IP transit is a paid interconnection service in which one network carries traffic to and from the entire Internet on behalf of a customer network. The transit provider announces the customer’s IP prefixes to the rest of the world and gives the customer routes to every reachable destination, usually through a Border Gateway Protocol (BGP) session on a dedicated port.
Among Internet fundamentals, IP transit is the commercial relationship that ties the network of networks together. Access providers, hosting companies, content networks and enterprises with their own autonomous system buy transit so that their users and servers can reach any address, not only the networks they connect to directly.
How IP transit works
IP transit works through a BGP session between the customer’s router and the provider’s router, normally over a physical port in a data center or a leased circuit. Over this session, two flows of routing information move in opposite directions.
- Customer to provider: the customer announces its own prefixes, and the provider propagates them to its peers, other customers and upstream networks.
- Provider to customer: the provider sends either the full Internet routing table, a default route only, or a partial table with selected routes.
The full IPv4 routing table contains roughly one million prefixes, and the IPv6 table adds a smaller but steadily growing number. Taking the full table lets the customer make precise path decisions when it has several transit providers, but it requires routers with enough memory and forwarding capacity.
Each participant in BGP is identified by an autonomous system number (ASN). A transit provider tags and filters announcements so that it accepts only the prefixes the customer is entitled to announce, typically verified against Internet routing registry records and RPKI route origin authorizations.
Commercial terms and technical parameters
An IP transit contract defines capacity, price model and service quality. The most common billing method is the 95th percentile: traffic is sampled every five minutes, the top 5 percent of samples in the month are discarded, and the highest remaining value is billed. Short bursts therefore do not raise the bill.
| Parameter | Typical values | What it affects |
|---|---|---|
| Port speed | 1, 10, 100 or 400 Gbit/s | Maximum burst capacity |
| Commit rate | A share of the port, such as 2 Gbit/s on 10 | Minimum monthly payment |
| Billing method | 95th percentile or flat rate | Cost of traffic peaks |
| Routes received | Full table, default or partial | Router load and path control |
| Service level | Availability, latency and repair targets | Compensation for outages |
| Extras | BGP communities, DDoS filtering | Traffic engineering and protection |
Prices are quoted per megabit or gigabit per second per month and have fallen for decades. They vary strongly by region: transit in major interconnection hubs is far cheaper than in markets with few providers or expensive long-distance capacity.
IP transit compared with peering
The essential difference between IP transit and peering is scope of reachability. Transit gives access to the whole Internet, whereas peering exchanges traffic only between the two networks and their own customers.
| Model | Who pays | Routes exchanged | Typical location |
|---|---|---|---|
| IP transit | Customer pays provider | The entire Internet | Data centers, private circuits |
| Settlement-free peering | Neither party | Each network’s own and customer routes | Internet exchange points, private links |
| Paid peering | One party pays | Only the seller’s own and customer routes | Private interconnects |
| Partial transit | Customer pays provider | A defined region or set of networks | Regional hubs |
Networks usually combine both models. Peering at an Internet exchange point offloads traffic toward large content and access networks at low cost, while transit covers everything that peering cannot reach. Only a small group of Tier 1 networks reach the entire Internet through peering alone and buy no transit at all.
Designing a resilient transit setup
Relying on a single transit provider creates a single point of failure. Most networks with their own ASN therefore practice multihoming: they buy transit from at least two providers and use BGP to balance and fail over traffic.
| Design | Resilience | Typical user |
|---|---|---|
| Single provider, single link | Low | Small office without its own ASN |
| Single provider, two links | Protects against link failure | Small hosting company |
| Two providers in one site | Protects against provider failure | Regional access provider |
| Several providers in several sites | High, also against site loss | Content and cloud networks |
Traffic engineering uses tools such as AS path prepending, local preference and provider BGP communities to steer inbound and outbound traffic. Choosing providers with different upstream paths and physical routes matters more than their number, because two providers sharing the same cable gives little real protection.
Risks and common mistakes
Problems with IP transit usually stem from routing hygiene and capacity planning rather than from the physical connection.
- Route leaks: announcing routes learned from one provider to another can turn a small network into an accidental transit path.
- Missing filters: without prefix and RPKI filtering, mistaken announcements can hijack other networks’ addresses.
- Port saturation: a commit sized for average traffic leaves no headroom for peaks or attacks.
- Hidden shared risk: two providers may rely on the same upstream or the same fiber route.
- Ignoring latency: the cheapest provider may route traffic through distant cities, which slows users.
A non-technical view
No single company owns the whole Internet. Thousands of separate networks connect to each other. A small network cannot link directly to all of them. It pays a larger network to carry its traffic everywhere else, and that service is IP transit.
It works much like a parcel company that hands packages to a national courier. The courier delivers to any address in the country. The small company pays for that reach, so its customers can send and receive data anywhere in the world.
A concrete example
A regional internet service provider with about 30,000 subscribers operates its own autonomous system. It buys IP transit from two different providers, each on a 10 Gbit/s port with a 3 Gbit/s commit, billed at the 95th percentile, and receives the full routing table from both.
At a nearby Internet exchange point, it also peers with large video and cloud networks, which carry close to half of its evening traffic. Transit therefore handles mainly the remaining destinations. When one transit provider suffers a fiber cut, BGP moves all traffic to the second provider within seconds, and subscribers notice only a brief slowdown. The combination keeps monthly transit costs predictable while removing a single point of failure.