Bandwidth Metering, Usage Caps and the Optical Upgrade Cycle Talk….

Historically users have preferred flat rate pricing, which is often in the favor of the service provider.  I have some thoughts on this concept, but I would state up front that my thinking is based on these two papers from Andrew Odlyzko and this one by Peter Fishburn.

The net of the decision for users to prefer flat rate versus usage billing has to do with (1) insurance effect as people like to have capacity available in case of unforeseen circumstances, (2) over estimation of what they typically use and (3) what Nick Szabo termed as “mental transaction costs” but I would tell you in that in real life it has more to do with the prospect theory.  The prospect theory was developed by Amos Tversky and Daniel Kahneman were pioneers in the study of cognitive science – how humans handle risk and the manner in which cognitive powers can be applied to economics.

In short, the prospect theory suggests that people are more motivated by losses than by gains.  They developed a model which showed that the way in which an investment is framed affects the choice that people will make.  The term “frame” is being used in the economic context, but you would call it “marketing” or “positioning.”  A foundation of behavioral economics is that framing biases affect how investment and lending decisions are made.  Investors are more likely to be motivated by losses, than by rational choices, hence it is better to buy or overbuild than get caught short in a networking world.  It turns out that the prospect theory is true for buying a service…like a network connection…as well as making a financial investment.  Humans by a large margin tend to want to over buy and the few cases that you find when people under buy, it is because they do not have the financial means to avoid points 1-2-3 above.

There is a long list of failures in the world of bandwidth management…companies like Enron Broadband…which does not mean that it is a poor business path – but rather that the few successful business examples played on the three themes from above.  A few examples:

• Akamai Dynamic Site Accelerator:  This is an extremely profitable business for AKAM…what are they selling?  Peace of mind.

• Nortel 10G LH: When Nortel commercialized their 10G optical systems, they were betting that the growth of the internet would rapidly drive telecom service providers to deploy 10G systems faster than projected – but when they went to sell the 10G system, they did not find any buyers for their more expensive 10G system.  John Roth (NT CEO) made deal with Joe Nacchio.  Nortel would sell 10G systems to Qwest at 2.5G prices and when the traffic on the 10G systems exceeded the capacity of the 2.5G systems Qwest was considering, Qwest would pay Nortel a premium for use of the capacity above 2.5G.  It was a win, win deal.  If Qwest’s network never needed the capacity that Nortel’s 10G systems provided, then they had not over paid – but if the network usage grew, Qwest had the bandwidth installed in the network to support the unanticipated demand.  What was NT selling?  Peace of mind.

I can come up with half dozen more examples, but this is really a long introduction into some thoughts about bandwidth metering that I started writing about a few weeks ago.  One of the developments I see going forward is really two networks.  The human network is going to have usage caps and the machine network will not.  Humans are living with them today in the mobile market.  Just go to any wireless service provider website and try to buy a 3G or 4G device without a data plan.  Not going to happen, unless it is wifi only.

One of the reasons I am not in the exaflood camp, is I cannot find a reason why service providers (SPs) would let that happen.  The SPs are already in a “bits R us” business, so why would they want to continue to roll down the price curve and let users consume/create more and more bits for less?  It is simplistic to think that OTT services will ruin the network.  The real question to consider is how will OTT services be monetized and where in the network will value be created?

On a day in which CIEN is (16%) after reporting earnings and a variety of analysts are shouting “only a pause in the optical upgrade cycle” I think it is prudent to really think about what upgrade cycle are they referring too.  This is not a blog post about CIEN, I think there is a long life to service provider network upgrades to 40G, 100G, SONET/SDH replacement, evolution to packet, etc…the question I am exploring is who or what is driving the network upgrade/evolution and should we expect the winners to be the companies before us today because that is precisely what many of Wall Street analysts what you to believe.

I think what is missing is an understanding of how I/O virtualization is going to affect network design.  We continue to move into a persistent connected device network in which we want to consume more long form or big data content.  If the next evolution of statistical gain is not achieved in the switch fabric, in the metro or core of the network, then where will it be found?  Will it be achieved in the I/O and the CPU?  If that assumption is true then how will the network interface change?  If I/O virtualization is going to enable the CPU to run at higher efficiency rates, then that has the potential to put pressure on the network.

The exaflood debate and how it affects the network is really a two part debate: (1) machine or (2) human.  Where will the flood occur if it occurs at all?  How will devices such as routers, L2 switches, optical switches, WDM elements and network appliances interface with the server which I call a compute point?  Do these devices now need to be I/O and CPU aware?  Something tells me it is all about how the network interfaces with the compute element and I would submit that the number of companies thinking this way is very few.  Most companies have been focused on the 2009 catch-up spend.

/wrk

* It is all about the network stupid, because it is all about compute. *

** Comments are always welcome in the comments section or in private.  Just hover over the Gravatar image and click for email. **