x
Routing

Larry's Router Rant

Larry Roberts, one of the "Fathers of the Internet," is at it again. The Internet could collapse, he writes on Internet Evolution, because of the inability of routers to keep up with the growth of bandwidth consumption.

Now, we've heard these Internet Apocalypse stories before. But this latest one comes with a different spin. It's economic, mostly, Roberts says. Routing costs aren't descending rapidly enough to keep up with the growth of bandwidth. In other words, at some point soon, moving packets will become a massively unprofitable enterprise.

The first question you have to ask when you hear the latest "Internet Collapse" scenario is: What's he selling? Of course, Larry's diatribe transitions smoothly into a big fat sales pitch for his latest flow-based routing company, Anagran Inc.

Also, keep in mind, this is not the first time Roberts has made the flow pitch. After all, he chewed through about a quarter of a billion in investor capital in the pursuit of the same holy grail at Caspian Networks. Didn't work.

But Roberts does have a point. Innovation in routing technology appears to have stagnated. We're still waiting for something dramatic, such as, say, an optical router. What's the best blockbuster innovation to come from the routing world? As Roberts points out, most routers are relying on the increased speed of chipsets to improve calculation power to do the same thing they've been doing for decades: routing packets. Most adjustments to the technology, whether it be something like DiffServ or Multiprotocol Label Switching (MPLS), appear to be incremental in nature. Have we really made any huge leaps forward in the networking paradigm since Ethernet and Internet Protocol (IP) were first invented? That was more than 30 years ago!

The mistake that Roberts is making is assuming that there won't be any new innovation, or that his particular innovation is the answer. You would think that if flow-based routing were the answer, somebody at AT&T Inc. (NYSE: T) or NTT Communications Corp. (NYSE: NTT) would have caught on by now.

I believe that carriers are actually moving in the opposite direction: away from routing. They are looking to engineer more circuits on their networks at the cheaper Layer 2 layer, with carrier-class Ethernet and optical technology, and they want to push routing further out to the edge.

Larry Dennison, the CTO of Avici Systems Inc. (Nasdaq: AVCI; Frankfurt: BVC7), which recently exited the routing business, told me that large carrier customers he has been speaking with (among them, AT&T) have been asking for circuit-based technology in the core, not packet-based technology. "They don't want to open up complexity," he said. "They need a mechanism that's not general IP." And, he noted, going with Layer 2 is just plain cheaper.

I don't think Roberts is necessarily wrong about the problem – it just seems as if he's being unduly pessimistic and narrow-minded about the solution. If the situation is indeed as dire as he says, the technology community is likely to engineer a solution. What is it? Well, we don't know yet – and that's the point.

In Nassim Nicholas Taleb's book, The Black Swan: The Impact of the Highly Improbable, he points out that most earth-shaking technological innovations – such as electricity, nuclear power, and the personal computer – came in a completely unexpected fashion. Not even the scientific community anticipated them.

I think this is how the next wave of Internet switching/routing/moving technology will unfold: out of the blue. When it does come, it will likely not involve an incremental improvement to the router. It will be a huge paradigm shift. It will mean that routers are being replaced with something completely different.

— R. Scott Raynovich, Editor in Chief, Light Reading

Page 1 / 3   >   >>
thausken 12/5/2012 | 3:00:06 PM
re: Larry's Router Rant Let's start with fundamentals. What is the problem that needs solving? If there is demand for more capacity (more users? more traffic per user?), then that should mean more money to providers to provide those services, and more money to box companies to make routers. In other words, Internet capacity is not a fixed resource, and the total cost should scale with the total demand. That is a good thing.

If the growth in demand doesn't grow revenues, then the implicit suggestion is that somehow the users shouldn't or won't pay more for getting more capacity. That can only happen if engineers find cheaper solutions or the providers and box companies cut their margins.

I just don't see the connection from the status quo to a collapse of the Internet. If someone can explain that, please do.

Tom Hausken
stephencooke 12/5/2012 | 3:00:04 PM
re: Larry's Router Rant Hi Scott,

Great article! I completely agree with your analysis. So the question becomes: "Where will that next innovation come from?" Here are some thoughts on that:

- Today the main limiting factor is access bandwidth. The eventual technology of choice has to provide a substantial increase in bandwith however FTTx is too costly, takes too much time to deploy and will only be rolled out to a limited percentage of the network, if at all. Therefore the real access solution has to be based on something that works over the existing copper infrastructure.

- Capital requirements for FTTx architectures are prohibitive and the return is uncertain. Yes higher bandwidth is possible but bandwidth is becoming a commodity so spending $Billions to deliver more commoditization is brain dead! This capital expenditure has to be phased in time based on customer demand alone (ie: not build-it-and-they-will-come). Do you think cable companies are just going to sit there and let telcos take their customers back...?

- The solution has to be economically deployable in rural & sub-urban areas. If not, you are spending the $Billions to deploy the commodity into an area that likely is saturated by the competition...who already have a significant head start.

- FTTx architectures seem to only address the bandwidth issue. How are managed services (ie: what carriers are telling their shareholders will provide the revenue in the years to come) going to be deployed without QoS or efficient multicast even if there is a bunch of bandwidth available (ie: How is the network going to become a revenue generator once again?)? Make no mistake, the technology that solves these issues will drive the change into the network.

- Carriers need new wired technology that uses the existing infrastructure to provide a substantial bandwidth increase and facilitate the offering of QoS-based managed services...Oh, and they need it yesterday.

Bottom Line: ALL of these points have to be dealt with at once or you are wasting your money. There is a combination of existing technologies that CAN deliver exactly this, but not today (DSL Rings).

Steve.
rjmcmahon 12/5/2012 | 3:00:04 PM
re: Larry's Router Rant Let's start with fundamentals.

Probably a good idea.

What you're addressing is price elasticity.

http://en.wikipedia.org/wiki/P...

Internet bandwidth is perceived as being price elastic. An engineer's solution to this, which I think you stated, is to increase supply by increasing efficiency and driving out competitors. Electricity followed this approach for 50 or more years and we ended up with centralized production where peak/average load converged, low cost transmission, and energy in a canonical form has been supplied in relative abundance to our entire society.

Now, the real problem is that engineers don't run the country (nor our industry) but bankers do. In that context the primary problem isn't how to advance society but rather is how to make a buck, literally. With a fiat currency a buck (a government backed IOU) gets "printed" when a bank receives a promissory note that the government agrees to enforce.

Let's look at an example of cars. A car mfg builds a car but no money is printed. A car dealer then sells the car to you. You could pay with cash on hand but no new money is printed and the price would be much less than if they bank and car dealer worked together to grow the money supply. So instead the money changers issue a promissory note (much greater than cash you have on hand) backed by your promises to make future payments. Here's the rub, the government won't enforce the promissory note unless there has been some indication that the income stream used to back the notes was legitimate in the first place. In car terms, the bank must receive three payments and only then is the new money backed by the state. At this time, the bank splits the financing fees with the car dealer, the mfg gets paid for the full value of the promissory note, and any failure to make payments is somebody else's problem.

The primary internet problem is that the bankers can't print money for unsecured revenue streams and the incumbents are already in debt up to their eyeballs meaning they can't borrow anymore to build the real stuff, particularly stuff which is price elastic.

New entrants with secure revenue streams could change all of this (real entrants would cause the incumbents to go bankrupt) but the FCC has decided to protect the incumbents (lawyers married to bankers) instead of assist engineers towards building real infrastructure.

So flow switching is a bit of a red herring as the internet problem is a financial one.
thausken 12/5/2012 | 3:00:03 PM
re: Larry's Router Rant rjmcmahon writes:
What you're addressing is price elasticity.

It's not about price elasticity. It's simpler than that.

Whether prices are elastic or not, if you sell more services, you receive more revenues. Period. For example, more users means more revenues, which means more money to buy more routers. This status quo approach is pretty robust.

I don't see how that could lead to a collapse of the Internet.

Tom Hausken
rjmcmahon 12/5/2012 | 3:00:03 PM
re: Larry's Router Rant This shows a strong bias toward AT&T's VDSL approach vs Verizon's FTTH approach. How can you explain Verizon's much higher subscription and growth rate?

I don't know if this explains it but VZ started much earlier and hence one would expect higher subscription numbers. It takes a lot of marketing material to convince and average person to switch service providers. Also, I believe FiOS internet service is better than what u-verse offers so folks interested in internet would prefer FiOS.

Regardless, the fundamental issue is can they earn an ROI for these projects? Having to offer TV to justify the business case suggests to me these projects are doomed. If that's the case, spending the least amount of money required to keep the ideologues at the FCC happy and their fantasy of facilities based competition alive is the prudent course of action.
opticalwatcher 12/5/2012 | 3:00:03 PM
re: Larry's Router Rant "FTTx is too costly, takes too much time to deploy and will only be rolled out to a limited percentage of the network, if at all. Therefore the real access solution has to be based on something that works over the existing copper infrastructure."

This shows a strong bias toward AT&T's VDSL approach vs Verizon's FTTH approach. How can you explain Verizon's much higher subscription and growth rate?
opticalwatcher 12/5/2012 | 3:00:02 PM
re: Larry's Router Rant Some questions to ask about Verizon's numbers:

- How many of the new FiOS customers were originally Verizon DSL customers?
---
Steve Burke, Comcast's chief operating officer said recently "Verizon is taking video customers from us."

- How many of the new FiOS customers are subscribing to Verizon's video service? The last I heard was that the video take rate was <10%. This was the main reason for doing FiOS in the first place.
---
The press release of a couple days ago annouced 500,000 FiOS video customer, adding 2600/day.

- How many of the new FiOS customers are actually Verizon employees/contractors? There was a rumor that FiOS was originally deployed in areas where literally 1 in 200 homes had a Verizon employee in it.
---
How many of the 500,000 subscribers do you think are Verizon employees?


AT&T's approach doesn't supply the necessary QoS or efficient multicast either.
---
I'm not sure how Verizon or AT&T are tranmsitting video through their backbone to their customers. I assume that there are video servers sprinkled around, especially in AT&T's case, which is true IPTV. Of course, live broadcasts would need multicast, as you mentioned.
rjmcmahon 12/5/2012 | 3:00:02 PM
re: Larry's Router Rant Stephen,

The ILECs maximize profits by increasing the price to consumers and by reducing what they pay to vendors. Adding more bandwidth (via DSL loops or whatever) really doesn't help in the above equation as milking the existing plant makes the most money.
rjmcmahon 12/5/2012 | 3:00:02 PM
re: Larry's Router Rant Whether prices are elastic or not, if you sell more services, you receive more revenues. Period. For example, more users means more revenues, which means more money to buy more routers. This status quo approach is pretty robust.

Uh mm, nope not necessarily. If this were the case there would be no problem with monopolies as they would serve the entire market. In the real world they don't unless they are regulated to do so.

http://en.wikipedia.org/wiki/M...

"If a monopoly can set only one price, it will produce a quantity where marginal cost (MC) equals marginal revenue (MR). The monopolist will then set the highest price at which that quantity can be sold. This, the optimal price according to supply and demand theory, is above the competitive price (Pc) and below the competitive quantity (Qc).

As long as the price elasticity of demand for most customers is less than one in absolute value, it is advantageous for a firm to increase its prices: it then receives more money for fewer goods. With a price increase, price elasticity tends to rise, and in the optimum case above it will be greater than one for most customers.

The economy as a whole suffers when monopoly power is used in this way because the extra profit earned by the monopoly will be smaller than the loss in consumer surplus. This difference is known as a deadweight loss."
stephencooke 12/5/2012 | 3:00:02 PM
re: Larry's Router Rant Some questions to ask about Verizon's numbers:

- How many of the new FiOS customers were originally Verizon DSL customers? In this case the FiOS business case evaporates as they have spent $Billions to gain a relatively small amount of ARPU because only the DSL/FiOS delta counts towards that business case.

- How many of the new FiOS customers are subscribing to Verizon's video service? The last I heard was that the video take rate was <10%. This was the main reason for doing FiOS in the first place.

- How many of the new FiOS customers are actually Verizon employees/contractors? There was a rumor that FiOS was originally deployed in areas where literally 1 in 200 homes had a Verizon employee in it.

AT&T's approach doesn't supply the necessary QoS or efficient multicast either. QoS and multicast simply don't make sense on point-to-point links unless they are implemented throughout the shared upstream network. In an ideal implementation there would only be a single instance of the Super Bowl in HD traversing the network instead of one per STB; even if there are multiple STBs in the same house.

Steve.
Page 1 / 3   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE