Differentiated Service (DiffServ)

Differentiated Service (DiffServ) is a relatively simple means for prioritizing
different types of traffic. The DiffServ protocol is described in RFC 2475,An
Architecture for Differentiated Services. Basically, DiffServ makes use of
the IPv4 Type of Service (TOS) field, contained in the IPv4 header and the
equivalent IPv6 Traffic Class field. The portion of the TOS/Traffic Class
field used by DiffServ is known as the DS field. The field is used in specific
ways to mark a given stream as requiring a particular type of forwarding.
The type of forwarding to be applied is known as per-hop behavior (PHB), of
which DiffServ defines two types. These are expedited forwarding (EF) and
assured forwarding (AF).
EF is specified in RFC 2598. It is a service whereby a given traffic stream
is assigned a minimum departure rate from a given node, one that is
greater than the arrival rate at the same node, provided that the arrival
rate does not exceed a pre-agreed maximum. This ensures that queuing
delays are removed. Since queuing delays are a major cause of end-to-end
delay and are the main cause of jitter, this ensures that delay and jitter are
minimized. In fact, EF can provide a service that is equivalent to a virtual
leased line.
AF is defined in RFC 2597. This is a service whereby packets from a
given source are forwarded with a high probability, provided that the traffic
from that source does not exceed some pre-agreed maximum. AF defines
four classes, with each class allocated a certain amount of resources (buffer
space and bandwidth) within a router. Within each class, a given packet
may have one of three drop rates. At a given router, if congestion occurs
within the resources allocated to a given AF class, then the packets with the
highest drop rate values will be discarded first so that packets with a lower
drop rate value receive some protection. In order to work well, it is necessary
that the incoming traffic does not have packets with a high percentage
of low drop rates. After all, the purpose is to ensure that the highest-priority
packets get through in the case of congestion, and that cannot happen if all
the packets have the highest priority.