TABLE OF CONTENTS Classification in the process of identifying flows of packets and grouping individual traffic flows into aggregated streams such that actions can be applied to those streams; the actions that may be applied after classification could be other than QOS actions, for example packet filtering. To complete this definition we define the terms flow, stream and class:
Flows. An IPv4 flow is typically defined by the 5-tuple of source IP address, destination IP address, source TCP/UDP port, destination TCP/UDP port and the transport protocol (e.g. TCP or UDP). Fragmentation, encryption or tunneling can make some flows difficult to classify, as some of these fields may be unavailable. For example, assume that a flow is classified using this 5-tuple, but that the sizes of packets in the flow exceed the size of packets that can be supported (defined by the maximum transmission unit or MTU) on some of the underlying links which the flow transits. Those packets, which exceed the MTU of transit links, will need to be fragmented using IP level fragmentation [RFC 791]. When a packet is fragmented, the TCP/UDP port information is included in the first packet only; hence, it may not be possible to identify uniquely non-first fragments of a packet as belonging to the same flow. IPv6 introduced a Flow Label field which can be used to overcome this problem.
More complicated criteria than this 5-tuple may also be used to define a flow, such as using deep packet...
