Two Products, Same Billing Model, Different Jobs
Lumen sells two NaaS products with confusingly similar names. Internet on Demand (IoD) and Ethernet on Demand (EoD) both ride on the same Lumen NaaS platform, both bill hourly by bandwidth tier, and both let you change tiers in minutes through the NaaS API. After that, they diverge — and picking the wrong one is one of the more expensive networking mistakes an enterprise can make in 2026.
The short version:
- • **IoD** gives you on-demand internet access at Layer 3 (public internet routing).
- • **EoD** gives you on-demand private Ethernet circuits at Layer 2 (point-to-point or point-to-multipoint).
If you need to reach the public internet, you want IoD. If you need a private path between two known endpoints — your data center to AWS, branch to headquarters, colo to another colo — you want EoD. The rest of this guide is about the details that determine which one fits your actual workload.
The Layer 2 vs. Layer 3 Distinction
This is the core difference and it drives everything else.
Internet on Demand is Layer 3. Your circuit terminates at a Lumen edge, gets a public IP block, and routes to the internet. Whatever is on the other side of "the internet" is whatever the internet routes to. You don't control the path past Lumen's edge. You do get full internet reachability — every public destination, every cloud provider's public endpoint, every SaaS service. Best-effort delivery, no path-level QoS, but global reach.
Ethernet on Demand is Layer 2. Your circuit is a private Ethernet Virtual Circuit (EVC) between two endpoints you choose at order time. There is no public IP routing involved — Lumen carries your Ethernet frames between the two endpoints over its private backbone. The destination is fixed. The path is private. You can run any Layer 3 protocol on top (BGP, static routes, MPLS-style behavior).
Why this matters: a 1 Gbps EoD EVC from your DC to AWS Direct Connect is fundamentally different from a 1 Gbps IoD circuit pointed at AWS's public endpoint. The EoD path is private, isolated from the public internet, and supports Class of Service tiers. The IoD path traverses public internet routing and is best-effort. Both can deliver 1 Gbps; only one is appropriate for production cloud connectivity at scale.
Use Case Matrix
A practical mapping of common scenarios:
| Scenario | Pick | Why |
|---|---|---|
| Branch office internet access | IoD | Public internet at a branch — Layer 3, best-effort, global reach |
| Burst capacity for a data migration to internet | IoD | Public internet, just needs more bandwidth temporarily |
| Branch to HQ private link | EoD | Layer 2 private Ethernet — predictable path |
| Colo to AWS / Azure / GCP via Direct Connect | EoD | Direct Connect–style architecture, private path with CoS |
| Colo to colo private link | EoD | Layer 2 between known endpoints |
| Public website hosting at colo | IoD | Internet-facing, needs public connectivity |
| DR failover internet circuit | IoD | Internet access during failover |
| DR private replication path | EoD | Replication traffic between two known DCs |
| SD-WAN underlay (commodity internet) | IoD | Layer 3 internet underlay |
| SD-WAN underlay (private overlay) | EoD | Layer 2 between SD-WAN endpoints |
If your traffic needs to reach the public internet, IoD. If your traffic needs a private path between two endpoints you control, EoD. There are edge cases (a single circuit doing both via a hub topology, customers using IoD with a tunneled overlay), but the table covers the vast majority of decisions.
Pricing Differences That Drive the Decision
Both products bill hourly by bandwidth tier. The rate structures are different.
Internet on Demand pricing. 14 bandwidth tiers from 1 Mbps to 100 Gbps. No Class of Service tiers — IoD is best-effort by definition. Reference rate: 1 Gbps runs roughly $1.06–$1.08/hr. See [Lumen IoD Pricing: Complete Hourly Rate Guide](/blog/lumen-iod-pricing-per-hour) for the full structure.
Ethernet on Demand pricing. 24 bandwidth tiers from 10 Mbps to 100 Gbps. Three Class of Service tiers (Basic, Enhanced, Dedicated). Two pricing schedules — Customer Premise (terminating at a customer site) and Datacenter Expansion (terminating at a partner DC or cloud on-ramp). Reference rate: 1 Gbps Customer Premise Basic runs roughly $2.21/hr; 1 Gbps Datacenter Expansion Basic runs roughly $1.44/hr. See [Lumen Ethernet on Demand Pricing](/blog/lumen-ethernet-on-demand-pricing) for the full table.
[CONFIDENCE: MEDIUM — Reference rates above reflect publicly observed Lumen pricing; your contracted rates may vary by metro and term.]
The headline: EoD is materially more expensive per Mbps than IoD at the same bandwidth tier. That is the price of a private path with CoS. Don't pay for an EoD EVC if your workload only needs internet reachability — you would be paying for guarantees you don't need. Don't try to use IoD for a workload that needs a private path with predictable performance — you'll pay less but the architecture won't hold up.
When Both Products Show Up in the Same Network
It is normal for an enterprise to have both products on the same Lumen NaaS account. A typical pattern:
- • **Branch offices:** IoD for general internet access (employee web, SaaS, video conferencing)
- • **Headquarters:** IoD for inbound public traffic (web, customer-facing endpoints)
- • **HQ to colo:** EoD Customer Premise EVC at Enhanced CoS
- • **Colo to AWS Direct Connect:** EoD Datacenter Expansion EVC at Enhanced CoS
- • **Colo to backup DC:** EoD Customer Premise EVC at Basic CoS for replication traffic
Apptifi schedules both product types from the same console. The scheduler discovers your IoD circuits and EoD EVCs through the NaaS inventory API, shows them side by side with their current tier and CoS, and lets you build schedules across the whole portfolio. Same client ID, same client secret, same /modifyBandwidth endpoint — the only difference at the API level is the EoD calls also carry the EVC type and CoS context.
What CoS Actually Buys You on EoD
Class of Service is one of the things people pay extra for without fully understanding. The three EoD CoS tiers are:
- • **Basic.** Best-effort across the Lumen private backbone. Suitable for replication, backup, file transfer, and any workload that tolerates jitter and reordering. Cheapest of the three.
- • **Enhanced.** Prioritized delivery with bounded latency and jitter. Suitable for converged voice and video alongside data. The most common CoS for production WAN traffic.
- • **Dedicated.** Premium CoS for real-time workloads with the tightest latency and jitter bounds. Use for voice-only EVCs, real-time video, and anything where a 50ms delay is a problem.
The CoS selection is locked at EVC order time — it is not a tier-change variable. Pick the CoS that matches the workload, then schedule the bandwidth tier to track demand. Most enterprises over-pick CoS (paying for Dedicated when Enhanced would do) and under-schedule bandwidth (running 1 Gbps 24/7 when off-peak demand is 100 Mbps).
Scheduling Both Products Together
The scheduling math works the same way for both:
1. Identify peak vs. off-peak windows for each circuit 2. Drop to a lower tier during off-peak windows 3. Snap back before peak demand returns 4. Capture the difference between contracted-tier hours and ordered-tier hours as savings
The biggest difference is degree of freedom. EoD has more dials (EVC type, CoS, 24 tiers) but only the bandwidth tier is on-demand. IoD has fewer dials (no CoS, 14 tiers) but the same on-demand bandwidth lever. In both cases, the lever you control is the bandwidth tier.
For the EoD scheduling patterns specifically, see [Ethernet on Demand Scheduling Guide](/blog/ethernet-on-demand-scheduling-guide). For the IoD patterns, see [IoD Bandwidth Scheduling Best Practices](/blog/iod-scheduling-best-practices).
The Decision Framework
When a network team is evaluating which product to use for a new circuit, the questions are:
1. Does this traffic need to reach the public internet? Yes → IoD. No → EoD. 2. If EoD, where does it terminate? Customer site → Customer Premise schedule. Partner DC or cloud on-ramp → Datacenter Expansion schedule. 3. What CoS does the workload need? Real-time voice/video → Dedicated. Mixed UC + data → Enhanced. Replication, backup, file transfer → Basic. 4. What's the demand pattern? If there is any predictable off-peak window of 4+ hours per day, scheduling will pay back the management software cost in the first month. 5. Is this a permanent circuit or short-term? Both products are no-commitment, so for short-term needs (data migration, event-driven capacity), IoD or EoD work equally well — pick based on Layer 2 vs. Layer 3 needs.
When You Genuinely Don't Know
If the product decision is unclear, the right move is to model the cost of both options at your expected utilization and look at total cost of ownership, not just hourly rates. The [EoD cost calculator](/features/ethernet-on-demand) and the [IoD cost calculator on the homepage](/) handle the math. Plug in your peak hours, your tier needs, and your CoS requirements; the calculators surface the realistic monthly cost for each option.
For most enterprises, the answer is some combination of both — IoD where the internet is the destination, EoD where a private endpoint is. The two products are complementary, not competitive. The mistake is using one where the other fits.