Big Routers - Few Prefixes
Most Internet hosting platforms only communicate with a limited set of prefixes (10-50K) while the Internet routing table contains around 600K prefixes today.
Optimising only the important portion of the Internet routing table enable to replace expensive BGP routers with cheap/high bandwidth L3 switches.
NSI provides the ability to execute this FIB reduction automatically, dynamically while optimising routing of these prefixes.
Virtual Router (VR) – FIB optimization
As of today (2016), the Internet consists of nearly 600k prefixes. Classic edge BGP routers need to load this number of prefixes into their FIB memory. Since FIB memory is very expensive, such routers tend to be costly.
But actually, most platforms do not swap data with all of these prefixes. In our experience, most BGP platforms exchange up to 100% of traffic using as little as 0.2% to 8% of their routing table.
Given the above, NSI can help in tremendously reducing the cost of routing hardware, by performing FIB optimization via its Virtual Router (VR) feature.
The implementation requires Selective Route Download feature set: The BGP edge routers would announce your local prefixes to your upstream transits, accept a default route and receive the full routing table but do not leak these routes into the FIB. The NSI platform then identifies and optimises only the Top volume prefixes that are finally passed to the L3 switches FIB, reducing the number of entries to less than 16 or 32K prefixes.
Instead of using full-featured BGP routers with plenty of expensive FIB memory, network admins are now able to make use of much more cost-effective L3 switches with a BGP-aware firmware, such as Cumulus Networks