How the Border Gateway Protocol (BGP) Determines the Optimal Path Between Autonomous Systems

Every time a user requests a web page, their data traverses a fractured landscape of competing corporate networks, arriving in milliseconds because a 35-year-old protocol mathematically negotiated the cheapest, most efficient route. The internet is not a single entity; it is a collection of over 70,000 independent networks, known as Autonomous Systems (AS), operated by telecom providers, tech giants, and universities.[6]

Stitching these competing domains together requires a mechanism that can evaluate millions of potential routes and select exactly one best path for any given destination. That mechanism is the Border Gateway Protocol version 4 (BGP-4), defined by the Internet Engineering Task Force (IETF) in 2006 under RFC 4271.[1]

As the IETF specification states, "BGP is an inter-Autonomous System routing protocol," designed to "exchange network reachability information with other BGP systems." This reachability information includes a list of AS numbers that routing information has traversed, preventing loops and enabling policy-based routing across the globe.[1]

Unlike interior routing protocols that simply seek the shortest physical path in milliseconds, BGP is fundamentally an instrument of corporate policy and economics. Network operators configure BGP over TCP port 179 to prefer routes that cost less money, honor peering agreements, or avoid specific geographic regions, making the optimal path a business decision rather than a purely technical one.[4][6]

To enforce these policies, BGP relies on a deterministic, 13-step path selection algorithm. When a router receives multiple routes to the same destination, it does not average them or load-balance by default; it eliminates candidates one by one until a single winner remains.[2][5]

The decision process often begins with the highest Weight, a proprietary attribute injected by hardware vendors. As Cisco's documentation notes, "Weight is a Cisco-defined attribute that is local to a router." It is not transmitted to other routers, allowing an administrator to force traffic down a specific link regardless of what the rest of the network advertises.[2]

If Weight is equal or unsupported—as is the case on non-Cisco hardware like HPE Aruba switches—the algorithm evaluates Local Preference. This is the primary tool for outbound traffic engineering within an Autonomous System. A route with a Local Preference of 200 will always beat a route with 100, allowing an AS to prioritize a paid transit link over a congested free peering connection.[3][4]

Only if Local Preference is tied does BGP look at the AS_PATH length. This is the closest BGP gets to a traditional distance metric. The protocol counts the number of distinct networks a route passes through; a path traversing three networks is preferred over one traversing four.[1][5]

However, the AS_PATH is easily manipulated. Network engineers frequently use a technique called "AS path prepending," artificially lengthening their own path by repeating their AS number up to 10 times to discourage inbound traffic from using a specific 10-gigabit or 100-gigabit link.[4][6]

If the path lengths are identical, the router examines the Origin attribute, preferring routes generated internally over those learned via older protocols or redistributed from other sources. This step ensures that native network routes take precedence over imported ones.[1][5]

The next major tiebreaker is the Multi-Exit Discriminator (MED). While Local Preference controls outbound traffic, MED attempts to influence inbound traffic from a directly connected neighbor. A lower MED value is preferred, signaling to the neighboring AS which of several redundant links it should use.[1][4]

As the Study CCNP curriculum published in February 2024 explains, "The BGP path selection algorithm is a complex process that evaluates multiple attributes in a specific order." If MED fails to break the tie, the router prefers External BGP (eBGP) paths over Internal BGP (iBGP) paths, pushing traffic out of the local network as quickly as possible.[5]

Finally, if all policy and distance metrics are identical, the protocol resorts to arbitrary tiebreakers: the lowest interior gateway protocol (IGP) metric to the BGP next hop, the lowest BGP router ID, and finally, the lowest neighbor IP address. These final steps ensure the algorithm always reaches a definitive conclusion.[2][5]

The resilience of this strict elimination sequence is the reason the internet functions at its current scale. Despite the lack of a central authority, BGP's decentralized decision-making allows competing entities to route terabytes of data per second while strictly enforcing their own commercial interests.[6]