As the amount of data increases, so too does the amount of computing needed to manage all that data. In fact, Intel estimates that up to 20% of a data center’s energy consumption is simply due to software inefficiency. We have reached the point where we now need to leverage software to optimize workloads, to maintain performance and lower power.
This is precisely the conversation I had recently with Jen Huffstetler, chief product sustainability officer, Intel, who says the future of sustainable computing is largely in the hands of AI (artificial intelligence) software developers.
“The power of AI is looking for trends and looking for patterns that your eye might not easily see,” Huffstetler says. “With their forecasting, after analyzing these trends, and looking at upcoming workload needs, it can actually fine tune the system-level solution to even change individual component level power like a processor and that can be increased, or it can be decreased automatically. When you do that, you can conserve electricity.”
Telco Leverages AI
To demonstrate the potential this technology can offer, Huffstetler points to three examples of how AI is being adopted widely across many companies and in all industries. The first case study is telco KDDI.
“They have implemented a proof-of-concept program that utilizes AI to predict the network traffic, and this reduces their electricity by even going to the extent of idling CPUs that are not needed and then saving up to 35% of their energy consumption,” she explains.
She then goes on to say we can take that same concept of predicting network traffic and envision it in other industries. As another instance, this same technology could be used for idle time for tools in a factory. Technology can certainly help with predictive maintenance and predictive down powering of tools.
AI in Healthcare
As another, very different example, Huffstetler points to what Siemens Healthineers is doing with radiation therapy, explaining that one of the risks of radiation therapy is the damage that can happen to the organs or the tissue surrounding the target area.
“Something they do is called contouring the organs so they can target radiation therapy,” she says. “This is tedious and time consuming and when done by humans it can be prone to error. It is not as accurate.”
For a long time, AI-based auto-contouring held the promise of providing more consistent results faster, but it was still limited in its ability to have an impact.
“With our latest Fourth Gen Intel Xeon processors, image generation is 35 times faster and uses 20% less power,” says Huffstetler. “Not only are you improving healthcare outcomes for the patient, but you are actually lowering healthcare costs at the same time by reducing the manual labor and the time needed for that contouring and you are improving the outcomes for the environment.”
AI and Massive Data
Another example Huffstetler gives is SK Telecom, which is one of the largest mobile operators in South Korea. The organization manages 400,000 cell towers, has 27 million subscribers, and the network handles 1.4 million records every second.
“That is a massive amount of data. The SK Telecom and Intel engineers looked at this problem and built an end-to-end network AI pipeline,” says Huffstetler.
“What is most impressive about this—and AI is the talk of the day—is in the tests that were conducted by SK, this AI pipeline outperformed their legacy GPU-based implementation by up to four times and up to six times for the deep learning training and inference, respectively for those,” she says.
She goes on to explain this now enables SK Telecom to more quickly forecast and detect that network degradation and the abnormal changes in the quality so they can take proactive action to deliver 5G service quality.
At the end of the day, AI has the ability to improve energy consumption, and these are simply a few examples. Green software can help reduce the carbon emissions of software by using less physical resources and less energy more intelligently.
With all this in hand, customers can ultimately have a more holistic view to make better decisions—and the data center operator will be leading the charge.
“The power of energy consumption both in the data center and in the network, they really take that move to digital optimization,” says Huffstetler. “In the hands of the software developers, the AI data scientists, the combination together to optimize your workloads with your hardware is really what we are going to need for the future to lower your energy consumption today and more tightly couple them over time.”
Want to tweet about this article? Use hashtags #IoT #sustainability #AI #5G #cloud #edge #futureofwork #digitaltransformation #green #ecosystem #environmental #circularworld #IntelInfluencer