What is IP transit?
IP transit is a service in which one network pays another to carry its traffic across the rest of the internet. The transit provider announces the customer’s IP ranges to every network it connects to, and gives the customer routes to everywhere else. With transit, a network can reach any address online, not only its direct neighbours.
Think of a small courier firm that only drives in its own town. For parcels going anywhere else, it pays a national carrier to take them further and to bring incoming parcels back. IP transit is that national carrier, for packets.
How does IP transit work?
Transit runs over BGP sessions between the customer’s routers and the provider’s routers. The service has two halves.
Inbound: announcing your routes
The customer announces its prefixes to the provider, and the provider passes them on to its own peers and upstreams. Other networks learn that your addresses can be reached through that provider, so traffic for you starts to arrive.
Outbound: receiving the internet’s routes
The provider sends routes back to the customer. That can be a single default route, meaning “send everything here”, or the full routing table with every prefix on the internet. A full table lets the customer pick the best exit when it has more than one provider.
IP transit vs peering
Peering is a direct exchange between two networks, often at an internet exchange point or over a private link. Each side only reaches the other network and that network’s own customers, not the networks further upstream. Most peering is settlement-free, so neither side pays for the traffic.
Transit reaches the whole internet, and the customer pays for it. A network that reaches every other network through its own customers and settlement-free peering, without paying anyone for transit, is called a Tier 1 network.
How is IP transit priced?
Transit is usually sold per Mbit/s per month, with a minimum commitment and a contract term. Most providers bill on the 95th percentile:
- Traffic is sampled, usually every five minutes, and turned into a rate in bits per second.
- At the end of the month the top 5% of samples are discarded, about 36 hours in a 30-day month.
- The highest remaining sample is the billable rate for the whole month, often measured per direction with the higher one billed.
How enum handles IP transit
The enum network runs as its own autonomous system, AS215998, and announces its own IP ranges from a RIPE NCC allocation. Two carriers are active at the same time, plus peering, with several paths each. When an upstream fails, BGP moves traffic to the others in seconds, and enum can change carriers without any IP address changing.
FAQ
Do I need IP transit to run workloads in the cloud? No. Your cloud provider buys transit and peering for its network, and you usually pay for outgoing traffic instead. You only buy transit yourself when you run your own network with its own address space.
Is buying transit from two providers better than one? Usually, yes. With two providers announcing the same prefixes, the failure of one provider or link only shifts traffic to the other. That setup is called multihoming and usually needs your own ASN and IP space.
Why does peering reduce transit costs? Traffic sent over a peering link does not cross the paid transit link. Networks peer with the networks they exchange the most traffic with, which lowers the 95th percentile rate billed on their transit ports.