Network peering explained: BGP, internet exchange points & benefits

Key takeaways

• Peering is a direct exchange of traffic between two networks, without paying a transit provider.
• BGP is the protocol that makes it work. It exchanges reachability information between autonomous systems.
• Not all peering is free. Settlement-free is one type; paid peering also exists.
• Internet exchange points let many networks peer through one port rather than many cross-connects.
• Peering and transit are complements. Transit reaches everywhere; peering optimises the routes that matter most.

Network peering is a direct connection between two networks that lets them exchange traffic without using a transit provider for those destinations. It can reduce latency, lower transit costs, and give networks greater control over routing.

BGP enables those networks to exchange reachability information and apply routing policies to that traffic. This guide explains how peering, BGP, and internet exchange points work together, and when peering makes sense alongside IP transit.

What is network peering?

Network peering is a direct interconnection between two networks. They use it to exchange traffic between their own customers.

Neither network pays the other to carry that traffic in most arrangements. The alternative is IP transit, where you pay a provider to reach the rest of the internet.

Peering removes the intermediary for the destinations covered by the agreement. Your traffic to that network takes one hop instead of several. Latency falls, and so does the transit bill.

Peering only covers the peer's own network and its customers: it does not give you routes to the wider internet.

What is BGP (Border Gateway Protocol)?

BGP is the routing protocol networks use to exchange reachability information across the internet. It is defined in RFC 4271 and runs over TCP on port 179.

Its job is inter-domain routing. BGP tells other networks which IP prefixes you can reach, and learns the same from them.

BGP works in units of autonomous systems. Each is a network under one administrative control, with a consistent routing policy. Each carries an autonomous system number allocated by a regional internet registry.

Any network wanting to peer independently needs an ASN. The ASN is how the internet identifies you in routing decisions.

How BGP routing works

Each network advertises the IP prefixes it can reach. Those advertisements carry a list of autonomous systems the route has passed through.

That list is the AS path. It prevents loops, because a network rejects any route already containing its own number.

BGP then selects among available routes using policy, not just distance. Shortest AS path is one factor. Operators can also weight decisions by cost, capacity or preference.

The protocol supports classless inter-domain routing. Networks advertise prefixes of any length and aggregate them where possible.

How BGP selects a route

BGP rarely picks the shortest path. It picks the path your policy prefers. That is why two networks can route the same destination differently.

Each advertised route carries attributes. BGP compares available routes in sequence, using these attributes to determine the preferred path:

• Weight: A Cisco-specific local value on a single router. Higher wins, and it never leaves that router.
• Local preference: Shared across your own autonomous system. Higher wins, and it decides which exit point your traffic uses.
• AS path length: Shorter wins. This is the closest BGP gets to a distance metric.
• Origin: Routes with an IGP origin are preferred over those with an EGP or incomplete origin.
• Multi-exit discriminator: Lower wins. MED suggests which of your entry points a neighbouring network should use.
• eBGP over iBGP: When other route attributes are equal, BGP prefers a route learned from an external peer over one learned internally.

Local preference and MED influence different sides of traffic engineering. Local preference influences which exit point your network uses for outbound traffic.

MED can influence which entry point a neighbouring network chooses when it has multiple routes into your network. BGP normally compares MED values among routes received from the same neighbouring AS, and the neighbouring network can choose whether to use MED in its routing policy.

What is BGP peering?

BGP peering is the session between two routers that agree to exchange routes. The term describes the technical relationship, whereas network peering describes the commercial one.

Establishing a session requires both sides to configure the other's IP address and AS number. Once established, each side sends the prefixes it is willing to advertise.

You choose which prefixes to advertise to each peer, and which to accept. That control is why BGP peering improves performance. You can prefer a direct route over a transit route for specific destinations.

What is an internet exchange point (IXP)?

An internet exchange point is a shared physical location where many networks interconnect. Each member connects once to a common switching fabric.

Without an IXP, peering with twenty networks means twenty separate cross-connects; through an IXP, one port can reach all twenty, so the cost of adding a peer drops to a configuration change.

Members still negotiate individually. Presence at an exchange does not oblige anyone to peer with anyone.

Types of peering: Public, private, settlement-free, and paid

Peering comes in four common arrangements, defined by where networks connect and how they handle commercial terms:

• Public peering: Happens over a shared internet exchange fabric, which suits moderate volumes across many peers.
• Private peering: Uses a dedicated link between two networks, usually a cross-connect in the same facility.
• Settlement-free peering: Means neither party pays, which works where volumes are roughly balanced.
• Paid peering: Applies where the benefit is uneven, typically when one network sends far more than it receives.

All four use BGP. Only the commercial terms change.

Peering vs IP transit

Peering and IP transit differ in reach, cost, routing control, setup and the destinations they serve.

Networks rarely choose one exclusively: they buy transit for universal reach, then peer where direct routes can shorten paths and reduce transit costs.

Benefits of network peering for enterprises

Peering can shorten paths to important destinations and reduce the amount of traffic carried over paid transit:

• Lower latency: Fewer hops to major destinations, which matters for interactive applications.
• Reduced transit cost: Traffic moving over peering does not count against a transit commitment.
• Better routing control: You decide which paths carry which traffic.
• Improved resilience: Multiple paths mean a single provider failure is not total.
• Predictable performance: Direct paths vary less than routes crossing several networks.

Most enterprises consume these benefits through their provider rather than peering directly. Running your own peering needs an ASN, address space and engineers to operate it.

Challenges in network peering

Peering is harder to arrange than to operate. Most of the difficulty sits in the negotiation, not the configuration.

• Negotiation and ratios: Peering is a commercial agreement before it is a technical one. Larger networks often set traffic ratio conditions, and refuse outright when the exchange looks unbalanced.
• Routing table growth: The global table keeps expanding as networks advertise more prefixes. Older routers run out of memory to hold it, which turns a routing problem into a hardware refresh.
• Configuration risk: A misconfigured advertisement can leak routes well beyond your own network. Route leaks have taken large services offline, and the damage lands on networks that made no mistake.
• Operational reach: Each session needs monitoring and each relationship needs maintaining. Peering also requires equipment where the peer is, which limits who you can realistically reach.

Peering and IP transit from Tata Communications

Peering only helps if a provider is present where your traffic needs to go.

IP Transit runs across 250+ IP points of presence in 40+ countries and 80+ cities. Edge capacity exceeds 300 Tbps. It reaches 70 per cent of the world's top 20 content providers, including all of the top five.

Much of your content traffic therefore takes a short path. Service objectives target 99.99% availability and packet loss under 0.1%.

Latency and loss separate a good route from a poor one. They are not the same thing as bandwidth. For multi-country estates, our guide to building a global network infrastructure covers how reach shapes design.

Reach is the part you cannot change later, so it is worth checking before you commit. Schedule A Conversation to check our routes and points of presence against your own locations.

See how global network reach and content proximity shape real routes. Read The Case Study

Discuss your routing and reach requirements with our team. Schedule A Conversation

Benchmark your internet setup on resilience and readiness. Take The Internet Maturity Assessment