Artificial intelligence has reached data centers and automated systems have been installed in server halls around the world to ensure equipment maintenance and smooth service execution. As data centers evolve to meet the demands of new technologies such as cloud computing and 5G and edge networks, further automation may increase over the next few years, but operators continue to lack AI skills. It’s becoming more and more pessimistic to prove the silver bullet.
As the demand for data center space soars as a result of extensive digital transformation and many companies are switching to the cloud, improving data center efficiency can maximize resource utilization, cloud computing and more. It is the key to maintaining the carbon footprint of other digital systems. To a minimum. At the same time, the sector faces a serious shortage of skilled workers, which could hinder the expansion of centers around the world.
AI can play a role in addressing both issues, and by 2025, half of cloud data centers will have advanced robots with AI and machine learning (ML) capabilities. As a result, we anticipate a 30% improvement in operations. efficiency.
Gartner Research Vice President Sidnag said: “The risk of doing nothing to address these shortcomings is significant for enterprises. Data center operations allow organizations to move more diverse workloads to the cloud, which combines additional technologies such as edge and 5G. It just gets more complicated as it becomes a platform for use. “
How is AI deployed in the data center?
AI is commonly found in software running in data centers, but automated systems also play a greater role in maintaining the center itself. “We are beginning to see growth in automation in terms of solving IT infrastructure problems,” said Paul Bevan, Bloor’s IT infrastructure research director. “As an example, the performance management system may indicate a problem with the storage device because the database backup was not performed properly and the cache needs to be cleared. This can be automated, and It’s a very simple task that is automated. ”
However, Bevan says this improvement in automation is not seen in other areas of data center management. “When we return to areas controlled by data center infrastructure management (DCIM) or building management systems, the level of automation is much lower,” he says.
The closer you are to the application, the more automated it will be and the faster the application will move. The rest is catching up with some work to do.
Paul Bevan, Blower
This is partly because many older data centers do not have the necessary interfaces to connect to modern AI systems. “There is skepticism about AI among some operations managers at that level, which means it’s taking some time to automate physical buildings,” he added. increase. This is because technology failures can lead to costly outages.
“Companies like Schneider Electric are doing a lot of work on heating and power automation, but the actual adoption is pretty slow,” continues Bevan. “It’s not cheap, so part of it is related to the investment cycle. But generally speaking, the closer you are to the application, the faster it’s automated and the faster the application moves. The rest should be done. Something is catching up. ”
Will more AI help improve the data center?
Despite staff reservations, the DCIM system has the potential to become the next frontier of AI in the data center, primarily due to the impetus for reducing carbon emissions. “Data centers are energy-intensive facilities that account for about 1% of global electricity consumption,” said Dr. Moises Levy, chief analyst for data center physical infrastructure at Omdia.
Tech Monitor Earlier reported on significant carbon dioxide emissions in cloud computing data centersPublic cloud providers are trying to mitigate by using more renewable energy or buying renewable energy to offset their emissions. Dr. Levy states that AI can be used to reduce power consumption by making servers operate more efficiently. “Data center energy consumption is all about workloads,” he says. “The server consumes energy and dissipates heat. Both depend on the workload being processed. The heat management system cools the facility by extracting heat. Therefore, the thermal behavior is processed. It is affected by the workload. ”
AI helps us understand, measure, and predict workloads in real time, so we can allocate resources accordingly, says Dr. Levy. “To be energy efficient, resource-based workload processing can be shifted over time and even across locations,” he adds. “We are living in an era where AI-enabled tools can help automate workload management and learn information directly from data without pre-determined equations.”
Google has already made great strides in this regard through its Deepmind AI division. We used an algorithm to analyze historical data and reduce power consumption in the company’s data center by 40%...
Stefanie Williams, Senior Research Analyst at 451 Research, is starting to catch up with other companies and making significant investments in this area. “DCIM’s idea is to move from simply monitoring equipment and analyzing usage to actually controlling the infrastructure to cool the facility efficiently based on those analyzes.” She says. “This is done by adjusting a variety of equipment to reduce machine wear, extend machine life, effectively deal with hotspots, and reduce the risk of human error in the facility to improve cooling efficiency. It involves enhancing. ”Williams adds that the same principles could apply to AI-driven power distribution and space planning.
Can AI solve the data center staffing crisis?
As the demand for data centers grows, so does the staffing requirement. According to a survey by the Uptime Institute The number of full-time staff required for data centers around the world will increase from 2 million in 2019 to 2.3 million by 2025... However, another Uptime Institute survey of data center operators found that 47% struggled to attract talent to play an open role, with 32% of respondents retaining existing talent. States that it is difficult. One-fifth (20%) said this was due to workers being poached by rivals, and 12% said staff were left in other industries.
“Stuffing is one of the biggest issues,” said Ed Galvin, CEO of DCByte, a research and analytics company that covers the data center market. “The industry is booming, but there is no clear career path for college people who want to enter this field. Therefore, they may follow the roots of computer science and engineering, but this huge No one really knows that there is an area of infrastructure. This is a great opportunity. ”
Many people from the first wave of data centers 20 to 25 years ago are approaching retirement, which presents a real problem.
Ed Galvin, DC Byte
The lack of a route to this industry has caused another problem. It is the aging of the workforce. “Companies are preying on each other’s staff,” Galvin said. “That is, everyone is getting older, from the first wave of data centers 20 to 25 years ago to the point where many are approaching retirement. This is a real problem.”
Is AI the answer? Data center operators suspect that they will be bailed out in the short term. Fifty percent of respondents to the Uptime Institute’s survey say they don’t expect AI to reduce staff levels from 42% for at least the next five years. When I had the same question in 2019.
Galvin believes this reflects the fact that getting humans out of the loop is not easy given the expertise of people working in the industry. “Data centers are usually not that crowded,” he says. “A large 30 MW data center can be operated by a team of 10 to 15 people, but everyone is very skilled. Therefore, there is some automation because there are applications that can be aggregated and outsourced to the machine. It can be useful, but in reality there are many things you can do [without a human] That is why people working in the industry can order high salaries. ”
In the long run, robotics may provide the answer. Earlier this year, data center provider NTT developed a robot that could perform maintenance tasks on data center servers.
Galvin said the need for automation is increasing due to the trend towards smaller, more distributed data centers known as “dark sites” for applications such as edge devices and 5G networks. I am. “We can’t place people economically,” he says. “Therefore, it can be either remote controlled or self-controlled. In these dark places, it is a patrol center equipped with an infrared camera, a” roaming robot “that can be discovered and found with a little machine learning. Will be seen. Flag the hotspot. This is usually the first sign of a short circuit in the equipment. You can then pass it on to a human operator who can check it manually. ”
For now, Bloor Research’s Bevan wants to mitigate the effects of lack of skills in other ways, rather than relying on relatively unproven AI. “We are still quite early in the AI and automation cycle in this area,” he says. “Many people still see it and say,’I’m not going to go first. I want to see what the risks are and how those risks are mitigated.'” ”
But CIOs and other tech leaders need to be aware of the changing faces of data centers, he says. “Whole piece of automation [in data centres] “It’s getting very hot,” he says. “There is still the danger of hype creeping in, but given the growth pressure and the changes that have taken place in the real-time digital economy since Covid-19, we need to study. These issues with automation and AI and they are now. How does it apply to your business? Otherwise, you risk being left behind. ”
Matthew Gooding Tech Monitor..
https://techmonitor.ai/technology/data-centre/data-centre-ai-artificial-intelligence-cloud-computing Data Center AI Doesn’t Solve Sector Skill Lack