Choosing Between Single-Mode and Multi-Mode Fiber for 100G Data Center Links

Problem: Planning a 100G link between two racks

When you need to connect servers or switches in adjacent rows, the first decision is often what type of fiber to pull. Pulling the wrong cable can mean either unnecessary expense or a link that fails to reach the target distance.

Thesis: Match fiber type to reach and budget

For most data‑center interconnects the choice boils down to single‑mode fiber (SMF) for longer reaches and multi‑mode fiber (MMF) for short, cost‑effective links. The right pick depends on the actual distance, the transceiver you plan to use, and how much headroom you want for future upgrades.

Understanding the fundamentals

Single‑mode fiber has a core diameter of about 9 µm and supports only one propagation mode, which eliminates modal dispersion. Attenuation is low—≈0.35 dB/km at 1310 nm and ≈0.22 dB/km at 1550 nm—allowing tens of kilometers without regeneration for 10 G/25 G/100 G Ethernet.

Multi‑mode fiber uses a larger core (typically 50 µm) and carries many modes. Modal dispersion limits the bandwidth‑distance product, so standards define reach limits: OM3 supports 10 G up to 300 m and 100 G up to 100 m; OM4 extends those to 400 m and 150 m respectively.

When MMF makes sense

If the link stays inside a row or between adjacent racks (≤300 m for 10 G, ≤150 m for 100 G), OM3/OM4 with SR4 transceivers is usually the cheapest option. The transceiver optics are VCSEL‑based, the connectors are often MPO‑12, and the cable cost per meter is lower than SMF.

When SMF is preferable

Beyond those distances, or when you anticipate moving to 400 G or need a longer stretch between buildings, SMF with LR4/ER4 transceivers becomes the practical choice. The tighter tolerances of the laser source raise the transceiver price, but the fiber itself is cheap and the link budget provides ample margin for splices, connectors, and future upgrades.

Worked example: 200 m 100 G link

Suppose you need to connect two switches located 200 m apart in a campus‑style data center. You evaluate both options.

# Link‑budget numbers taken from typical vendor datasheets (values are illustrative)
# SMF LR4 (1310 nm)
Tx power: -2.3 dBm
Rx sensitivity: -10.6 dBm
Fiber attenuation: 0.35 dB/km × 0.2 km = 0.07 dB
Connector loss (2 × 0.5 dB): 1.0 dB
Splice loss: 0 dB
Available margin = Tx - Rx - (fiber + connectors) = -2.3 - (-10.6) - (0.07 + 1.0) = 7.23 dB

# MMF OM4 SR4 (850 nm)
Tx power: -7.6 dBm
Rx sensitivity: -9.5 dBm
Fiber attenuation: 3.5 dB/km × 0.2 km = 0.7 dB (typical MMF higher)
Connector loss (2 × 0.5 dB): 1.0 dB
Margin = -7.6 - (-9.5) - (0.7 + 1.0) = 0.2 dB

The SMF calculation shows a comfortable 7 dB margin, well above the usual 3 dB design reserve. The MMF SR4 link leaves only ~0.2 dB, which is vulnerable to any extra loss from dirty connectors or bends. In this scenario SMF is the safer pick despite the higher transceiver cost.

Trade‑off and limitation

MMF wins on upfront cable and connector cost, but its reach is capped by modal dispersion. If you later need to extend the link or upgrade to a higher‑speed standard that assumes SMF, you would have to replace the fiber plant. SMF avoids that limitation but requires more expensive laser‑based transceivers and careful handling to avoid exceeding the minimum bend radius (typically 10 mm).

Actionable steps

1. Determine the maximum link distance and any planned future upgrades.
2. Collect transceiver specifications (Tx power, Rx sensitivity, operating wavelength) for both SMF and MMF options.
3. Perform an optical loss test using a calibrated power meter and light source at the transceiver’s wavelength; note the measured loss and compare it to the calculated link budget.
4. If the measured loss leaves at least a 3 dB margin, the link is viable; otherwise consider upgrading the fiber type or adding repeaters.
5. Label the installed fiber (SMF vs MMF) and update documentation to avoid mixing connector types.