BGP Fundamentals and Route Selection

BGP stands for Border Gateway Protocol. It is the postal service of the internet. When you send an email, watch a video, or load a website, your data is broken into packets and sent across multiple different networks. These networks are called Autonomous Systems (AS). An Autonomous System is a collection of routers under the control of a single organisation, like an internet service provider (ISP) or a large company. BGP is the protocol that allows these different Autonomous Systems to exchange routing information so that packets can find their way from your computer to a server on the other side of the world.

Let us start with the basics. A router is a specialised computer that forwards packets between networks. Each router has a routing table, which is like a list of directions. BGP helps build that routing table by telling a router about the paths to different destination networks. Every network on the internet has a unique identifier called an IP prefix, which is a block of IP addresses. When a BGP router learns about a new IP prefix from a neighbour, it adds that information to its BGP table, which is a more detailed list of all possible paths.

BGP has two main flavours: Internal BGP (iBGP) and External BGP (eBGP). eBGP is used between different Autonomous Systems, like two ISPs exchanging routes. iBGP is used within the same Autonomous System, to make sure all routers inside that AS have consistent routing information. This distinction is important because eBGP and iBGP have different rules about how they pass on routes.

Now, how does BGP decide which path to use when there are multiple possible routes? This is called the BGP route selection process. BGP does not just pick the shortest path in miles. Instead, it uses a list of attributes, which are like scores or properties attached to each route. The most important attribute is the Weight. Weight is a Cisco-specific attribute that is local to a single router. A higher weight makes a route more preferred. Next is Local Preference, which is a value shared across all routers in the same AS. A higher Local Preference means the route is preferred for traffic leaving the AS. After that, the router looks at whether the route was learned via eBGP or iBGP. eBGP routes are generally preferred over iBGP routes. Then comes the AS Path length. This is a list of all the AS numbers a route has passed through. A shorter AS Path is preferred. The Origin attribute tells how BGP learned about the route (IGP, EGP, or incomplete). IGP is preferred over EGP, and EGP is preferred over incomplete. Finally, the router looks at the Multi-Exit Discriminator (MED) value, which is a hint from one AS to another about which entry point to use. The lowest MED is preferred. If all these are equal, BGP falls back to tie-breaking rules like preferring the route from the router with the lowest Router ID.

Why does BGP need all these attributes? Because the internet is not a single network. It is a collection of thousands of networks with different business agreements. Some ISPs pay others for transit. Some exchange traffic for free in peering arrangements. BGP attributes allow network operators to implement these business policies. For example, an ISP might set a higher Local Preference for a route that comes through a cheaper transit provider, making that path preferred for most traffic.

BGP also uses a state machine with several states: Idle, Connect, Active, OpenSent, OpenConfirm, and Established. A BGP session goes through these states as it tries to connect to a neighbour. If the session is stuck in one state, that indicates a problem, like a misconfigured IP address or a firewall blocking the connection.

BGP replaced older protocols like EGP (Exterior Gateway Protocol) because EGP could not handle the complex topology and policy requirements of the modern internet. BGP is scalable, flexible, and supports features like route aggregation and policy-based routing, which are essential for service providers.

In summary, BGP is the protocol that makes the internet work as a single, globally connected network. It allows different organisations to exchange routing information while keeping full control over how their traffic flows.

An IT professional working for a medium-sized internet service provider (ISP) uses BGP every day to manage how customer traffic reaches the wider internet. Let us walk through a typical scenario.

One morning, the network operations centre (NOC) receives an alert that some customers in London are experiencing slow internet speeds when accessing a popular video streaming service. The engineer, Sarah, begins to investigate. She logs into the main border router, which is the router that connects the ISP's network to the rest of the internet. Sarah uses a command called 'show ip bgp' to examine the BGP table for the destination IP prefix belonging to the streaming service. She sees that there are two paths to that prefix: one through a transit provider called GlobalConnect and one through a peering agreement with another ISP, RegionalNet.

Sarah checks the BGP attributes for each path. She notices that the route through GlobalConnect has a longer AS Path length (meaning it traverses more intermediate networks) and a lower Local Preference. The route through RegionalNet has a higher Local Preference because the ISP has a free peering agreement with them. Sarah suspects that the problem is with the RegionalNet peering link. She checks the interface statistics for that link and sees a high error rate. This is causing packet loss, which slows down the streaming service.

Sarah decides to temporarily change the BGP attributes to steer traffic away from the problematic link. She adjusts the Local Preference on the router so that the GlobalConnect route becomes the preferred path for this particular prefix. This is a quick fix that restores performance for customers. The next day, Sarah coordinates with the RegionalNet engineers to fix the hardware issue on the peering link. Once the link is stable, she removes her temporary configuration change, and traffic flows back to the preferred path.

Another common task is implementing a new peering agreement. When the ISP signs a peering agreement with a content delivery network (CDN), like a video streaming site, the engineer must configure BGP to accept the CDN's routes and advertise the ISP's customer routes to the CDN. This involves configuring a BGP session with the CDN's router, setting the correct AS number, and applying route maps to filter which prefixes are exchanged. A route map is a set of rules that match certain prefixes and then change their attributes. For example, the engineer might set a MED value on routes sent to the CDN to indicate which of the ISP's multiple connection points is preferred.

Real-world BGP work also involves troubleshooting. Common problems include:

The engineer uses tools like ping, traceroute, and BGP-specific commands to diagnose these issues. They also use network monitoring systems that alert them when BGP session state changes. Understanding BGP fundamentals is crucial because a misconfiguration can affect thousands of customers or even cause a global internet outage.