Monday, October 29, 2018

Quick thoughts on 5G

I've been doing a lot of work - and events - on 5G recently. 
I've noticed a few recent shifts in perception and focus amongst vendors, regulators and operators. Some quick take-outs (a few more than appear on my similar LinkedIn post, as I'm not limited to 1300 characters!)
  • 5G smartphones launch in 2019, but will be low-volume until 2020/21. Expect the first 5G iPhone towards the end of 2020
  • Fixed-wireless use cases for 5G are high on the agenda in some markets (eg US, S Korea, Turkey, Germany), but seemingly almost absent in others.
  • Commercial, large-scale, automated network slicing only becomes real from around 2023 onwards. A few "hand-carved" slices will be sooner, for example for internal use by MNOs' own business units, or perhaps public safety
  • URLLC (ultra-reliable low latency) use-cases seem to have shifted from sci-fi fantasies around automated vehicles and surgical robots, to industrial IoT and factory automation... 
  • ... but industrial use will often be controlled by industry itself, via one of several forms of private network, either using shared spectrum, private cores or private slices / enterprise MVNOs. MNOs' role may be minor
  • Some claim that NB-IoT is the 5G version for "massive IoT", despite it being developed as a 4G variant. This is revisionist nonsense; if it was true then DT, VF and others would have been putting out PR 2+ years ago, claiming to be first to launch 5G
  • 3.5GHz should be OK-ish outdoors but will struggle with outdoor-to-indoor coverage. mmWave will be worse. Beware of demos showing good indoor performance - ask about uplink from inside-out, or whether signals penetrate double-glazing, or at oblique angles to walls/windows. In any case, #WiFi will continue to dominate in the home.
  • There will be some small-cells and neutral-host deployments for 3.5GHz (and similar bands) in enterprises and other large buildings, but this will take a long time to become widespread. 
  • Existing in-building DAS systems will need some serious upgrades to support higher 5G frequency bands - most of today's top out at 2.6GHz and can't handle MIMO very well.
  • Despite it not being an "official" 5G candidate band, 28GHz seems to be the most popular option, at least for test networks. This is partly because of chipset support, notably Qualcomm's X50. The European-proposed 26GHz hasn't seen much action yet
  • Two of the largest 5G "verticals" associations, for Automotive (5GAA) and Industrial (5GACIA) seem to be heavily driven by German companies - and the German regulator looks like it's going to award 100MHz of spectrum to verticals directly (not 100% certain but getting clearer). In other countries apart from the US (CBRS) and China (Huawei's enterprise LTE), there doesn't seem to be as much action from large firms knocking on the regulator/governments doors.
  • The 5G New Core is getting a lot of discussion and attention... but given that some of the existing NFV deployments have been slow, and the cost-savings somewhat illusory, I don't expect much near-term action on this.
  • Some of the visions for 5G seem to lean heavily on automation and AI back-office for optimising radio, core, user-plane etc. Yet those are also still at an early stage - and few telcos have many skilled engineers -  so could act as a brake. There are also some emerging questions on security of network AI, and whether the algorithms might be single points of failure, especially when used for networks used for critical national infrastructure. 
  • Connected-car companies are interested in 5G, but not as enthusiastic as some might imagine. One told me "it's a nice-to-have" - especially as vehicles will need to be able to work offline, and have prodigous on-board compute capabilities.
  • I'm more positive about some of the discussion around Cloud RAN for 5G. In many ways, it's going to be necessary, given the complexity of NR. That said, there's some serious practical challenges about the radio, such as the size/weight/cost of the massive-MIMO antennas.
  • There's lots of talk about network-slicing for 5G, but nobody has really thought about whether today's MNO wholesale departments are up to the task of selling "slice as a service". Speaking to some of today's MVNOs, it seems like they will have to do a lot of homework before they can become effective slicemongers.
That's a quick list of things off the top of my head. Plenty more observations and comments to come, or on my Twitter feed from various events I've attended.

If you'd like me to give an unvarnished presentation at an event, on "5G opportunities, realities & myths", please get in touch via:  information AT disruptive-analysis DOT com
And if you're interested in my last point, on 5G+MVNOs+Slicing+Wholesale, please look at my upcoming workshop doing a deep-dive on this (link)

No comments:

Post a Comment