An emerging function in the provision of services in IP-enabled networks is Service Function Chaining (SFC). This talk will look at the motivation for SFC with special focus on its deployment in data centres, and then will describe the architecture developed by the IETF's SFC working group. We will examine the different functional components of an SFC system and how they are combined to direct traffic off its normal path s that it can be acted on by service functions hosted at different places in the network.
The SFC working group has developed a protocol-independent encapsulation called the Network Service Header (NSH). The NSH allows packets to be marked according to the service functions that they must traverse, and can be applied to any payload and with any transport protocol. But two alternate approaches have been suggested in the MPLS and SPRING working groups to achieve the same function as the NSH in existing forwarding systems using either MPLS or Segment Routing encapsulations, thus allowing legacy routers to participate in SFC. Furthermore, there is a proposal to achieve SFC by extensions to L3VPN techniques being progressed by the BESS working group.
Lastly we will consider the control and management necessary for SFC. We will look at approaches for central control using an SDN-based central controller, BGP southbound instructions, or distributed BGP information exchange.