Arista Networks, Inc.

Chen Jia Jang

Ganesan Venkataraman

Ramakrishna Shivaramaiah Paduvalli

Muhammad Khalid Yousuf

Suhas Raghunath Joshi

Techniques for improving the prefix summarization scale of the longest prefix match (LPM) tables of a network device are provided. In one set of embodiments, these techniques involve deduplicating sub-tries of LPM trie data structures that the network device uses to program LPM entries into the LPM tables. This sub-trie deduplication avoids the creation of duplicate LPM entries in the LPM tables across different virtual routing and forwarding (VRF) instances, thereby increasing the LPM tables' effective capacity.

Prefix summarization scale by deduplicating longest prefix match (LPM) sub-tries

A logical table is configured with a first set of memory banks, where each logical row in the logical table comprises a corresponding memory row from each of the memory banks. Lookup instructions to access a logical row includes a bank set that lists the memory banks associated with that logical row. In response to a range of memory rows of one of the memory banks being reallocated to another logical table, a new memory bank is identified to store the data in the reallocated memory rows. Logical rows associated with the reallocated memory rows are mapped to the new memory bank. Bank sets in the lookup instructions that refer to the remapped logical rows are updated to list the new memory bank in place of the 'old' memory bank.

Growing community of inventors

Burnaby, Canada

Chen Jia Jang