PARTNER CONTENT: Integrated AI-based technology enables networks to adapt autonomously which helps operators optimize their service delivery to consumers and enterprises. As communication networks increase in complexity, operators seek to simplify operation, maintenance activities and reduce operating costs. Network automation and AI could help operators achieve this objective. Automation is not a new concept: it has been around under different names for over 10 years. It is integral to digital transformation that’s reshaping industries by increasing productivity and efficiency. Yet, network automation has seen limited application for different reasons. One of the reasons is that the process of automation was sufficiently disruptive to operators’ processes without a clear value proposition. This may have been the case when networks were simple, but today’s networks are an order of magnitude more complex than their predecessors. For this reason, I had a close look at automation when applied to power consumption which is a major challenge facing operators as they upgrade to 5G.
The Power Challenge in Perspective
There’s one open secret about 5G: it consumes a lot of power to provide the gigabit speed it promises. A 5G site could consume double the power of an LTE site. Operators seeking to deploy 5G find themselves in a bind, especially in controlling operational expenditures that chip at their margins.
To illustrate, I like to use Vodafone as an example for a perspective on power consumption in telco networks. Their average site power consumption is 2,560 W. Base stations account for 66% of the total power used in a year. In 2019, base stations used 3,684 GWh [1]. In comparison, the rest of their network including switching and data centers used less than half that amount: 1,571 GWh. If the cost of power is $0.1/kWh, Vodafone’s annual power bill for base stations alone would be $368 million. If they decide to upgrade all the sites to 5G, we could add 60% additional power draw, or about 1,500 kW per site, for additional $221 million. This is $1,350 per site per year in additional opex.
[2]
Power consumption in wireless base stations. [Source: Enable Autonomous Driving Network [3], Huawei, 2019]
Complexity and Optimization across Layers
Managing cell site power consumption is an old problem that becomes more challenging with the introduction of 5G. Power management was part of the Self-Organizing Network (SON) [4] concepts which dates back more than 10 years. Back then, LTE networks based on a single frequency carrier. Today, Docomo in Japan, for example, runs 4-carrier aggregation which includes FDD and TDD modes spanning 800 MHz, 1700 MHz, 2100 MHz and 3400 MHz bands. 5G will introduces yet another layer of spectrum. Each of these bands has its own performance characteristics for coverage and capacity. One can optimize performance across one layer, but as the layers increase it becomes necessary to optimize across layers.
As network complexity increases, taking that holistic network approach to optimization becomes ever more important to ensure that any local gain is also translated into global improvement in performance. The absence of such holistic approach could be not too different from the Whac-A-Mole game where improvement in one area degrades another. This is where automation and application of AI technologies can play a leading role as otherwise, the task would be impossible.
Applying Automation and Predictive Analytics
Network automation comprises different aspects of the network lifecycle including provisioning and configuration, maintenance and optimization. Power optimization is one of many use cases where one could apply AI techniques and automation to minimize power consumption. Service providers would need the system starting with open interfaces and APIs to gather and streamline the appropriate data, the tools to develop and train the predictive algorithms, the inference engine to continually monitor performance and the overarching engine to send optimization instructions back to the network.
Service providers have been adapting to increase complexity of networks by using different tools and process to optimize networks. But we are yet to reach a stage where networks have the autonomy to make decisions, independent of humans, and configure their operation depending on conditions such as traffic load. This would be an important step to deal with ever increasing network complexity.
Automating Power Optimization
The main question of applying automation to power optimization is how can one realize high energy efficiency across multiple radio technologies while maximizing subscriber experience. Answering this requires coordinated power saving mechanisms through precise traffic prediction.
User traffic which drives base station activity is relatively predictable parameter that varies within known patterns over the course of a day, week and month. This makes it possible to optimize power consumption not only for a certain site, but across an entire network by taking a wholistic view of underlying utilization patterns.
As an example, Huawei’s PowerStar is an AI-based, three-level energy saving solution for legacy wireless network that includes network level, site level and equipment level optimization. At the network-level, an energy-saving AI Engine called Power Turbo allows cell sites to turn off certain carriers according to predicted traffic patterns within priority rules. The engine bases its decisions on a large data set that could include up to 200 parameters. It offers different granularity of power optimization where the engine could turn off a radio technology (e.g. 4G), a frequency carrier, a power amplifier, a time slot, or a symbol. The system shows between 10-15% reductions in power consumption, and could save even more power in future 5G networks.
The solution does not require hardware replacement or upgrade and will not cause network KPI deterioration. It does however contribute to industry standards for fulfilling the social responsibility of energy saving and emission reduction.
Huawei PowerStar solution helps multiple operators save energy for wireless network.
Service Provider
Power Savings
kWh/Day/Site Saved
China Mobile
12%
1.89
MTN
12%
6
Inwi
17.9%
8.5
Axiata
11%
7
Source: Huawei
An operator with 2,500 W 4G cell site, would realize savings are on the order of 300 W, or $263/year. With 5G at additional 1,500 W, the savings are close to 500 W, or $420/year. If this does not sound large, multiply the savings by tens of thousands for each service provider. Or even better, considering there are over 7 million base stations in the world, the power savings could reach 2 GW, which is about what a medium size nuclear reactor generates!
Metrics to Gauge Automation
To deliver the ‘extreme broadband’ speeds, 5G uses power-hungry massive 32x or 64x MIMO antennas and 100 MHz frequency carriers. But 5G does not need to be active at all times. Rather, one can optimize cell site power consumption by selectively powering systems to meet underlying traffic demand. This brings to my mind the need for a metric that measures the benefits of automation. Would it not be appropriate to use a metric such as bits/Joule and make it mainstream similar to $/Mbps? This can help frame the benefits for automation in the proper perspective for everyone.
Concluding Thoughts
The application of AI and automation technologies to power optimization highlights the benefits for a specific use case. Automation could be applied for a host of different use cases, such as optimizing VoLTE or MIMO beamforming performance. This requires a well-integrated infrastructure with open interfaces and APIs necessary to enable automation. Corner cases would need to be identified and addressed. Operators can start with the features that provide the maximum benefit and least risk, and then expand to automate additional use cases.
Frank Rayal
Xona Partners
[1] https://www.vodafone.com/content/dam/vodcom/sustainability/pdfs/sustainablebusiness2019.pdf
[2] https://www.mobileworldlive.com/wp-content/uploads/2020/05/h1.png
[3] https://carrier.huawei.com/~/media/CNBGV2/download/adn/Huawei-NAIE-White-Paper.pdf
[4] https://frankrayal.com/2014/05/19/son-progress-report-a-lot-still-to-be-done/
![\"IT電腦補習](\"http://itteacheritfreelance.hk/test/wordpress/wp-content/uploads/2016/05/logo2-2.png\"){kind=logo}
www.itsec.hk
www.itsec.vip
www.itseceu.uk
Leave a Reply