Data centers, which power the apps, websites and services that billions of people use every day, can be dangerous places for the workers who build and maintain them. Employees sometimes need to maintain a data center’s electrical equipment while it is live. And they can be exposed to chemicals such as chlorine, which is used as a sterilizer for the water circulating through liquid cooling systems for computers and servers. In June 2015, five people had to go to a Hopital following a chlorine gas leak at an Apple data center in Maiden, North Carolina.
Data centers are more secure than they used to be. But looking for forward-looking solutions, some tech giants say they are exploring how AI can be applied to prevent security problems. For example, Microsoft is developing an AI system that will analyze data from various sources and generate alerts for data center construction and operations teams to “prevent or reduce the impact of security incidents.” An additional but related system, also under development, seeks to detect and predict implications for data center construction schedules.
“These initiatives are both in early testing and are expected to begin expanding to our production environments later this year,” a Microsoft spokesperson told TechCrunch via email.
Meta also claims to investigate how AI can anticipate how its data centers will operate under “extreme environmental conditions” that can lead to unsafe work environments. The company says it has developed physical models to simulate extreme conditions and introduced this data into the AI models responsible for optimizing power consumption, cooling and airflow across its servers.
“We have significant operational data from our data centers, in some high-frequency areas with sensors built into servers, racks, and in our data halls,” a Meta spokesperson told TechCrunch. “Each server and network device, which takes on different workloads, consumes different amounts of power, generates different amounts of heat and provides different amounts of airflow in the data centers. U.S [infrastructure] team collects all data from each server and then develops AI models that can map our servers and racks in the data centers and send workloads to these servers to optimize [for] performance and efficiency.”
In addition to security, companies also have other motivations to ensure that data centers remain in top condition. Outages are expensive – and are becoming more common. According to a 2020 questionnaire by the IT Uptime Institute, an IT consultancy, a third of data center owners and operators admit they have experienced a major outage in the past 12 months. One in six claimed their outage cost them more than $1 million, compared to one in ten in 2019.
Meta operates more than 20 data centers around the world, including new projects in Texas and Missouri estimated together cost $1.6 billion. Microsoft now operates more than 200 data centers, and say it is on track to build between 50 and 100 new data centers each year for the foreseeable future.
AI also promises to find opportunities for energy and therefore cost savings in the data center that normally fly under the radar, another attractive aspect for businesses. In 2018, Google claimed that AI systems developed by its DeepMind subsidiary were able to deliver an average of 30% energy savings compared to the historical energy consumption of its data centers.
Reaching for comment, DeepMind said it had no updates to share after the initial announcement. IBM and Amazon did not respond to questions. But both Meta and Microsoft say they are now using AI for similar energy tuning purposes.
Microsoft launched AI “anomaly detection methods” by the end of 2021 to detect unusual power and water consumption events within the data center, using telemetry data of electrical and mechanical devices. The company also uses AI-based approaches to identifying and resolving power meter issues in the data centerand to identify ideal places to place servers to minimize waste of power, network and cooling capacity.
Meta, for its part, says it has used reinforcement learning to reduce the amount of air it pumps into data centers for cooling purposes. (At a high level, amplification learning is a type of AI system that learns to solve a trial and error problem.) Most of the company’s data centers use outdoor air and evaporative cooling systems, making optimizing airflow a high priority. has.
The reduced carbon footprint is an added benefit of energy-regulating AI systems. Data centers consumed about 1% of global electricity demand and contributed to 0.3% of all CO2 emissions in 2020, according to to a report from the Environmental Research Bureau. And the typical data center uses 3 million to 5 million gallons of water per day, the same amount of water as a city of 30,000 to 50,000 people.
Microsoft has previously said it plans to run all of its data centers on 100% renewable energy by 2025. Meta claimed to have accomplished the feat in 2020.