Statistical learning technique in a multi-port bridge for a local area network

Abstract

A statistical learning technique in a multi-port bridge for a local area network (LAN). In response to data packets received by the multi-port bridge, learning operations are performed for updating a look-up table in the multi-port bridge. The look-up table is also utilized for identifying an appropriate destination port for each data packet by performing a look-up operation. A learning operation, however, is performed only in response to selected ones of the data packets. A statistical learning controller determines whether a received data packet is a broadcast or uni-cast packet. If the packet is a broadcast or multi-cast packet, the statistical learning controller forwards a request for a learning operation. Upon being granted, the look-up table is updated. It the packet is a uni-cast packet, the statistical learning controller forwards a request for a learning operation only if (m) uni-cast packets have been received since a prior learning operation was performed for a uni-cast packet. In the preferred embodiment, (m) is sixteen. Accordingly, the look-up table is preferably updated in response to every broadcast and multi-cast packet and in response to every sixteenth uni-cast data packet. The invention minimizes the need to preoccupy the look-up table for performing learning operations and, thus, increases the availability of the look-up table for performing look-up operations.