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
The Broadcom BCM4705 processor is the highest performance chip in the 802.11n/MIMO Intensi-fi processor family. It integrates a powerful 300 MHz MIPS32 core estimated at 500 DMIPS with 4-way set associative 32 KB instruction cache, a 32 KB 2-way set associative data cache, and a 4 KB prefetch cache. An enhanced 10/100/1000 GigabitEthernet MAC controller with TCP Segmentation and checksum Offload (TSO) hardware acceleration engine substantially enhances system throughput.
The BCM4705 integrates two USB2.0 EHCI Host ports and one USB2.0 Device port. The USB2.0 device port allows connectivity to PC or consumer electronics (CE) devices equipped with USB host ports. The BCM4705 supports a PCI 2.3 host interface and a USB host/device interface for adding application-specific peripherals including BCM4321 2x2 MIMO, BCM4318 single-chip 802.11g, BCM4309 802.11 a/b/g, BCM4320 USB2.0 WLAN, BCM5397 Gigabit Ethernet switch, and future generations of Broadcom wired, wireless, audio/video codec and multimedia processing devices.
if you want to have a discussion about the concepts of utility and substitution i am happy to do so. but i am not interested if the real agenda is to simply launch in to the same old harangs (sp?).
Actually, it is something I have an interest in discussing though I like to better organize my thoughts on the issues. I probably won't have a chance to do that until after Halloween. Hope the holiday is a good one for you.
i understand the world, markets, companies, consumers, analysts, regulators, governments and models are not perfect. however, i did not present any models, i presented basic concepts. marginal utility is almost always present, even in monopoly markets, which is why the general and simple form of a demand curve presented in text books is sloping down. to think otherwise is to suggest the general form is horizontal. even substitution effects can be present in a monopoly market. certainly from the consumers opportunity cost and spending perspective. that said, in many markets today it is not a monopoly. whether it becomes so in the future is a matter of opinion - and you are certainly entitled to yours.
if you want to have a discussion about the concepts of utility and substitution i am happy to do so. but i am not interested if the real agenda is to simply launch in to the same old harangs (sp?).
The problem I have is when real world outcomes contradict the possible solutions predicted by the human models, many tend to become stronger believers in their models instead of questioning them in the first place. This can be observed in all walks of life. For example, [de]regulator Alfred Kahn predicted in 1981 that there would be many cable providers competing with duplicate facilities in the face of natural monopoly characteristic and completely missed. Today there has been no adjustment to his beliefs from what I can tell and he's become a sock puppet for a "deregulated" phone industry (which has trended towards Ma Bell) and still denies the existence of natural monopoly. Many engineers force fit linear solutions into nonlinear problem domains because that's the only way their models can come up with any "answers" at all. Others see an eight year old die from leukemia (or the father of two young children from pancreatic cancer) and that brings up the question of theodicy. No longer so for me.
All I'm really saying is if that the models aren't getting to the outcomes desired then I don't want to lean on faith. This may not ever resolve the dialectics assumed by "markets vs. regulation" or "individual freedoms vs societal coercions" but I'm ok with that.
As far as Wednesday's costume, I don't really believe in "Evil Operators" so I'll probably go as some sort of treat as hard to believe as that sounds ;-)
so is this going to be another one of those highly entertaining and enjoyable (from my perspective at least) dialogs that ends in you twisting what I say until you are convinced it is an ultra extreme libertarian view point and ending in you attacking the twisted view you have created in stead of what I actually said?
1) what is your argument that consumers will consume an infinite amount of bandwidth at the same price per bit?
2) what is your argument that operators will purchase an infinite amount of routers at the same price per bit?
Answers to 1 and 2 can be used to discuss marginal utility.
3) what is your argument that there are no substitution effects between different vendors and technologies?
4) what is your argument that there are no substitution effects between the multiple suppliers of video and internet services?
5) what is your argument that there are no substitution effects between video and internet services and all the other things in the world consumers can spend their money on?
your answers to 3,4, and 5 can be used to discuss marginal rates of substitution.
6) what is the marginal cost of bandwidth? you assert with confidence that it is extremely low? so pressumably you know.
7) just out of curiousity, which operator are your dressing up as on Wednesday night ;-)
"It's really strange to me that some still believe in competition as the guiding hand with respect to digital computer networks."
pressumably this is derived from my long dissertation on competition during my last post, or did you just throw that in for my amusement?
I don't really buy Larry's argument that changing the way packets are delivered (flow-based vs. stateless) is going to make a lot of difference.
We've seen a bunch of new forwarding technologies in the past few years and none of them has made an impact, because the competitive advantage that Cisco and Juniper have is not in the forwarding plane. It's in the control plane. They are the only two companies that can make a control plane that is halfway stable in the face of ever-growing and ever-more-volatile routing tables.
Routing is still a relatively small component of the total transmission mix when you add up digging fibre into the ground, lighting it, and billing for the end result.
If and when routing becomes too expensive, it is my belief that carriers will push more of their transmission through non-routed paths, using more direct optical paths and cheaper switching-based forwarding technologies (i.e. bypass the routers) rather than fiddling at the margins and risking stability with new routing paradigms.
This will have the unintended effect of making the routing topology that much more complex, entrenching Juniper and Cisco in the routing market, which will probably start to shrink as a share of the total market.
This assumes a functioning market with substitution which is an ideal.
On pricing, the first glass of water that sustains life is worth much more than the water we use to spray dogs humping in the yard. The access network owners know this and have been using their distribution position to strangle the market and keep bandwidth prices from reaching marginal costs.
Transport providers haven't had this luxury. They can use technological advantages for a brief time but they can't really exclude their competition (duplicate providers) from doing the same. Also, the access networks and transport networks aren't decoupled despite human efforts to model them as such (any traceroute will show this.) The marginal cost to move a bit is extremely low and pricing power moves to those who monopolize the bottlenecks. Sprint selling out to VZ and AT&T to SBC are expected outcomes in such conditions.
At least with the rest of our utilities we build out regional monopolies that serve a community in its entirety. It's really strange to me that some still believe in competition as the guiding hand with respect to digital computer networks.
There is still the raw cost of just pushing bits around at high speeds. SERDES, buffers, power/cooling, switch fabric. No matter how creative the routing engine is, you still need to do the boring, yet expensive work of pushing all those bits through the interfaces. Larry needs to buy the same expensive optics as everyone else.
The blogs and comments are the opinions only of the writers and do not reflect the views of Light Reading. They are no substitute for your own research and should not be relied upon for trading or any other purpose.
To save this item to your list of favorite Light Reading content so you can find it later in your Profile page, click the "Save It" button next to the item.
If you found this interesting or useful, please use the links to the services below to share it with other readers. You will need a free account with each service to share an item via that service.
Digg
Del.icio.us
Reddit
