Amid a global economic slowdown, companies are doing everything they can to cut costs. Energy consumption for data centers is topmost on everyone’s minds across the globe. So, optimising data center costs amid rising demand for digital services is going to be crucial for 2020.

A data center building is a place where there are tens of components like servers, cooling gear, storage devices, workloads, and networking, among others. Running of the data center is controlled by a combined functioning of all data center components, allowing several designs to coordinate together.

About 10% of data center operating expenditure is electricity, with Gartner expecting that it will rise to about 15% by 2021. This, in turn, needs vital means to keep them cool. In the past, this has driven businesses to make them at sea or picking locations with lower temperatures, like in Europe. At the same time, AI can also help decrease the amount of power data centers use for cooling, regardless of the location.

Why This Is Important

Energy consumption by data centers may become a costly affair for the tech industry. Along with these data centers, companies also need to hire skilled specialists to maintain and monitor data centers. Operating data centers and hiring staff can be expensive for every organisation. 

Furthermore, supervising and managing teams is an extra task. Therefore, organisations are continually looking for better options for conventional manual systems. As an option, organisations can deploy AI in the data center to autonomously manage multiple jobs, like server optimisation and equipment monitoring.

The expanding interest for superior performance computing over the domain is driving an expansion in dense servers, and the utilisation of GPUs, just as specialised AI chips. These frameworks produce more heat than conventional CPUs, making heat dispersal an undeniably crucial issue in data centers, which can drive up electricity bills.  

Even though there are cooling systems, nonetheless, they cannot run at the ideal effectiveness point. New AI-based systems leverage deep learning to figure out how to draw the fitting relationships between different types of cooling gear with IT workloads and condition factors. 

AI frameworks accomplish this by breaking down a huge measure of historical data and their effect on energy utilisation to produce predictive modeling, using specific parameters that are transmitted to different control frameworks. By gathering data from the power supply framework, AI-fueled systems can foresee approaching gadgets and part failures to caution operations and maintenance (O&M) staff early.

What Google Did

In this regard, Google has also figured out the best way to alter cooling frameworks — fans, ventilation, and other hardware — to bring down power utilisation. Google’s system recently made proposals to its data center administrators, leading to around 40% cost cuts in those cooling frameworks. 

Now, Google says it has delivered control to the algorithm, which is running cooling at many of its data centers all by itself. The project shows the capability of AI to handle infrastructure—and reveals how advanced AI systems can operate in tandem with humans. Even though the algorithm runs autonomously, a staff member manages it and can intercede if it appears to be doing something unsafe.

Developed in collaboration with DeepMind, Google’s algorithm employs a method known as reinforcement learning, which learns by trial and error. This led to AlphaGo, the DeepMind program which defeated human professionals of the board game Go.

DeepMind served its new algorithm data collected from Google data centers and allowed it to discover what cooling configurations would decrease energy expenditure. The project could produce millions of dollars in electricity savings and may help the company reduce its carbon footprint, according to Google.

The post Why Data Centers Should Utilise AI To Optimise Power Costs appeared first on Analytics India Magazine.