Data centers use an estimated 200 terawatt hours (TWh) of electricity annually, equal to roughly 50% of all electricity currently used for all global transport, and a worse-case-scenario model troublingly predicts that data centers alone could account for roughly 8% of global electricity consumption by 2030.
This may seem shocking to some. But what may be equally shocking (to some) is that edge computing and machine learning will actually play a key role in reducing the carbon footprint of data centers. Through edge computing and machine learning, we can actually significantly reduce the amount of time—and hence, power—data centers need to use to process data.
Let’s explore why this is so important and how and why it will happen.
5G And The Data Explosion
While our new era of 5G communications will drive unprecedented new opportunities, it’s also expected to drive data volumes—and the related amount of electricity required to process that data—to unprecedented levels.
Data centers already have the same carbon footprint as the aviation industry. While data center energy consumption has been held relatively level for the past few years via super-efficient computing architectures (“hyperscaling”), advanced cooling systems and increased use of renewable energy, much of these efficiencies have already been realized.
How can data center energy consumption possibly be held in check in our impending new information age, in which data processing volumes are expected to explode?
The (Rapid) Rise Of Edge Computing And Machine Learning
Edge computing—the moving of processing power closer to data sources at the edge of the network—has rapidly evolved into a vital enabling technology for our impending new 5G-powered information age.
Telcos are increasingly working edge computing strategies into their 5G deployments to ensure they will keep up with new applications and devices requiring real-time processing, capitalizing on real-time data before it goes stale. IDC recently estimated the global edge computing market will continue its dramatic growth and exceed $275 billion in 2025—an 80% increase from 2021.
Combined with edge computing, machine learning is also a must-have technology to reap the full business potential of 5G. New applications at the edge will constantly capture and generate massive amounts of data, often requiring real-time responses. Using the machine learning model at the edge will be essential for such intelligent decisioning for continuously improving on its decisions and actions.
Here are some of ways 5G is driving up data volumes in specific sectors and how edge computing and machine learning are beings used to handle it.
• Healthcare: “On-device” data analysis in healthcare is absolutely vital whenever there is an immediate need to make decisions based on patient data. This becomes even more critical if health providers are in a remote location and have limited connectivity to transfer patient data to a server. As hardware and machine learning methods become more sophisticated, more patient data can be collected by devices located at the edge of the network, including such key metrics as brain activity and heart rate.
• Visualization Algorithms: Convolutional neural networks are a type of artificial neural network designed to analyze and recognize visual imagery. Key use cases include facial recognition to identify known “bad actors” or to detect activity within sensitive regions and borders; and quality control, including applying deep learning algorithms to recognize product defects during automated assembly processes.
• Internet Of Things (IoT): The video cameras just mentioned above certainly also fall into the category of IoT, as do a vast array of sensors and monitors designed to detect and recognize motion, sound, gestures and/or virtually any measurable, observable activity. What all IoT devices share in common is they require on-demand inferences in order to trigger actions. Any delay in action—even a “round trip delay” from API calls to a remote server—can adversely impact the functionality of the system. 5G can connect billions of sensors to the edge data center for efficient, real-time processing that saves data center energy.
The Importance Of Aggregating Edge Data
Only use what’s needed. It’s a common refrain, but one that’s becoming increasingly important in the world of data processing. Real-time edge applications are capturing, accumulating and producing more data than ever before—and at a faster rate. How and where this data is stored can either bolster or diminish the speed and efficiency of the necessary decisioning processes and eventually the machine learning iterations.
Relying on traditional central data centers for storing and retrieving raw, edge-generated data will prove detrimental in many ways.
1. Moving huge volumes of raw data from the edge to the data center consumes network bandwidth.
2. Storage resources must retain the data until accessed by machine learning process(es).
3. The more data that machine learning models have to sift through to be effective, the more inefficient (and slow) your systems and applications will be.
Enabling intelligent decisions at the edge is vital for enabling the uninterrupted functionality of real-time industry-specific applications themselves, which generally demand ultra-low latency.
But doing this effectively requires certain processes, one of which is data aggregation.
To take full advantage of the compute resources at the edge, enterprises must be able to locally store, use and, most importantly, aggregate data, meaning discard any unnecessary data and send only what’s necessary from the edge to the data center. By doing this, enterprises can extract the full value of their data in less time, enabling quicker decisions and the most efficient possible use of central data center power, thus stretching both the IT and the energy budget.
Conclusions
The active use of machine learning with edge computing is a vital win-win for fully realizing the breakthrough potentials of 5G without increasing power consumption on the part of data centers.
Working together, machine learning and edge computing will significantly reduce data center energy usage by ensuring data processing occurs as close to the end user as possible so that less data is transmitted over shorter distances.
Less work. Less electricity used. That’s the sustainability power of edge computing and machine learning, and we’re just getting started.
About USDC Technology
Universal Smart Data Center Technology was established by people with a broad vision of Information and Communication Technology. They are a professional and leading company in technology construction for Smart Data Center services in Vietnam and the region. Its commitment to take total ownership of projects has resulted in an enviable client portfolio, featuring some of the most renowned brands. Its mission to deliver society the most optimal products and services by applying the latest technologies.
USDC Technology Data Center
Launched on December, 2020, USDC Technology Data Center was built on class Tier III. Located on Sai Gon Hitech Park, District 9, Ho Chi Minh City, Vietnam (so called Vietnam’s Silicon Valley). The data center is available connection with all large networks, locate in strategic site to cover the East of Ho Chi Minh city. Our world-class data centers provide full-scale services. At USDC Technology, nothing is of greater importance to us than keeping your applications online and your data secure.
News Contact:
Universal Smart Data Center Technology
Phone: (+84) 28 73080708
Email: info@usdc.vn
