How Deploying AI in Data Centers Can Save You $ Millions
Artificial intelligence is moving away from the confines of sci-fi movies and novels and into the mainstream, with autonomous vehicles, AI in devices, and superiority in cerebral games like chess and Go increasingly in the news and entering public consciousness.
McKinsey believes that AI will add US$13 trillion to total economic output by 2030 and boost global GDP by about 1.2% per year. As a general-purpose technology for building a fully connected, intelligent world, AI will be the core driving force behind the fourth industrial revolution, spurring profound and disruptive changes in the world.
Moving forward, its deployment in data center (DC) infrastructure is essential to ensure powerful and green DCs that meet the future digital needs of individuals, homes, and organizations.
Read more: Huawei’s AI Strategy and Product Portfolio
What’s Powering AI?
The changes caused by AI aren’t appearing out of thin air.
AI utilization can be maximized only when it’s deeply integrated into industry verticals. However, it needs robust infrastructure to support its development, not only in core elements like algorithms, data, and computing power, but also in the technology that forms the very basis of AI technology: ICT infrastructure, especially data center infrastructure serves as AI’s underlying facilitator.
With the arrival of Industry 4.0, data centers will become a key part of enterprise production and management systems and integral to competitiveness. As AI is deployed more, the value of DC infrastructure will become more prominent. In turn, more powerful DC infrastructure can promote AI applications – AI and DC infrastructure are mutually complementary.
So, what opportunities does AI bring for DCs? What can we do with these opportunities?
Key Opportunity 1: Proactive Prevention for DC Safety and Reliability
One example of industry practices of using AI to improve data center reliability and availability is collecting information about the power supply and distribution system to generate warnings for device and component faults, which helps O&M personnel with decision-making. However, we need to consider how to use AI’s self-learning ability to ensure that the power supply and distribution system runs safely and reliably. This may be a good opportunity for AI adoption to achieve a higher degree of intelligence and proactive avoidance.
Huawei developed its iPower intelligent power supply and distribution technology, which integrates AI performance to improve DC availability. iPower uses intelligent hardware for fault location and warning management, applying big data to intelligently analyze massive amounts of daily O&M data to identify potential risks in the equipment room and ensure DCs run reliably.
With a modular UPS at its core, iPower ensures the availability of DC energy infrastructure by full-link monitoring, warnings, and automatic fault isolation, and uses AI for predictive maintenance.
iPower achieves ms-level fault detection and isolation and minute-level recovery from faults, eliminating fire risks and greatly improving the reliability and availability of DC energy infrastructure.
To prevent battery failures, iPower uses AI to predict battery life and health, providing a maintenance and decision-making basis for users to handle battery strings with potential risks. This changes post-event remedies to pre-event prevention and passive response to active maintenance, greatly improving the power supply and distribution safety of DCs. iPower enables the Huawei modular UPS to automatically cut off the battery string when the battery temperature increases, ensuring incidents like fires don’t happen.
Key Opportunity 2: Intelligent O&M for Maximum Automation and Efficiency
The traditional maintenance approach is error-prone and costly in terms of both time and people, with low-level human errors potentially causing great losses to the data center, errors which traditional O&M cannot solve.
The AI-based iManager serves as the brain of a data center, using intelligent hardware and sensors for precise sensing. Automation reduces routine repetitive tasks, such as manual inspections, and pools expert resources and capabilities. It also incorporates O&M experience into the O&M process, with digitalized O&M covering inspections, maintenance, and emergency drills. Processes and operation guides are provided online so that O&M quality shifts from relying on employees to relying on process management. Through E2E electronic O&M tracing, information that previously couldn’t be quantified can now be. For example, electronic O&M improves inspection execution capabilities, enhances O&M quality, and predicts faults, greatly improving DC reliability.
Edge computing results in more edge DCs, which provide services and process computing near to users. However, that also means that DCs for branches and sites cannot be centrally managed, resulting in slow response to faults and low O&M efficiency.
On the Edge with DCs
Huawei iManager supports E2E network management, including centralized monitoring, unified management, and preventive maintenance for DC infrastructure across multiple sites. GIS positioning technology accelerates fault location and O&M efficiency, while mobile apps simplify management, allowing users to understand the running status of a large number of DCs in different sites. There’s no need for manual O&M, so site visits and costs are reduced and efficiency increased.
iManager also uses AI to intelligently track assets, protect critical data, and optimally allocate resources like power, cooling, space, and bandwidth, maximizing efficiency and cost savings.
Key Opportunity 3: Reducing Energy Consumption and Improving Energy Efficiency
The average PUE – the Power Usage Effectiveness – of a traditional DC is 1.8, far more than the ideal of 1.0. And PUE is set to increase due to the increasing deployment of AI, which demands more from GPU computing servers. The heat generated by accelerated computing servers is several times higher than traditional CPUs.
For heat dissipation, intelligent DC energy saving technology uses sensors to obtain data from key nodes and then optimizes the energy consumption of all systems and devices to lower PUE. Moreover, Huawei’s iCooling is an intelligent thermal management solution that uses deep learning to enable cooling system devices, such as the precision air conditioner, chiller, cooling tower, and water pump, to associate with IT loads and environment variables.
iCooling analyzes historical big data, explores what affects energy consumption, and outputs a PUE prediction model. Its optimization algorithm obtains and delivers optimization parameters to control the system to optimize PUE.
In Huawei Langfang DC, the iCooling@AI solution has cut annual PUE by more than 0.1, lowering it to 1.3 and saving potentially millions of dollars in electricity costs per year.
A Greener Future with Innovative Tech
Huawei focuses on both business development and operational sustainability. According to the 2018 Huawei Global Industry Vision (GIV) report, the average carbon emissions per ICT connection will drop to 15 kg thanks to continuous innovation in ICT and energy saving technologies applied in infrastructure – a whopping 80% lower than 2015 (75 kg). As the foundation for a better connected world, DC infrastructure can cut PUE with innovative tech like AI, electric power electronics, thermodynamics, digital information, telecommunications, and IoT.
Currently, Huawei has deployed over 800 large low-PUE DCs worldwide, lowering energy use and carbon emissions.
Click the link for more information about Huawei’s energy-saving data center solutions. And make sure you subscribe to this blog – Huawei’s 2019 GIV report, which examines 10 key converged tech trends that will shape the future up until 2025, will be released later this month.
Disclaimer: Any views and/or opinions expressed in this post by individual authors or contributors are their personal views and/or opinions and do not necessarily reflect the views and/or opinions of Huawei Technologies.