Artificial system

Microsoft and Meta start testing artificial intelligence to help with data center operations


Data center workers who power billions of apps, websites and services are at risk. Data center electrical equipment must sometimes be serviced by data center employees. Computer and server liquid cooling systems are often disinfected with chlorine. A chlorine gas leak at Apple’s data center in Maiden, North Carolina in June 2015 resulted in the hospitalization of many people.

The level of security provided by data centers is unmatched. Leading technology companies around the world are looking for ways to reduce the likelihood of accidents through the use of artificial intelligence (AI). Microsoft is developing an artificial intelligence system capable of identifying and reporting any security issues that arise during the design and operation of data centers. Data center performance can be predicted using this type of technology.

By the end of the year, both programs should be ready for use.

According to Meta, they are trying new approaches. Under “harsh environmental conditions,” AI can predict the performance of its data centers. As a result of this research, the company developed physical models of hostile environments and used the data from these models to improve server power consumption.

Various sensors are installed in Meta’s data centers to capture operational data from racks, servers and hallways. Each server or network device generates a different amount of electricity, heat, and airflow in a data center. Data from servers and AI algorithms help our infrastructure team organize servers, racks, and workloads in the most efficient way possible.

In addition to providing security, data centers also provide a variety of other functions. Having frequent breakdowns costs money. A third of data center owners and operators have experienced a major outage in the past year, according to research by 2020, an IT consulting organization. A 10% increase in 2019 to 60% of customers saying their outage cost over $1 million is encouraging.

$1.6 billion of data centers in Texas and Missouri are part of Meta’s global network of more than 20 data centers. The number of data centers managed by Microsoft has grown to more than 200, with 50 to 100 new data centers being built each year.

Artificial intelligence (AI) can uncover previously unrecognized energy and cost saving opportunities in the data center. According to Google, DeepMind’s AI reduced data center energy consumption by 30% in 2018.

As for the first version, DeepMind did not say anything. Neither Amazon nor IBM took a look. Optimizing energy usage is a claim made by both Meta and Microsoft.

To monitor and reduce electricity and water consumption anomalies in data centers, Microsoft developed “anomaly detection algorithms” in late 2021. Data center energy meters are scanned artificial intelligence (AI) help to avoid waste of energy, network and cooling capacity.

To reduce the amount of cooled air it pumps into data centers, Meta uses reinforcement learning. In artificial intelligence, a process known as reinforcement learning, mistakes are made and subsequently lessons are learned. Enterprise data centers rely heavily on outside air and evaporative cooling, so airflow management is critical.

The environmental effect of AI energy management systems is reduced. This report summarizes the findings of the EIA. By 2020, data centers will represent 1% of global electricity consumption and 3% of global CO2 emissions. A typical data center uses as much water per day as a city of 30,000 people.

By 2025, Microsoft expects all of its data centers to be powered entirely by renewable energy. According to Meta, this will be the case in 2020.