The Network is at Apogee
The network is at apogee. There I said it. After thirty years the complexities that inhibit the network have reached the point that we must now begin the process of taking apart the complexities that we have built. There are armies of people who will be opposed to what I am writing. They are the people rooted in the past. They are comforted by the familiar, the complex, the specialized knowledge they possess. It is a closed society in which entry is obtained by knowledge of closed system. I believe we are at point of lowest entropy for the network and thus the lowest utility to man and that is the point or points when revolutions can begin.
I have taken some motivation for this post from David Isenberg’s “Rise of the Stupid Network.” If you take the time to read this paper, two epiphanies may occur to you as they did to me. The first is how much of what he wrote in 1997 is applicable today, which I will cover in the subsequent text. The second is that we became our own worst enemy in attempting build the stupid network.
We based the stupid network on IP, complex control planes and protocols for each solution. In building the internet, web 1.0, web 2.0, social media, video, virtualization, cloud and all connected services we enjoy and pontificate about; we built the complex network. The complex network is so intelligent it is dysfunctional and has no place in the future. What we are about to embark on is the creation of a new network paradigm that for the first time in thirty years can be enabled with technology found in compute and storage.
In reading through the Rise of the Stupid Network, many of Isenberg’s arguments against the telephone company are applicable to the network we live with today. I have taken elements of his paper and adopted the concepts to our present day condition:
Design by assumption works as long as assumptions hold. It seems a week does not go by that we do not hear of service outage caused by a network outage. Recently, Blackberry’s data service suffered a series of cascading failures due to a switch or router failure. Why do we allow network elements take down a service or the network itself? If we are powerless to prevent the fault and the fault requires days to repair; can we be proud of what we built? I am sure the team at Blackberry never assumed this type of failure could take down their network as they had engineered and tested the network to respond to these types of failures by employing complex backup alternatives, but the complex network did fail.
Network design is guided by the principal of scarcity – not plenty. For the past fifty-one (51) years the network was a scarce resource. The result of concern for the scarcity of network resources is the unbalanced, complex network we have today. It is full of stacked over-subscribed elements that have been partitioned and divided and connected many times over and over until we have no idea how it is working and that is why we are fearful of any change in the network. We operate on the principal of do not fix what is not broken and if it is working, do not change it.
Network design is guided by the needs of the network – not the needs of the compute function. The architectural manner in which we have been designing networks has changed little in thirty years. We design networks today based on principals from a bygone era.
The network works as long as engineering assumptions do not change. The network has been and continues to be inflexible to the changing needs of compute and storage. We would have no issues with the network if the other technology elements had remained static, but they are not static, they are dynamic.
Who are the main beneficiaries of the current network? If people who use networks to enable compute and storage are not benefiting from network, who is? My answer to that question is the incumbent suppliers of networking technology. If you look carefully, you can see solutions that are rooted in the past and sustained by the self-affirmation that these solutions are meeting the needs of customers.
New customer demands are homogenized into the existing collective. When is someone going to offer a new solution? A solution that breaks from the past and a solution that is bold and truly revolutionary. When I look at what customers want and then look for the solution in the market all I find are homogenized solutions, drawing on the past to extend what we have been doing for decades. New solutions are increasingly complex and we are made to feel better about this complexity because we are given a sedative of tools to help engineer the traffic flows and provide security for the network. This is great until the complex network breaks and we have to try to fix it.
The network to date has missed the age of plentiful computing. The network today is like a time capsule. It is a beautiful thing. We can open the capsule and pull out a network design from 2001 or 1991 or 1981 and it looks the same as a network design from 2011. We would also find in the time capsule:
- Kodak Instamatic 100 (1963)
- MITS Altair 8800 (1975)
- Atari Video Computer System (1977)
- Milton Bradley Simon (1978)
- Texas Instruments Speak & Spell (1978)
- Sony Walkman TPS-L2 (1979)
- Commodore 64 (1982)
- Radio Shack TRS-80 Model 100 (1983)
- Cisco AGS (1989)
- Nintendo Game Boy (1989)
- Netware 3.0 (1989)
- OS/2 2.0 (1990)
- Telematics X.25 Packet Switch (1990)
- Cabletron MMAC-8 (1992)
- Iomega Zip Drive (1995)
- PalmPilot 1000 (1996)
- Fore Systems ATM Switch (1999)
- Sony PlayStation 2 (2000)
- Apple iPod (2001)
A new network philosophy and architecture will replace the complex mess we have created. The early signs of this trend can be seen in Hadoop and Cassandra clusters. Buy low cost compute, cluster the technology; purchase more than you need because it is inexpensive, if some of it fails fix it in the morning or after the weekend because you bought enough to handle the work load. No one ever has excess amounts of routers standing by because you cannot afford to buy excess amounts of routers to share the work load because the complexity of engineering traffics loads over clusters of routers requires teams of people performing complex calculations using elaborate software tools.
Over the past weekend I posted a 17k word post on revolutions and how they start. It had nothing to do with technology, but it is applicable because all of the required elements for a revolution in networking are now in place. We have a network dominated by the old regimes of networking. They are no longer the upstarts; the lean, hungry, evangelical revolutionaries of their youth, but rather they have grown up and become their father’s IBM of 1991, the Wang of late 1980s and DEC of early 1990s.
The network is in a financial crisis. Financially the architectural legacy of past thirty years cannot support the demands of the network going forward. The network is starting to show signs of structural weakness as incumbents are losing control of thought leadership. New sources of thought leadership are emerging outside of the areas controlled by the old regimes of networking. Politics is entering the technology buying cycle process as IT leaders are questioning the path forward. These CIOs and CTOs as well as the operations people are showing signs of receptivity to new ideas and new solutions; because they know that past practices have been exhausted.
These past practices have served us well for many years, but the time has come to start anew. Thought leaders are emerging in the technology ecosystem from AMZN, GOOG, MSFT, Facebook and others. They are seeking a new network. Someday soon when the revolutionaries emerge these thought leaders and intellectuals will transfer their allegiance from the old regimes to the revolutionaries and that is when the revolution will begin. Class antagonism between teams in IT organizations already exists. The sycophant router era is coming to an end. All that is needed is the spark to start the revolution. We need our Bastille Day, our Shot Heard Round the World.
* 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. **