Steve Largent on Spectrum
Steve Largent authored a piece in Politico a couple of days ago that is worth reading, if you’ve not had a chance (and I’m not just saying that because he quoted our fair editor).
He refuted recent arguments that suggest the spectrum crisis is exaggerated. Spectrum is “the key element to addressing the ‘hockey stick’ of demand we face in terms of usage, users and use. If there were a more effective solution, which would allow carriers to avoid spending billions of dollars at auction, our carriers would be deploying it.”
Quoting Richard, he adds, “if there were a cheaper solution that would perform in any spectrum bands, ‘any such company would rule the market for smartphone service in short order. … The reason this hasn’t happened is that it’s simply not possible with the technology we have today. … The communications industry has made steady progress toward improving radio efficiency for a hundred years. … but the constraints are very real.’”
Again, it’s all worth a read.
A few points . . .
There are signs the hockey stick is inverting, and mobile broadband data growth is saturating. 35% year to year growth by some operators, including Telefonica and Vodafone. In this blog post I cite a recent analysis by 4G Americas, a US-based wireless trade association, that has a projection of flattening-growth based on Cisco data:
http://stevencrowley.com/2012/05/16/4g-americas-scenario-shows-mobile-data-growth-tapering-off/
(Yes, we’re all using lots more wireless data, but a smaller percentage of it is coming from mobile broadband macrocell networks. More is coming from Wi-Fi, small cells, etc.)
Others, including I, are more optimistic in the near-term prospects for technology to improve wireless capacity. One good example is a recent report for Ofcom that projects technology improvements out to 2030:
http://www.ofcom.org.uk/static/uhf/real-wireless-report.pdf
CTIA’s VP of Technology, John Marinho, is familiar with spectrum engineering. I expect he knows the above is reasonable. If CTIA let him talk I think we’d get more informative comments on spectrum issues from CTIA.
Like many analysts, you make the mistake of confusing utilization with demand, Steve. Projections of future demand based on rates of change in utilization can only be rational if we assume that networks have always had the capacity to carry all the traffic that users want them to carry. This assumption is reasonable to make in many Wi-Fi scenarios, but it’s rarely rational where mobile networks are concerned.
What you’re really modeling is the rate of growth in *network capacity* on the mobile side, and the rate of *demand growth* only on the indoor, stationary side. So neither of these figures actually enables us to make rational projections about mobile network demand growth.
In order to project demand growth we need to know the extent to which current demand outstrips current supply, and we know that there has been more demand for capacity than supply in many locales since the introduction of the iPhone in 2007. This can only be measured by examining the average transmit queue depth on smartphones and base stations, and nobody has that data. Cisco doesn’t, because it doesn’t have its software running in smartphones and base stations, so it does the best it can by measuring traffic on its routers and switches, where capacity is fairly elastic since they’re wireline devices.
In the past, you’ve accused Cisco – incorrectly, as an empirical matter – of pumping up forecasts to sell more equipment, despite the fact that Cisco doesn’t make or sell base stations. But now you’re treating the Cisco projections as gospel. So which is it?
Ultimately, the demand for mobile capacity is driven by applications. The apps that people seem to be using most on today’s smartphones are social apps like Twitter, Facebook, and Instagram that are not highly bandwidth-intensive. The forecast for demand growth is inherently a bet about the applications that mobile smartphone users will want to use in the future, not simply a matter of projecting past performance into the future.
Do you feel confident that you know what these apps are? I certainly don’t.
The history of demand growth shows that new applications are its principal driver. Consider the growth of Internet traffic that came about when the Web was introduced. It was in fact exponential for several years. Then look at the increase in demand for Internet bandwidth that came about with the advent of YouTube and Netflix. Once again, exponential growth for a period of time.
The dynamics of bandwidth demand are thus step functions rather than a smooth curve.
Richard, I’m confused by what you are trying to say here. You start off by saying we should project demand not supply (utilization) but then conclude by saying that you don’t know what demand is going to be.
Projecting “demand” in an open loop sense (i.e. in the absence of pricing and available capacity) is fundamentally nonsensical. Are you saying that the step change in demand from YouTube and Netflix represented “demand” that didn’t materialize until sometime in the last few years? Surely people would have loved to do this 10 years ago (remember broadcast.com) but the network cost and available capacity simply didn’t support that.
So I think its much better to look at demand in a constrained sense of “what is feasible for a particular set of pricing and capacity trends” (something that is usually a bit easier to predict than “what is the next popular app”). We can look at the cost per bit of a network, consider optimizing the mix (voice & SMS subsidize cheap contended data) and come up with a reasonable estimate of what is affordable for the consumer and therefore viable. For example, watching HDTV on your iPad over wireless simply isn’t affordable at $10 per Gbyte (or even $1 per Gbyte for most people) – you will find WiFi or download a movie in advance. As a result, forecasts (like Cisco’s) that have the majority of traffic coming from applications that generate very little revenue should be treated with a pinch of salt.
Going back to the original article, I’d note that CTIA would have more credibility in this area if it wasn’t the “boy who cried wolf”. For example, here is the CTIA President back in July 2001 saying “We’re in a spectrum crisis right now” (http://www.businessweek.com/magazine/content/01_31/b3743089.htm). It would be ironic if this stance now caused the administration to ignore them and go with Google’s latest proposal of “agile radio technologies”.
I tried putting responses in your response. Hope this works …
“Like many analysts, you make the mistake of confusing utilization with demand, Steve. Projections of future demand based on rates of change in utilization can only be rational if we assume that networks have always had the capacity to carry all the traffic that users want them to carry. This assumption is reasonable to make in many Wi-Fi scenarios, but it’s rarely rational where mobile networks are concerned.”
>> I don’t see where I’m confusing utilization with demand. 35% year to year growth by Vodafone and Telefonica is measured, not projected.
“What you’re really modeling is the rate of growth in *network capacity* on the mobile side, and the rate of *demand growth* only on the indoor, stationary side. So neither of these figures actually enables us to make rational projections about mobile network demand growth.”
>> I’m not modeling anything. The Ofcom technical report does model demand.
“In order to project demand growth we need to know the extent to which current demand outstrips current supply, and we know that there has been more demand for capacity than supply in many locales since the introduction of the iPhone in 2007. This can only be measured by examining the average transmit queue depth on smartphones and base stations, and nobody has that data. Cisco doesn’t, because it doesn’t have its software running in smartphones and base stations, so it does the best it can by measuring traffic on its routers and switches, where capacity is fairly elastic since they’re wireline devices.”
“In the past, you’ve accused Cisco – incorrectly, as an empirical matter – of pumping up forecasts to sell more equipment, despite the fact that Cisco doesn’t make or sell base stations. But now you’re treating the Cisco projections as gospel. So which is it?”
>> I’m not treating the Cisco projections as gospel. They skew high, compared to most, in part because of the lowballing of offload. The Ofcom report also shows how Cisco’s numbers skew high. Even with those numbers, the 4G Americas scenario shows saturating growth. Cisco does indeed cell mobile broadband hardware, mainly in the form of core network equipment designed to help operator with data increases. (It may still sell a WiMAX base station.) Cisco’s core-network marketing department is in charge of the VNI.
“Ultimately, the demand for mobile capacity is driven by applications. The apps that people seem to be using most on today’s smartphones are social apps like Twitter, Facebook, and Instagram that are not highly bandwidth-intensive. The forecast for demand growth is inherently a bet about the applications that mobile smartphone users will want to use in the future, not simply a matter of projecting past performance into the future.”
>> I think I agree with some of this. The social networking apps you mention are indeed bandwidth intensive not because of user data but because of signalling demands — constant pinging for updates.
“Do you feel confident that you know what these apps are? I certainly don’t.”
>> I am not confident that I can pick what the popular apps are going to be many year from now.
OK, we’ve got a debate.
You say: “I don’t see where I’m confusing utilization with demand. 35% year to year growth by Vodafone and Telefonica is measured, not projected.”
35% year to year growth in what? In actual traffic, not in gross demand. If the demand exceeds network capacity year after year – as it has in places like New York and San Francisco on the AT&T network ever since the iPhone – what you’re seeing in year-after-year increases in actual traffic is actually an increase in capacity in a scenario where the actual demand is unknown.
You say: “Cisco does indeed cell [sic] mobile broadband hardware, mainly in the form of core network equipment designed to help operator with data increases. (It may still sell a WiMAX base station.)”
Cisco does not sell any cellular gear at all, period, end of story. The switches and routers it sells are wireline Internet systems, and historically its estimates have come in under actual measurements when measurements can be made. That’s the beauty of predictions, they can be tested. Cisco’s have historically been conservative, it’s a simple fact.
And you finally say: “I am not confident that I can pick what the popular apps are going to be many year from now.”
That’s my point, nobody’s projections can be taken too seriously because nobody knows what the killer app is going to be in 2014 and beyond. If we have the capacity, we can probably bet it will involve social and video, but who knows what the combination will look like?
The 35% growth I cite is mobile data growth year to year.
Cisco does sell cellular gear through it’s service provider division, such as the ASR 5000 series multimedia core platform for mobile packet core networks:
http://www.cisco.com/en/US/products/ps11072/index.html
Here’s the WiMAX base station:
http://www.cisco.com/en/US/prod/collateral/wireless/ps9577/ps9602/data_sheet_c78-470405.html
None of that is particular to cellular networks, Steve. Their latest forecast for mobile data growth is 78% CAGR from 2011-2016.
If demand exceeds network capacity as happened to AT&T due to iPhone introduction is something that should be confirmed in what extend. Network capacity, nowadays more important than before, must be divided in control plane capacity and user plan (data traffic) capacity. Since smartphone introduction, signaling storm deeply affects network performance and probably many networks (these that pionering smartphone deployment) were not prepared for this so problems were meant to appear (now we know its effects).
Due to more smart devices attached to MBB networks (HSPA & LTE), sinagling and traffic will grow and this is a fact. Whether they will grow linearly or exponentially will depends of factors like the type of technology, pricing, coverage availability, spectrum resources, new busisness birth, how the users make profic of network resources and the services (OTT Apps, Video, etc.) associated, among others. However, I think that make forecasts, even “open loop”, is needed to plan investments in network expansion and/or evolution.
Nowadays I have seen some networks that make expansions based not only to bear data traffic but mostly to overcome signaling storm, so a new dimension to network planning must be added.
Wait until network crashes to expand is a bad idea and operators now it. User perception now is more important than ever, so MONETIZING every bit of traffic is the key to survive.
In the other hand, goverment can also make profit of this managing and pricing the spectrum and service licenses smart.
Finally, about the spectrum crisis, this may be happening in a number of markets as many operators request for more resources, but are they optimizing their actual resources in the right way?