- Facebook is doubling down on using AI to support its data centers and network infrastructure.
- The social networking firm has formed a new infrastructure team, Network.AI, according to a recent job posting.
- The team is now hiring as it looks to build out hardware and software that supports Facebook's AI infrastructure.
- The effort underscores the ever-growing importance of AI to Facebook, and in networking more broadly.
- "As we continue to use our advances in AI and ML more and more, it's naturally becoming more intensive on our networking infrastructure," a Facebook spokesperson told Business Insider, in part.
- Visit Business Insider's homepage for more stories.
Facebook has quietly formed a new team focused on incorporating more AI capabilities in its massive network infrastructure and data centers.
The Network.AI unit is part of Facebook's broader infrastructure team and is working to develop hardware and software to build out Facebook's expansive fleet of data centers around the world that power its various products, Business Insider has learned.
A recent Facebook job listing seeking a engineer disclosed the existence of the Network.AI group, describing it as "[spanning] the design and operations of the AI networking [infrastructure]...work. Network Engineers at Facebook are a hybrid software/network engineers who design, build and operate our worldwide data center network. This team owns the complete life-cycle of the AI network in the data center from planning, design, product definition, QA, deployment and monitoring."
The new team illustrates how Facebook is keen to continue innovating as it builds the infrastructure that supports its more than 3 billion users across Facebook, Instagram, WhatsApp, and Messenger around the world, and the ever-increasing importance of AI to its strategy.
In an email, a Facebook spokesperson said: "We develop and deploy AI and machine learning [ML] technologies in our products and services to connect, protect and empower people with new abilities. As we continue to use our advances in AI and ML more and more, it's naturally becoming more intensive on our networking infrastructure. As a result, we're expanding our existing network engineering team to drive our continued growth."
'This team owns the compete lifecycle of the AI network'
The job listing says that the engineer will "(Re)Design, deploy, manage and maintain the Facebook data center networks for AI infrastructure worldwide." It's not clear how many people are currently working on Network.AI.
Here's the key section of the listing, that details some of the hardware and software work the team is focused on (emphasis added):
"The Network.AI group is a new team within Facebook Infrastructure. The charter of the new group spans the design and operations of the AI networking Infra including the network switches and the host side systems, as well as forward-looking projects such as transport evolution. Network Engineers at Facebook are a hybrid software/network engineers who design, build and operate our worldwide data center network. This team owns the complete lifecycle of the AI network in the data center from planning, design, product definition, QA, deployment and monitoring. Simple and scalable network design, automation and data analytics are the keys to meeting our demands. In this role, you will be responsible for conceiving, developing and deploying network software, systems and tools that keep the AI data center network operating at maximum reliability, scalability and efficiency.
AI's growing importance for hyperscalers
Facebook's new team underscores the growing importance of automation and AI in networking, especially for so-called hyperscalers, tech behemoths that operate massive data centers, such as Google, Microsoft and Amazon.
Investing in more AI technology would enable Facebook to have a data center network that "more resilient, scalable and probably more secure," IDC President Crawford Del Prete told Business Insider. "Given the scale the Facebook operates, I suspect that they need a fairly high degree of automation in these areas as the job is too big to get done manually."
More automation would mean substantial cost savings and would also allow Facebook to flag network glitches faster and more efficiently, said Michael Dortch, principal analyst of DortchOnIT.com.
"I can't overemphasize the value of AI-powered site design and infrastructure deployment," he added. "Once that infrastructure is deployed, AI can help to identify and isolate operational problems, and quickly tell technicians precisely what needs fixing or replacing and how best to do it."
Facebook is no stranger to ambitious data center plans
Facebook has taken the lead in industry-redefining data center practices before. In 2011, it spearheaded the Open Compute Project, a plan to design data centers along open source principles — an initiative that is now also followed by companies including Google, Microsoft, Cisco, and Alibaba.
The company has also explored using robotics in its data centers, and now has a fleet of robots patrolling its facilities and a dedicated team working on automating the multi-billion-dollar locations, as Business Insider first reported in February 2020.
Data center sustainability is another area of investment for Facebook. It says its facilities are 80% more water-efficient than a typical data center, and that "in 2020, we will have committed to enough new renewable energy resources to match 100% of the energy used by every data center built by Facebook, and always in the same state or power grid as the data center itself."
Do you work at Facebook? Contact Business Insider reporter Rob Price via encrypted messaging app Signal (+1 650-636-6268), encrypted email (email@example.com), standard email (firstname.lastname@example.org), Telegram/Wickr/WeChat (robaeprice), or Twitter DM (@robaeprice). We can keep sources anonymous. Use a non-work device to reach out. PR pitches by standard email only, please.