5G-hyperscale convergence holds a lot of promise for businesses, cities, and consumers. But engineers must collaborate to deliver on that promise.
Despite talk of 5G versus hyperscalers, 5G networks and hyperscalers are absolutely a match made in heaven.
5G’s massive machine-type communications (mMTC) feature set enables private and public networks to support up to one million devices per square kilometer. That’s just one example of how and why IoT and other connected devices will generate about 73 zettabytes annually by 2025, IDC predicts.
All that data must go somewhere for analysis and action, which is where hyperscalers come in. The designation “hyperscaler” reflects both the scale of facilities under their responsibility — at least 5,000 servers in a building that’s 10,000 square feet or larger, according to IDC’s definition — and their ability to rapidly scale up to meet demand. The latter includes horizontal scaling through the addition of hardware and square footage or vertical scaling through improved bandwidth and efficiency using existing hardware.
IoT is transforming and disrupting virtually every aspect of electronic engineering, a trend that 5G will turbocharge with gigabit speeds, ultra–low latency, mMTC connectivities, ultra–densification with beamforming, and other more advanced features. 6G will offer even greater capabilities with frequencies in the TeraHz bands, Internet of Space Things and Internet of Bio–Nano Things, all controlled by the most advanced artificial intelligence (AI) engines the world has seen. Yet, all these transformations will never reach full potential without hyperscale computing.
Like 5G’s mMTC, enhanced mobile broadband (eMBB) and ultra–reliable low–latency communications (URLLC) feature sets, hyperscalers have their own next–gen technologies for keeping up with IoT’s unprecedented traffic loads. For example, optical transceivers in the 800G range offer double or quadruple the amount of data center interconnect (DCI) capacity between data centers as well as intelligent edge computing centers. Distributed edge computing is key because it reduces system latency by bringing user plane applications and network functionality closer to IoT devices. It’s also the perfect location to implement advanced AI analytics and automation (closer the action planes).
Enabling new applications and use cases with 5G and hyperscalers
Combining 5G networks and beyond with hyperscalers enables a variety of next–gen applications that wouldn’t be practical or possible with older technologies. A few examples include advanced driver–assistant systems (ADAS), Industry 4.0, unmanned data centers and metaverse applications.
Unmanned data centers. Similar to lights-out factories, next-gen data centers — particularly at the edge — will be overwhelmingly unmanned. This new model has helped accelerate development of the data center IoT vertical with its own unique and dedicated network slice. These data centers will use the same real-time 5G IoT sensing and remote automation that’s redefining manufacturing plants, ports, retail centers and cities. Robots or drones can effectively perform important surveyance tasks within the center, while URLLC can be used to queue automatic links to service dispatch. Much like smart port sensors within shipping containers, strategically deployed temperature and humidity sensors can feed back important environmental data to automate and expedite hardware and HVAC adjustments.
With new technologies comes new challenges
A lot of work remains to achieve these and other benefits. For example, the resource constraints associated with hyperscale computing underscore the challenges of scalability as 5G emerges and the demand for telecom, ICP and big data storage applications rise exponentially. Even with innovations such as 800G DCI, continuing to scale horizontally has proven to be physically and environmentally unsustainable.
Each data center deployment or expansion also requires a proportional increase in fiber count, introducing additional exposure to fiber damage, contamination, and vandalism, which can lead to service disruption, steep SLA–violation penalties, and costly repairs. Indeed, fiber installers spend up to 20 percent of their time troubleshooting, with DCI issues being notorious for high mean–time–to–repair and recovery costs.
Some of the more arduous hyperscale challenges for the years ahead will be driven by the intricacy of distributed, disaggregated and cloudified 5G networks. As the triple constraint theory assures, data center costs will continue to escalate so long as scale and complexity continue along their current trajectories. For example, virtualized RANs, massive MIMO, and antenna beamforming all have enormous benefits for enterprises and other end users. But those technologies also further complicate RF and network performance testing, introducing new spectrum analysis, demodulation, and SLA–conformance challenges.
To overcome these challenges and ensure that 5G-hyperscale convergence lives up to its full potential, end–to–end network slicing must be orchestrated seamlessly for each unique vertical. This immense challenge resists any “swivel–chair” or siloed approaches to network management. Outdated modes of data center and network testing and assurance run contrary to the objectives of fully automated and programmable network slicing and edge computing. Critical 5G IoT use cases leave no margin for error with respect to SLA conformance and reliability.
The bottom line is that 5G–hyperscale convergence holds a lot of promise for businesses, cities, consumers, and others. But it’s going to take a lot of engineering work to deliver on that promise.
This article was originally published on EE Times.
Sameh Yamany is the chief technology officer at Viavi Solutions.