Dynamic structuring services. Introduction to Service Function Chaining (SFC)

The implementation of value-added services relies on the activation of various service functions. The trend is for operators to progressively rely on shared infrastructures, particularly in the core network. However, to meet specific customer requirements, it is often necessary to differentiate traffic processing by activating specific functions. For example, to meet Quality of Service (QoS) requirements, it may be necessary to activate TCP or http acceleration functions, while traffic associated with other applications, for example, will need to be inspected by a DPI (Deep Packet Inspection) function. Today's deployment options for customer traffic to be handled by specific functions are based on fairly heavy routing configurations, or even the implementation of specific architectures, sometimes redundant in the most extreme cases. Network topology and the difficulty of dynamically implementing new value-added services within acceptable timescales are also an obstacle to a faster opening-up of advanced customer services, in line with operators' marketing options.

The aim of the new Service Function Chaining (SFC) functionalities is to respond to these constraints and open up prospects for greater flexibility and dynamism in the implementation of services tailored to customer needs, while ensuring that this does not lead to (greater) complexity, particularly in the management of networks and services. Moreover, functional chaining can also be advantageously implemented to satisfy traffic engineering constraints specific to each operator.

Functional chaining is a technique for sequentially calling up elementary functions, such as NAT (Network Address Translation), firewall and DPI. This technique makes it possible to set up traffic routing policies differentiated according to the nature of the service associated with the different types of traffic. Classification rules are typically installed at the edge of the network to determine the service function chain to which a flow is associated, and which will be used to process all the packets belonging to this flow. Classification rules are specific to each deployment case. For example, classification may be based on a customer identifier, a network identifier, a combination of several fields in an IP packet, or any other parameter.

Once the motivations and objectives of service function chaining have been described, this article presents some of the most typical use cases, in particular those linked to fixed and mobile networks and data centers. Based on these use cases, the main requirements to be considered for the implementation of an SFC system are described; types of SFC architectures to meet the requirements of these use cases are proposed. Some typical implementation examples are then...