Decision Optimization

 View Only

How will AI Drive the Development of Future Data Centers?

By Tim Stone posted Thu June 02, 2022 05:18 AM

  


Artificial Intelligence and Machine Learning continue to make great strides in their development, and they are now having a real impact on data center processes and IT management.

Today, we see Artificial Intelligence and Machine Learning applied to functions that range from energy and cooling to resource control and allocation. To that end, we have seen data- and algorithm-driven technologies deployed in fast failure detection/prediction, root cause investigation, power usage optimization, and help capacity allocation optimization; all in the quest to confirm that data centers are operating as efficiently as probable.

Artificial Intelligence in action

One magnetic current application of Artificial Intelligence in data centers, is the use of assessment robots. The second-generation robots are AI-powered and can operate without human intervention to return any faulty hard drives automatically. The whole disk relief process - including automatic review, defective disk discovering, disk replacement, and setting - can be achieved quickly and smoothly, with the disk replaced within four minutes.

Also, ML-based temperature warning systems have also been deployed in data centers, with hundreds of temperature detectors monitoring information in real-time and using an ensembled graph model to fast and precisely recognize a temperature event due to cooling facility defects. The developed alerts give valuable insights in real-time and provide the data center's functions team with the time required to respond to the fault and avert any possible disasters.

Artificial Intelligence has been around for quite some time now, disrupting companies and sectors with its powers to boost performance and bring in operational usefulness. The data center business is no exception. In today’s time, data carries a huge significance for any community, and what’s equally significant is managing that data effectively. Once screened and crunched, the gathered data proves vital to making strategic company decisions for businesses. Hence, companies invest in advanced automation instruments for data processing and relocate to hyperscale data centers to boost their IT infrastructure. The explosion of data in current years has led hyperscalers to innovate and deploy Artificial Intelligence technologies in their facilities to manage tasks autonomously.

The use of mechanization technologies in data centers is hardly unique. For instance, Google has explained the use of DeepMind Artificial Intelligence for cooling. Nevertheless, businesses are yet to leverage AI/ML to a whole. Elements such as distrust in technology have obstructed many organisations to take a leap toward Artificial Intelligence. While the most known use cases for Artificial Intelligence deployment in data centers are temperature management and predictive maintenance, AI’s possible to enhance the efficiency of a data center infrastructure is far more than widely comprehended. Let’s look at some use cases of Artificial Intelligence in data centers that will transform the industry's future.

Managing Workloads

As data center workloads move upward with increased data, many companies are looking toward Artificial Intelligence to boost efficiency and cut costs. Artificial Intelligence can be used to define the workload movement in a hybrid environment in real-time to the most efficient infrastructure, which could be cloud, on-premise or edge infrastructure. As Artificial Intelligence makes its way into the data center industry, organisations are embracing innovative approaches to manage their data to permit more use of robust Artificial Intelligence methods and analytics.

Mitigating people shortage

Automatic technologies in data centers promise less human intervention in routine and repetitive jobs. It frees up staff from mundane activities such as storage optimisation, cooling allotment, security settings and so on and lets them time to concentrate on more critical problems. It gains greater efficiencies and reduces the risk of human mistakes while handling difficult and diverse workloads. For instance, at our Yotta NM1 data base, in case of a leakage in a chiller pipe, a sensor-enabled Leak Detection System diagnoses the leakage and mitigates the situation in real-time without manual intervention. Upon detecting leakage in the chiller pipeline, the system automatically slows the water flow from an alternate pipeline. All this can be managed without bearing the risk of downtime in the data center. Mechanization is creating a pathway for data centers to go from reactive to preventative, leading to predictive.

Maximising power efficiencies

Power consumption is one of the most essential topics for data centers worldwide. Energy prices surge by at least 10% every year; its expanded use in high-density servers is also not sustainable for the environment. Deploying AI/ML technologies can be a key to improved energy use in data centers. Methods in data centers cause significant heat; traditionally, air conditioners, chillers, water pumps are controlled by Building Management System to keep the temperature in check. Nevertheless, it is not energy efficient. AI-based energy management can help optimize cooling strategies by analysing historical data and creating a Power Usage Effectiveness prediction model, cutting power costs and enhancing efficiency.

Enhancing safety

In a hyperscale data center, where several occasions occur together, it is nearly impossible for humans to watch and alert everyone in case of threat conditions. AI-powered instruments have proven useful in such areas. For instance, image and sound credit capabilities are being used widely to improve the physical security of a data center building; Artifical Intelligence analytics is being used as a video management solution to make sense of data gathered by security cameras. ML techniques are also being leveraged for anomaly detection, where the procedure is trained to determine usual patterns and catch the irregular ones.

Data centers of the future will undoubtedly be more Artificial Intelligence dominant, and almost all positions in the facility will be automated. Though these technologies are only in the hands of a few big hyperscalers and businesses, they will soon trickle down to other data center participants as technology upgrades, trust grows, and costs are cut down. Moreover, given the digital adoption revved by the pandemic, these improvements will be seen in training sooner than later.

Artificial Intelligence and Machine Learning for all?

One intriguing question is what kinds of data and at what kind of scale do companies need to start creating their own AI/ML for data center management? It will depend on each use matter, but monitoring data in the data center would be an ideal place to start when creating AI/ML methods. A model can be prepared with a couple of months of data collection with a selection rate of around a few minutes. Some current data center equipment already delivers structured monitoring data. 

We think that it would be helpful to establish some industry standards for monitoring data structures for major data center gear manufacturers to follow; this, in turn, will accelerate the adoption of AI/ML technologies. In addition, data center operators can consistently install different IoT devices - such as simple temperature detectors or sound or image collectors - to enhance the variety and dimensions of data for more advanced Artificial Intelligence functions.

Given that data centers are filled with mechanical and electrical tools, one concern is whether they are complex environments to enable the creation of data and insights and the following embedding of automated systems. We would like to see the drive embrace some changes to handle this. This includes being more open-minded to major internet technology trends and embracing an overall mindset toward software-controlled programmability and flexibility. 

Artificial Intelligence & data analytics are often useful during the early phases of data center planning and creating with Building Information Modelling and Building Performance Simulation. Nevertheless, not all facilities are new, so many ask whether AI/ML can be applied in existing buildings and whether it is hard to ‘retrofit’ AI/ML into older buildings with existing operations. The good information is that external data collection devices can always be established to retrofit an old facility into an AI-driven condition. This is entirely possible, and we have background of successfully completing this.

Building a technology ecosystem

In fact, it’s on that issue where other technologies can bring value and insights to data center design and control. We believe digital simulation abilities are essential for dedicated data center management, especially for difficult and large-scale scenarios. Often actual tests or trials could not be conducted in these facilities due to their sophistication and the risks of random failures to the existing services. 

An Artificial Intelligence and Machine Learning driven future

The interplay between Artificial Intelligence and DCIM is also worth monitoring, and it will be interesting to see whether the two will join or whether there will always be some break. As it currently stands, we believe Artificial Intelligence technologies will be incorporated into DCIM and become an essential feature for control software in order to provide enhanced functionality and process reliability.

Given the real and vital service data means perform – and how key they are to megatrends, such as driving one’s infrastructure to the cloud — they must always adopt the latest technologies and procedures to continue providing the service their clients demand. That’s why I’m sure that data centers will consistently be earlier adopters of many technologies that later filter through to the rest of our everyday lives.

0 comments
6 views

Permalink