6G convergence or "network of networks" must be bi-directional, not assume a 3GPP umbrella

This post originally appeared on my LinkedIn feed, which is now my main platform for both short posts and longer-form articles. It can be found here, along with the comment stream. Please follow / subscribe to receive regular updates (about 1-3 / week)

 Following on from my (rather controversial) post the other day about #6G and #IMT2030 needing to be indoor-primary and also have an IEEE / #WiFi candidate, I'm now going to *further* annoy various people.

There's a lot of talk about 6G being a "network of networks". This follows on from previous similar themes about #convergence and #HetNets. At one level I agree, but I think there needs to be a perspective shift.

There has been a long string of attempts to blend Wi-Fi and cellular, going all the way back to UMA in the 2G/3G era around 2005. (I was a vociferous critic).

There's been a alphabet-zoo of acronyms covering 3GPP gateway functions or selection/offload approaches - GAN, ANDSF, TWAG, N3IWF, ATSSS - and probably others I've forgotten. From the Wi-Fi side there's been Hotspot 2.0 and others. More recently we've seen an attempt to bridge fixed and mobile networks, even going as far as pitching 3GPP-type cores for fixed ISPs.

Pretty much all of these have failed to gain traction. They've had limited deployments and successes here and there, but nobody can claim that true "converged wireless" is ubiquitous or even common. 99% of WiFi has no connection to cellular. Genuine "offload" is tiny.

But despite this, the 6G R&D and vision seems to be looking to do it all over again. This phrase "network of networks" cropped up regularly at the 6GWorld #6Gsymposium events I attended this week. It now usually includes integrating #satellite or non-terrestrial (NTN) capabilities as much as Wi-Fi.

But there's a bit of an unstated assumption I think needs to be challenged. There seems to be unquestioned acceptance that the convergence layer - or perhaps "umbrella" sheltering all the various technologies is necessarily the 3GPP core network.

I think this is a problem. Many of the new and emerging 6G stakeholders (for instance enterprises, satellite operators, or fixed providers) do not understand 3GPP cores, nor have the almost religious devotion to that model common in the legacy cellular sector.

So I think any "convergence" in IMT2030 must be defined as bi-directional. Yes, Wi-Fi and satellite can slot into a 3GPP umbrella. But satellite operators need to be able to add terrestrial 6G as an add-on to their systems, while Wi-Fi controllers (on-prem or cloud based) should be able to look after "naked" (core-free) 3GPP radios where appropriate.

This would also flow through to authentication methods, spectrum coordination and so on. Also it should get reflected in government policy & regulation.

My view is that 3GPP-led convergence has largely failed. Maybe it gets fixed in 5G/6G eras, but maybe it won't. We need #5G and 6G systems to have both northbound and southbound integration options.

I also think we need to recognise that "convergence" is itself only one example of "combination" of networks. There are numerous other models, such as bonding or hybrids that connect 2+ separate networks in software or hardware.

 

Local networks: when telecoms becomes "pericoms"​

Published via my LinkedIn Newsletter - see here to subscribe / see comment thread

"Telecoms" or "telecommunications" is based on the Greek prefix "tele-".

It means "at a distance, or far-off". It is familiar from its use in other terms such as telegraph, television or teleport. And for telecoms, that makes sense - we generally make phone calls to people across medium or long distances, or send then messages. Even our broadband connections generally tend to link to distant datacentres. The WWW is, by definition, worldwide.

The word "communications" actually comes from a Latin root, meaning to impart or share. Which at the time, would obviously have been done mostly through talking to other people directly, but could also have involved writing or other distance-independent methods.

This means that distant #communications, #telecoms, has some interesting properties:

• The 2+ distant ends are often (but not always) on different #networks. Interconnection is therefore often essential.
• Connecting distant points tends to mean there's a good chunk of infrastructure in between them, owned by someone other than the users. They have to pay for it, somehow.
• Because the communications path is distant, it usually makes sense for the control points (switches and so on) to be distant as well. And because there's typically payment involved, the billing and other business functions also need to be sited "somewhere", probably in a #datacentre, which is also distant.
• There are a whole host of opportunities and risks with distant communications, that mean that governments take a keen interest. There are often licenses, regulations and internal public-sector uses - notably emergency services.
• The infrastructure usually crosses the "public domain" - streets, airwaves, rooftops, dedicated tower sites and so on. That brings additional stakeholders and rule-makers into the system.
• Involving third parties tends to suggest some sort of "service" model of delivery, or perhaps government subsidy / provision.
• Competition authorities need to take into account huge investments and limited capacity/scope for multiple networks. That also tends to reduce the number of suppliers to the market.

That is telecommunications - distant communications.

But now consider the opposite - nearby communications.

Examples could include a private 5G network in a factory, a LAN in an office, a WiFi connection in the home, a USB cable, or a Bluetooth headset with a phone. There are plenty of other examples, especially for IoT.

These nearby examples have very different characteristics to telecoms:

• Endpoints are likely to be on the same network, without interconnection
• There's usually nobody else's infrastructure involved, except perhaps a building owner's ducts and cabinets.
• Any control points will generally be close - or perhaps not needed at all, as the devices work peer-to-peer.
• There's relatively little involvement of the "public domain", unless there are risks like radio interference beyond the network boundaries.
• It's not practical for governments to intervene too much in local communications - especially when it occurs on private property, or inside a building or machine.
• There might be a service provider, but equally the whole system could be owned outright by the user, or embedded into another larger system like a robot or vehicle.
• Competition is less of an issue, as is supplier diversity. You can buy 10 USB cables from different suppliers if you want.
• Low-power, shared or unlicensed spectrum is typical for local #wireless networks.

I've been trying to work out a good word for this. Although "#telecommunications" is itself an awkward Greek / Latin hybrid I think the best prefix might be Greek again - "peri" which means "around", "close" or "surrounding" - think of perimeter, peripheral, or the perigee of an orbit.

So I'm coining the term pericommunications, to mean nearby or local connectivity. (If you want to stick to all-Latin, then proxicommunications would work quite well too).

Just because a company is involved in telecoms does not mean it necessarily can expect a role in pericoms as well. (Or indeed, vice versa). It certainly can participate in that market, but there may be fewer synergies than you might imagine.

Some telcos are also established and successful pericos as well. Many home broadband providers have done an excellent job with providing whole-home #WiFi systems with mesh technology, for example. In-building mobile coverage systems in large venues are often led by one telco, with others onboarding as secondary operators.

But other nearby domains are trickier for telcos to address. You don't expect to get your earbuds as an accessory from your mobile operator - or indeed, pay extra for them. Attempts to add-on wearables as an extra SIM on a smartphone account have had limited success.

And the idea of running on-premise enterprise private networks as a "slice" of the main 4G/5G macro RAN has clearly failed to gain traction, for a variety of reasons. The more successful operators are addressing private wireless in much the same way as other integrators and specialist SPs, although they can lean on their internal spectrum team, test engineers and other groups to help.

Some are now "going the extra mile" (sorry for the pun) for pericoms. Vodafone has just announced its prototype 5G mini base-station, the size of a Wi-Fi access point based on a Raspberry Pi and a Lime Microsystems radio chip. It can support a small #5G standalone core and is even #OpenRAN compliant. Other operators have selected new vendors or partners for campus 4G/5G deployments. The 4 UK MNOs have defined a set of shared in-building design guidelines for neutral-host networks.

It can be hard for regulators and policymakers to grasp the differences, however. The same is true for consultants and lobbyists. An awful lot of the suggested upsides of 5G (or other forms of connectivity) have been driven by a tele-mindset rather than a peri-view.

I could make a very strong argument that countries should really have a separate pericoms regulator, or a dedicated unit within the telecoms regulator and ministry. The stakeholders, national interests and economics are completely different.

A similar set of differences can be seen in #edgecomputing: regional datacentres and telco MEC are still "tele". On-premise servers or on-device CPUs and GPUs are peri-computing, with very different requirements and economics. Trying to blur the boundary doesn't work well at present - most people don't even recognise it exists.

Overall, we need to stop assuming that #pericoms is merely a subset of #telecoms. It isn't - it's almost completely different, even if it uses some of the same underlying components and protocols.

(If this viewpoint is novel or interesting and you would like to explore it further and understand what it means for your organisation - or get a presentation or keynote about it at an event - please get in touch with me)

The Worst Metrics in Telecoms

 (This post was initially published as an article on my LinkedIn Newsletter - here - please see that version for comments and discussion)

GDP isn't a particularly good measure of the true health of a country's economy. Most economists and politicians know this.

This isn't a plea for non-financial measures such as "national happiness". It's a numerical issue. GDP is hard to measure, with definitions that vary widely by country. Important aspects of the modern world such as "free" online services and family-provided eldercare aren't really counted properly.

However, people won't abandon GDP, because they like comparable data with a long history. They can plot trends, curves, averages... and don't need to revise spreadsheets and models from the ground up with something new. Other metrics are linked to GDP - R&D intensity, NATO military spending commitments and so on - which would needed to be re-based if a different measure was used. The accounting and political headaches would be huge.

A poor metric often has huge inertia and high switching costs.

Telecoms is no different, like many sub-sectors of the economy. There are many old-fashioned metrics that are really not fit for purpose any more - and even some new ones that are badly-conceived. They often lead to poor regulatory decisions, poor optimisation and investment approaches by service providers, flawed incentives and large tranches of self-congratulatory overhype.

Some of the worst telecoms metrics I see regularly include:

• Voice traffic measured in minutes of use (or messages counted individually)
• Cost per bit (or increasingly energy use per bit) for broadband
• $ per MHz per POP (population) for radio spectrum auctions
• ARPU
• CO2 savings "enabled" by telecom services, especially 5G

That's not an exhaustive list by any means. But the point of this article is to make people think twice about commonplace numbers - and ideally think of meaningful metrics rather than easy or convenient ones.

The sections below gives some quick thoughts on why these metrics either won't work in the future - or are simply terrible even now and in the past.

(As an aside, if you ever see numbers - especially forecasts - with too many digits and "spurious accuracy", that an immediate red flag: "The Market for Widgets will be $27.123bn in 2027". It tells you that the source really doesn't understand numbers - and you really shouldn't trust, or base decisions, on someone that mathematically inept)

Minutes and messages

The reason we count phone calls in minutes (rather than, say, conversations or just a monthly access fee) is based on an historical accident. Original human switchboard operators were paid by the hour, so a time-based quantum made the most sense for billing users. And while many phone plans are now either flat-rate, or use per-second rates, many regulations are still framed in the language of "the minute". (Note: some long-distance calls were also based on length of cable used, so "per mile" as well as minute)

This is a ridiculous anachronism. We don't measure or price other audiovisual services this way. You don't pay per-minute for movies or TV, or value podcasts, music or audiobooks on a per-minute basis. Other non-telephony voice communications modes such as push-to-talk, social audio like ClubHouse, or requests to Alexa or Siri aren't time-based.

Ironically, shorter calls are often more valuable to people. There's a fundamental disconnect between price and value.

A one-size-fits-all metric for calls stops telcos and other providers from innovating around context, purpose and new models for voice services. It's hard to charge extra for "enhanced voice" in a dozen different dimensions. They should call on governments to scrap minute-based laws and reporting requirements, and rejig their own internal systems to a model that makes more sense.

Much.

the

same

argument...

.... applies to counting individual messages/SMS as well. It's a meaningless quantum that doesn't align with how people use IMs / DMs / group chats and other similar modalities. It's like counting or charging for documents by the pixel. Threads, sessions or conversations are often more natural units, albeit harder to measure.

Cost per bit

"5G costs less per bit than 4G". "Traffic levels increase faster than revenues!".

Cost-per-bit is an often-used but largely meaningless metric, which drives poor decision-making and incentives, especially in the 5G era of multiple use-cases - and essentially infinite ways to calculate the numbers.

Different bits have very different associated costs. A broad average is very unhelpful for investment decisions. The cost of a “mobile” bit (for an outdoor user in motion, handing off from cell to cell) is very different to an FWA bit delivered to a house’s external fixed antenna, or a wholesale bit used by an MVNO.

Costs can vary massively by spectrum band, to a far greater degree than technology generation - with the cost of the spectrum itself a major component. Convergence and virtualisation means that the same costs (eg core and transport networks) can apply to both fixed and mobile broadband, and 4G/5G/other wireless technologies. Uplink and downlink bits also have different costs - which perhaps should include the cost of the phone and power it uses, not just the network.

The arrival of network slicing (and URLLC) will mean “cost per bit” is an ever-worse metric, as different slices will inherently be more or less "expensive" to create and operate. Same thing with local break-out, delivery of content from a nearby edge-server or numerous other wrinkles.

But in many ways, the "cost" part of cost/bit is perhaps the most easy to analyse, despite the accounting variabilities. Given enough bean-counters and some smarts in the network core/OSS, it would be possible to create some decent numbers at least theoretically.

But the bigger problem is the volume of bits. This is not an independent variable, which flexes up and down just based on user demand and consumption. Faster networks with more instantaneous "headroom" actually create many more bits, as adaptive codecs and other application intelligence means that traffic expands to fill the space available. And pricing strategy can basically dial up or down the number of bits customers used, with minimal impact on costs.

A video application might automatically increase the frame rate, or upgrade from SD to HD, with no user intervention - and very little extra "value". There might be 10x more bits transferred for the same costs (especially if delivered from a local CDN). Application developers might use tools to predict available bandwidth, and change the behaviour of their apps dynamically.

So - if averaged costs are incalculable, and bit-volume is hugely elastic, then cost/bit is meaningless. Ironically, "cost per minute of use" might actually be more relevant here than it is for voice calls. At the very least, cost per bit needs separate calculations for MBB / FWA / URLLC, and by local/national network scale.

(By a similar argument, "energy consumed per bit" is pretty useless too).

Spectrum prices for mobile use

The mobile industry has evolved around several generations of technology, typically provided by MNOs to consumers. Spectrum has typically been auctioned for exclusive use on a national / regional basis, in fixed-sized slices in chunks perhaps 5/10/20MHz wide, with licenses often specifying rules on coverage of population.

For this reason, it's not surprising that a very common metric is "$ per MHz / Pop" - the cost per megahertz, per addressable population in a given area.

Up to a point, this has been pretty reasonable, given that the main use of 2G, 3G and even 4G has been for broad, wide-area coverage for consumers' phones and sometimes homes. It has been useful for investors, telcos, regulators and others to compare the outcomes of auctions.

But for 5G and beyond (actually the 5G era, rather than 5G specifically), this metric is becoming ever less-useful. There are three problems here:

• Growing focus on smaller areas of licenses: county-sized in CBRS in the US, and site-specific in Germany, UK and Japan for instance, especially for enterprise sites and property developments. This makes comparisons much harder, especially if areas are unclear.
• Focus of 5G and private 4G on non-consumer applications and uses. Unless the idea of "population" is expanded to include robots, cars, cows and IoT gadgets, the "pop" part of the metric clearly doesn't work. As the resident population of a port or offshore windfarm zone is zero, then a local spectrum license would effectively have an infinite $ / MHz / Pop.
• Spectrum licenses are increasingly being awarded with extra conditions such as coverage of roads, land-area - or mandates to offer leases or MVNO access. Again, these are not population-driven considerations.

Over the next decade we will see much greater use of mobile spectrum-sharing, new models of pooled ("club") spectrum access, dynamic and database-driven access, indoor-only licenses, secondary-use licenses and leases, and much more.

Taken together, these issues are increasingly rendering $/MHz/Pop a legacy irrelevance in many cases.

ARPU

"Average Revenue Per User" is a longstanding metric used in various parts of telecoms, but especially by MNOs for measuring their success in selling consumers higher-end packages and subcriptions. It has long come under scrutiny for its failings, and various alternatives such as AMPU (M for margin) have emerged, as well as ways to carve out dilutive "user" groups such as low-cost M2M connections. There have also been attempts to distinguish "user" from "SIM" as some people have multiple SIMs, while other SIMs are shared.

At various points in the past it used to "hide" effective loan repayments for subsidised handsets provided "free" in the contract, although that has become less of an issue with newer accounting rules. It also faces complexity in dealing with allocating revenues in converged fixed/mobile plans, family plans, MVNO wholesale contracts and so on.

A similar issue to "cost per bit" is likely to happen to ARPU in the 5G era. Unless revenues and user numbers are broken out more finely, the overall figure is going to be a meaningless amalgam of ordinary post/prepaid smartphone contracts, fixed wireless access, premium "slice" customers and a wide variety of new wholesale deals.

The other issue is that ARPU further locks telcos into the mentality of the "monthly subscription" model. While fixed monthly subs, or "pay as you go top-up" models still dominate in wireless, others are important too, especially in the IoT world. Some devices are sold with connectivity included upfront.

Enterprises buying private cellular networks specifically want to avoid per-month or per-GB "plans" - it's one of the reasons they are looking to create their own dedicated infrastructure. MNOs may need to think in terms of annual fees, systems integration and outsourcing deals, "devices under management" and all sorts of other business models. The same is true if they want to sell "slices" or other blended capabilities - perhaps geared to SLAs or business outcomes.

Lastly - what is a "user" in future? An individual human with a subscription? A family? A home? A group? A device?

ARPU is another metric overdue for obsolescence.

CO2 "enablement" savings

I posted last week about the growing trend of companies and organisations to cite claims that a technology (often 5G or perhaps IoT in general) allows users to "save X tons of CO2 emissions".

You know the sort of thing - "Using augmented reality conferencing on your 5G phone for a meeting avoids the need for a flight & saves 2.3 tons of CO2" or whatever. Even leaving aside the thorny issues of Jevon's Paradox, which means that efficiency tends to expand usage rather than replace it - there's a big problem here:

Double-counting.

There's no attempt at allocating this notional CO2 "saving" between the device(s), the network(s), the app, the cloud platform, the OS & 100 other elements. There's no attempt such as "we estimate that 15% of this is attributable to 5G for x, y, z reasons".

Everyone takes 100% credit. And then tries to imply it offsets their own internal CO2 use.

"Yes, 5G needs more energy to run the network. But it's lower CO2 per bit, and for every ton we generate, we enable 2 tons in savings in the wider economy".

Using that logic, the greenest industry on the planet is industrial sand production, as it's the underlying basis of every silicon chip in every technological solution for climate change.

There's some benefit from CO2 enablement calculations, for sure - and there's more work going into reasonable ways to allocate savings (look in the comments for the post I link to above), but readers should be super-aware of the limitations of "tons of CO2" as a metric in this context.

So what's the answer?

It's fairly easy to poke holes in things. It's harder to find a better solution. Having maintained spreadsheets of company and market performance and trends myself, I know that analysis is often held hostage by what data is readily available. Telcos report minutes-of-use and ARPU, so that's what everyone else uses as a basis. Governments may demand that reporting, or frame rules in those terms (for instance, wholesale voice termination rates have "per minute" caps in some countries).

It's very hard to escape from the inertia of a long and familiar dataset. Nobody want to recreate their tables and try to work out historic comparables. There is huge path dependence at play - small decisions years ago, which have been entrenched in practices in perpetuity, even though the original rationale has long since gone. (You may have noticed me mention path dependence a few times recently. It's a bit of a focus of mine at the moment....)

But there's a circularity here. Certain metrics get entrenched and nobody ever questions them. They then get rehashed by governments and policymakers as the basis for new regulations or measures of market success. Investors and competition authorities use them. People ignore the footnotes and asterisks warning of limitations

The first thing people should do is question the definitions of familiar public or private metrics. What do they really mean? For a ratio, are the assumptions (and definitions) for both denominator and numerator still meaningful? Is there some form of allocation process involved? Are there averages which amalgamate lots of dissimilar categories?

I'd certainly recommend Tim Harford's book "How to Make the World Add Up" (link) as a good backgrounder to questioning how stats are generated and sometimes misused.

But the main thing I'd suggest is asking whether metrics can either hide important nuance - or can set up flawed incentives for management.

There's a long history of poor metrics having unintended consequences. For example, it would be awful (but not inconceivable) to raise ARPUs by cancelling the accounts of low-end users. Or perhaps an IoT-focused vertical service provider gets punished by the markets for "overpaying" for spectrum in an area populated by solar panels rather than people.

Stop and question the numbers. See who uses them / expects them and persuade them to change as well. Point out the fallacies and flawed incentives to policymakers.

If you have any more examples of bad numbers, feel free to add them in the comments. I forecast there will be 27.523 of them, by the end of the year.

The author is an industry analyst and strategy advisor for telecoms companies, governments, investors and enterprises. He often "stress-tests" qualitative and quantitative predictions and views of technology markets. Please get in touch if this type of viewpoint and analysis interests you - and also please follow @disruptivedean on Twitter.

5G as an enterprise LAN / Wi-Fi replacement is a myth

Introduction

There are at least 10 reasons why 5G will not be a viable WLAN (wireless local area network) technology for mainstream enterprise. Despite recent claims to the contrary, it is not an alternative to Wi-Fi in offices, hospitals, apartment blocks or similar locations. 

Enterprises, investors, policymakers and vendors should be extremely skeptical of assertions 5G will displace mainstream Wi-Fi uses. Indeed, they should question the credibility and honesty of those uttering such claims.

There will certainly more deployments of indoor cellular (including private and neutral host networks) in future, but these will almost all be incremental and not substitutional to Wi-Fi. They may be used for IoT/OT uses, but typically these will be entirely new. 

4G / 5G technologies will not be integrated into most laptops or tablets, despite hype.

This post looks at why 5G cannot replace enterprise Wi-Fi - including numerous obvious reasons - and then examines why the cellular industry (which mostly understands the problems) is pretending otherwise. What's the story behind the unrealistic fantasy?

(It's quite long. Get a coffee first.)


What is wireless used for in enterprises?

Before drilling into the specifics of 5G, it is worth looking at how and why wireless gets used in enterprises at a local area. Local area here means a single office or site - whether that's an office block, a supermarket, a hospital or a factory. Some locations have multiple LANs across several buildings on a campus network, such as a university, corporate HQ site or major new property development.

(By contrast, WANs run over wide areas, such as between a retail chain's stores and warehouse sites in different cities, or linking a multinational bank's offices in several countries. Huge locations like smart cities and airports are somewhere in the middle).

A top-level list of local wireless uses, often with separate infrastructure, includes:

• Local IT / Internet connectivity - this is the main LAN space, dominated by fixed ethernet and Wi-Fi. It connects PCs, laptops, deskphones, tablets, conferencing gear and various other computing products, either to the business's own servers or to the Internet and cloud. It is also good for private use of smartphones.
• Coverage extension of public mobile operator (MNO) services onto private property, where outdoor-to-indoor signals don't reach. This uses various forms of distributed antenna system (DAS), repeaters, small-cells etc, typically installed today only in the largest buildings. Essentially, this enables local smartphone connectivity both to the telco's services (eg telephony) and to the public Internet. In some cases Wi-Fi is used to "offload" data that could normally have gone over the carrier network. Note: genuine offload is a tiny fraction of total smartphone data traffic - see this post for more (link)
• Local IoT connectivity (connecting building automation systems, HVAC, entry control etc. This can be further divided into
• Static IoT - things that don't move around, such as sensors, door controls, CCTV cameras & aircon units
• Mobile IoT, for instance credit card terminals, robots, wearables, asset-tracking tags etc
• Local OT connectivity - operational technology, often business/safety-critical with a need for realtime deterministic control, such as industrial machinery, process controls, medical equipment and so on.
• Local voice connectivity - especially walkie-talkies and private two-way radio, which are now starting to be replaced with cellular alternatives.
• Other local uses - numerous sectors have their own niche wireless requirements, maybe linking to public safety / first-responders, broadcast, audiovisual systems etc

While some of these categories overlap (for example, smartphone connections), others remain pretty well-defined in practice. Yet often, they get conflated, especially in discussions about the future roles of 4G/5G cellular networks, whether run by mobile carriers, or new/specialist indoor operators and the enterprises themselves.

This post is specifically about the IT/LAN/Internet access use-case. I think cellular has a lot to offer IoT (especially mobile IoT) for enterprise, as well as OT in industrial settings. We will also see more indoor / premises neutral-host networks (NHNs) both for coverage and private onsite voice/smartphone access. However, none of those generally gets classed as "LAN" connections.

(SAVE THE DATE: I will be running a private workshop in London on July 9th about NHNs, looking at both indoor and wide-area / metro / rural uses and deployment scenarios)

Why can't 4G/5G be used for wireless LANs instead of Wi-Fi?

At the heart of this debate is whether 3GPP cellular technologies can be used for local-area computing networks, especially for laptops, tablets and private smartphone use. Can it replace fixed ethernet and especially Wi-Fi connections? Will future PCs connect to the Internet via 5G? I'm being asked this by various of my clients, so it's worth going in to some detail here (and obviously, more detail & analysis for paying advisory customers)

This is not a new discussion - the 3GPP vs. IEEE standards war has waged for decades. I've addressed the topic multiple times, whether that's been about in-home usage (link), debunking the "5G will kill WiFi" myth (link) or discussing the important role for private cellular in industry and the need for local spectrum licensing (link). 

I can see at least 10 reasons why cellular (whether provided by MNOs as a service, or owned by the business or an NHN provider) is not a suitable IT LAN technology for enterprise:

• 13bn installed base of actively-used Wi-Fi devices today, of which only smartphones typically have cellular radios. Some have shortish replacement cycles (eg tablets) but others will last for 10+ years. They will need to be supported in-building.
• A tiny fraction of laptops & tablets have in-built 4G radios today. Despite the hype, this will not change significantly with 5G. Customers won't pay more for them, and manufacturers don't want the margin hit. We might see 10-20% penetration, but I'm doubtful (This is a whole other "10 reasons" post in its own right...) Wi-Fi remains primary.
• Plenty of other devices will never have cellular connections (5G printers? Servers?!). The number of Wi-Fi devices is exploding in IoT as well - from smart-speakers to lighting to interactive screens and terminals. Add in new low-power Wi-Fi for things like headsets (and a separate battle between Wi-Fi / BLE / ZigBee which cellular doesn't even have a toe in). Again, Wi-Fi remains primary. High-end/critical IoT devices may actually adopt passive optical LAN connections, rather than any wireless technology.
• Private 4G/5G networks are not just radios. They need core networks, control software, and maybe SIMs/eSIMs. The average IT department does not want to, or have the skills, to deal with all this, compared with installing some ethernet wiring and some Wi-Fi APs.
• Almost no businesses want to deal with the complexities of private / public cellular interconnection, roaming, regulation (lawful intercept?!) and so forth
• Even if some IT departments want to go towards 4G/5G connectivity, they still have BYOD policies, and guests, contractors and tenants who will keep desiring (& often expecting) Wi-Fi
• In-building 5G is going to be hideously complex anyway, especially for mmWave frequencies. Installing small cells also needs fibre backhaul, power etc. in the right places, whch may be different locations to Wi-Fi APs.
• Ironically, in-building 4G small cells usually need wired LAN connections to connect them. In future, it might even be possible to use Wi-Fi6 as backhaul, as it should have good-enough deterministic QoS for the time-sync requirements.
• The world would need, I estimate, 100-300,000 more enterprise cellular specialists for designing, installing, maintaining & operating 5G LANs. And AFAIK there aren't even proper training programmes, or certification schemes. That's a decade or more on its own (and probably a big opportunity for some) 
• Indoor wireless coverage is difficult and variable. Radio is absorbed by interior walls, partitions, furniture, insulation, pipework etc. Giving QoS guarantees is almost impossible. Few design, planning and testing tools are available.
• Device-to-device use cases for Wi-Fi are not easily replicable with cellular. Maybe in the future.
• User perceptions of Wi-Fi and cellular, and behaviour around it, are entrenched and will take years to change, if ever.
• Patent & royalty costs for cellular are higher, as well as the extra chipset costs.
• Unknowable new security / threat surfaces (and the fact that Wi-Fi security is often integrated with the enterprise's identity & threat-management systems today)

I could go on. Some of these will change, some will have rare exceptions, and some industries will have particular local requirements for whatever reasons. But the underlying story is clear: 5G is not a Wi-Fi replacement for the enterprise. 
 
This should not really be a surprise to anyone. I honestly find it hard to believe that most people involved in networks/telecoms don't realise at least 4 or 5 of these points off the top of their heads.

Historical note: I've been skeptical of cellular-enabled laptops since 2006 (link). I wrote a full report in 2008 (link), which was actually far too optimistic (I predicted 30% attach-rate by 2011) despite being criticised as too-negative by the cellular industry. Most of the arguments remain valid for 5G.


So why the hype?

What's a bit baffling is why the 5G/WiFi replacement fantasy is becoming more common. Even AT&T's CEO was quoted at the company's financial results event (link) as saying "It’s serving as a LAN replacement product". Other 5G-centric commentators have said similar things.

To be fair, in some cases it will be genuine ignorance, although frankly anyone that clueless about enterprise networks shouldn't be making pronouncements anyway. Another more important issue is the conflation of all the different use-cases for connectivity (as above), and people conflating the LAN, offload and IoT domains in particular. Through that lens, the AT&T statement could (very generously) be considered applicable to some IoT scenarios.

Yes, 5G has a long-term role in some industrial verticals, especially with time-sensitive networking and private control of core and/or radio & spectrum. Neutral-host cellular will be important indoors too. But controlling robots & process machinery, or doing asset-tracking in a hospital, is not the same as accessing SaaS applications from a laptop or tablet, or local IoT applications from billions of devices with local gateways. Neither is using a new 4G/5G CBRS or local-spectrum network for "reverse roaming" or "MVNO onload" really a LAN business either.

But I think there are a few other more cynical reasons in play too:

• Embarassment over mmWave's poor indoor penetration (despite the rhetoric), meaning Wi-Fi is an essential in-building complement for any 5G FWA deployment. This applies in residential use, but also for businesses too. Pretending that some sort of 5G outdoor - 5G indoor hybrid could fix this might spare a few blushes.
• Cost & complexity of future indoor 5G deployments: Reality is biting. Existing indoor systems are going to be hard to upgrade to even 3-4GHz bands, let alone adding mmWave and massive-MIMO support too. It's not just the radio elements, either - how exactly are carriers going to offer QoS / network-slicing over someone else's indoor wiring and antenna infrastructure? See this eBook I recently wrote for iBwave (link) for more details. Basically, if the telcos are going to help pay for 5G indoor connectivity, then new use-cases/revenues are desperately sought, beyond just MBB coverage. A "managed 5G LAN" line on a spreadsheet likely looks appealing, even if it's an exercise in wishful thinking.
• Bluster & hype aimed at regulators considering the 6GHz or other bands for unlicensed use (& thus mostly more Wi-Fi). The US FCC and various European regulators seem minded to add another large band (500-1200MHz) to the unlicensed systems arsenal. Taking a public stance of saying "Oh, 5G could do all those use-cases as well - how about normal exclusive licenses for that band instead?" fits the political narrative, even if it doesn't fit reality. 
• Some 3GPP fundamentalists' dislike of Wi-Fi and unlicensed spectrum generally, or non-telco controlled networks, explains some of the comments. I've seen posts on LinkedIn saying "I wish Wi-Fi would go away", and similar. They have long fantasised about MNO-managed cellular LANs , in the same way that some Wi-Fi (and satellite) fundamentalists think they can replace mobile networks. They're all wrong. (And so are the 5G FWA folk claiming it's a mainstream alternative that could replace fibre or cable).

Conclusion

So to sum up:

• There's lots of different uses for wireless networks in enterprise, whether in individual buildings or across larger campuses. Ignore anyone who groups them all together.
• IT-centric LAN connections, for normal computing devices connecting to the Internet, cloud or local servers, are dominated by ethernet - either using fixed cabling, Wi-Fi or occasionally optical LAN. Smartphones connect both by Wi-Fi and cellular, where indoor connection is good enough.
• A handful of laptops and tablets can use 4G connections today, although few owners even bother to sign up for data plans. A slightly larger handful will have integral 5G in 5 years time, but most will just stick to Wi-Fi only. They will need to be supported in all the same locations as today, plus many new ones (eg public transport & retail).
• Private 4G and 5G networks come in many varieties, with a huge range of shared/local spectrum options being considered by regulators (link). Most, however, are not aimed at LAN use-cases, but more oriented to IoT/OT/indoor cellular coverage requirements. In those instances, Wi-Fi has limitations, for instance in applicability to robots moving around a large factor or warehouse.
• Some industrial/critical use-cases are not ideally suited to unlicenced spectrum, even with the better performance of new WiFi6 deployments. Given that the WiFi industry doesn't (yet) have a licensed-band version, then cellular is a likely option instead. 
• Neutral host cellular networks are very exciting future developments, both indoors and out. But they're not going to be LAN replacements either.
• Operator 4G and 5G networks are very important to extend in-building, especially if telcos want to offer new network-slicing or QoS products that don't just work outdoors. However, upgrading existing in-building coverage solutions to 5G is hard, expensive and has many unknowns. Many small buildings don't have indoor coverage solutions at all today. The mobile industry is casting around for new revenues, as well as costs. One takeout: end-to-end network slicing is largely mythical, and will need to work over 3rd-party Wi-Fi indoors.
• We will also see various forms of Wi-Fi + cellular bonding, with devices connecting to both networks simultaneously. That's for another post.
• As with all areas of 5G hype, there's an "it'll solve world hunger", "one-size fits all" pitch to politicians, regulators, investors and media. It needs to be called out for its disengenuousness.

Overall, the key takeout: 

Private/enterprise 4G and 5G networks have lots of potential future use-cases & market opportunities. Replacing Wi-Fi for IT/Internet access LANs is not one of them.

Note: if you're interested in deeper analysis, or a private workshop / advisory engagement on this topic, please drop me an email at information at disruptive-analysis DOT com, or contact me via LinkedIn or Twitter.

Also - on July 9th, I'm running a London private workshop on Neutral Host Networks, together with Peter Curnow-Ford of Viatec Associates. Drop me a message if you're interested, and look out for full details & registration coming very soon.

MEC and network-edge computing is overhyped and underpowered

I keep hearing that Edge Computing is the next big thing - and specifically, in-network edge computing models such as MEC. (See here for a list of all the different types of "edge"). 

I hear it from network vendors, telcos, some consultants, blockchain-based startups and others. But, oddly, very rarely from developers of applications or devices.

My view is that it's important, but it's also being overhyped. Network-edge computing will only ever be a small slice of the overall cloud and computing domain. And because it's small, it will likely be an addition to (and integrated with) web-scale cloud platforms. We are very unlikely to see edge-first providers become "the next Amazon AWS, only distributed".

Why do I think it will be small? Because I've been looking at it through a different lens to most: power. It's a metric used by those at the top- and bottom ends of the computing industry, but only rarely by those in the middle, such as network owners. This means they're ignoring a couple of orders of magnitude.

(This is a long post. You might want to grab a coffee first....)


How many zeroes?

Cloud computing involves huge numbers. There are many metrics that you can use - numbers of servers, processors, standard-sized equipment racks, floorspace and so on. But the figure that gets used most among data-centre folk is probably power consumption in watts, or more commonly here kW, MW & GW. (Yes, it's a lower-case k for kilo). 

Power is useful, as it covers the needs not just of compute CPUs and GPUs, but also storage and networking elements in data centres. It's not perfect, but given that organising and analysing information is ultimately about energy it's a valid, top-level metric. [Hey, I've got a degree in physics, not engineering. Helloooo, thermodynamics & entropy!]

Roughly speaking, the world's big data centres have a total power consumption of about 100GW. A typical one might have a capacity of 30MW, but a number of the world's largest data centres already use over 100MW individually, and there are enormous plans for locations with 600MW or even 1GW (link). No, they're not all running at full power, all the time - but that's true of any computing platform.

This growth is partly driven by an increase in the number of servers and equipment racks needed (hence growing floor-space for these buildings). But it also reflects power consumption for each server, as chips get more powerful. Most equipment racks use 3-5kW of power, but some can go as high as 20kW if that power - and cooling - is available.

So, to power "the cloud" needs 100GW, a figure that is continuining to grow rapidly. We are also seeing a rise in smaller, regional data-centres in second- and third-tier cities. Companies and governments often have private data-centres as well. These vary quite a bit, but 1-5MW is a reasonable benchmark.


How many decimal places?

At the other end of the computing power spectrum, are devices, and the components inside them. Especially for battery-powered devices, managing the power-budget down to watts or milliwatts is critical. This is the "device edge".

• Sensors might use less than 10mW when idle & 100mW when actively processing data
• A Raspberry Pi might use 0.5W
• A smartphone processor might use 1-3W
• An IoT gateway (controlling various local devices) might be 5-10W
• A laptop might draw 50W
• A decent crypto mining rig might use 1kW

New innovations are pushing the boundaries. Some researchers are working on sub-milliwatt vision processors (link). ARM has designs able to run machine-learning algorithms on very low-powered devices.

But perhaps the most interesting "device edge" is the future top-end Nvidia Pegasus board, aimed at self-driving vehicles. It is a 500W supercomputer. That might sound a lot, but it's still less than 1% of the engine power on most cars. A top-end Tesla P100D puts over 500kW to the wheels in "ludicrous mode", or 1000x that figure. Cars' aircon might use 2kW, to give context.

Of course, all of these device-edge computing platforms are numerous. There are billions of phones, and hundreds of millions of vehicles and PCs. Potentially, we'll get 10s of billions of sensors. Most aren't coordinated, though. 


And in the middle?

So we have milliwatts at one end of distributed computing, and gigawatts at the other, from device to cloud.

So what about the middle, where the network lives?

There are many companies talking about MEC (multi-access edge computing) and fog-computing products, with servers designed to run at cellular base stations, network aggregation points, and also in fixed-network nodes and elsewhere. 

Some are "micro-data-centres" capable of holding a few racks of servers near the largest cell towers. The very largest might be 50kW shipping-container sized units, but those will be pretty rare and will obviously need a dedicated power supply.

It's worth noting here that a typical macro-cell tower might have a power supply of 1-2kW. So if we consider that maybe 10% could be dedicated to a compute platform rather than the radio (a generous assumption), we get 100-200W, in theory. Or in other words, a cell tower edge-node will be less than half as powerful as a single car's computer.

Others are smaller server units, intended to hook into cellular small-cells, home gateways, cable street-side cabinets or enterprise "white boxes". For these, 10-30W is more reasonable.

Imagine the year 2023

Let's think 5 years ahead. By then, there could probably be 150GW of large-scale data centres, plus a decent number of midsize regional data-centres, plus private enterprise facilities.

And we could have 10 billion phones, PCs, tablets & other small end-points contributing to a distributed edge, although obviously they will spend a lot of time in idle-mode. We might also have 10 million almost-autonomous vehicles, with a lot of compute, even if they're not fully self-driving. 

Now, imagine we have a very-bullish 10 million "deep" network-compute nodes, at cell sites large and small, built into WiFi APs or controllers, and perhaps in cable/fixed streetside cabinets. They will likely have power ratings between 10W and 300W, although the largest will be numerically few in number. Choose 100W on average, for a simpler calculation. (Frankly, this is a generous forecast, but let's run with it for now).

And let's add in 20,000 container-sized 50kW units, or repurposed central-offices-as-datacentres, as well. (Also generous)

In other words, we might end up with:

150GW large data centres
50GW regional and corporate data centres
20,000x 50kW = 1GW big/aggregation-point "network-edge"
10m x 100W = 1GW "deep" network-edge nodes
1bn x 50W = 50GW of PCs
10bn x 1W = 10GW "small" device edge compute nodes
10m x 500W = 5GW of in-vehicle compute nodes
10bn x 100mW = 1GW of sensors & low-end devices

Now admittedly this is a very crude analysis. And a lot of devices will be running idle most of the time, and may need to offload functions to save battery power. Laptops are often switched off entirely. But equally, network-edge computers won't be running at 100%, 24x7 either.


The 1% edge

So at a rough, order-of-magnitude level, we can see that the total realistic "network edge", with optimistic assumptions, will account for less than 1% of total aggregate compute capability. And with more pessimistic assumptions, it might easily be just 0.1%. 

Any more will simply not be possible to power, unless there are large-scale upgrades to the electricity supply to network infrastructure - installed at the same time as backhaul upgrades for 5G, or deployment of FTTH. (And unlike copper, fibre can't even power small devices on its own). And haven't seen announcements of any telcos building hydroelectric power stations anywhere.

Decentralised, blockchain-based edge "fogs" are unlikely to really solve this problem either, even if they also use decentralised, blockchain-based power supply and management.

Now it could be argued that this 0.1-1% of computing workloads will be of such pivotal importance, that they will bring everything else into their orbit and indirect control. Could the "edge" really be the new frontier? 

I think not.

In reality, the reverse is more likely. Either device-based applications will selectively offload certain workloads to the network, or the webscale clouds will distribute certain functions. Yes, there will be some counter-examples, where the network-edge is the control point for certain verticals or applications - I think some security functions make sense, for instance, as well as an evolution of today's CDNs. But will IoT management, or AI, be concentrated in these edge nodes? It seems improbable.


Conclusion & TL:DR

In-network edge-computing architectures, such as MEC, will become more important. There are various interesting use-cases. But despite that, they will struggle to live up to the hype. 

There will be almost no applications that run *only* in the network-edge - it’ll be used just for specific workloads or microservices, as a subset of a broader multi-tier application. The main compute heavy-lifting will be done on-device, or on-cloud. As such, collaboration between edge-compute providers and industry/webscale cloud will be needed, as the network-edge will only be a component in a bigger solution, and will only very rarely be the most important component. 

One thing is definite: mobile operators won’t become distributed quasi-Amazons, running image-processing for all nearby cars or industry 4.0 robots in their networks, linked via 5G. 

Yes, MEC nodes could host Amazon Greengrass or other functions on a wholesale basis, but few developers will want to write directly to telcos' distributed-cloud APIs on a standalone basis, with or without network-slicing or 5G QoS mechanisms.

Indeed, this landscape of compute resource may throw up some unintended consequences. Ironically, it seems more likely that a future car's hefty computer, and abundant local power, could be used to offload tasks from the network, rather than vice versa.

Comments and feedback are very welcome. I'm aware I've made many assumptions here, and will doubtless generate various comments and detailed responses, either on my blog or LinkedIn posts. I haven't seen an "end to end" analysis of compute power before - if there's any tweaks to my back-of-envelope calculations, I'd welcome suggestions. If you'd like to contact me about projects or speaking engagements, I can be reached via information at disruptive-analysis dot com.

5G & IoT? We need to talk about latency

Much of the discussion around the rationale for 5G – and especially the so-called “ultra-reliable” high QoS versions – centres on minimising network latency. Edge-computing architectures like MEC also focus on this. The worthy goal of 1 millisecond roundtrip time is often mentioned, usually in the context of applications like autonomous vehicles with snappy responses, AR/VR headsets without nausea, the “tactile Internet” and remote drone/robot control.

Usually, that is accompanied by some mention of 20 or 50 billion connected devices by [date X], and perhaps trillions of dollars of IoT-enabled value.

In many ways, this is irrelevant at best, and duplicitous and misleading at worst.

IoT devices and applications will likely span 10 or more orders of magnitude for latency, not just the two between 1-10ms and 10-100ms. Often, the main value of IoT comes from changes over long periods, not realtime control or telemetry.

Think about timescales a bit more deeply:

• Sensors on an elevator doors may send sporadic data, to predict slowly-worsening mechanical problems – so an engineer might be sent a month before the normal maintenance visit.
• A car might download new engine-management software once a week, and upload traffic observations and engine-performance data once a day (maybe waiting to do it over WiFi, in the owner’s garage, as it's not time-critical).
• A large oil storage tank, or a water well, might have a depth-gauge giving readings once an hour.
• A temperature sensor and thermostat in an elderly person’s home, to manage health and welfare, might track readings and respond with control messages every 10 minutes. Room temperatures change only slowly.
• A shared bicycle might report its position every minute – and unlock in under 10 seconds when the user buys access with their smartphone app
• A payment or security-access tag should check identity and open a door, or confirm a transaction, in a second or two.
• A networked video-surveillance system may need to send a facial image, and get a response in a tenth of a second, before they move out of camera-shot.
• A doctor’s endoscope or microsurgery tool might need to respond to controls (and send haptic feedback) 100 times a second – ie every 10ms
• A rapidly-moving drone may need to react in a millisecond to a control signal, or a locally-recognised risk.
• A sensitive industrial process-control system may need to be able to respond in 10s or 100s of microseconds to avoid damage to finely-calibrated machinery
• Image sensors and various network sync mechanisms may require response times measured in nanoseconds

I have not seen any analysis that tries to divide the billions of devices, or trillions of dollars, into these very-different cohorts of time-sensitivity. Given the assumptions underpinning a lot of 5G business cases, I’d suggest that this type of work is crucial. Some of these use-cases are slow enough that sending data by 2G is fine (or by mail, in some cases!). Others are so fast they’ll need fibre – or compute capability located locally on-device, or even on-chip, rather than in the cloud, even if it’s an “edge” node.

I suspect (this is a wild guess, I'll admit) that the proportion of IoT devices, for which there’s a real difference between 1ms and 10ms and 100ms, will be less than 10%, and possibly less than 1% of the total. 

(Separately, the network access performance might be swamped by extra latency added by security functions, or edge-computing nodes being bypassed by VPN tunnels)

The proportion of accrued value may be similarly low. A lot of the IoT examples I hear about are either long time-series collections of sensor data (for asset performance-management and predictive maintenance), or have fairly loose timing constraints. A farm’s moisture sensors and irrigation pumps don’t need millisecond response times. Conversely, a chemical plant may need to alter measure and alter pressures or flows in microseconds.

Are we focusing 5G too much on the occasional Goldilocks of not-too-fast and not-too-slow?