Data hall improvements

Abstract

A method of designing a data centre layout with improved energy efficiency comprising the steps of: A) creating plural model physical layouts for the data hall each representing many N server cabine

Description

Field of the Invention The field of the invention relates to hyperscale data centres. The invention is applicable to, but not limited to data centres with improved energy efficiency. Background of the Invention Data centres are facilities housing servers, processors, and other electronic equipment, along with backup power systems like UPS. They are essential for supporting internet services and other infrastructure demands. The servers are typically maintained in cabinets, which are known as racks. These racks are organized into parallel rows. The arrangement of these rows, facing front-to-front or back-to-back, forms aisles that serve as pathways for cooled air and provide space for operational tasks. Typical cooling systems provided are based around air cooled chillers feeding chilled water to cooling equipment, such as a Computer room air handler, CRAH, Computer room air conditioning, CRAC and Fan wall units, FWU located within the data halls. Server halls within data centres are configured to align with end-user requirements and are rigorously monitored to adhere to precise server inlet temperature guidelines. These guidelines, defined in tenant-specific and typically confidential Service Level Agreements (SLAs), necessitate that data centres maintain a buffer between operational temperatures and the SLA limits. One of the more popular cooling arrangements in data centres to minimise recirculation problems is the so-called hot aisle containment. These supply cold air and contain the return warm air from the servers and are divided into two regions as shown in the Figure 1.Figure 1. Illustration of Hot and Cold Aisle Containment (John N, et. al., 2011) As illustrated in Figure 1, the hot aisle zone 104 includes server racks 130, 132, 134, 136, control system area 118, entrance 122, tape library 116, storage 14, Similarly, the cold aisle zone 102 includes server racks 106, 108, 110, 112. The hot aisle containment zone 104, separates the back of its respective server cabinet rows, while the cold aisle containment zone 102 separates the row of racks in front. As such, cold air will be drawn in front of the server racks 106, 108, 110, 112 from the cold aisle containment zone 102 and hot air is exhausted from the back of these server racks. They are positioned with the server cabinet front facing the cold aisle and the back facing the hot aisle. The Data Hall is a climate-controlled area housing heat-generating IT and telecommunications equipment. Due to the equipment's sensitivity to temperature and humidity, strict power and cooling conditions are essential for ensuring its integrity. This space is often referred to as "whitespace" by manufacturers. In this study we have considered traditional CRAG unit will supply cold air from the bottom and takes hot air return from top of the unit. Data Hall optimisation is important in the high-density modern data centre design. Data centres play a vital role in today's digital world, providing support for numerous services and applications. With the growing need for more computing power, these facilities are adapting to include high-density IT racks, which require effective power management and cooling systems. One key aspect to consider is how to set up UPS to back all critical cooling units with the data hall. It is known in the field that UPS systems are typically provided in highly resilient data centre facilities to ensure continuous cooling is maintained during transient and failure events. As per the white paper by Schneider, detailed literature work has been carried out on whether all mechanical cooling equipment requires critical power back up from a dedicated Mechanical UPS system (MUPS), if this back up can be omitted or combined with the critical IT UPS system. Although there are several research can be conducted in the field to explain how these cooling systems can be backed by UPS. But there is not clear publicly available research material act as guide to future Data centre designer or operators (Schneider Electric 2014, data centre temperature rise during a cooling system outage [Schneider Electric, 2014]). While general guidelines and common layouts exist, data hall configurations are fundamentally driven by client-specific needs. Factors such as power density, cooling requirements, and future connections, a bespoke approach can be needed to ensure optimal performance and efficiency. Data hall air limits are based off ASHRAE TC9.9 A2 Class, this is slightly less stringent than A1, however still suitable for most modern data centre equipment. Figure 2 is an illustration of the 2 ASHRAE 2015 Thermal Guidelines (ASHRAE, 2015) Summary of the Invention According to an embodiment there is provided a computer implemented method for determining a configuration of a data hall in a data centre with improved energy efficiency comprising the steps of: creating a plurality of model domains for the data hall; wherein each of the plurality of model domains is provided with: a representation of a plurality, N, of server cabinets; a representation of a plurality, M, of cooling units for the data hall, where M<N; a representation of a plurality, P, of backup power supply units for the M cooling units, where P=M / 2; using computational fluid dynamics CFD, equations to model a temperature in the data hall at each of the server cabinets under a standard operating conditions of the cooling units and the server cabinets for each of the plurality of model domains; modelling the temperature in the data hall at each of the server cabinets for a first time period when power is no longer provided to one or more of the M cooling units, for each of the plurality of domains; modelling the temperature in the data hall in each of the server racks in the data hall for a second time period when one or more of the M / 2 cooling units are provided with power from the P backup power supply units, for each of the plurality of domains; modelling the temperature in the data hall at the N server cabinets when power is restored to the M cooling units, and the M / 2 backup power supplies are turned off, for each of the plurality of domains; using results of the temperature modelling from all of the above stages to determine overall energy usage in each of the data halls in the plurality of model domains, and using the temperature modelling results and the overall energy usage to provide an output of the configuration of the server cabinets, cooling units and backup power supply units in the data hall with optimum energy efficiency. Preferably, the temperature modelling at all different stages is modelling using computational fluid dynamics. Further preferably, the CFD equations to model temperature are used to provide information about airflow within the data hall. In a preferred embodiment, power is no longer provided to the M cooling units, for each of the plurality of domains, and the M / 2 cooling units are provided with power from the P back up power supply units, for each of the plurality of domains. Preferably, the first time period when the power is not provided to the one or more M cooling units for a period between 0-60 seconds. Further preferably, the first time period when power is not provided to the one or more M cooling units for a period between 0-30 seconds. In an example, when power is resumed, the power is provided to M / 2 units for a period between 30-150 seconds Further preferably, standard operating conditions of the cooling units are when the cooling unit provides a maximum temperature in the data hall of 35°C. Preferably, the cooling unit provides a temperature in a range 10-35°C. In a preferred embodiment, the plurality of model domains is divided into a plurality of mesh cells, and the CFD equations for the temperature modelling stage are determined for each of the mesh cells. Preferably, the cooling units comprise one or more of: Computer Room Air Conditioning, CRAC; Computer Room Air Handler, CRAH; and Fan Wall Units, FWU. Preferably, each of the plurality of model domains includes a representation of a floor configuration in each of the data halls. Further preferably, the floor configuration includes one or more raised sections of floor in each of the model domains of the data hall. In an embodiment, in the plurality of model domains of the data hall a representation of one or more colling units is positioned in a floor void under a raised section of the floor. These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. Brief Description of The Drawings Further details, aspects and embodiments of the invention will be described, by way of example only, with reference to the drawings. In the drawings, like reference numbers are used to identify like or functionally similar elements. Elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale. Figure. 1 Illustration of Hot and Cold Aisle Containment (John N, et. al., 2011) Figure 2 ASHRAE 2015 Thermal Guidelines (ASHRAE, 2015) Figure 3(a): An example methodology according to an embodiment of the invention; Figure 3(b) and 3(c) are an example flowchart according to an embodiment of the method: Figure 4: illustrates an example of a Scenario Data Hall Floor plan according to an embodiment of the invention Figure 5: illustrates an example of a raised Floor Section according to an embodiment of the invention; Figure 6: Scenario 01 Server Row (SR) Temperature results in failure condition Figure 7: Scenario 02 Data Hall Floor plan Figure 8: Scenario 02 Server Row (SR) Temperature results in failure condition Detailed Description In general Data Halls need to be resilient to Power outage conditions, failure conditions typically arise when cooling system malfunctions lead to excessive heat buildup, potentially causing equipment shutdowns or reduced capacity. One critical failure scenario is a power failure affecting data hall cooling units. If cooling units are not backed up by UPS, their fan speeds will immediately begin to drop, reducing airflow and diminishing cooling efficiency. During a power failure, if backup generators do not start quickly, the situation worsens as cooling performance declines further. Even when power is restored, CRAG fans require time to ramp back up to full speed. This delay in airflow recovery can lead to hotspots forming in the data hall, particularly in high-density areas where heat loads are significant. Without immediate cooling intervention, prolonged exposure to elevated temperatures can trigger IT equipment shutdowns as they exceed safe operating thresholds. This project aimed to improve data hall UPS control strategy to maintain a healthy operational conditions in power failure conditions. In this study, the UPS backed cooling systems for CRAG are studied and understood 50% UPS backed cooling units sufficient in comparison with traditional 100% UPS backed. Hence, this approach expects to revolutionise data hall design by making it operation in 50% UPS backed control strategy. This will reduce the number UPS units required though reduce the energy consumption needed for over all Data centre site. In an embodiment, a computer implemented method for determining a configuration of a data hall in a data centre with improved energy efficiency is described. The comprising the steps of: creating a plurality of model domains for the data hall; wherein each of the plurality of model domains is provided with: a representation of a plurality, N, of server cabinets; a representation of a plurality, M, of cooling units for the data hall, where M<N; a representation of a plurality, P, of backup power supply units for the M cooling units, where P=M / 2; using computational fluid dynamics CFD, equations to model the temperature in the data hall at each of the server cabinets under a standard operating conditions of the cooling units and the server cabinets for each of the plurality of domains; modelling the temperature in the data hall at each of the server racks for a first time period when power is no longer provided to one or more of the M cooling units, for each of the plurality of domains; modelling the temperature in the data hall in each of the server racks in the data hall for a second time period when one or more of the M / 2 cooling units are provided with power from the P back up power supply units, for each of the plurality of domains; modelling the temperature in the data hall at the N server cabinets when power is restored to the M cooling units, and the M / 2 backup power supplies are turned off, for each of the plurality of domains; using the result of the temperature modelling from all of the above stages to determine overall energy usage in each of the data halls in the plurality of model domains, and using the temperature modelling results and the energy usage to provide an output of the configuration of the server cabinets, cooling units and backup power supply units in the data hall with the optimum energy efficiency. The technological advancement sought was the improvement of knowledge and capabilities towards data hall configurations through the development of a solution for mechanical UPS (MUPS) systems to be removed or altered. In traditional data centre server rooms (Data Hall), UPS systems are available for both IT servers and Server cooling units such as FWU, CRAH and CRACs. However, have these two separate power back up units can cause space constraints and in efficient way in data centre design. The critical cooling equipment that was supported by this UPS system can be moved onto a combined IT server and mechanical system allowing for a more effective solution with spatial savings. Specifically, the removal of the two separate systems reduced the quantity of physical equipment required and can also reduce switch room sizes as the required footprint is smaller. This innovative approach looks to provide UPS backed solution to the building cooling system, circulation pump, servers and data hall cooling system fans to continue to run the equipment during a utility power failure scenario to maintain temperatures in the interim period when the building generator systems are energised. Although this is very innovative approach towards cost and space saving, lack of awareness in this design strategy still makes it hard for DC sector. Additionally, if the cooling system lacks redundancy or if airflow distribution is poor, certain zones within the data hall may experience down time in server operation. In extreme cases, this can result in permanent hardware damage. Industry best practices, such as ASH RAE guidelines, recommend ensuring continuous cooling and sufficient UPS-backed cooling operation to prevent critical failures during power transition events. Keeping above things in agenda, study conducted with the 0% cooling units backed UPS against the 50% UPS backed. To assess the impact of UPS-backed cooling, a study was conducted comparing two different example scenarios in data halls with server cabinets and an arrangement of cooling units: Scenario 1: A data hall with no UPS-Backed Cooling (0%) 0 Sec: Power failure occurs in the data hall. 10 Secs: 100% of cooling units lose power, and fans begin ramping down. 23 Secs: Backup generators come online. 24 Secs: Cooling units regain power and start operating. 46 Secs: Cooling units reach 100% operational capacity. Scenario 2: A data hall with partial UPS-Backed Cooling (50%) 0 Sec: Power failure occurs in the data hall. 0 Sec: 50% of cooling units lose power, while the remaining 50% continue operating on UPS. 10 Secs: Fans in the non-UPS-backed cooling units ramp down. 23 Secs: Backup generators come online. 24 Secs: Remaining 50% cooling units regain power and start operating. 54 Secs: Cooling units reach 100% operational capacity. In each of the two scenarios, a data hall model is constructed by creating a plurality of model domains for the data hall; wherein each of the plurality of model domains is provided with: a representation of a plurality, N, of server cabinets; a representation of a plurality, M, of cooling units for the data hall, where M<N; a representation of a plurality, P, of backup power supply units for the M cooling units, where P=M / 2. In an embodiment, power is no longer provided to the M cooling units, for each of the plurality of domains, and the M / 2 cooling units are provided with power from the P back up power supply units, for each of the plurality of domains. Of course, other scenarios are also possible, and could be modelled according to the methodology described below. Computation Fluid Dynamics Computational fluid dynamics (CFD) simulations are a one of the essential tools in conducting due diligence and validation studies for data centres in lieu of experimental data. Preferably, temperature modelling at all different stages of possible scenarios according to an embodiment is modelled using computational fluid dynamics. Further preferably, the CFD equations to model temperature are used to provide information about the airflow within the data hall. The core of CFD is based on the Navier-Stokes equations, which describe the motion of fluid substances. These equations account for various factors such as velocity, pressure, density, and temperature of the fluid. To solve the governing equations numerically, the fluid domain is divided into a finite number of small control volumes or elements. This process is known as discretization. Common methods include the Finite Volume Method (FVM). CFD solvers use numerical methods to solve the discretized equations iteratively. In this context a mathematical model or representation of a part of data centre such as a data hall can be simulated allowing accurate performance predictions. The governing equation of the velocity, temperature and pressure can be described below as per (Versteeg &Malalasekera, 1995): Continuity equation: (pu) = 0 ; if incompressible V ■ (u) = 0 In terms three dimension the momentum equation for the three components x, y, and z are shown below: x-momentum u p (Du / Dt) d(pu) / dt + 7 ■ (pu u) y-momentum V p (Dv / Dt) d(pv) / dt + 7■(pv u) z-momentum w p (Dw / Df) dtpw^ / dt + 7 ■ (pwu) energy E p (DE / Dt) d(pE} / dt + 7 ■ (pFu) The energy equation which characterises temperature change as one components: of its ar , . p cp — + r • {pcp t u) = -p r • u + k r • (r r) + q Where: p is the fluid density, Cp is the specific heat capacity at constant pressure, T is the temperature, t is time, u is the velocity vector, p is the pressure, k is the thermal conductivity, Q is the heat source term, 7 • represents the divergence operator, 7 represents the gradient operator. The CFD analysis according to an embodiment consists of 3 different stages such as Pre-Processing, Processing and Post-Processing. The pre-processing stage involves making a domain for a proposed design of the data hall within the data centre and meshing the domain. In the meshing process, the domain will be divided into number of mesh cells and the discretised governing equations will solve the solution on each nodal points of the cells divided. The accuracy of the solution depends on the number of cells in the total domain. More cells mean much better accuracy, and vice versa. The processing stage is where the 6Sigma CFD Tool will solve the discretised governing equations in each nodal points of the cell (finite control volumes). The solution method will be solved in more detail in an upcoming section. In the post processing stage, the processed solution, which were presented in results section. In this current research following a methodology flow is used to evaluate the thermal behaviour inside the model of the data hall in the data centre. Evaluating the max accepted temperature uplift at the server cabinet level were evaluated in this study along with the simulating time dependant power failure scenario. Transient CFD (Time dependent) analysis captures the fluid flow over time allowing time varying conditions to be accounted for. This was employed to simulate mains power failure scenario in the data hall, evaluating if temperatures exceed maximum permissible temperature (typically 35°C) before backup generator power is established in the data hall. The simulations showed that temperature breaches during the period and necessitated implementing UPS systems for critical IT equipment to prevent thermal-related failures and ensure continued operation. However as in this proposed study, combination of IT backed UPS and Meeh back UPS in one has examined with the 50% of Meeh systems will be backed by UPS. This clever approach has reduced the need for all mechanical systems needs to be in UPS backed control. This reduction would apply to every Primary Critical System (PCS) supporting a data centre. For example, if a data centre has 50 PCSs, this change would eliminate 50% units need. In achieving the advance, additional savings were made through less equipment being required to be procured, installed, commissioned, and maintained. Technological uncertainty lay in how to develop a solution for a data hall environment that is a highly complex space with various parameters and moving parts that impact one another. One of the main factors to consider in this complex design is Temperature for server inlet. Temperature for server will be evaluated in all scenarios conducted and compared against. Methodology for Research Conducted Figure 3(a) is a flowchart for the: Methodology for Research Conducted. The above flowchart represents a structured methodology 300 for conducted research according to an embodiment. Start, 302: Marking the initiation of a systematic investigation into a control strategy for CRAG inside data hall is addressed with design suggestions. Literature, 304: The research reviewed the papers on current cooling techniques, temperature, air conditioning methods, energy optimisation along with the UPS operational methods in data hall. The need for further development on UPS control strategy is also being addressed. Involves reviewing existing research and publications to understand current trends, methodologies, and gaps in the field. Finding Gap in the Field, 306: Based on detailed literature review key gap is identified, as although there are many publicly available sources on UPS and their control strategy in Data Centre operations, yet there is not any clear information available publicly on how to control these UPS can be controlled effectively and with limited space constrains. At step 306, Setting Up the Number of Scenarios Required: determines the different test cases or scenarios needed to analyse varying conditions and ensure a comprehensive evaluation. This includes the 0% UPS backed cooling units in server hall only and 50%.Model Built-Up 310: Once the research gap is identified, the next step is model built up, which serves as an initial framework for further refinements and scenario testing in CFD. This preliminary model helps establish benchmarks for comparison on ensure working principles of data hall in power failure operations. Setting Up Boundary Conditions, 310: involves defining constraints and parameters to ensure the model operates within realistic and applicable limits. Proceeding with CFD simulations, 314. Convergence, 316: A critical step in the methodology, where the model is checked for stability and consistency in producing reliable results. If the model does not converge, adjustments are made, and the process loops back to refining the baseline model 326. Once convergence is achieved, the next step is Checking Results Against Project Requirements 318, where the outcomes are evaluated to determine if they meet the desired objectives. If the results fail to meet the project criteria 320, the process moves to changing the boundary condition to identify suitable Scenario, where a different case is analysed until a satisfactory outcome is found . If the results meet the project requirements 322, the data is Saved and Post Processed for further analysis. Finally, the comparison of results 324to find the best solution is conducted by analysing the data from multiple tested scenarios and selecting the most effective and optimised outcome. This structured methodology ensures a systematic and iterative approach to research or computational modelling, facilitating thorough testing, refinement, and the selection of the best possible solution. Figures 3(b) and 3(c) are flowchart of an example process 350. In some implementations, one or more process blocks of Fig. 3(b) and (c) may be performed by a device . As shown in Fig. 3(b), process 350 may include creating a plurality of model domains for the data hall (block 352). For example, device may create a plurality of model domains for the data hall, as described above. As also shown in Fig. 3(b), process 350 may include where each of the plurality of model domains is provided with: a representation of a plurality, N, of server cabinets; a representation of a plurality, M, of cooling units for the data hall, where M<N; a representation of a plurality, P, of backup power supply units for the M cooling units, where P=M / 2 (block 354; using computational fluid dynamics CFD, equations to model the temperature in the data hall at each of the server cabinets under a standard operating conditions of the cooling units and the server cabinets for each of the plurality of domains (block 356; modelling the temperature in the data hall at each of the server racks for a first time period when power is no longer provided to one or more of the M cooling units, for each of the plurality of domains (block 358); modelling the temperature in the data hall in each of the server racks in the data hall for a second time period when one or more of the M / 2 cooling units are provided with power from the P back up power supply units, for each of the plurality of domains (block 360); modelling the temperature in the data hall at the N server cabinets when power is restored to the M cooling units, and the M / 2 backup power supplies are turned off, for each of the plurality of domains (block 362). For example, device may where each of the plurality of model domains is provided with: a representation of a plurality, n, of server cabinets; a representation of a plurality, m, of cool units for the data hall, where m<n; a representation of a plurality, p, of backup power supply units for the m cooling units, where p=m / 2; using computational fluid dynamics cfd, equations to model the temperature in the data hall at each of the server cabinets under a standard operating conditions of the cooling units and the server cabinets for each of the plurality of domains; modelling the temperature in the data hall at each of the server racks for a first time period when power is no longer provided to one or more of the m cooling units, for each of the plurality of domains; modelling the temperature in the data hall in each of the server racks in the data hall for a second time period when one or more of the m / 2 cooling units are provided with power from the p back up power supply units, for each of the plurality of domains; modelling the temperature in the data hall at the n server cabinets when power is restored to the m cooling units, and the m / 2 backup power supplies are turned off, for each of the plurality of domains, as described above. As further shown in Fig. 3(b) and (c), process 350 may include using the result of the temperature modelling from all of the above stages to determine overall energy usage in each of the data halls in the plurality of model domains (block 362), and using the temperature modelling results and the energy usage to provide an output of the configuration of the server cabinets, cooling units and backup power supply units in the data hall with the optimum energy efficiency (block 364). For example, device may use the result of the temperature modelling from all of the above stages to determine overall energy usage in each of the data halls in the plurality of model domains, and using the temperature modelling results and the energy usage to provide an output of the configuration of the server cabinets, cooling units and backup power supply units in the data hall with the optimum energy efficiency, as described above. Although Figures 3(b)and 3(c) shows example blocks of process 350, in some implementations, process 350 may include additional blocks, fewer blocks, different blocks, or differently arranged blocks than those depicted in Figures 3(b) and 3(c). Additionally, or alternatively, two or more of the blocks of process 350 may be performed in parallel. Results and Key Findings: A study investigated on both Scenario 1 and Scenario 2 evaluates, the impact on server cabinet inlet temperatures of a data hall with one model of the data hall supported by no UPS and an alternative model of the data hall has 50% of CRACs supported by UPS. This study modelled a mains power failure to the data hall with a 23-second outage until generator activation and power restoration, followed by an additional 23 seconds for non-U PS-backed units to reach 100% capacity. Results indicated that even without UPS support, the scenario remained within acceptable limits at 34.80°C, though close to the maximum server cabinet inlet temperature. The scenario with 50% UPS-backed CRACs resulted in a server cabinet inlet temperature of 29.20°C. Extensive computational fluid dynamics (CFD) modelling was performed for various scenarios testing of the conditions within the computer room. To assess the impact of data hall configuration on system resilience, distinct data hall configurations were employed across the two simulated scenarios. Scenario 1 utilised a smaller data hall configuration, reflecting a potential failure scenario under a reduced IT load. Scenario 1, using a smaller configuration with reduced IT load, modelled potential failure. It was hypothesized that similar temperature profiles would occur in larger, IT denser data halls. Conversely, Scenario 2, involving a larger data hall with 50% CRAC units powered by UPS, was designed to evaluate the system's performance under increased capacity and redundancy. Successful simulation of Scenario 2 suggests the viability of this configuration for smaller data halls, given its demonstrated robustness in a more demanding environment. Two transient simulations were conducted to evaluate the impact of UPS for CRACs on data hall thermal stability during power failure. In each of the scenarios, a data hall model is constructed by creating a plurality of model domains for the data hall; wherein each of the plurality of model domains is provided with: a representation of a plurality, N, of server cabinets; a representation of a plurality, M, of cooling units for the data hall, where M<N; a representation of a plurality, P, of backup power supply units for the M cooling units, where P=M / 2. In an embodiment, power is no longer provided to the M cooling units, for each of the plurality of domains, and the M / 2 cooling units are provided with power from the P back up power supply units, for each of the plurality of domains. In a preferred example, the first time period when the power is not provided to the one or more M cooling units for a period between 0-60 seconds. Further preferably, the first time period when power is not provided to the one or more M cooling units for a period between 0-30 seconds. In an example embodiment, when power is resumed, the power is provided to M / 2 units for a period between 30-150 seconds. In an example, standard operating conditions of the cooling units in the plurality of data halls, are when the cooling unit provides a maximum temperature in the data hall of 35°C. Further preferably, the cooling unit provides a temperature in the range 1035°C. In an example, the cooling units comprise one or more of: Computer Room Air Conditioning, CRAC; Computer Room Air Handler, CRAH; and Fan Wall Units, FWU. Further preferably, each of the plurality of model domains includes a representation of the floor configuration in each of the data halls. In an example, the floor configuration of the one or more data halls, includes one or more raised sections of floor in each of the data halls. Preferably, in the model domain of the plurality of data halls a representation of one or more colling units is positioned in a floor void under a raised section of the floor. Figure 4 illustrates a cross section of a data hall 400 with raised floor configuration-442, and ceiling 434, for both of scenarios 1 and 2. The raised floor 442which facilitates the distribution of cooled air from CRACs-420 located in a dedicated cooling corridor 404. The CRAC units introduce conditioned air into the raised floor via the fans-441 located in the floor void 442. Perforated grilles, strategically positioned near server cabinet inlets-430, allow for the intake of this cooled air, while simultaneously expelling hot air. This hot air is then recirculated to the CRAC units for reconditioning via the Hot Aisle containment 443, maintaining a consistent temperature within the data hall. During a simulated mains power failure, all CRACs are deactivated until generator power is established and restored. Scenario 1 - 0% of CRAG units backed by UPS (0.855MW Data-hall) Figure 5: Scenario 01 Data Hall Floor plan Figure 5 shows the data hall 500 configuration for Scenario 1, the above is similar what was discussed in Figure 1. To add to that, 502 refers to the white space in data hall means cold air space which is supplied by CRACs. 520 refers to an CRAC corridor configuration means it’s the series of CRAC cooling units along the front wall-504, aligned with the cold aisles, to deliver chilled air directly to server intakes. This design optimises airflow management and cooling efficiency while maintaining a structured hot aisle / cold aisle arrangement. This CRAC corridor usually connected from false ceiling-134 where the hot air is drawn from the exhaust duct of the servers-543. All CRACs are reliant on mains power. The data hall is equipped with CRAC positioned on opposing sides to ensure uniform air distribution and mitigate the formation of thermal hotspots. Figure 6: shows a graph of the temperature results with time, in failure condition for Server Rows (SR) 1-5 in Scenario 01. The graph represents temperature variations at different row cabinet inlets (SR1 to SR5, whereas SR represents each server Row) over time during a power failure and subsequent recovery. The analysis reveals both qualitative and quantitative trends that highlight the impact of cooling recovery delays. During the power failure phase, there is an immediate increase in inlet temperatures across all monitored server rows SR1-SR5. The steepest rise is observed in SR3, suggesting that this row is in a critical location with higher heat density or poorer airflow circulation when CRAC fans stop working. Other rows, such as SR1,2,4 and SR5, show a more gradual increase in temperature, indicating they may be benefiting from residual airflow or better thermal inertia. From a numerical perspective, the highest peak temperature is approximately 34.5°C in SR3, representing a 7.5°C increase from its baseline (27°C). Other server rows show peak temperatures between 29°C and 31°C, indicating varying thermal impacts across the data hall. The temperature rise duration for most racks is around 20-30 seconds, aligning with a delay in generator startup and CRAC recovery. Additionally, the server cabinet inlet temperature profiles are also obtained during the simulation for Figure 5. As discussed above, the graph of Figure 6 depicts temperature fluctuations at selected server rack locations over time. Notably, a maximum server cabinet inlet temperature of 34.8°C is observed at 30 seconds post-power failure. This temperature spike is attributed to the absence of cooling during the power outage, leading to a rapid thermal increase. Upon generator activation and power restoration, a duration of 46 seconds is required for the server cabinet inlet temperatures to return to acceptable operating levels. This delay is due to the CRACs needing to ramp up to 100% capacity to compensate for the preceding thermal excursion. As power is restored and CRAC units resume operation, temperatures in the server racks begin to drop, but the rate of cooling is not uniform, over the server racks. Some server rows, particularly SR3, experience prolonged temperature elevation before fully stabilizing, likely due to airflow distribution delays. Additionally, secondary temperature fluctuations indicate intermittent cooling efficiency or staggered fan recovery, rather than an immediate return to optimal conditions. These observations suggest that certain areas in the data hall experience higher thermal stress, leading to potential risks for IT equipment. As the cooling system stabilises, temperatures start decreasing within 30-40 seconds, but full recovery to pre-failure levels takes approximately 80-100 seconds. Notably, SR3, which had the highest temperature spike, takes the longest to stabilize, indicating a thermal lag in that section. The cooling fluctuations observed after initial recovery suggest that CRAC units take time to fully restore airflow and cooling capacity. Additionally, the 5°C temperature variation between SR3 and SR5 demonstrates that airflow distribution is uneven across the data hall. Scenario 2 - 50% of CRAC units backed by UPS. (2.880MW Data-hall) Figure 7:illustrates a Data Hall Floor plan 700 for scenario 1. The data hall has CRAC units 720, server racks 734, white space 704 cold aisle zone 702. Figure 8 shows temperature variations at multiple row cabinet inlets (SR1 to SR11) of figure 7, for scenario 2, over time during a power failure event and subsequent recovery. The analysis provides insights into the impact of cooling system disruptions on thermal conditions across different server rows. In this scenario 50% of the CRACs 720 are supported by UPS. During mains power failure these UPS backed CRACs will continue to operate while the remaining CRACs will deactivate until the generator is active As power is restored and cooling systems begin recovering, temperature responses vary across the data hall. Certain server rows, such as SR3 and SR5, show extended periods of elevated temperatures before stabilising, highlighting non-uniform cooling recovery. The fluctuations observed in several rows post-recovery suggest that airflow redistribution is not instantaneous, and some areas experience delayed cooling due to staggered CRAC fan ramp-up or inconsistent airflow patterns. However, temperatures recorded all under 32°C compared with Scenario 1. Figure 8 demonstrates the stabilising effect of partial UPS backed CRACs as these maintain cooling significantly reducing the maximum server cabinet inlet temperature and preventing thermal runaway. The results show that partial UPS back has a significant impact of the room temperature as prevents exceeding 35°C.The utilisation of 50% UPS-backed units, as opposed to 100%, offers notable advantages in terms of equipment and operational cost savings. Numerically, the peak temperature recorded is approximately 29.2°C in SR3, representing a 3.5°C increase from its baseline (~25.7°C). Other server rows show peak temperatures ranging between 26.5°C and 28.8°C, reflecting variations in thermal impact based on location. The temperature rise occurs within 20-40 seconds, aligning with the generator startup delay and CRAC fan inactivity. During recovery, most server rows begin cooling down within 40-60 seconds, but the rate of cooling varies. Full recovery to post-failure conditions takes up to 200-250 seconds for certain rows, particularly SR3 and SR5, which exhibit prolonged thermal instability. The temperature differences between server rows, such as the ~2°C gap between SR3 and SR10, indicate airflow inefficiencies and possible recirculation effects in some zones. Conclusion Scenario 1 exhibits sharper and higher temperature peaks compared to Scenario 2. The maximum recorded temperature in Scenario 1 reaches 34.5°C in some critical rows, representing a 6-8°C rise from the baseline temperature (~26°C-28°C before failure). The temperature rise occurs within 20-30 seconds, closely aligning with the time required for the power generator to start up. The cooling recovery phase takes approximately 200-250 seconds, with certain server rows showing delayed stabilization beyond this timeframe. In Scenario 2, the highest recorded temperature is around 29.2°C, which is still a notable increase but less severe compared to Scenario 1. The temperature rise is slightly more gradual, taking 30-50 seconds to reach peak values, suggesting some level of airflow retention or delayed CRAG fan shutdown. The cooling recovery in Scenario 2 is also faster and more uniform, with most server rows returning to acceptable temperature levels within 150-200 seconds—a noticeable improvement over Scenario 1. Moreover, the temperature differential between server rows is wider in Scenario 1, with some rows experiencing much higher peaks than others. For example, SR3 and SR5 show a ~6-7°C deviation from their lower-temperature counterparts (SR10 and SR11), indicating uneven cooling distribution. In Scenario 2, this differential is reduced to around 2-3°C, pointing to a more balanced airflow system that prevents extreme temperature disparities across different zones. Peak Temperature Increase: Scenario 1: up to 34.5°C in Scenario 1 indicating complete airflow loss and slower cooling recovery. Scenario 2: max 29.2°C, suggesting some cooling support or improved airflow management. Time to Peak Temperature: Scenario 1: 20-30 seconds (rapid and abrupt rise). Scenario 2: 30-50 seconds (slightly delayed rise, indicating airflow retention). Recovery Time: Scenario 1: 200-250 seconds, with delayed stabilization in some rows. Scenario 2: 150-200 seconds, indicating faster cooling recovery. Post-Recovery Stability: Scenario 1: Significant temperature fluctuations post-recovery, highlighting airflow inconsistencies. Scenario 2: More gradual and stable return to normal operating temperatures. Temperature Uniformity Across Server Rows: Scenario 1: Large temperature difference (6-7°C between different rows). Scenario 2: Smaller differentials (2-3°C), showing better thermal distribution. In conclusion, it is evident that backing 50% of cooling units within Data Hall using UPS is beneficial rather than not backing all units under UPS backed. As seen in non-UPS backed, it is hard to get temperature under 32C to meet ASH RAE standards whereas 50% UPS backed does. There may have many similar practices are in place in Data centre design but there is not very particular publicly available material in Data centre industry. This study will be acting benchmark for those similar strategies. Although the present invention has been described in connection with some example embodiments, it is not intended to be limited to the specific form set forth herein. Rather, the scope of the present invention is limited only by the accompanying claims. Additionally, although a feature may appear to be described in connection with particular embodiments, one skilled in the art would recognize that various features of the described embodiments may be combined in accordance with the invention. In the claims, the term ‘comprising’ does not exclude the presence of other elements or steps. Furthermore, although individually listed, a plurality of means, elements or method steps may be implemented by, for example, a single unit or processor. Additionally, although individual features may be included in different claims, these may possibly be advantageously combined, and the inclusion in different claims does not imply that a combination of features is not feasible and / or advantageous. Also, the inclusion of a feature in one category of claims does not imply a limitation to this category but rather indicates that the feature is equally applicable to other claim categories, as appropriate. Furthermore, the order of features in the claims does not imply any specific order in which the features must be performed and in particular the order of individual steps in a method claim does not imply that the steps must be performed in this order. Rather, the steps may be performed in any suitable order. In addition, singular references do not exclude a plurality. Thus, references to ‘a’, ‘an’, ‘first’, ‘second’, etc. do not preclude a plurality. Acronyms: ASHRAE-American Society of Heating, Refrigerating and Air-Conditioning Engineers CFD - Computational Fluid Dynamics CRAC - Computer Room Air Conditioning CRAH - Computer Room Air Handler FVM - Finite Volume Method FWU - Fan Wall Units 5 IT - Information technology MUPS - Mechanical Uninterruptible Power Supplies PCS - Primary Critical System SLA - Service Level Agreement UPS - Uninterruptible Power Supplies 10 SR-Server Row

Claims

1. A computer implemented method for determining a configuration of a data hall in a data centre with improved energy efficiency comprising the steps of: creating a plurality of model domains for the data hall;wherein each of the plurality of model domains is provided with:a representation of a plurality, N, of server cabinets;a representation of a plurality, M, of cooling units for the data hall, where M<N;a representation of a plurality, P, of backup power supply units for the M cooling units, where P=M / 2;using computational fluid dynamics CFD, equations to model a temperature in the data hall at each of the server cabinets under a standard operating conditions of the cooling units and the server cabinets for each of the plurality of model domains;modelling the temperature in the data hall at each of the server cabinets for a first time period when power is no longer provided to one or more of the M cooling units, for each of the plurality of domains;modelling the temperature in the data hall in each of the server racks in the data hall for a second time period when one or more of the M / 2 cooling units are provided with power from the P backup power supply units, for each of the plurality of domains;modelling the temperature in the data hall at the N server cabinets when power is restored to the M cooling units, and the M / 2 backup power supplies are turned off, for each of the plurality of domains;using results of the temperature modelling from all of the above stages to determine overall energy usage in each of the data halls in the plurality of model domains, and using the temperature modelling results and the overall energy usage to provide an output of the configuration of the server cabinets, cooling units and backup power supply units in the data hall with optimum energy efficiency.

2. The computer implemented method as claimed in claim 1 wherein the temperature modelling at all different stages is modelling using computational fluid dynamics.

3. The computer implemented method as claimed in claim 2 wherein the CFD equations to model temperature are used to provide information about airflow within the data hall.

4. The computer implemented method as claimed in any preceding claim wherein power is no longer provided to the M cooling units, for each of the plurality of domains, and the M / 2 cooling units are provided with power from the P back up power supply units, for each of the plurality of domains.

5. The computer implemented method of any preceding claim wherein the first time period when the power is not provided to the one or more M cooling units for a period between 0-60 seconds.

6. The computer implemented method of claim 4 wherein the first time period when power is not provided to the one or more M cooling units for a period between 0-30 seconds.

7. The computer implemented method as claimed in claim 5 or claim 6 wherein, when power is resumed, the power is provided to M / 2 units for a period between 30-150 seconds8. The computer implemented method as claimed in any preceding claim wherein standard operating conditions of the cooling units are when the cooling unit provides a maximum temperature in the data hall of 35oC.

9. The computer implemented method as claimed in claim 8 wherein the cooling unit provides a temperature in a range 10-35oC.

10. The computer implemented method as claimed in any preceding claim wherein the plurality of model domains is divided into a plurality of mesh cells, and theCFD equations for the temperature modelling stage are determined for each of the mesh cells.

11. The computer implemented method as claimed in any preceding claim wherein the cooling units comprise one or more of: Computer Room Air Conditioning, CRAC; Computer Room Air Handler, CRAH; and Fan Wall Units, FWU.

12. A computer implemented method as claimed in any preceding claim wherein each of the plurality of model domains includes a representation of a floor configuration in each of the data halls.

13. A computer implemented method as claimed in claim 12 wherein the floor configuration includes one or more raised sections of floor in each of the model domains of the data hall.

14. A computer implemented method as claimed in claim 12 or claim 13, wherein in the plurality of model domains of the data hall a representation of one or more colling units is positioned in a floor void under a raised section of the floor.

Images