Next-Generation

Data Center Facility

White Paper
01 Next-Generation Data Center Facility White Paper Foreword

Foreword

The world is undergoing a rapid digital transformation. Innovative services, such as smartphones,
smart homes, smart manufacturing, and autonomous driving, are reshaping our work and life. At the
same time, with climate change accelerating and a global consensus on carbon neutrality, the pursuit
of green and sustainable development has become a common goal.

Data centers are fundamental for the digital economy. However, data centers are currently drawing
much attention from all segments of society due to their high energy consumption and carbon
emissions. With surging demand for data center construction, information technology evolution
and low-carbon requirements, the data center industry is undergoing profound changes, and its
development will enter a new era.

What is a "next-generation data center facility" that meets the requirements of the new era? Huawei
has worked with global data center industry leaders and technology experts to conduct extensive
discussions on industry and technology trends, and reached many important consensuses, including
the definition of next-generation data center facilities.

The future is here. It is believed that next-generation data center facilities, defined by the industry's
wisdom, will play an important role in the high-quality and sustainable development of the industry.
 Contents Next-Generation Data Center Facility White Paper 02

Contents

Foreword01

Intelligent and Low-Carbon Technologies Drive the Rapid and

Sustainable Development of Data Centers03

1.1 Digital Economy Drives Fast Growth of Data Centers04

1.2 Carbon Neutrality Raises New Requirements for Data Center Sustainability05

Next-Generation Data Center Facility06

2.1 Sustainable07
2.1.1 All Green: Utilizing Green Sources for Harmony with Nature 07
2.1.2 All Efficient: PUE → xUE, One Dimension to Multi-dimensional System 08
2.1.3 All Recyclable: Maximizing Resource Recovery Throughout the Life Cycle 09

2.2 Simplified10
2.2.1 Simplified Architecture: Innovative Forms of Buildings and Equipment Rooms 10
2.2.2 Simplified Power Supply: Redefined Components and Links 12
2.2.3 Simplified Cooling: Maximized Heat Exchange Efficiency 13

2.3 Autonomous Driving14

2.3.1 O&M Automation: Unmanned Inspection 15
2.3.2 Automatic Energy Efficiency Optimization: Enables Smart Cooling 15
2.3.3 Operation Autonomy: Maximizes Resource Value 16

2.4 Reliable17
2.4.1 Proactive Security: Early Warning and Quick Fault Closure 18
2.4.2 Secure Architecture: Safeguards All Ranging from Components to Data Centers 18

Summary19
 Intelligent and Low-Carbon Technologies Drive the
03 Next-Generation Data Center Facility White Paper Rapid and Sustainable Development of Data Centers

01
Intelligent and Low-Carbon Technologies
Drive the Rapid and Sustainable
Development of Data Centers

Digitalization and low-carbonization are two unstoppable trends in the world. Data centers are
the foundation of the digital world. By integrating digital and power electronics technologies,
building smart and low-carbon data centers will become the mainstream trend.
 Intelligent and Low-Carbon Technologies Drive the
Rapid and Sustainable Development of Data Centers Next-Generation Data Center Facility White Paper 04

1.1 Digital Economy Drives Fast Growth of Data Centers

Currently, digital technologies such as AI, cloud computing, In the USA, the National Science Board has recommended
big data, IoT, and 5G are evolving at a fast pace, fueling increased investments in data, software, computing,
global GDP growth. According to China Academy of and networks to maintain competitiveness in the digital
Information and Communications Technology (CAICT), economy in Vision 2030 .
the digital economy accounted for 43.7% of global GDP
in 2020 and has expanded at an astonishing speed. Nowadays, digital technologies have been evolving
and reshaping the way we live and work. In the post-
Major economies like China, the EU, and the USA have pandemic era, the digital economy has unleashed a
released digital economy visions. powerful force to reshape the global economic structure,
and reconstruct the global competitive landscape. Data
China has set the digital economy as the main developing center construction is expected to develop rapidly.
direction in the 14th Five-Year Plan (2021–2025) for
National Economic and Social Development . Data has become a key factor of production now, and the
data volume keeps increasing dramatically. According to
The European Union has set out digital ambitions for the Huawei's Intelligent World 2030 , computing power will
next decade in 2030 Digital Compass: the European way increase ten times, and AI computing power will increase
for the Digital Decade , which states that by 2030, 75% of 500 times. The skyrocketing demand for computing power
European enterprises will use cloud computing, big data will unlock the significant potential of the data center
and AI; more than 90% of European small and medium- market. According to the report of Synergy Research
sized enterprises (SMEs) will reach at least a basic level Group, which was released in March 2022, the number of
of digital intensity, and Europe will increase investment in global ultra-large data centers will be more than 1,000 in
renewable energy and digital infrastructure. the coming three years.
 Intelligent and Low-Carbon Technologies Drive the
05 Next-Generation Data Center Facility White Paper Rapid and Sustainable Development of Data Centers

1.2 Carbon Neutrality Raises New Requirements for Data

Center Sustainability

With global warming problems, carbon neutrality has industry associations have announced that data centers
become a common mission. Until now, more than 130 will be carbon neutral by 2030.
countries have declared their commitment to carbon
neutrality. Energy conservation and carbon emissions China has released the Guidance on Accelerating the
reduction have become key trends across all sectors. Construction of the Collaborative Innovation System for
National Integrated Big Data Centers to promote the
In recent years, the data center industry is booming, construction of national integrated big data centers. And
its market space has been growing rapidly, and power the "Eastern Data and Western Computing" program has
consumption has increased dramatically. According to been launched to channel more computing resources
Uptime Global Data Center Report 2021 , from 2014, from the eastern areas to the less developed western
the PUE of large data centers around the world has regions, which promotes the green and sustainable
remained around 1.6 for seven consecutive years. The development of data centers, and accelerates the R&D
resource utilization of data centers is low and needs and application of energy-saving and low-carbon
to be upgraded urgently. To advance the sustainable technologies. It is required that the PUE of newly-built
development of the data center industry, a number large data centers be lower than 1.3 by 2025.
of countries and international organizations have
released data center-related policies. For example, the Low carbonization of data centers will become an
U.S. government adopts the Data Center Optimization inevitable trend. Large-scale application of clean energy,
Initiative (DCOI), which requires that the PUE of new- development of carbon-reduction technologies including
built data centers be lower than 1.4 and that of upgraded PV and energy storage, and waste heat recovery, will
old data centers be lower than 1.5. In the Climate Neutral effectively advance data centers to achieve sustainable
Data Centre Pact , European data center operators and development.
 Next-Generation Data Center Facility Next-Generation Data Center Facility White Paper 06

Next-Generation
Data Center Facility 02
In the face of the booming development of data centers complexity of data centers, the architecture and key
as well as green and low-carbon requirements, Huawei subsystems will be minimalist through hardware
and global data center industry leaders and technical convergence.
experts have conducted in-depth and extensive 3. Autonomous driving: Digital and AI technologies
discussions and reached many important consensuses, will be leveraged to address the operation and
summarizing the four characteristics as Sustainable, maintenance challenges of massive data centers and
Simplified, Autonomous Driving, and Reliable. automate data center facilities.
4. Reliable: Reliable operation is the cornerstone of a
1. Sustainable: Innovative measures will be taken to data center. Through modular redundant architecture
build green and highly energy efficient data centers and AI active prevention, we can guarantee
benefiting the society. the reliability of data centers at all levels from
2. Simplified: In response to the increasing scale and components and equipment to systems.
07 Next-Generation Data Center Facility White Paper Next-Generation Data Center Facility

2.1 Sustainable

Looking back on the development

of global modernization, we went All Green: Utilizing Green Sources for Harmony with Nature
through three industrial revolutions
from "Steam Age", to "Electrical
Age", and to "Information Age". Green growth has become a global consensus. It is a sustainable development
These three industrial revolutions model based on green, low-carbon, and recycling technologies, which means
mainly relied on fossil energy. that while maintaining sustained economic growth, it can reduce the damage
With the rapid economic growth, to the environment.
carbon dioxide emissions increased
exponentially, which caused global Based on the concept of green growth, the next-generation data center construction
climate change and brought great must utilize green sources.
challenges to the survival and
development of human society.

The next-generation data center

facilities will be all green, efficient,
and recyclable. Electricity: Water:
Renewable energy should be used Reducing the consumption of water
on a large scale for data centers. The resources requires minimal use
development and utilization of PV, of clean water, maximum use of
wind power, and hydropower should reclaimed water, or even without
be given a high priority in order to using water. As water resources are
reduce fossil energy dependence. increasingly scarce in many areas
In addition, distributed PV can be of the world, data centers need
deployed in a data center campus to save precious water resources.
to fully utilize the rooftop and land For example, all data centers in
resources of the campus. Distributed Ulanqab city of China are banned
PV can reduce the carbon emissions from using any groundwater for
of the data center by 1% to 2%. cooling systems.

Land: Climate:
The data center scale is increasing Data centers should use natural
and the construction will use land resources, such as external cold air
resources intensively. Land should and cold water. We can build a data
be efficiently used to support more center in an appropriate area or use
computing power on each meter technical measures, such as increasing
square of land. the supply and return temperature
to use more external cold air.

Harmonious coexistence between data centers and nature can be ensured by

utilizing green resources.
 Next-Generation Data Center Facility Next-Generation Data Center Facility White Paper 08

All Efficient: PUE → xUE, One Dimension to Multi-dimensional System

The Green Grid proposed PUE as an indicator to measure evaporative cooling systems are also different. With a
the power efficiency of data centers in 2007, which certain grid power capacity, the number of IT racks that
has been widely accepted and used in the industry. can be deployed varies according to the solutions.
However, PUE alone cannot fully reflect the resource
utilization of data centers. For example, even with the Therefore, the system evaluating whether resources are
same PUE, the carbon emissions from thermal power used efficiently should be changed from PUE-only to xUE,
and green power are completely different, and the water that is, a multi-dimensional system covering CUE, PUE,
resources consumed by chilled water systems and indirect WUE, GUE, etc.

Carbon Usage Effectiveness (CUE) Water Usage Effectiveness (WUE) Grid Usage Effectiveness (GUE)
is a quantifiable indicator for measures the water consumption indicates the maximum number
the carbon emissions of data of IT devices in a data center of IT devices that can be deployed
centers. Different energy sources during cooling. It can be used to with a certain grid power capacity.
emit different amounts of carbon restrict solutions and products that This indicator helps the industry to
dioxide. For example, coal emits consume a large amount of water. optimize products and deploy more
1023 grams of carbon dioxide per IT devices.
kWh, and photovoltaic emits only
30 grams per kWh. Therefore, CUEs
of different sources of electricity in
data centers vary greatly.

Different regions and industries pay different importance

to the corresponding indicators. Therefore, each indicator
can be measured based on the characteristics of regions
and industries, and finally an optimal evaluation scheme
can be selected based on comprehensive consideration.
09 Next-Generation Data Center Facility White Paper Next-Generation Data Center Facility

All Recyclable: Maximizing Resource Recovery Throughout the Life Cycle

A data center is a big energy consumer and heat emitter. (1) The data center itself, for heating water in office areas
The energy consumed by the data center is converted into and heating a diesel generator room
heat and discharged into the air, which is not effectively (2) Ancillary businesses around the data center, such as
utilized or recycled. In addition, hot spots can be formed aquaculture or heating of a commercial complex
around the data center, affecting the cooling and thus the (3) The municipal heating network.
PUE. Waste heat recovery is an important technology in
the low-carbon era and it is one of the important features In addition to the waste heat recovery of the data center,
of next-generation data centers as well. the recyclability of the data center equipment and the
materials of the facilities should also be considered.
The European Climate Neutral Data Centre Pact In the future, more and more regions will promote
(CNDCP) has clearly stated that heat recovery is one of new prefabricated green buildings and green building
the five key measures to achieve carbon neutrality in materials, with the recyclability rate exceeding 80%.
2030. The European Sustainable Digital Infrastructure Lead-free and halogen-free green recyclable materials will
Alliance (SDIA) has put heat recovery on the roadmap, be used to replace traditional harmful substances, such
and the proportion of heat recovery in data centers as lead in components, boards, and auxiliary materials, to
should be greater than 60%. The recovered heat can be achieve a higher recovery rate, and promote a low-carbon
used for: circular economy for data centers.
 Next-Generation Data Center Facility Next-Generation Data Center Facility White Paper 10

2.2 Simplified

To cope with problems such as

Simplified Architecture:
slow construction and difficult
Innovative Forms of Buildings and Equipment Rooms
O&M of traditional data centers,
the continuous pursuit of simplicity
drives the convergence of
components, devices, systems, and GB 50174-2017 Code for Design of Data Centers defines a data center as
data center architectures. "a building that provides a running environment for centralized electronic
information devices. It can be one or several buildings or part of a building,
including the computer room, auxiliary area, support area, and administration
area." Data centers are complex engineering buildings. Traditionally, civil
engineering, power supply, and cooling works are conducted one after another.
The construction process is lengthy and subject to weather and design changes.
In addition, different vendors are responsible for the implementation of different
procedures. As a result, on-site construction companies and construction
personnel are complex, which increases on-site management difficulty and
makes it hard to ensure the construction progress and final quality.

In the future, with the increasing scale of data centers, convergent and
simplified buildings and equipment rooms will meet fast service rollout
requirements and come in line with the industry trend.

Building prefabrication: A data center is divided into parts, changing the

building form and construction mode. The traditional "reinforced concrete +
onsite construction" is transformed into "fabricated steel structure building
+ prefabrication". The engineering is product-oriented, enabling parallel
data center construction, fast delivery, and on-demand deployment to meet
requirements for fast service rollout in the cloud data center era.

For the traditional building construction, the whole process includes the serial
works of the foundation, building, decoration, power supply and distribution,
heating ventilation & air conditioning (HVAC), fire extinguishing, and
monitoring systems plus installation and commissioning. The prefabricated
modular construction can realize all-round parallel construction, which is
embodied in two levels. The first is the parallel processes of foundation works
and manufacturing, pre-integration, and pre-commissioning of functional
modules. At the second level, standard modules are used, and multiple
modules can be produced simultaneously in the factory. The efficiency is greatly
improved through mass production processes. Modules are pre-integrated
and pre-commissioned in the factory, which greatly improves the quality and
supports quick onsite deployment and joint commissioning and delivery.

In China, for example, when the prefabricated modular construction mode

is used to construct a 1000-rack data center, the construction period can be
reduced from more than 18 months to 6–9 months. This difference is more
11 Next-Generation Data Center Facility White Paper Next-Generation Data Center Facility

obvious in the Middle East. Due to climate conditions, that in other regions. In the Middle East, the construction
May to October is the hottest time. Outdoor construction period of prefabricated modular data centers is shortened
is not allowed from 13:00 to 17:00 every day. The by 60% from 30 months to 12 months compared with
construction time of a large data center is longer than the traditional construction mode.

The Wuhan AI computing center uses the prefabricated modular data center solution. The building was constructed in four
months, and the project was completed in five months. The rollout time was shortened by more than 50%.

Data center modularization: A modular

architecture can be used to reshape data
center facilities. In traditional equipment room
construction, parts such as racks, air conditioners,
UPSs, power distribution cabinets, batteries, fire
extinguishing devices, and cables are purchased,
installed, and tested separately, which brings
great challenges to construction, O&M, and
energy efficiency management. The modular
architecture integrates the racks, cooling, power
supply, monitoring, and fire extinguishing
subsystems into one module and isolates the hot
and cold air. By changing the architecture of the
equipment room, the delivery period is shortened,
the O&M difficulty is reduced, and the energy
efficiency is improved.
 Next-Generation Data Center Facility Next-Generation Data Center Facility White Paper 12

Simplified Power Supply: Redefined Components and Links

The trends of data centers are high-density and large- on-site delivery and commissioning, the system space is
scale. The power supply system is the heart of data centers. reconstructed and optimized by all-in-one integration of
To meet the increasing power density and large-scale originally distributed devices and innovative components
requirements, the future innovation of data centers is to without changing power supply links. For example, the
integrate and innovate power supply links to deliver a switch miniaturization technology is used to reduce the
simplified power supply system. number of switch cabinets without reducing the number
of switches. On the other hand, the topology pooling
Component integration: In the past, the power supply and component optimization technologies are used to
system of a large-scale data center mostly incorporates improve the power density of UPS modules, effectively
the UPS parallel system and lead-acid battery solution. reducing the footprint and simplifying onsite delivery.
There are many complex devices, which result in
challenges such as complex on-site installation and Simplified link: In response to the call for carbon
commissioning, and a large footprint. Using a data center neutrality, clean energy applications, peak-valley electricity
with 1500 racks and 8 kW/rack as an example, the area price differences, and virtual power plants (VPPs) will
for the power supply system (including transformers, drive data centers to deploy PV and energy storage.
UPSs, power distribution cabinets, and batteries) is about In a traditional solution, PV systems (inverters and PV
1800 square meters, accounting for 15% to 20% of the modules) and energy storage systems (converter boxes
total data center area. As the power density of IT racks and energy storage boxes) are deployed on power supply
increases in the future, this proportion will continue to links. More devices increase the complexity of the power
increase if the power supply system remains unchanged. supply system, which leads to challenges such as longer
links and more difficult maintenance. A next-generation
To meet the challenges of large footprint and complex data center integrates all power links innovatively

CTTIC Cloud uses Huawei's PowerPOD solution, which saves the power supply system space by more than 40% compared with
a traditional solution, helping the customer deploy 350 more IT racks and save more than 16,000 meters of power cables.
13 Next-Generation Data Center Facility White Paper Next-Generation Data Center Facility

to make them simple. For example, the PV inverter used to connect the mains, PV, and energy storage at the
and energy storage converter are integrated with the same time, integrating the mains rectifier and inverter, PV
uninterruptible power system to provide a new medium- inverter, and energy storage functions. The complexity of
voltage uninterruptible power solution. A single system is the power supply link is greatly reduced.

Simplified Cooling: Maximized Heat Exchange Efficiency

To achieve low carbonization, energy saving and multiple heat exchanges to one, shortening the cooling
consumption reduction in data centers lies in the simplified link, and improving the cooling exchange efficiency.
integration of cooling systems and the maximization of
heat exchange efficiency. Air-liquid convergence: As the power density of
IT devices increases, especially in ultra-high power
Simplified cooling link: In the past, large data centers computing scenarios involving AI and supercomputing,
mainly used mechanical cooling systems. Taking a the servers and chips require more capability and higher
chilled water system as an example, which includes efficiency of the cooling system. Close-coupled cooling
seven devices (chiller, cooling tower, cold storage tank, has become an important trend in the development
indoor unit, cooling water pump, plate heat changer, of cooling systems. The cooling mode ranges from
and management system), four heat exchanges take traditional room-level dispersive cooling, to module-level
place in the whole cooling system. The project delivery cooling in cold/hot aisle containment, to cabinet-level
lasts for more than half a year and is hard to implement. and chip-level liquid cooling, and finally to direct cooling
The construction quality affects the cooling efficiency. In for chips. In the future, in scenarios where the power
the future, a data center cooling system will integrate density per rack is more than 20 kW, liquid cooling will
multiple components into one module, shortening the be used as the primary and air cooling as the auxiliary
delivery period and reducing O&M difficulty. In addition, approach. Converged air-liquid cooling will become the
a free cooling mode is available in this system, changing mainstream.
 Next-Generation Data Center Facility Next-Generation Data Center Facility White Paper 14

2.3 Autonomous Driving

As the scale of a single data center expands from 3. H

 igh operation gap. Firstly, the actual operation PUE
thousands of racks to ten thousands of racks, the number is not as good as the designed value, even though
of devices and their operation status to be monitored are the designed one is attractive. Secondly, the average
increasing exponentially, causing great challenges in data resource utilization of data centers is usually lower than
center facility management. 65%, causing much waste of resources.

1. R
 elying on manual inspection. Generally, 15 to To address the above challenges, data centers need
30 professional O&M engineers are required in a more advanced management methods and systematic,
1000-rack data center to perform 6 to 12 on-site automated, and intelligent technologies to improve
inspections every day. The O&M cost is high and the management efficiency, and to realize autonomous
inspection quality is difficult to guarantee. driving in the data centers with O&M automation, energy
efficiency optimization, and operation autonomy.
2. P
 assive response. A data center generates massive
alarms every day, including common, redundant,
and root alarms. One fault may trigger hundreds of
alarms, and it is difficult to find root alarms among
them. As a result, fault identification becomes difficult.
Traditional data center O&M relies on manual alarm
priority identification, alarm root cause identification,
and fault analysis, which leads to low O&M efficiency
and may cause risks in manual operations.
15 Next-Generation Data Center Facility White Paper Next-Generation Data Center Facility

O&M Automation: Unmanned Inspection

As the scale of a single data center becomes larger, the Traditional manual O&M cannot meet complex O&M
O&M difficulty increases. requirements of data centers. Automatic O&M will
become an important feature of next-generation data
Take a data center with 1500 racks as an example. There centers. To build a digital visualized foundation, AI
are more than 100 types of devices and most of them are technologies will be applied to implement unmanned
dumb, resulting in huge difficulty for manual inspection inspection through intelligent sensing, sound recognition,
and much time for fault locating. In addition, O&M relies and image recognition. In addition, the O&M process will
on labor mainly, leading to a growing proportion of labor be standardized, and expert experience will be shared
cost. According to the report of Uptime Institute global on a cloud platform and incorporated it into the process,
supply-side survey 2021 , the data center O&M labor quickly improving the skill sets of O&M engineers.
cost increased from 4.5% in 2015 to 10% in 2020. On
the other hand, the aging population of the world keeps In the past, it took two hours for one engineer to finish
increasing and labor shortage becomes a prominent an on-site inspection for a data center with 2000 racks.
issue, making it even harder for enterprises to employ In the future, by using automatic inspection methods,
proper O&M engineers. According to the above report such as indicator collection, camera image analysis, and
of Uptime Institute, 47% of data center enterprises have infrared sensing, engineers can complete the inspection
difficulty in finding qualified O&M engineers. of 2000 racks in 5 minutes remotely.

Automatic Energy Efficiency Optimization: Enables Smart Cooling

Cooling system is the second largest power consumer In the future, AI dynamic modeling technology will be
in a data center, next only to IT devices. Challenges used in data center energy efficiency optimization to build
in traditional building automation (BA) systems a machine learning model between energy consumption
include less than 10 collection parameters, less than 3 and varying parameters such as IT load, climate, and
adjustable parameters, slow and inaccurate adjustment, device running parameters. In this way, the energy
2 hours for each optimization, and short effect of consumption of each subsystem can be diagnosed in
adjustment. In addition, the traditional BA systems rely real time while device and system reliability is ensured.
heavily on manual optimization and expert experience, The technology automatically and accurately infers and
with high skills required. For increasingly complex configures the optimal control logic for a data center,
cooling systems, manual adjustment based on expert and builds a PUE model through deep neural network
experience cannot be performed in real time based on (DNN) training to obtain appropriate instructions. An
environmental parameters (such as temperature and optimal cooling policy can be inferred from 1.4 million
humidity) and IT load. In addition, to ensure system original combinations within 1 minute, and feedback
reliability, cooling requirements often expand, resulting on the effect is provided after the policy is delivered,
in energy waste. achieving optimal energy efficiency of the data center.
 Next-Generation Data Center Facility Next-Generation Data Center Facility White Paper 16

AI energy efficiency optimization is used in Huawei Cloud Langfang Data Center. There are more than 700 collection points. After
optimization, the annual average PUE decreases from 1.42 to 1.25, saving 33.36 million kWh or US$3.83 million ($0.115/kWh).

Operation Autonomy: Maximizes Resource Value

Resource Optimization @AI assets based on asset management. The technologies

will comprehensively analyze the available space,
During data center operation, a large number of power, cooling, and network (SPCN), and intelligently
devices need to be added, removed, or changed to recommend an optimal position for devices on the racks.
match service changes, causing great challenges to the In this way, data center resources can be visualized and
management of rack space resources. In a data center, managed, preventing capacity stranding. This feature
some racks are not fully used or even idle, and some maximizes the utilization of data center resources,
racks are overloaded. How to manage data center improves resource usage, and boosts operation revenue.
resources more efficiently and maximize their utilization
should be a big consideration in the next-generation Energy scheduling @AI
data centers.
With the increasingly diversified energy input and usage
AI-based resource optimization technologies can be of future data centers, an AI-powered autonomous
used to manage data center resources throughout the driving platform will enable flexible scheduling of various
life cycle, establish a management model and analysis energy resources and implement on-demand scheduling
platform centered on device management, automatically in terms of green power, PV deployment, and energy
count data center resource status through AI simulation storage, reducing manual calculating and operations and
and service prediction, and accurately locate data center maximizing resource value.
17 Next-Generation Data Center Facility White Paper Next-Generation Data Center Facility

2.4 Reliable

Reliability of data center facilities is still jeopardized by For example, in 2021, a fire broke out in the data center of
many factors. a cloud service provider in Europe. As a result, 3.6 million
websites were down and some data was permanently lost,
According to the Uptime Institute global supply-side survey causing huge losses.
2021 , failure in the power supply system and cooling
system are still the main causes of data center operation Next-generation data center facilities will be sustainable,
interruption, accounting for 57% of them, of which 43% simplified, and supports autonomous driving. However,
are from the power supply system and 14% are from the these features are feasible only when data centers run
cooling system. reliably.

Facility reliability problems would cause huge consequences Proactive security and architecture security should be built
and impact. to ensure the reliability of next-generation data centers.
 Next-Generation Data Center Facility Next-Generation Data Center Facility White Paper 18

Proactive Security: Early Warning and Quick Fault Closure

According to the Heinrich's Law, any accident can be to implement predictive maintenance from components
prevented; for every serious accident, there are 29 minor to data centers, based on the visibility and perception of
accidents, 300 near-misses, and 1000 potential accidents. all domains in the data centers.

Based on real-time fault detection, visualized panoramic For example, for key components and vulnerable
display, and AI-assisted locating, faults can be quickly components, such as capacitors and fans, big data
detected, analyzed, and troubleshot, and services can be collection and AI model training can be used to predict
quickly recovered. their service life, calculate the failure time of the
components in advance, and remind O&M personnel
Fault prevention is part of high-level reliability. to replace the components. This feature avoids failure
before it happens.
Traditional data center maintenance usually relies on
manual passive response, which makes it difficult to detect On the other hand, the fault response mechanism
minor risks and prevent faults in advance. should change from work order-driven manual response
to automatic fault response.
Proactive security means using big data and AI technologies

Secure Architecture: Safeguards All Ranging from Components to Data Centers

In the future, ICT technologies will be more widely point of failure (SPOF), ensuring that devices run
used in the data center facility field. They will be deeply without any interruption. At the system level, an end to
integrated with power electronics technologies to ensure end (E2E) visualizable, manageable, and controllable
reliability at various levels, such as components, devices, platform enables the system availability to reach "Five
and systems, to effectively enhance facility resilience Nines" (99.999%).
and build a comprehensive security defense line. For
example, at the component level, key components To sum up, the architecture-level optimization design
that adopt modular design are hot-swappable to reduces energy conversion layers and removes extra
quickly recover from faults. At the device layer, the components. In addition to reducing fault points, hitless
full redundancy design enables seamless switching to and interruption-free switching ensure that systems are
redundant modules within 0 ms delay after a single always online and secure.
19 Next-Generation Data Center Facility White Paper Summary

Summary

In the pursuit of next generation data center facilities, technological innovation will
be a key force in ensuring sustainable development. Looking towards the future,
Huawei will keep making breakthroughs in products and technologies through
continuous investment in R&D and with the all-round cooperation with customers,
ecosystem partners, and industry organizations. We can jointly usher in a new era of
data center development.
Tradememark Notice
, , are trademarks or registered trademarks of
Huawei Technologies Co., Ltd.
Other Trademarks, product, service and company names mentioned are the
property of thier respective owners.

General Disclaimer
The information in this document may contain predictive statement including,
without limitation, statements regarding the future financial and operating
results, future product portfolios, new technologies, etc. There are a number of
factors that could cause actual results and developments to differ materially
from those expressed or implied in the predictive statements. Therefore, such
information is provided for reference purpose only and constitutes neither an
offer nor an acceptance. Huawei may change the information at any time
without notice.
Copyright © 2022 HUAWEI TECHNOLOGIES CO., LTD. All Rights Reserved.
No part of this document may be reproduced or transmitted in any form or by
any means without prior written consent of Huawei Technologies Co., Ltd.

HUAWEI TECHNOLOGIES CO., LTD.

Huawei Industrial Base
Bantian Longgang
Shenzhen 518129, P. R. China
Tel: +86-755-28780808
www.huawei.com