Setting bits in the TOS octet at network edges and administrative boundaries Using those bits to determine how packets are treated by the routers inside the network Conditioning the marked packets at network boundaries in accordance with the requirements of each service.
The diff-serv architecture is composed of a number of small functional units implemented in the network nodes. This includes the definition of a set of Per-Hop Behaviors (PHBs), packet classification and traffic conditioning functions like metering, marking, shaping and policing. The resource allocation for each service type adds a new dimension to the problem, for which the Bandwidth Brokers are being considered.
The diffserv model is scalable because of a few reasons that are listed below:
Diff-serv suggests that the more expensive tasks like multi-flow classification, policing, shaping and marking be done at the border routers of the ISP networks. This is because the border routers deal with the customer links that are slow as a result of which it has time to do the costly functions like MFC and traffic conditioning. The core routers, on the other hand simply does the forwarding based on the diff-serv code point (DSCP), which is the first six bits in the TOS byte in the IP header. Since the core routers need not maintain any per-flow state, this model is more scalable.The granularity of service provisioning is a class in diff-serv, as opposed to being a flow in int-serv. Multiple flows may be mapped on to a single per-hop behavior (PHB), which is indicated by the value in the DSCP. This too, ensures the scalability of the diff-serv model.