Improved quality-of-service/traffic control
While quality of service can be controlled by use of a control protocol such as RSVP, IPv6 provides for explicit priority definition for packets by using the priority field in the IP header.
A node can set this value to indicate the relative priority of a particular packet or set of packets, which can then be used by the node, one or more routers, or the destination to make choices concerning the packet (that is, dropping it or not).
IPv6 specifies two types of priorities, those for congestion-controlled traffic, and those for non-congestion-controlled traffic. No relative ordering is implied between the two types.
Congestion-controlled traffic is defined as traffic that responds to congestion through some sort of "back-off" or other limiting algorithm. Priorities for congestion-controlled traffic are:
| Item | Description |
|---|---|
| 0 | uncharacterized traffic |
| 1 | "filler" traffic (for example, netnews) |
| 2 | unattended data transfer (for example, mail) |
| 3 | (reserved) |
| 4 | attended bulk transfer (for example, FTP) |
| 5 | (reserved) |
| 6 | interactive traffic (for example, Telnet) |
| 7 | control traffic (for example, routing protocols) |
Non-congestion-controlled traffic is defined as traffic that responds to congestion by dropping (or simply not resending) packets, such as video, audio, or other real-time traffic. Explicit levels are not defined with examples, but the ordering is similar to that for congestion-controlled traffic:
- The lowest value that the source is most willing to have discarded should be used for traffic.
- The highest value that the source is least willing to have discarded should be used for traffic.
This priority control is only applicable to traffic from a particular source address. Control traffic from one address is not an explicitly higher priority than attended bulk transfer from another address.
Flow labeling
Outside of basic prioritization of traffic, IPv6 defines a mechanism for specifying a particular flow of packets. In IPv6 terms, a flow is defined as a sequence of packets sent from a particular source to a particular (unicast or multicast) destination for which the source desires special handling by the intervening routers.
This flow identification can be used for priority control, but might also be used for any number of other controls.
The flow label is chosen randomly, and does not identify any characteristic of the traffic other than the flow to which it belongs. This means that a router cannot determine that a packet is a particular type by examining the flow label. It can, however, determine that it is part of the same sequence of packets as the last packet containing that label.