3GPP to implement one-millisecond latencies and network slicing in mid-June specs.
SAN JOSE, Calif. — In June, the 3GPP is expected to finish two standards that, together, form the next big step in 5G cellular. The low latency and network slicing specifications pave the way for distributed wireless networks that some of the world’s largest carriers plan to build.
The ultra-reliable low-latency communications (URLLC) spec will enable a one-millisecond round-trip latency between a user device and a base station, compared to tens of milliseconds today. The next-generation core (NGC) standard opens the door to a mix of virtual services that can share a carrier’s underlying hardware.
Observers expect that the 3GPP will complete both Release 15 specs by their mid-June deadline. However, implementing them and their Release 16 follow-ons will be a multi-year process.
Today’s cellular networks generally require processing at one of perhaps a dozen large central office locations that big carriers maintain. The two new specs will let carriers build hundreds of smaller distributed nodes that can support a mix of new, old, and more responsive services still being defined.
“So far, everyone has been monetizing data rate,” said John Smee, a vice president of engineering for corporate R&D at Qualcomm. “The challenge of 5G is monetizing latency with new services.”
The edge clouds will vary in size from a small closet to a warehouse, containing mostly servers and storage and a cellular base station. For its part, AT&T described its plans for edge clouds in November.
“Carriers globally are at different stages that range from not thinking about this yet to having all their decisions made and being on an execution path … I expect hundreds, if not thousands, of them to be built starting this year,” said Mike Murphy, the CTO for Nokia in North America.
Market watchers at Dell’Oro group disagree. “Our view is that the enhanced mobile broadband use case will be driving more than 99 percent of the 5G investments over the next five years, meaning low latency use cases will not be a significant capex driver over the near-term,” said Stefan Pongratz, a senior analyst at Dell’Oro.
For carriers, the shift requires a leap of faith in new services still being defined “and finding partners to make them useful,” said Smee.
The cost decisions loom large. “There are big decisions for CTOs about how to restructure their networks; do they have faith in URLLC or not — it takes years to build these networks, and if you don’t build them and your competitors do, you could be left behind,” said Murphy.
Technical issues are smaller but not insignificant. For example, network slicing requires using IDs to define and secure services and route them through optimal paths. Besides the new 3GPP standards, edge clouds will need to implement years of work on software-defined networks and ETSI’s network functions virtualization specs.
“It’s relatively complex because, end-to-end, you want to tune the radio, transport, and core networks, and coordinating all that needs to stay simple,” said Murphy. “The implementation will take time due to this complexity.”
The hunt is on for low-latency services
Carriers are trying to figure out what low-latency services will be commercially successful and how they map to the spectrum that they have available.
For example, carriers in China are most interested in services that ride their 2.4- to 3.5-GHz mobile broadband networks, offering broad coverage. They hope to enable whatever in 5G might be the equivalent of successes such as the ride-sharing services enabled in part by widespread LTE, said Smee.
German carriers see 5G as an enabler for highly automated wireless factories using their 3.4- to 3.8-GHz bands. Qualcomm will demonstrate at Mobile World Congress wireless Ethernet links over 5G using the industrial ProfiNet protocol that could enable links on a factory floor with sub-millisecond latencies.
In the U.S., such midband spectrum is relatively scarce, but millimeter-wave frequencies are available. They will be used to launch residential internet access services to challenge cable operators. Others will explore 600- to 700-MHz bands that offer wide area coverage at low data rates for the Internet of Things.
Among other apps under consideration, edge clouds may support augmented or virtual reality services based on remote banks of GPUs. AT&T’s first test for an edge cloud targets AR/VR services in Palo Alto, California, the home of Stanford University.
“We do anticipate carriers will gradually invest in their networks to ensure they are better prepared when AR/VR become mainstream technologies. No one truly knows when new low latency IoT use cases beyond video gaming will reach mass adoption,” said Pongratz of Dell’Oro.
The June URLLC spec will lay down a foundation. The 3GPP will build on that base a set of more comprehensive features and optimizations with its Release 16 standards due in 2019.
Whatever happens with the new services, today’s 4G nets will hit limits by 2024 in the capacity and data rates that they can offer, a phenomenon that Murphy calls LTE exhaust.
“Peak data rates will hit 1.2 Gbits/second this year, and we see a path to upwards of 2 Gbits/s using shared, unlicensed 5-GHz and 3.5-GHz spectrums,” he said. “But going beyond that is very difficult and really needs 5G.”
In terms of capacity, 5G with massive MIMO antennas promise gains beyond the 2x to 3x increases expected in LTE over the next few years. In addition, 5G could deliver 50% gains in spectral efficiency over LTE, lowering costs of providing services.
The gains are strategic at a time when T-Mobile is offering U.S. users costs of $1.40/GByte compared to rival Verizon at $5/GByte. “I think the only way you solve this problem is by going to the next generation,” said Murphy.
— Rick Merritt, Silicon Valley Bureau Chief, EE Times