Monday, June 27, 2011

Wither (Non Social) Networking?

10 gigabit router
There are a lot of threads in play right now in the networking space. From an infrastructure perspective networks are the foundation of information technology. When people talk about the cloud they think about huge data centers storing information and providing applications and other services to people everywhere and at any time. To make that dream work you need sophisticated and complicated networks. In some ways networks should be thought of as utilities, at least from the perspective that any business should assume that the physical plant is in place to service their employees needs, but from an architectural and maintenance perspective care and feeding is increasingly important and difficult.

There is a theory that states that networking is rapidly becoming a mature market. A mature market is one in which innovation is relatively slow, product cycles are measured in many years and thus industry revenues are relatively stagnant. Think Refrigerators for instance or washers and dryers. The problem with mature markets is that they don't have much opportunity for growth. Investors aren't real excited by mature markets. While they can be good sources of cash in the form of dividends stock prices tend to be stagnant and the excitement factor low.

Personally I think we're at least a decade away from having to worry about networking falling into this category. I have two reasons for thinking this way.

The first is the shift to "The cloud", or highly centralized model for IT. The cloud is really just a new label for an old concept. In IT everything old is new again about every twenty years. We're essentially shifting back from a highly decentralized model to one where pretty much everything lives in the data center. This shift back to a more centralized model has a number of implications in the network world, not the least of which is network design. Data centers, already bandwidth hogs become even bigger consumers of network resources.

Locality plays a part as well. If you're a big company you're going to have your own data center(s). In the past you could more or less get away with just building bigger and bigger pipes as your needs increased. Now you need not only bandwidth but sophisticated resource management as well. Virtualization and resource management don't happen in a vacuum, they depend very heavily on and need support from the network and impact the network architecture.

Data center consolidation will not be cheap and network equipment and service vendors are going to be picking up a significant chunk of the money being invested here.

If you're a small to medium sized business you may outsource your data center all together. The networking infrastructure you had on your edge that was previously adequate no longer is at this point. You'll need a lot more bad width and a much higher level of reliability. If you connection to the Internet goes down in the cloud based world you're not just inconvenienced, your dead in the water. This scenario is mitigated somewhat by smartphones and cellular wireless hotspots but there is only so much bandwidth available in those scenarios. Distributed work forces don't have to worry so much about this but I think we're going to see a fair amount of investment in network edges as companies come to realize that getting rid of their data centers may save them money but only if they make the right strategic investments to assure that their employees have a high probability of being able to get their work done. Perimeter routers and security devices figure prominently into this equation.

The other area I see for growth is at the campus level. Specifically in the need to move from IP and weak identity based security management to strong identity based authentication and access.  Both computers and users can be grated access based on security mechanisms such as smart cards and 802.1x. The thing about these technologies, particularly 802.1x is that they can't easily be done with existing access layer switches. This means a substantial investment needs to be made to get into the game. Initially not everyone is going to need or want 802.1x, but given the sophistication of attackers and highly mobile work forces I don't think there will be many companies that can afford to take the risk of not going this route.

Another issue with access layer switches is management. Managing switch infrastructure is complicated and messy. Currently the majority of switch configuration lives on each switch. This means changes can only be made by connecting to a switch and issuing commands. This isn't a big deal if you only have a handful of devices but for campus networks that often include hundreds of switches it is infeasible. What is really needed is an architecture that requires just enough configuration information on each switch so that it can phone home on boot up and get its configuration from a central server or controller. This is already done in the wireless world.

Of course service providers aren't immune. Data has to get from point A to point B and they provide the pathways that allow this to happen. Bandwidth requirements will continue to increase while the need to minimize latency is increasingly important all of which means continued investment in new equipment.

One way of looking at all this is that for large companies IT spend on desktops and departmental IT is likely going to decline but at least some of that money is going to need to be invested in the network. For small to medium sized companies the story is slightly different but the punchline is the same. we'll likely be seeing less money spent on data centers and personal computers but again there will need to be a larger investment in network infrastructure to support this shift to the cloud.

So basically what I'm saying is that the shift to cloud computing in conjunction with ever increasing security needs combine to actually increase the need to invest in network infrastructure and thus networking is unlikely to become a mature market for several more years at the very least.

Image via Wikipedia
Enhanced by Zemanta

No comments:

Post a Comment