For 30 years or more, networks have been defined as a set of interconnected hardware boxes, primarily switches and routers. But in today's dynamic environment, it makes more sense to think of networks as a collection of things that interact to perform a service. The endpoints, the people, the devices and the data they generate are the stuff that really matters.
When defined in this way, it becomes clear that networks are complex infrastructures and that, as a result of developments like BYOD, cloud computing, M2M and the Internet of Things, they are rapidly becoming more complex. Thinking of networks in this way is a more meaningful definition for businesses, whose critical processes are executed on a network of devices, applications, and sensors that may be virtualized in their own data centers and distributed across different clouds and managed infrastructure providers.
Network complexity increases as more users join the network with more devices, each of which devices has more sensors and apps, all sharing more information with each other. This means that complexity, not bandwidth, is now likely to be the barrier to network growth and scale.
This scenario also indicates the need for some kind of tool that will enable network management to measure network complexity, to quantify and understand the problem before attempting to manage it.
Complexity can be judged by the number of interconnected endpoints, the number of connections between these endpoints, and the way information moves through these connections to perform business processes. The more endpoints that are deployed to support a business, the more interactions between endpoints and the more critical business processes that employ those endpoints, the higher the level of complexity.
The Network Complexity Index
Stu Bailey, CTO of Infoblox, and Professor Robert Grossman of the University of Chicago, have developed a mathematical equation for measuring network complexity. Their Network Complexity Index (NCI) evaluates the number of endpoints on a network and how they interact to perform key business functions. The traditional approach would count the number of network infrastructure devices and map the interconnecting wires, but the Bailey-Grossman approach goes further: a lot further.
The Bailey-Grossman equation reveals the NCI of corporate networks by measuring the interactions between the endpoints--the users, devices, sensors and apps--using Domain Name Systems (DNS) as a data source. However, companies do not record these interactions, so for the equation to work, it would need to be implemented by software that could access and capture this data in real time. And the software tool that Infoblox has developed in order to generate a NCI figure is a free, open-source software application known as Tapestry, available for download here.
Infoblox markets DNS, IP Address Management and Network Automation Tools, so at first sight it might appear that the NCI was conceived and developed in order to sell more products. At the end of the day that might very well happen, but Bailey says the intention behind Tapestry is simply to raise awareness of network complexity, an issue that will otherwise limit the ability of networks to scale in future.
The ultimate aim of uncovering a business's NCI is to better understand the change in the level of reliance that a business has on its network: how intertwined the network is with all of the core functions of the business. By taking regular measurements, a business can gain information to understand just how fast its network complexity is growing. According to Bailey, this rate of growth is more important than a single, isolated measurement.
A Simple Example
The Bailey-Grossman equation could be applied to a record of every phone number that called every other phone number at some point, including conference calls with multiple numbers connected, and would be totalled into a single figure. The resulting NCI reflects the complexity of the network, in this case a network of phones with people talking to each other.
If you were to compare the NCI of a town's phone network with that of a comparatively sized corporation, the town's figure would probably be higher than that of the corporation. Even though both networks might have a similar number of endpoints, the corporate calls would be more uniform and that would result in a less complex network.
Reducing Complexity
As indicated earlier, proprietary hardware boxes--switches, routers, and load balancers--define today's networks. However, knowing a network's NCI should encourage managers to think along different lines, i.e. focus on endpoints, users, devices and the data they generate. And a logical consequence of thinking differently is a new perspective on solutions that will reduce network complexity: automated network management today and Software-Defined Networking (SDN) in future.
Automation allows network managers to offload mundane day-to-day tasks and focus on those that will add value to their organization. In turn this makes the network management operation more efficient and more cost-effective.
Network complexity is expected to increase over time in most modern organizations, with the Internet of Things adding meters, sensors and apps to networks. Organizations will need to manage this complexity carefully, with a particular emphasis on ensuring that rising complexity does not translate into increased cost. Tapestry is designed to help businesses cope with growing complexity and enable networks to scale effectively, while keeping costs down as the network continues to expand and complexity rises.
SDN is a hot topic and the evaluation of a network's complexity index can be seen as a related development. SDN introduces programmability to the network, and one of the benefits is the ability to view networks as communication fabrics that connect devices and that are partitioned in certain ways so that they are easier to manage and control. The network therefore becomes a network of networks. Managing lots of small networks is easier than managing one large network.
It is worth noting that FlowForwarding.org, the site from which Tapestry can be downloaded, is an open source community promoting software defined networking based on OpenFlow. However this indexing application runs on regular hardware-defined networks: if it didn't, then evaluations would not be possible. But Tapestry itself can be deployed on a network white box, thereby providing an easily accessible starting point for organizations interested in SDN.
Conclusion
It's hard to predict when SDN will break into the mainstream but, when it does, it has the potential to reduce both network complexity and associated cost. The market correction that SDN promises could occur quickly, and those businesses that have planned for its arrival will be the first to gain a competitive edge. Calculating the network's NCI is the logical first step in that plan.
