One of the biggest dilemmas for operators which are adopting 5G at an early stage (before 2022) is when to move from the Non-Standalone (NSA) mode, which continues to use the LTE core, to the Standalone (SA) mode, which requires its own 5G core. Some operators, such as those in Singapore, are under pressure from governments to move quickly to SA because it will enable a far wider variety of services, and look more like ‘real 5G’, than NSA. But most are protesting that SA systems, though standardized, are not ready for primetime yet, and moving too early risks excessive cost and complexity.
When the first generation of 5G standards, 3GPP Release 15, was split into two phases under pressure from AT&T and others, the NSA mode was fast-tracked as a simpler set of specifications (because it did not involve the core) which could be finalized and commercialized quickly. That brought deployment forward, for the very impatient MNOs like AT&T, by about six months without having to resort to pre-standard equipment, as Verizon did.
Critics pointed out that NSA was not ‘true 5G’ as the full promises of the platform, such as the dynamic support for a wide range of use cases and network behaviors, could not be achieved with just a RAN upgrade. The benefits on which so many national 5G visions rest would only be possible with an end-to-end solution with RAN and (preferably cloud-native) core. NSA turbocharges the RAN’s speed by using wider bandwidth and new spectrum, including millimeter wave, but it does not change the way the network operates because it is still using the 4G core.
But it gave many operators a relatively simple way to bow to consumer, shareholder and government pressures to at least be able to claim a 5G network. Even China Mobile, the only early mover MNO which had said it would go straight to SA immediately, changed its plans and deployed NSA first, to avoid being left behind its rivals (though it has plans to move to SA from 2020).
However, there are very significant architectural decisions to be made before deploying SA, which were not needed for the relatively simple NSA. These include:
- how to implement a cloud-native 5G core, a huge change in thinking for most MNOs;
- whether to do that on a private or public cloud, and how much to distribute to an edge cloud;
- even more challenging, how and when to deploy a virtualized RAN, and whether it makes more sense to wait to move to SA until the operator is confident it is the right time to place the RAN’s digital functions in the cloud;
- in a disaggregated RAN, which of the 13 standardized functional splits between the virtual and physical functions, and the centralized and distributed ones, to adopt;
- how to ensure that 4G and 5G can coexist efficiently in the long term (NSA uses Dual Connectivity, which allows both to link to the same core, and devices to connect simultaneously to 4G or 5G base stations).
With all that, and more, to work out, it is no surprise that many operators believe they will wait five years or more to move to SA, if they have deployed NSA at an early stage (some, of course, will live with 4G for years and then go straight to SA). They know the main benefits of 5G, such as slicing, will come with SA, but the business case and return on investment is not clear given the disruptive nature of the migration, and the immature state of commercialized products (every vendor claims a cloud-native 5G core, in fact they are all still ‘on a journey’ to that end point).
In addition, many of the new use cases most associated with 5G, like autonomous vehicles, will also require the next release of the 3GPP standards, which will enable IoT services with ultra-low latency, critical availability and support for shared spectrum.
In Singapore, the operators are resisting the government’s demand that they deploy SA, and Release 16, next year. SingTel, the market leader, told regulator IMDA that it wanted to wait until there was fully commercialized, and therefore affordable, SA equipment supporting millimeter wave spectrum. For now, it argued: “There is no confirmed roadmap for chipset and network equipment for mmWave devices on a Standalone 5G network architecture.”
In launching a consultation on 5G licensing in May, the IMDA stipulated that to “achieve the full benefits of 5G, Standalone 5G networks that do not ride on existing 4G networks will be required”, and it wants to mandate SA from day one, for any operator winning the 3.5 GHz, 26 GHz and 28 GHz licenses it will issue this year (it will only allocate the spectrum to two MNOs).
It argued that “5G deployments in many countries today are based on ‘Non-Standalone’ specifications, which leverage existing 4G networks for connectivity. Such deployments are limited to higher speeds and will not support the full suite of 5G capabilities.”
It will award the two licenses on a beauty contest basis and only the four existing MNOs will be eligible to apply (M1 and TPG are the other two). Its selection process will give a 40% weighting to network design and resilience, 30% to roll-out and performance, 15% to financial capability and 15% to spectrum price.
In its submission to IMDA, SingTel said that, so far, all 5G deployments, and the application development and device ecosystems, have been based on NSA and called on the agency to “take NSA into consideration and facilitate a roadmap to transition from NSA to SA network architecture”.
In its own filing, second operator StarHub warned that the business case for SA is as yet unclear. “If the authority wants 5G to flourish, the winning bidders must be allowed to deploy 5G based primarily on commercial considerations,” it wrote. “If onerous regulatory obligations are imposed on the 5G winners from the outset, this will undermine the business case for 5G.”
The GSMA supported its Singaporean members in its own submission, saying that an SA-only rule would be “a roll-out handicap” for licensees. “We encourage IMDA to consider the cost benefits of imposing such a license condition, and recommend issuing full technology neutral licenses,” it said.
SingTel and StarHub are taking different stances on other aspects of the proposed licensing policy, especially the size of the allocations. SingTel believes that awarding just 100 MHz per MNO in the 3.5 GHz band “would result in two sub-optimal 5G networks that are not considerably superior to 4G.” It proposes allocating more spectrum to the first network and staggering the deployment of the second network, “which can commence when more 3.5 GHz spectrum becomes available”.
StarHub opposes that, presumably believing it would be the second network. It told IMDA that, “to allow meaningful competition for 5G, both spectrum parcels must be equal in size.”
The IMDA has separately initiated another consultation related to legacy spectrum in the 800 MHz, TDD 1.9 GHz and FDD 2.1 GHz bands. It proposes setting aside the 800 MHz band for enterprise private networks.