aisle containment system is the chosen system implement for this project. According to Gavin Banks, HACS is significantly more cost-saving, compare to cold aisle containment system (CASC), it can save 43% in energy cost, which could relate to a 15% reduction in PUE [43]. This is also supported by Schneider Electric, in the report showing HACS provides 40% more energy cost-saving annually compare to CACS [40]. The legacy/traditional cooling system could be more expensive due to inefficiency and uneven cooling, and it may also require more and oversized equipment to accomplish the task. When looking at the bigger picture and all the considerations, other IT equipment in the same room that would need cooling as well, cold aisle containment system might make the server room area extremely hot. It will become a challenge for those who require to be inside the server room for maintenance or servicing, as well as for other IT equipment work under high temperatures. Therefore, hot aisle containment would be the appropriate choice (Figure 13).

Figure 13.

Hot aisle containment [39].

4.4 Space allocation

A proposed layout is as follows (see Figures 14 and 15).

Figure 14.

Proposed layout [44].

Figure 15.

Tier 4 N + N cooling system [45].

4.5 Physical security

A breach of physical security may cause unimaginable damage to a data centre. Given the growing need to protect valuable information, any loss of data or even the incapability to comply with mandatory regulatory requirements may result in obloquy, loss of customers, fines and loss of revenue. Interoperability is a critical building block for the physical security of a data centre. The entire ecosystem of manufacturers and integrators serving the data centre physical security market needs to ensure that the products work together to provide a scalable, layered physical security solution [46].

The prime purpose of implementing physical security is to protect the information, devices and IT infrastructure of the data centre from any threat that could disrupt the operation of the data centre. It could be caused by any illegal activity, such as theft, leakage of data or damage by any physical involvement in the data centre. Building a layered approach to data centre security helps to customise the solution to the needs of a data centre. The organisation needs to determine the right layered approach, and understand the current system, the working environment and future needs (Figure 16).

Figure 16.

Security map showing the depth of security [47].

A practical, layered approach requires all systems to function coherently. Generally, the security architecture consists of multiple layers of physical security that need to be considered to protect the data centre as a whole and to comply with the data centre protection guidelines (Figure 17).

Figure 17.

Layers of physical security in data Centre [48].

4.5.1 Layer 1: perimeter defence

The first layer of physical security is perimeter defences, a physical boundary or fence at the property edge to deter external threats, which is controlling and restricts the access to the data centre property. There are three D’s to describe the purpose of perimeter security, Deter, Detect, and Delay [49]. Usually, there are only two doors that are allowed to enter the data centre, the front door and the loading bay. The perimeter fence detection system can integrate with trespassing alarms, high-definition CCTV system, limited access control points, and motion-activated security lighting.

4.5.2 Layer 2: clear zone

The second layer called as Clear Zone creates a buffer zone between the perimeter, and the data centre to have better detect physical invasion [48]. The clear zone is also a large area containing critical infrastructures such as fuel containment, generators, and main power supplies [46]. This zone needs security measures that provide a total awareness of the situation.

4.5.3 Layer 3: facility entrance and reception

Control visitor’s access to the data centre and validate authorised access. All the employees and visitors before entering the data centre must check-in or register at the front desk, visitors, must obtain a temporary pass in order to access the secured areas.

4.5.4 Layer 4: services Corridor (Escorted areas and Grey space)

Validate the rights of authorised persons to access specific areas within the building. The corridors, grey spaces and escorted areas that head to the data centre floor are often where the proper security measures are overlooked [46]. This could lead to unauthorised access to critical mechanical and electrical infrastructure.

4.5.5 Layer 5: data centre room

In this layer, it is further restricting access through various forms of authentication, and monitoring all authorised access. Implement high-security electronics to prevent general staff or trespassers from accessing sensitive areas. To prevent unauthorised persons from entering white space, access control such as dual-factor biometrics is essential for the control of authorised access to the data centre.

4.5.6 Layer 6: date centre cabinet

Establish the protection of sensitive electronics (servers) that contain crucial data. The security measures to accomplish include cabinet locking, audit trails and an intelligent infrastructure strategy. This layer is especially essential and effective in reducing the critical and frequently forgotten the insider threat.

Most data centres did an excellent job of accomplishing the first few layers. Still, the absence of reliable control of the cabinet may result in costly data breaches caused by a malicious or disgruntled employee, or maybe even unknowing and unintentional access to data.

5. Synthesis

Power: The data centre for the food ordering application will be using the grid electricity as the main power source to power up all the equipment and components in the data centre. In case of a power break or power down all the equipment’s will be powered with the UPS backup battery for a short time and by that time the diesel generator will replace the power source with it. For every data centre uptime for its servers are highly crucial since the clients should have uninterrupted services. And 99.9% uptime will ensure uninterrupted services for all the customers and keep the data secure and ongoing.

5.1 Server racks and computing resources

The server racks will consist of all the necessary components for the servers to compute the desired function. Which will process placing an order, processing the payment and creating data for the customer. Which should be done within no time and should keep things fast are reliable. The servers are responsible for doing all the backend processes of the application and should be running all the time without any issues. All of the components should be working properly. And should be connected with each other to compute the tasks.

5.2 Storage infrastructure

Every customer detail and data should be kept in a storage device. The storage system consists of many storage components which will be connected to the servers all the time and store the necessary data in the storage. They will compute the data to the server interaction with it. Hard disk drive, tape drive and other forms of internal and external storage devices make process and stores the computed data. There will be a backup for all the data in case of storage devices get interrupted and to keep it working depending on necessary times. The storage utility software keeps monitoring [50] the process for uninterrupted processes.

6. Evaluation

6.1 PUE and efficiency

Data centres use significant amounts of power to operate. Majority of the power is consumed by the cooling systems. A successful data centre must be efficient. One of the metrics for calculating the efficiency of the data centre is Power Usage Effectiveness (PUE). PUE is calculated by using the total amount of power consumed by the energy used by the IT equipment [51, 52].

In order to calculate the PUE, values for power consumption and IT energy needs are to be determined. The Table 1 shows the estimated power usage for different components and equipment of the data centre based on enterprise IT equipment.

Components	Qty	Power (W)
IT
Compute Rack	8	8 × 5250 W = 42,000 W
Storage Rack	2	2 × 5250 W = 10,500 W
Switch (Fibre Channel and Ethernet)	10	10 × 300 W = 300 W
Router	4	4 × 500 W = 2000 W
Computer	2	2 × 300 W = 600 W
Printer	1	1 × 250 W = 250 W
CCTV Camera	14	14 × 10 W = 140 W
Total		55,790 W
Non-IT
Cooling	1	1 × 40000 = 40,000 W
Lighting	20	20 × 50W = 100 W
Smoke Detectors/Alarms	14	14 × 2W = 28 W
Misc.	1	1 × 500W = 500 W
Total		40,628 W
Total Power for Facility		55,790 + 40,628 = 96,418 W

Table 1.

Estimated power usage for different components and equipment of the data centre based on enterprise IT equipment.

PUE = Total Power Consumption/IT Energy Needs.

Based on the estimates above, the PUE for the data centre is calculated by Eq. (1),

$$\mathbf{PUE}=\mathrm{96,418}/\mathrm{55,790}=\mathbf{1.73}\mathbf{.}$$E1

The PUE value lies between efficient and average according to Figure 18. One of the main reasons why the efficiency is slightly below efficiency is that cooling system requires more energy due to the geographical location of Malaysia. In other countries such as Iceland, the cooling system will not need that much power because they are also closer to the poles and they experience the winter season which Malaysia does not. However, the efficiency is still optimal for a data centre in this region.

Figure 18.

Level of efficiency relative to the PUE value [37].

6.2 Expandability

A data centre is expected to meet future business needs and expand accordingly. The food company already has 5 million users, in the next few years, they estimate their user base will increase given the current popularity. Therefore, the data centre design must be able to take this into consideration and allow future expansions. According to [53], many data centre expansions result in failure. As for expandability in this data centre, it will initially occupy 10 racks which are about 25% of the floor space available to avoid the mistake of oversizing and wasting resources. This allows the facility to not be overcrowded when demands rise. Starting with few racks lowers the cost to build (CapEx). In addition to this, the usage of rack enclosures/cabinets helps with cooling. Therefore, in the future cooling systems will not be overburdened with additional racks in the facility since rack enclosures have better airflow and cooling. The data centre is designed with a modular approach. Designs that are modular and flexible are the key to long-term success [53]. For example, increasing the storage capacity is trivial since the data centre storage infrastructure is based on Storage Area Network (SAN) which offers great scalability and expandability compared to other architectures. Finally, through proper planning using the total cost of ownership (TCO) approach and flexibility of the facility, the data centre can meet the requirements of recent market demand.

7. Conclusion

We may sum up by noting that because the IT industry is continually expanding, there will always be a need for new and improved solutions. There is no doubt that the solution and tools selected for this work will remain the same in the future. The intended data centre is meticulously thought out, from security to smart execution. Considerations for the design of a data centre in terms of power efficiency, cooling systems and protection should include scalability, power effectiveness, CO2 reduction, system resilience, sustainability, the use of machine learning, and other cutting-edge technology. Additionally, by renting space in the data centre and selecting their own equipment, clients using the co-location system can locate their data. Finally, through proper planning using the total cost of ownership approach and flexibility of the facility, the data centre can meet the requirements of today and tomorrow.

Conflict of interest

The authors declare no conflict of interest.