In order to prevent or lessen the impact of safety incidents, Microsoft is developing an AI system that generates alerts for data center and operations teams by analyzing data from a variety of sources.
Data center build schedule impacts are the goal of a related but complementary system that is currently in development.
Workers who build and maintain the apps, websites, and services that are used by billions of people every day may be at risk in data centers, which house those applications.
While working in a data center, employees may occasionally be required to handle electrical equipment.They may be exposed to chemicals like chlorine, which is used to disinfect water that circulates through computer and server liquid cooling systems.
A chlorine gas leak that occurred in June 2015 at Apple's data center in Maiden, North Carolina, necessitated the hospitalization of five individuals.
Despite the fact that data centers are now safer than they were in the past.But some tech giants say they're looking into how artificial intelligence can be used to avoid safety problems in the future.
A Microsoft representative stated:The initial testing of these initiatives has just begun.Later this year, it is anticipated to begin expanding into our production environments.
Meta also says it is looking into ways AI can predict how its data centers will work in harsh environments that could make working there unsafe.
According to the company, it creates physical models to simulate extreme conditions and feeds this data to AI models that are in charge of maximizing server power consumption, cooling, and airflow.
A spokesperson for Meta stated:Our data centers hold important operational data.Servers, racks, and data halls all have sensors incorporated into them.Every server and organization gadget handles various responsibilities and consumes various measures of force.In data centers, it also makes different amounts of heat and different amounts of airflow.
He added, "Our framework group gathers every one of the information from every server and afterward creates computer based intelligence models that can modify our servers and racks in server farms and send responsibilities to these servers to further develop execution and effectiveness."
Beyond safety, companies like Microsoft, Google, and Meta want to keep their data centers in top shape because outages are expensive and getting more frequent.
In the previous year, a significant outage was reported by a third of data center owners and operators.Six out of ten people stated that the outage cost them more than $1 million.
Around the world, Meta operates more than 20 data centers.including brand-new initiatives that have a total cost of $1.6 billion.
In the meantime, Microsoft owns and operates over 200 data centers.It claims that for the foreseeable future, it is on schedule to construct 50 to 100 new data centers annually.
Another feature that makes AI appealing to businesses is its promise to provide data centers with energy-saving opportunities.In 2018, Google claimed that AI systems developed by its subsidiary DeepMind were able to reduce data center energy consumption by 30 percent on average.
Microsoft and Meta now employ AI for power tuning purposes that are comparable.A method for measuring and mitigating unusual energy and water use events in the data center was also launched by Microsoft at the end of 2021.Telemetry data from mechanical and electrical devices are used for this.
Additionally, the company employs AI-based strategies to identify and resolve data center power issues.Additionally, you use it to optimize server positioning in order to reduce network, cooling, and power consumption.
Energy-regulating AI systems also have the advantage of reducing their impact on the environment.In 2020, data centers were responsible for 0.3% of all carbon dioxide emissions and consumed approximately 1% of the global demand for electricity.
Microsoft previously stated that it intends to use renewable energy to power all of its data centers by 2025.Meta said it would have reached this point in 2020.
0 Comments