First AI-enhanced 5G Networks Unveiled through NTT DOCOMO and Nvidia Collaboration

Highlights:

• In Japan, the recently deployed network features a robust 5G virtual RAN (vRAN) developed by Fujitsu, utilizing Nvidia’s Aerial 5G vRAN stack and integrated accelerators, ensuring exceptional performance.
• According to NTT DOCOMO, this integration has the potential to slash overall costs by up to 30% and elevate network design utilization by an impressive 50%.

NTT DOCOMO, a leading Japanese telecom company, achieved a significant milestone at MWC Las Vegas this week by commercially implementing the inaugural 5G network powered by GPU-accelerated wireless tech.

The launch resulted from a 5G Open RAN Ecosystem Experience (OREX) service brand partnership. This collaborative effort involved 13 tech firms, including notable names like Nvidia Corp., Fujitsu Global, and Wind River Systems Inc.

These companies collectively tackle challenges tied to Open RAN (Open Radio Access Network), encompassing integration, interoperability, performance, and operational complexity to enhance the network infrastructure. This advancement is crucial as Open RAN facilitates seamless collaboration among multiple vendors, dismantling the monopoly of traditional single-vendor configurations. This marks a shift towards collaborative deployment of wireless networks, where various companies can contribute and collaborate instead of a single entity having complete control. The prevalent vendor lock-in within the telecom industry has historically led to heightened operational expenses and arguably stifled innovation.

In Japan, the recently deployed network features a robust 5G virtual RAN (vRAN) developed by Fujitsu, utilizing Nvidia’s Aerial 5G vRAN stack and integrated accelerators, ensuring exceptional performance. This network is entirely software and cloud-defined, allowing operators to dynamically align resources with mobile users’ data traffic—a significant advancement from older fixed network configurations.

Furthermore, the fusion of artificial intelligence and machine learning enables automated network slicing, elevating service quality, optimizing resource control, and enhancing the overall user experience. Upcoming systems can autonomously adapt in real-time using AI, responding to dynamic user traffic conditions. According to Masaki Taniguchi, Senior Vice President and Head of Fujitsu’s Mobile Systems Business Unit, these advancements, particularly in task automation, will be implemented through software updates, as mentioned in a pre-briefing announcing the 5G network launch.

Integrating accelerators such as GPUs into wireless networks offers significant advantages. According to NTT DOCOMO, this integration has the potential to slash overall costs by up to 30% and elevate network design utilization by an impressive 50%. Moreover, this integration can enhance power consumption and efficiency, akin to engineering a car that utilizes only half the fuel while maintaining high-performance levels.

Taniguchi said, “We are not only continuously innovating this breed of software but also building technology for the entire network using high-efficiency wireless products. This will allow us to reduce power consumption by more than 50% compared with current RAN products.”

In Nvidia’s case, their Aerial platform is meticulously crafted for software-defined, GPU-accelerated, and cloud-native 5G networks. Its flexibility allows GPU acceleration of the 5G stack, which can be used only for physical layer 1 or the whole stack over time. Ronnie Vasishta, senior vice president of telecom at Nvidia, says that more functions can be added with the help of tech partners like Fujitsu. This will make it possible for the hardware to run both vRAN applications and AI applications.

Vasishta stated, “We now have a live network completely software-defined with a GPU, enabling telecommunications acceleration on RAN. This is a significant milestone. We have also demonstrated AI applications, such as intelligent video analytics, that can run on exactly the same hardware. So, in addition to the network, there are benefits of having a GPU within the enterprise.”

Wind River is the other key partner, and it helps set up the OREX 5G vRAN by providing a cloud platform. Wind River Studio is a cloud, automation, and analytics solution based on open source that can be distributed. Operators use the solution to set up and manage their 5G networks worldwide.

As part of a multivendor partnership, the deployment in Japan represents a crucial first step in demonstrating the commercial viability of these key technologies. NTT DOCOMO is packaging them under the OREX brand and offering three packages to network operators: OREX RAN for virtualized radio base stations and wireless equipment, OREX SMO for software facilitating autonomous equipment operation, and OREX Services for Open RAN adoption.

Sadayuki Abeta, NTT DOCOMO’s global head of Open RAN solutions, stated that a key advantage for operators using OREX is the ability to select vendors without having to worry about integration, as it is wholly managed by NTT DOCOMO and its partners. This indicates not only the incorporation of cutting-edge technologies, but also a future in which networks are more efficient, flexible, and user-friendly.