Older: 
Megatrends Part 1 cont’d: Hyper-Connectivity - The Implications
Location: On Flight from London to Ottawa
Continuing the dialog on mega-trends from my last blog…
(First, let me just say that I sincerely appreciate the very thoughtful and insightful comments people have been making. It's definitely the start of what I hope will be an interesting and on-going dialog. With that in mind, I'm going to start to intersperse my more in-depth technology posts with responses to some of your great questions, comments, and observations. I think that might be the most effective way for me to handle some of the volume and to keep the conversation going on some of the topics you've raised.)
So, more on hyperconnectivity...
If the future is about a hyper-connected environment, where everything that can and should be connected is connected, then what will this mean to the telecommunications industry? What network technology will be needed to support this trend and its impacts?
I’ve been through this thinking exercise before, as I’m sure many of you have been. My best example was in the late 1990’s when a few industry colleagues and I were considering the fact that local area networks lacked an essential element of intelligence that would be critical to realizing the goals of policy-based networking. That missing intelligence was the need to understand who was actually using the network. In other words, what was missing was human identity as an element of network intelligence.
We collaborated on a technical paper, documenting a model that used the extensible authentication protocol over Ethernet (EAPOE) to bring authentication into the network in a flexible and effective way. When I first began talking about this in the late 1990’s, most people’s response was that there was no need for this because the network was doing just fine as a big, fast, cheap pipe and intelligence of this kind was unnecessary.
Well, a few years later, after we experienced the Code Red worm, Nimda, and a host of other denial-of-service attacks, along with a surge in regulation that required businesses to protect and control their IT systems from unauthorized users, the idea that an intelligent authenticated network might be needed was suddenly being widely discussed.
The reason I bring this up is not to suggest that I was smarter than any other IT person but rather to point out that the most important new technologies are the result of a need and, in many cases, that need is seen far in advance and solutions are formulated well ahead of the time when both the need and the solution become obvious. Had we not started work on IEEE 802.1X (what EAPOE became) far in advance of the need, we would have had a far less evolved toolset to deal with the inevitable security and control issues that some of us knew were a mega-trend that was inevitable.
Hyper-connectivity is as significant a mega-trend as security threats ultimately became to IT. Imagine, for example, what your IT systems and networks will look like if the number of connected nodes increases by 10 or 100 fold. Will the standard configuration process of element-by-element configuration via such primitive interfaces as CLI (command line interface) work any more? CLIs are a nice rudimentary way to make specific actions happen on a switch or another network device, but imagine if you had to define services in a dynamic network consisting of 100,000 ports with a wide range of end systems attaching to them dynamically. Could you even use the existing network management systems to deal with this?
Even the most robust policy-based management systems today will probably be inadequate because they only scale so far. Bear in mind that most IT budgets today are not set based on the size of the network, but rather on a percentage of the total organization’s gross budget and/or, in some cases, on a cost per employee. Neither of these funding calculations takes into account a hyper-connected network. Since there is usually little chance of increasing IT funding, and we do not have more hours in the day, we must look to new technologies to achieve operational efficiency as networks scale.
So how will we deal with this inevitable trend? I put forward that there are three technical areas of innovation that will help us scale to operate in the hyper-connected world. First, we must use intelligent automation in our management systems so that any task that can be automated is automated, freeing human operators to focus on the exceptions. Second, we must de-layer our architectures to make the network system simpler (fewer nodes to manage). And, third, we must make the core functions and protocols extensible so that we can use one method of service to address a very diverse environment. We must build networks that architecturally are prepared to deal with the op-ex part of the equation by simplifying the operating paradigm at a rate faster than the inevitable growth of the system.
Today, too many enterprise and carrier networks insert complexity into the network to deal with one type of end system or application and then repeat that process until dozens or hundreds of unique abstractions or overly complex protocols are running in parallel over a single network. At the risk of being called a heretic, some examples of this are the use of MPLS technologies to deliver Ethernet services over an Ethernet network (why not just make Ethernet services smarter and de-layer the system?) or using a proprietary routing protocol to build resiliency into one part of a network (constructed by one vendor’s gear) instead of using standard protocols to create resiliency throughout the entire network. At the time decisions like these were made, they made sense and, in some cases, may even make sense today. However, if we think about the future where large, complex heterogeneous networks with hugely diverse end systems and applications are attached, simplicity will matter more than we can imagine.
One of my jobs at Nortel is to make sure that we see the future trends in advance so that we can research and develop the solutions needed to make sure that our customers are prepared for that event.
Consider this blog entry as an early warning on a trend that, by most analysis, is inevitable as we see value in connecting more and more nodes and people. (Remember Metcalfe’s Law: The value of the network is equal to the square of the number of connected users, where a network with one user, for example, has a value of 1, a network with 2 users a value of 4, a network with 4 users a value of 16, and so on…).
In the new Nortel, this kind of dialog is now being internalized and the R&D community is focused on solving for this scenario. Will we get it completely right? Of course not, but I believe we see it clearer than many and will not be surprised by it.
Apologies for another long blog entry, but I thought I would take advantage of finally having some time to write after a very busy week and got a little deeper than originally intended. 
[…] Older: Megatrends Part 1 cont’d: Hyper-Connectivity - The Implications […]
February 11th, 2007 at 9:40 pm from John Roese’s Blog » Blog Archive » Discussion of Comments on Hyper-Connectivity