Currently, when we get a new set of routes, say from a BGP announcement, we compute the bestpath for each distinct destination prefix in the set and synchronize bestpath decisions to the ASIC one prefix at a time. This can be a bit slow for large announcements. For example, the lag is noticeable on a 2k NLRI BGP announcement.
Currently, when we get a new set of routes, say from a BGP announcement, we compute the bestpath for each distinct destination prefix in the set and synchronize bestpath decisions to the ASIC one prefix at a time. This can be a bit slow for large announcements. For example, the lag is noticeable on a 2k NLRI BGP announcement.