2km vs 40 km Data Center Interconnect

Data center interconnect spans distances from a few hundred meters to hundreds of kilometers. Two reach tiers anchor the short and medium segments of that spectrum: 2 kilometers for campus building-to-building links and 40 kilometers for metro and regional interconnect. Both use single-mode fiber and duplex LC connectors. Both are direct-detect PAM4 interfaces. The differences lie in the wavelength plan, the optical power budget, the laser design, and the physical environment each tier is built to serve.

The 2-kilometer tier is represented by FR interfaces. It connects buildings on a corporate campus, a university, or a data center campus with multiple halls. The fiber path may cross outside plant, but it remains within a controlled environment where the operator owns or controls the fiber. The 40-kilometer tier is represented by ER interfaces. It connects data centers across a metropolitan area or between cities in a region. The fiber path often traverses leased fiber, carrier patch panels, and utility infrastructure that the operator does not control.

The distance gap is twentyfold, and the engineering gap is shaped by the environment as much as the fiber loss. A 40-kilometer metro link must account for connector losses in carrier patch panels, splice losses in leased fiber, and the possibility of fiber re-routing over the life of the installation. A 2-kilometer campus link operates in a more controlled environment where the fiber path is known and stable.

1. What 2 km DCI Covers

A 2-kilometer interconnect connects buildings on a campus. The fiber runs between a data center and an office building, between two data halls in separate buildings, or between a campus core and an edge facility. The path is typically owned or leased by the same organization that operates the network. The fiber is single-mode, the connectors are duplex LC, and the link is engineered with a known fiber plant.

The 2-kilometer tier is defined by FR interfaces. A 400GBASE-FR4 module uses four CWDM wavelengths, each carrying 100G PAM4, multiplexed onto a single fiber pair. An 800GBASE-FR4 module uses eight wavelengths of 100G PAM4 or four wavelengths of 200G PAM4. The internal CWDM multiplexer combines the wavelengths onto one fiber for transmit and separates them from one fiber for receive.

1.1 2 km DCI Characteristics

• Reach: 2 kilometers.

• Wavelength: CWDM at 1271, 1291, 1311, and 1331 nm for FR4.

• Optical architecture: CWDM multiplexing on a single fiber pair.

• Connector: Duplex LC.

• Fiber count per link: 2 fibers.

• Power budget: 8 to 10 dB.

• Laser cooling: Uncooled or lightly cooled.

• Environment: Campus, controlled fiber plant.

2. What 40 km DCI Covers

A 40-kilometer interconnect connects data centers across a metropolitan area or between cities in a region. The fiber runs between two facilities that may be several tens of kilometers apart, traversing streets, carrier hotels, and utility infrastructure. The fiber is often leased from a carrier, and the operator may not control the patch panels or splice enclosures along the path. The link must be engineered with margin for the additional connector and splice losses that accumulate in a metro or regional fiber plant.

The 40-kilometer tier is defined by ER interfaces. A 100GBASE-ER4 module uses four LAN-WDM wavelengths around 1310 nm, each carrying 25G NRZ, multiplexed onto a single fiber pair. A 400GBASE-ER8 module uses eight wavelengths, each carrying 50G PAM4. An 800GBASE-ER8 module uses eight wavelengths of 100G PAM4. Some ER variants use wavelengths in the 1550 nm window to take advantage of lower fiber attenuation and the possibility of erbium-doped fiber amplification for extended reaches.

2.1 40 km DCI Characteristics

• Reach: 30 to 40 kilometers.

• Wavelength: LAN-WDM at approximately 1295, 1300, 1305, and 1309 nm, or 1550 nm.

• Optical architecture: LAN-WDM or 1550 nm on a duplex LC pair.

• Connector: Duplex LC.

• Fiber count per link: 2 fibers.

• Power budget: 18 to 25 dB.

• Laser cooling: Temperature-controlled.

• Environment: Metro and regional, leased or carrier fiber.

3. Wavelength Plan and Laser Design

FR uses CWDM wavelengths with 20 nm spacing. The wide spacing tolerates the wavelength drift of uncooled or lightly cooled lasers, which reduces module cost and power consumption. The CWDM multiplexer is relatively simple and inexpensive because the channels are far apart.

ER uses LAN-WDM wavelengths with approximately 4.5 nm spacing, or 1550 nm wavelengths for extended reach. The tighter spacing requires temperature-controlled lasers and a more precise multiplexer. The tighter spacing is necessary to keep the four or eight wavelengths within the 1310 nm window, where chromatic dispersion is near zero, while maintaining enough separation for the multiplexer to separate them. The 1550 nm variants trade dispersion for lower attenuation and amplification compatibility.

4. Power Budget and Link Engineering

The 2-kilometer link budget is sized for a controlled campus fiber plant. Fiber loss at 1310 nm over 2 kilometers is approximately 0.7 dB. Connector and splice losses add 1 to 2 dB. The CWDM multiplexer adds 1 to 2 dB. The total path loss is typically 3 to 5 dB, and the module provides 8 to 10 dB of budget, leaving ample margin for aging and temperature variation.

The 40-kilometer link budget is sized for a metro or regional fiber plant with more connector and splice losses. Fiber loss at 1310 nm over 40 kilometers is approximately 14 dB. At 1550 nm, fiber loss over 40 kilometers is approximately 8 dB. Connector and splice losses in a leased metro path can add 2 to 4 dB. The LAN-WDM multiplexer adds 2 to 3 dB. The total path loss is typically 12 to 20 dB, and the module provides 18 to 25 dB of budget.

5. Fiber Plant and Environment

FR and ER both use duplex LC connectors and two fibers per link. The physical connector type is identical. The difference lies in the fiber plant environment and the engineering discipline each tier requires.

A campus FR link is engineered with a known fiber path. The operator knows the length of the fiber, the number of patch panels, and the splice locations. The link budget calculation is straightforward, and the margin is generous because the path is stable.

A metro or regional ER link is engineered with less certainty. The fiber may be leased, and the operator may not know the exact number of patch panels or splice enclosures along the path. The link budget must include margin for unexpected losses. The temperature-controlled lasers and higher-power transmitters in ER modules provide that margin.

6. Application Boundaries

The 2-kilometer tier covers campus building-to-building links. It connects a data center to an office building, two data halls in separate buildings, or a campus core to an edge facility. The distance is short enough that no amplification is required, and the fiber plant is controlled enough that the link budget can be calculated with confidence.

The 40-kilometer tier covers metro and regional data center interconnect. It connects two data centers within a city or between nearby cities. The distance is long enough that the link budget must be engineered with margin, but still short enough that no amplification is required for the 1310 nm variants. The 1550 nm variants can be amplified if the reach needs to extend beyond 40 kilometers.

7. Module Cost and Power

FR modules cost less than ER modules. The CWDM multiplexer, uncooled lasers, and relaxed wavelength tolerances reduce the module cost. ER modules cost more because of the LAN-WDM multiplexer, temperature-controlled lasers, and higher-power transmitters. The premium for ER over FR is typically 40 to 70 percent depending on the data rate and the vendor.

Power consumption is higher for ER modules. A 400GBASE-FR4 module draws approximately 10 to 12 watts. A 400GBASE-ER8 module draws approximately 14 to 16 watts. An 800GBASE-FR4 module draws approximately 14 to 16 watts. An 800GBASE-ER8 module draws approximately 20 to 24 watts. The additional power in ER comes from the higher laser drive current, the temperature control, and the more sensitive receiver.

8. Standards and Ecosystem

FR is defined by IEEE 802.3cu for 100GBASE-FR1 and 400GBASE-FR4, and by IEEE 802.3df for 800GBASE-FR4. ER is defined by IEEE 802.3ba for 100GBASE-ER4, by IEEE 802.3bs for 400GBASE-ER8, and by IEEE 802.3df for 800GBASE-ER8.

9. Comparison Summary

10. Selection Framework

The choice between the 2-kilometer and 40-kilometer tiers is determined by the physical distance and the fiber plant environment.

• Distance under 2 kilometers on a controlled campus fiber plant: Use FR. The module is cheaper, the CWDM wavelength plan is sufficient, and the link budget is straightforward.

• Distance between 2 and 40 kilometers: Use ER. FR cannot reach, and ER provides the additional power budget and tighter wavelength control needed for the longer path.

• Distance beyond 40 kilometers: Use ZR or coherent optics. ER is rated for 30 to 40 kilometers and does not have the budget for longer reaches without amplification or coherent detection.

The fiber plant environment matters as much as the distance. A 20-kilometer link on a controlled campus fiber plant may be served by LR or ER with margin to spare. A 20-kilometer link on a leased metro fiber plant may require ER for the additional margin needed to accommodate unknown connector and splice losses. The choice is not just about distance; it is about how much uncertainty the fiber path introduces.

11. Common Misconceptions

• "2 km and 40 km DCI use different connectors." False. Both use duplex LC connectors and two fibers per link. The physical fiber plant is identical.

• "40 km DCI requires amplification." False. The 40-kilometer tier is still direct-detect PAM4 and does not require amplification for the 1310 nm variants. The 1550 nm variants can be amplified if the reach needs to extend beyond 40 kilometers, but amplification is not required at the rated distance.

• "FR modules can be used for 40 km links." False. FR is rated for 2 kilometers. For 40-kilometer links, ER is required.

• "ER modules are always more expensive per gigabit." Not necessarily. ER modules cost more per port, but both tiers use two fibers per link. The cost difference is in the module, not the fiber. The cost per gigabit depends on the data rate and the module vendor.

• "2 km DCI is obsolete." False. Campus building-to-building links are a permanent part of data center infrastructure. FR remains the standard interface for these links because it matches the distance and the controlled fiber environment.

12. Summary

The 2-kilometer and 40-kilometer DCI tiers serve different physical environments. The 2-kilometer tier, represented by FR interfaces, uses CWDM wavelengths with 20 nm spacing and uncooled or lightly cooled lasers. It connects buildings on a campus over a controlled fiber plant. The 40-kilometer tier, represented by ER interfaces, uses LAN-WDM wavelengths with 4.5 nm spacing and temperature-controlled lasers, or 1550 nm wavelengths for extended reach. It connects data centers across a metropolitan area or between nearby cities over leased or carrier fiber.

Both tiers use duplex LC connectors and two fibers per link. Both are direct-detect PAM4 interfaces. The differences lie in the wavelength plan, the laser design, the power budget, and the fiber plant environment. FR is cheaper and simpler for campus links under 2 kilometers. ER provides additional margin for metro and regional links up to 40 kilometers.

The choice between them is determined by the physical distance and the fiber plant environment, not by a performance comparison. A campus uses FR. A metro or regional area uses ER. The two tiers coexist in the same network, connected by ZR or coherent optics where the distance exceeds 40 kilometers.

13. Q&A