SDN Technology

Donovan: AT&T Beating Moore's Law

AT&T is seeing its best cost-per-megabyte gains for two decades, according to its top technologist, John Donovan.

In fact, Donovan said today that, after two decades in which its networks failed to match Moore's Law improvements in performance, AT&T has now caught and exceeded that standard and will continue to improve by moving things from hardware to software and doing smarter planning of capital outlays.

In a wide-ranging presentation at Citi's 2016 Global Internet, Media and Telecommunications Conference, the senior executive VP-technology and operations for AT&T Inc. (NYSE: T) said his company remains laser-focused on lowering its cost per megabyte and that focus is driving other strategies, including its view of 5G and small cells and its ongoing fiber deployment.

By integrating its wireless and wireline network planning, it is not only able to justify extending fiber deeper into its network, to make more efficient use of available wireless spectrum, but can also sell a wider range of services to enterprises, from WiFi to its NetBond and Network on Demand offers. (See AT&T: SDN Is Slashing Provisioning Cycle Times by up to 95%.)

Donovan kicked off the discussion by calling this "the most exciting time for technologists" in his long career because "all dimensions of the network are in movement," including the core, the access and the software and operations that glue everything together. Then, over the course of the 45-minute Q&A session, he calmly laid out the dramatic changes underway at AT&T that, he says, are letting the company meet capacity demands at much lower cost, roll out new services faster and plan its network investments more efficiently. (See Donovan Touts AT&T's Software Push.)

He referred to Netbond, AT&T's SDN-based secure VPN access to popular cloud platforms, as "a franchise changer," in that it gave enterprises the secure connections to the cloud they needed and prevented the virtual private network business from being decimated by a mass shift to basic Internet access to cloud. AT&T's other SDN-based offering, Network on Demand, is "the fastest growing thing I've seen, from a customer adoption standpoint," Donovan said. (See AT&T NetBond Getting Amazon Ties.)

Zoom in on carrier SDN strategies in our SDN section here on Light Reading.

The ability to beat Moore's Law actually happened about five years ago on the network side. The picture continues to improve as more things are moved into software and commodity hardware does more of the hard work within the network as well, enabling reductions in capex and opex, he stated.

"We have put out these ratios for percentages on the new platform to help give guidance to our supply community of how aggressive we want to be on this, and to the financial community as well, so you have an understanding of some of the impacts of these things," he said, referring to AT&T's aggressive plan to go all-IP and have 75% of its functions virtualized by 2020.

"When you go to software-defined networking for the delivery of fundamental services, opex is dramatically cheaper and capex is as well, because you go to more commodity hardware," he said. "So we will keep putting those numbers out -- the supply community needs them, because they need to be able to forecast how ambitious they need to be in directing their R&D investments."

Part of achieving those capex gains while continuing to meet rising demand for bandwidth is AT&T's integrated planning. While its Project VIP local fiber deployment initiative has wound down, the company is still able to push fiber more deeply into some areas, based on the need for business services or backhaul for cell towers and small cells, Donovan said.

"We have a really good cost curve on incremental costs for wireless," he said. "We are still putting fiber out where it is economic -- that is a big part of our program."

AT&T is "doing everything everyone else is doing on 5G" but its tight focus on cost-per-megabyte means that the carrier doesn't feel the urgency to be first to deploy the next generation of wireless -- unless it identifies cost advantages. "We haven't been overly public, because what we want to do is keep the option of being early, in the middle or at the back-end, depending on whether we decide to optimize speed, availability or cost."

What it comes down to is managing some massive transformations in customer behavior, Donovan said, as users have shifted to using Bluetooth and WiFi in the last 1,000 feet, cellular wireless in the last couple of miles and everything else on a densely packed wireline network.

"We need to seamlessly manage that stuff so that we get the right yield -- that is what we are after, to get customers on the right network, get the right cost structure for all of those networks and meet the demand with a minimum cost," he said. "Right now we have done a really good job of dealing with some of these pretty massive transitions in user behavior."

One place AT&T hasn't expanded as rapidly as expected is in small cell deployment, but Donovan views that from the broader lens of integrated network planning as well. On the heels of "fast shot clock" initiatives at the federal and local regulatory levels -- which insured more rapid approval of cell tower permits -- AT&T was able to deploy more cell towers faster and in some places that weren't previously available, cutting the need for small cells. But the company has been aggressive on in-building work and is selling WiFi and other services, alongside NetBond and Network on Demand, to a growing base of business customers.

— Carol Wilson, Editor-at-Large, Light Reading

Page 1 / 3   >   >>
rmazzi 1/12/2016 | 12:01:24 PM
Re: Is the cost trend enough? For the wireless big 4, IOT, real wireless business productivity solutions, and other services are critical to their success.  Only AT&T and VZ are positioned for success there and i give ATT the edge.  Where will that leave TMO and S?  They will continue to be the most dependent on subcription based service revenue and i don't see an easy path for them until one or the other disappears through a merger, bankruptcy or some other version of significant restructuring.
cnwedit 1/12/2016 | 11:55:52 AM
Re: Is the cost trend enough? That is the critical question - can AT&T and others operate their networks with such great efficiency that the network business itself remains profitable? I don't think they are counting on that - Donovan also talked about new services like NetBond and Network on Demand and I think they are counting on revenue from those things and things like their Smart Cities initiative, their IoT efforts with connected cars and more, and many other new revenue possibilities. 

But no one survival is guaranteed today and I think they are well aware of that. 
rmazzi 1/12/2016 | 11:40:56 AM
Is the cost trend enough? Rather than debate the use of Moore's Law as a measuring stick, the question is:  Will the rate of cost per megabyte reduction enough to maintain profitabiltiy if demand grows and revenue does not (or at least not rapidly enough)?  In other words ATT and down a great job, but is it enough to maintain or grow profitability

For the carriers, the revenue growth opportunities are tough.  There continues to be intense price competition.  Net Neutrality limits the opporunity for new rate plans and may preclude other revenue streams.  OTT players can drive demand without incurring incremental network costs.   Device players, notably Apple, control the network connections and customer experience.  Stickiness is to the device first and the network provider second.

For wireless carriers the cost of spectrum continues to rise via the auctions.  The cost for fiber placement required deeper in the network is not declining.  Gains are made in software, chip cost to performance and optical transmission costs on exisiing routes.  In the RF side, there are gains made with better codecs, more efficient schedulers and other technologies, but eventually they run into physics problems not solved by "Moore's Law".  (Although this last comment may provoke a debate on Shannon's law instead0).

AT&T is doing the right things aggessively investing to reduce unit costs, but will it be enough?
mhhf1ve 1/11/2016 | 4:44:31 PM
It's probably time to stop referring to Moore's Law... Moore's Law isn't actually a "law" in the sense that it can't be violated or anything... And we're already bending it with processors, anyway.

Let's agree to stop referring to Moore's Law as if it has any meaning outside of its original meaning for the processor development cycle.
brooks7 1/8/2016 | 1:07:53 PM
Re: What is software, what was the cost improvement, why hasn't the customer seen a benefit? "Moore's law (/mɔərz.ˈlɔː/) is the observation that the number of transistors in a dense integrated circuit doubles approximately every two years."


From: https://en.wikipedia.org/wiki/Moore%27s_law


So to beat Moore's law, it would say that the cost of bandwidth was dropping in half every two years.  That is why all the engineers here are reacting.  To say "we are on a great curve to lower cost and look what we have done" is GREAT!  We are impressed.  But then we get specific and conjure up Moore's law.  That is a great technology symbol and one that drove innovation from say 1970 until today.  It looks like that time is done and now innovation is coming almost exclusively from software.  

Here is what would satisfy us....publish a chart of the total bandwidth of AT&T and the cost per bit per second over the last 10 years and project 10 years into the future.  None of us believe that is a straight line on a consistent slope.  Which is why the super techie types are calling it fluff.  Great sentiment and great that AT&T is drivng down costs.  Of course if you are a vendor to AT&T that is horrific news.  It says that you are on a path to lowered revenues and profits forever.


cnwedit 1/8/2016 | 9:34:29 AM
Re: What is software, what was the cost improvement, why hasn't the customer seen a benefit? I don't know if I should be repsonding at this point, because I may be overstepping my boundaries a bit but it's my understanding that Donovan's cost per megabyte numbers include capex and opex, which would mean construction is factored in. 

I can certainly see why exact numbers would be helpful, and also very difficult. 
brooks7 1/7/2016 | 7:11:15 PM
Re: What is software, what was the cost improvement, why hasn't the customer seen a benefit? Carol,

The difference was that when Intel did it they posted numbers about their exact cost and density.  To be a direct comparison, one would have to show a graph with their cost per bit over the last 10 years and a projection for the next 5.

In terms of cost change, the network is a much more complicated beast. In Intel's case, they were talking about scaling a standard process at the time of Moore's Law.  For Bandwidth we are talking about Transmission and Switching of many kinds of elements.  To truly meet "Moore's Law" for Bandwidth, then POTS should be declining as well as high speed transport.

So, I suspect as an aggregate bandwidth at some measureable point has declined over a certain period of time.  However, we all talk about how construction costs have not declined.  How do we factor that all in?


cnwedit 1/7/2016 | 2:31:49 PM
Re: What is software, what was the cost improvement, why hasn't the customer seen a benefit? I get your point, but that wasn't the intent of this presentation. And I will press them for numbers but where speeches to the investment community are concerned, execs tend to be cautious and the lawyers tend to vet everything. 

They have quantified in the past the increase in bandwidth and their prices are their prices = and they aren't going up. So some of this is already out there. 

AT&T also has stated its provisioning margin improvements - that story is referenced above. And Donovan does provide some proof points - saying NetBond is the fastest growing service in his experience is a proof point. 

tegelad 1/7/2016 | 2:26:52 PM
What is software, what was the cost improvement, why hasn't the customer seen a benefit? I see lots of big words, and little details.  

Where is the data points showing we lowered our MB/$ cost from X to Y, and this helped improve the speed by Z and reduced user cost by W?

How does quality of service play (residential vs business)?

If the prices for bandwidth are similar between Verizon and AT&T why does Verizon continue to win all the quality awards?

Moore's Law in implementation about comparing benchmarked results of current performance against past.   This discussion states no real data.   For all that we know they are using Circuit Switched Data and hard lines (T-1's) as the basis versus shoving data off of a satellite from DirecTV ...  e.g. look at the 9 MBps we are delivering HD wise to all users in the US ...

Have they innovated yes, but how about some realism outside of stating we "are the strongest".

Duh! 1/7/2016 | 1:40:45 PM
Re: Exceeding Moore's Law To clarify: the Moore's Law claim is fluff.  The substantive part is, well... substantantive. The former detracts from the latter.
Page 1 / 3   >   >>
Sign In