IP transport is a carrier service that carries your IP traffic between locations you control, such as offices, data centers, colocation cages and cloud on-ramps, across the provider’s network instead of the public internet. The provider either routes your traffic for you (as in an MPLS IP VPN) or supplies an Ethernet or optical path that you run your own IP routing over. The term is used loosely in the industry, so it is worth checking exactly what a provider means when it appears on a quote.
At a glance
- IP transport buys a path between your own points; it does not, on its own, connect you to the internet.
- It is delivered as a Layer 3 service (the carrier routes your traffic) or over a Layer 1 or Layer 2 circuit that you route over yourself.
- Its value is predictability: latency, jitter and loss are engineered and usually covered by an SLA.
- It is kept separate from the public internet, but separation is not the same as encryption.
- It is often paired with IP transit or internet access at a hub location.
What problem it solves
When an organization has traffic that needs to move between its own sites, the cheapest option is to send it over the public internet, usually inside encrypted tunnels. For a lot of traffic that works. But the internet gives no guarantees about the path, the delay or what happens during congestion, and that shows up in voice and video quality, database replication, backups and anything sensitive to latency.
IP transport gives that traffic a managed path instead. The carrier engineers the route, can prioritize classes of traffic, and commits to performance targets. It also keeps internal traffic off the internet entirely, which simplifies some security and compliance arguments, and it gives you a predictable way to bring branch traffic back to a data center or a private cloud connection.
How it works
Layer 3 (routed) transport. In an MPLS IP VPN, each of your sites connects to the carrier’s network, and the carrier keeps your routes in a separate virtual routing table (a VRF) from other customers. The carrier forwards packets between your sites and can apply classes of service, so voice is handled ahead of bulk transfers. You see the carrier as one large router; adding a site means adding a connection.
Layer 1 and Layer 2 transport. With an Ethernet private line, an Ethernet LAN service, or a wavelength, the carrier provides a circuit, and your own routers handle IP routing over it. This gives you more control and is common between data centers and for very high bandwidth, but you take on more of the design and operations.
Performance and resilience. Providers commit to availability and often to latency, jitter and packet loss between specified points. Resilience comes from diverse physical routes, protected circuits or a second provider. Ask how “diverse” is defined, because two circuits in the same conduit fail together.
Where internet fits. Many designs bring branch traffic over IP transport to a data center or regional hub, then send internet-bound traffic out through internet access or IP transit at that hub. Cloud on-ramps work the same way: transport to a facility where a private cloud connection is available.
For more on combining private transport with internet underlays across regions, see our Global WAN Services solution page.
When it matters for buyers
- You run voice, video or other real-time traffic between sites and the internet path is causing quality complaints.
- You replicate data between data centers or to a disaster recovery site and need predictable throughput.
- You are adding a site or a region and need to decide between private transport, internet with SD-WAN, or a mix.
- Your MPLS contract is up for renewal. It is a natural point to check which sites still need private transport and which could move to internet-based options.
- A quote is ambiguous. “IP” on a carrier quote can mean transport, transit or internet access; confirm which before comparing prices.
Questions to ask vendors
- Is this a routed Layer 3 service, or a Layer 1 or Layer 2 circuit that we route over ourselves?
- What latency, jitter, packet-loss and availability commitments apply, and between which points are they measured?
- How many classes of service do you support, and how do our markings map to them?
- What physical diversity can you document for each site, and do any paths share conduits or building entries?
- Does the service include internet access at any point, or is that a separate service?
- Is traffic encrypted anywhere on your network, or should we encrypt it ourselves?
- How are sites outside your own footprint delivered, and who is responsible for faults on those connections?
How it differs from IP transit
IP transport is the path: a circuit or carrier service (for example a wavelength, Ethernet or MPLS service) that carries IP traffic between two or more of your own locations. It does not, by itself, connect you to the rest of the internet. IP transit is the internet piece: an upstream network that routes your traffic to, and announces your addresses to, every other network on the internet, usually priced per Mbps (often 95th-percentile billing on a committed rate). A common design uses transport to carry traffic from offices to a data center, and transit at that data center for internet access. When comparing quotes, transport is judged on the path between your sites; transit is judged on how well it reaches everyone else.
Problems this helps with
- Hauling traffic from offices back to a data center or cloud on-ramp
- Real-time apps suffer over the public internet between sites
- Not sure whether a quote is for internet access or a private path
