A wavelength service is a dedicated channel of light, at a fixed capacity such as 10G, 100G or 400G, carried between two points on a carrier's DWDM optical system. DWDM puts many wavelengths on a single fiber pair, each on its own colour of light, so you get full line-rate capacity that carries any protocol, without buying or lighting fiber yourself. It sits between an Ethernet private line and dark fiber in price, control and responsibility.
Waves used to be something only carriers bought from each other. Now data centre operators, financial firms, media companies, hospitals, universities and companies moving large data sets buy them directly. If someone has suggested one to you, this is what they meant.
How DWDM works
DWDM stands for dense wavelength division multiplexing. The idea is the same as radio stations sharing the air by using different frequencies. Each customer signal is converted onto a precise wavelength, a single colour of infrared light, and many of those colours travel together down the same fiber without interfering.
- The band. Most systems use the C-band, around 1550 nanometres, where fiber loses the least light and optical amplifiers work well.
- The grid. Channels sit on a standard ITU grid, traditionally 100 GHz or 50 GHz apart. Newer "flexible grid" systems size each channel to the signal, which is how 400G and faster waves fit.
- The equipment. Transponders at each end turn your signal into a wavelength and back. Amplifiers along long routes boost every channel at once. ROADMs (reconfigurable optical add-drop multiplexers) at network nodes let carriers add, drop or redirect individual wavelengths without disturbing the rest.
A single fiber pair on a modern system can carry dozens of wavelengths, each at 100G or more. That is why carriers can sell you a whole 100G channel without laying new glass: they are renting you one colour on fiber they already light.
What you actually receive
At each end, the carrier hands you an optical port at the agreed rate, typically a standard Ethernet interface such as 10GBASE-LR, 100GBASE-LR4, or one of the 400G variants, or an OTN (optical transport network) interface if you need one. Everything above the light is yours. The carrier does not switch, route or inspect your frames.
- Full line rate. A 100G wave carries 100G. There is no committed-rate-versus-port-speed question, and no oversubscription.
- Protocol transparency. Ethernet, Fibre Channel for storage, or OTN, whatever the handoff supports.
- Deterministic latency. The path is fixed, so delay is essentially the same every day. That matters for storage replication, trading and some media workflows.
More detail on how we quote them is on the wavelength services page.
Wave vs Ethernet vs dark fiber
| Ethernet private line | Wavelength | Dark fiber | |
|---|---|---|---|
| What you rent | Capacity on the carrier's switched Ethernet network | One optical channel on the carrier's DWDM system | The physical fiber strands |
| Typical capacities | From about 10 Mbps to 100 Gbps | 10G, 100G, 400G | Whatever your own optics can drive |
| Who lights it | The carrier | The carrier | You |
| Protocols carried | Ethernet | Ethernet, Fibre Channel, OTN | Anything |
| Latency | Low, with switching along the way | Fixed by the optical route | Fixed by the physical route |
| Your responsibility | Your equipment at each end | Your equipment at each end | Optics, amplification if needed, monitoring |
| Best for | Sub-10G site links, many locations | 10G and up between a few key points | Very high or growing capacity, total control |
For site-to-site links below 10G, Ethernet transport is usually the better buy, and our guide to EPL vs EVPL covers the choices there. At the other end, dark fiber makes sense when you would fill several waves and have the staff to run optics yourself.
Protected or unprotected
An unprotected wave follows one path. If that fiber is cut, the wave is down until it is repaired. A protected wave has a second path, and the system switches to it automatically if the first fails. Carriers deliver protection in different ways, sometimes with optical switching at the ends and sometimes on their network, so ask how it works and how quickly it switches.
The alternative is to buy two unprotected waves on physically diverse routes, perhaps from two carriers, and let your own equipment handle the failover. That costs more but removes dependence on one carrier's network. Either way, the protection is only as good as the separation between the paths. Ask for route maps, and check where the two paths share a bridge, a tunnel, a conduit or a building entrance.
Latency: the arithmetic
Light in fiber travels at roughly two-thirds of its speed in a vacuum, which works out to about 5 microseconds per kilometre, one way. Equipment adds a little on top. The number that matters is the fiber route length, not the straight-line distance, because fiber follows streets, rail lines and rights of way.
A worked example
A hypothetical design used to show the arithmetic. The distances, sites and requirements are invented.
Suppose a company runs its main systems in a data centre and wants a second data centre about 40 km away for replication and disaster recovery. It needs 100G between them, and its storage vendor lists a maximum round-trip time for synchronous replication.
Two carriers bid. Route A is 65 km of fiber. Route B takes a longer path around a river crossing and is 90 km.
- Route A: 65 km × 5 µs = about 0.33 ms one way, or about 0.65 ms round trip.
- Route B: 90 km × 5 µs = about 0.45 ms one way, or about 0.9 ms round trip.
Both are comfortably fast in everyday terms. If the storage vendor's limit were tight, though, the difference could matter, and it could matter more once the protection path is included, since a protected wave's backup route is often longer than its primary. The company should ask each carrier for the latency of both the working and the protection path before choosing.
If the company expected to need several 100G links within a few years, the same exercise would be worth running for dark fiber, where it would add its own optics and could grow capacity by upgrading equipment rather than ordering new waves.
Who buys wavelengths
- Data centre interconnect. Linking two data centres, or a data centre and a colocation site, for replication and backup.
- Cloud on-ramps. High-capacity links from your data centre to where a cloud provider connects. See cloud connectivity.
- Financial and trading firms that care about fixed, low latency on specific routes.
- Media and research organisations that move very large files between fixed points.
- Carriers and ISPs building their own backbone capacity.
If none of these describe you, you probably do not need a wave. An ordinary office rarely does.
Mistakes we see when buyers first order waves
- Comparing price without route. Two 100G waves between the same buildings are not the same product if one runs 30 km further or shares a conduit with your existing circuit.
- Assuming "protected" means diverse. Protection only helps if the backup path avoids the places where the primary is most likely to be cut. Look at the map.
- Forgetting the last few metres. In a data centre, the wave ends at the carrier's equipment, and a cross-connect to your cage is ordered separately, often from the data centre operator, with its own lead time and monthly fee.
- Mismatched optics. The handoff has to match the optics in your switch or router. Confirm the exact interface type before either side orders hardware.
- Buying one wave when growth is certain. If you already know you will need four, price dark fiber or a multi-wave commitment at the same time.
Timelines and what drives cost
Where both ends are already on the carrier's optical network, such as major data centres, a wave can often be delivered on a timeline similar to a lit-building order, typically 2–4 weeks. Where one end needs a new fiber lateral or new optical equipment, expect a construction-style schedule of 60–90 days or more.
The main cost drivers are capacity, distance, protection, the route requested, contract term, and whether both ends are on-net. Diverse or low-latency routes can carry a premium because fewer carriers can offer them. We do not publish prices, because the same wave between two different pairs of buildings can be priced very differently.
Questions to ask your carrier
- What is the fiber route length and the expected latency, for the working path and any protection path?
- Can I see a route map, and where might it share a path with my other circuits?
- Is the wave protected, and how does the switchover work and how quickly?
- What handoff interface and optics are supported at each end?
- Are both ends on your network today, or does either need construction?
- What does the SLA cover: availability, repair time, latency?
- If I need a second wave on the same route later, how long does that take?
Getting waves compared route by route
Wavelengths are one of the products where comparing carriers changes the answer most, because each has different routes, different diversity and different latency between the same two points. We put identical specifications out to several optical networks and bring back price, route and latency side by side. The carrier pays us, so there is no markup. The quote is free with no obligation and you usually hear back the same day. Send us both addresses, call 478-758-8091 or text (347) 870-0965.
