Aggregation Basics

Several methods exist to aggregate multiple physical transceivers into one logical transceiver with a higher bitrate. The layered architecture of xDSL and Ethernet transceivers, with clearly identified interfaces between the sublayers, is very aggregation-friendly; to allow aggregation, it suffices to either add the aggregation function to an existing sublayer (which will then interface with several lower-layer instances instead of just one), or to specify a new aggregation sublayer, which presents the same interface to the higher layer as the aggregated lower layer instances present to itself.

For Ethernet, there is a mechanism called link aggregation (also known as 802.3ad, specified in Clause 43 of IEEE Std. 802.3), which combines a number of identical full-duplex PHYs of a system into a single fat Ethernet pipe. This is done by dividing the stream of transmit frames from the attached MAC Client or MAC Relay into conversations, which are sets of frames sharing the same Source Address and Destination Address. By assigning all the frames of a given conversation to the same link, frame reordering and duplication within a conversation can be avoided. ^[18] The Aggregator is governed by a Link Aggregation Control function, which may be configured manually through dedicated management variables or automatically by means of the Link Aggregation Control Protocol (LACP). LACP continually monitors the eligibility of links to participate in the aggregation; as such, it provides a simple and rapid redundancy mechanism for links between multi-port Ethernet elements.

Where Ethernet PHYs typically...