What are the similarities between social concepts and the SDN model?

The social principle seems to have a strange connection with network engineering, design and software defined networking, perhaps as many people say a seemingly unrelated principle can often be applied in other scenarios. The principle of social subsidiarity has certain guiding significance for the use of network control planes and the design of SDN models.

[[243192]]

The principle of subsidiarity is primarily a social and legal concept, defined by the Oxford English Dictionary as “the principle that central authorities should have a subsidiary function, controlling only those matters which cannot be controlled locally.” But the concept can be applied in other contexts.

For example, when talking about SDN models and network design, it is best to remove its political and cultural colors and state this principle in a more general way as follows: The principle of subsidiarity is that decisions should be made as close as possible to the location of the information needed to make the decision.

Network operation and ideal state

In this regard, we can make some observations on how this principle can be applied to network engineering and SDN models. A network never has only one state; it has at least two states.

• The first state is the reality in which the system exists. This is the operational state of the network, including its connections, configuration, link utilization levels, queue depths, and other physical and logical states.
• The second state is the ideal state. This is the policy goal or desired state, which is a set of goals that the network is designed to achieve. The most direct expression of this source of truth is the idealized design, or how the network originally worked.

We need to further investigate the ideal state to determine how the principle of subsidiarity applies. The ideal state often results from a series of design decisions made in the face of a business problem that needs to be solved. For example, one application may need to be prioritized over another, or a specific type of traffic may need to be encrypted, or a specific type of traffic may be dropped in the event of a failure.

How the two states collect their information

The actual operational and ideal states receive their information from different places. The operational state is obtained from the network itself, including its links and devices. The ideal state is obtained from the network designers and business intent. The following figure illustrates these two states and how they interact.

In this example, business intent is driven through the design process to create the ideal state of the network. At the same time, the operational state can be understood by inspecting the network itself. Gaps between these two views are discovered through telemetry, while configuration and other information are imposed on the network in an attempt to align it with the ideal state.

Questions about this process might include the following: How long does the telemetry configuration cycle take? In other words, how long does it take for the ideal state of the network to be factored into the operational reality? And, how long does it take for the new ideal state to be imposed on the operational state? What happens when the ideal state and the operational state don’t match?

The first question needs to be answered based on the second question. When the operational state and the ideal state do not match, traffic flows through the network in a sub-priority manner, which wastes resources. Another result is that the network simply does not support business needs. Unfortunately, it often takes too long to impose the ideal or desired state on the operational state.

The trade-off between distributed and centralized control

Fully distributed control planes are built on the observation that the operational state of the network is what matters most. When the operational state changes, decisions about how to act are usually made close to the actual network and its components. Therefore, distributed control planes react faster to changes in the operational state of the network, but they react much slower to changes in business intent.

On the other hand, centralized control planes are usually designed to implement business intent. The operational state of the network is transmitted to the controller, where it can be merged with the business intent and then reflected to the network as operational state. This means that centralized control planes react quickly to business intent, but react slower to changes in operational state than distributed control planes.

So far, the subsidiarity principle has revealed a new way to look at network design and SDN models, exposing the pros and cons between centralized and decentralized control planes. For example, the main problem with SDN is its slow reaction to link and node failures. This intuitively seems to match the slower reaction of a centralized control plane, because software-defined networks are usually types of centralized control planes.

Network with two control planes

The Subsidiarity Principle explains a phenomenon that most engineers already understand. So how can it be used in current SDN models to improve the situation? The principle states that decisions should be made as close as possible to the information needed.

Because there are two real states in the network, there should be two control planes representing the two states for decision making. Each control plane should make different types of decisions based on business intent and the actual network. The results of these decisions should be sent to a central location and then fed back to the decision control plane.

It makes sense to allow the distributed control plane to make decisions locally based on operational changes. At the same time, the centralized control plane should be allowed to make decisions based on business intent and somehow reflect those intents into the business state. The ideal state becomes a merger of operational reality and business requirements.

The SDN model divides decision-making power based on the location of the information required for decision-making. It is a powerful network model for responding quickly based on actual network conditions and business needs.