Data center fiber cabling types: single-mode vs multimode selection

The fiber-type choice, and why it sets the link

Data center fiber cabling comes in two types, single-mode and multimode, and choosing between them, plus the right grade and the matching optic, is the cabling decision that sets how far a link reaches, how fast it runs, and what it costs. Single-mode has a tiny core, roughly 9 microns, that carries one light path a long way. Multimode has a larger core, roughly 50 microns, that carries many paths a shorter way for less money on the optic.

The type is not the whole job. The fiber rides inside the structured cabling system, the TIA-942 spaces and the hierarchical-star design the structured cabling work lays out, and it gets proven by the insertion-loss and OTDR testing the fiber certification work covers. This guide answers the one question those leave open: which glass, which grade, which connector, for which link.

Get the choice wrong and you find out at the worst time. A multimode link pulled to a distance or a future rate the optic cannot hold links intermittently or not at all, and the fix is a re-pull in a live room. Pick the type for where the speeds are heading, not just where they sit today.

How optical fiber carries the data

A fiber carries data as pulses of light down a thread of glass. The light travels in the core, the center of the fiber, which is surrounded by cladding, a layer of glass with a slightly lower refractive index. That index difference is the whole trick. Light hitting the boundary at a shallow angle reflects back into the core instead of escaping, total internal reflection, so the pulse stays trapped and follows the fiber around bends and over distance.

A transceiver at one end turns the electrical signal into light, switching it on and off fast enough to carry the bit rate, and a receiver at the far end turns it back. The glass is drawn to a precise core and cladding diameter, coated, and jacketed. The core diameter is the number that splits the two fiber types, because how wide the core is decides how many distinct light paths, called modes, the fiber will carry.

None of this works if the end face is dirty or the bend is too tight. The physics underneath is simple. Light, trapped in glass by a refractive-index step, carries the data from one optic to the other.

Single-mode fiber

Single-mode fiber has a core around 9 microns across, narrow enough that it carries just one mode, a single light path straight down the center. With only one path there is no modal dispersion, the spreading that happens when many paths arrive at slightly different times, so the pulse stays sharp over a long distance. That is why single-mode reaches kilometers where multimode reaches tens or hundreds of meters.

The light source is a laser, usually at 1310 nm or 1550 nm, focused into that tiny core. Single-mode glass has very low attenuation, well under 0.5 dB/km on good fiber, and effectively unlimited bandwidth for the distances inside and between data centers. The cost was never the glass. Single-mode fiber is cheap. The cost was the optic, because launching light into a 9 micron core takes tighter tolerances and a precise laser, which historically made single-mode transceivers several times the price of multimode.

That price gap has been closing fast, and it is the trend reshaping the whole choice. Single-mode is graded OS1 and OS2, covered below.

Multimode fiber

Multimode fiber has a much larger core, around 50 microns on the grades used today and 62.5 microns on the old OM1. A wide core accepts light at many angles, so it carries many modes at once, dozens of light paths down the same fiber. That makes coupling light in cheaper, which is the point of multimode, but it comes with a limit. The modes travel slightly different path lengths and arrive at slightly different times, smearing the pulse. That spreading is modal dispersion, and it is what caps multimode distance.

The source is a VCSEL, a vertical-cavity surface-emitting laser at 850 nm, far cheaper to make and drive than a single-mode laser. A short-reach multimode optic costs a fraction of its single-mode equivalent, which is why multimode owned the short links inside the data hall for years.

The catch is that as the bit rate climbs, modal dispersion eats the distance faster. The same OM4 fiber that carried 10G a few hundred meters carries 400G only about 100 m or less. Multimode is graded OM1 through OM5.

Single-mode vs multimode: the core difference

The split between the two comes down to core size and what it does to distance, cost, and the optic. Single-mode's narrow core kills modal dispersion and buys distance, at the price of a more expensive laser. Multimode's wide core makes the optic cheap and the coupling easy, at the price of distance that shrinks as the rate goes up.

For years the rule on the floor was simple. Multimode for the short links because the optics were cheap, single-mode for the long links because nothing else reached. That rule still describes the physics. What has changed is the economics, because single-mode optics have dropped in price to where the old cost argument for multimode gets weaker every refresh.

Read the table as the physics. Read the selection and cost sections for how the money now lands.

The multimode grades: OM1 through OM5

Multimode is graded by how much bandwidth the glass carries, which sets the distance at a given speed. The grades run OM1 through OM5, and the working world is OM3, OM4, and OM5. OM3 and OM4 are laser-optimized for the 850 nm VCSEL and jacketed aqua, with OM4 carrying more bandwidth and more reach than OM3. The headline figure is the effective modal bandwidth, about 2000 MHz-km for OM3 and about 4700 MHz-km for OM4 at 850 nm, but the number that matters on the job is the reach the optic supports at your rate, not the bandwidth-distance product itself.

OM5 keeps the 50 micron core and adds wideband performance across roughly 850 to 953 nm, jacketed lime green, so it can carry four wavelengths on one pair using short-wavelength division multiplexing. OM5 was meant to stretch multimode further, though much of the industry stepped past it toward single-mode for the high-speed links.

OM1 and OM2 are legacy. OM1 is the old 62.5 micron orange fiber, OM2 the older 50 micron, both built for LED sources and lower rates. You still find them in older rooms, and you do not mix them on a link with the laser-optimized grades. Confirm the grade by the jacket and the cable print, not by assumption, because a length of OM3 spliced into an OM1 run is a loss problem waiting to happen.

The single-mode grades: OS1 and OS2

Single-mode is graded OS1 and OS2, and the difference is the glass and where it goes. OS1 is conventional single-mode, built and tested for indoor and campus runs, with a higher maximum attenuation, commonly around 1.0 dB/km. OS2 is low-water-peak, sometimes called zero-water-peak, glass that suppresses the attenuation spike near 1383 nm caused by hydroxyl ions, so it stays low-loss across more of the spectrum and is rated tighter, commonly around 0.4 dB/km.

For the long and the outside-plant runs, OS2 is the one. The low attenuation and the open spectrum let it carry long distances and support wavelength multiplexing across the E-band that OS1 cannot. On a new data center build the single-mode you pull is almost always OS2, because the cost difference over OS1 is small and OS2 covers every reach and every wavelength plan you are likely to need.

The grades map to the ITU-T fiber categories, OS1 to G.652.A and B, OS2 to G.652.C and D, but the attenuation values move with the standard edition and the manufacturer, so verify the rated loss against the cable specification before you build a loss budget on it.

The fiber color code

Jacket color is the fast field identification of fiber type, and it is mostly consistent, with a couple of traps. Yellow is single-mode, OS1 or OS2, the long-reach glass. Aqua is laser-optimized multimode, OM3 and OM4. Lime green is OM5, the wideband multimode. Older multimode, OM1 and OM2, was orange. These follow the common TIA color convention, but color is a convention, not a guarantee.

The traps are real. Some manufacturers jacket OM4 in violet to tell it apart from OM3, since both are otherwise aqua, so a violet cable is OM4 but an aqua cable could be either OM3 or OM4. And the jacket tells you the fiber type, not every grade detail. The cable print, the legend running along the jacket, is the authority. Read it.

Connector boot and housing colors carry their own meaning, often beige or black for multimode and blue for single-mode UPC, green for single-mode APC, but that is the connector polish, not the fiber grade. Do not read the boot color as the fiber type.

Distance and speed by type and grade

Reach is the product of the fiber grade, the optic, and the data rate, and the same fiber reaches far less as the rate climbs. That last part is the trend that drives modern design. Multimode at 10G on OM4 reached a few hundred meters. The same OM4 at 100G with an SR optic reaches roughly 100 m, and at 400G the multimode reach lands near 100 m on SR4.2 and around 50 m on the shorter VR4. Push the rate and the distance collapses. Single-mode, by contrast, holds kilometers across all those rates, because it has no modal dispersion to spend.

The table gives representative reaches. Treat them as the shape of the trend, not as gospel. The actual supported distance is set by the specific transceiver and the IEEE Ethernet specification for that rate, and it depends on the fiber grade and the connector loss in the link. Always confirm the reach against the optic datasheet and the applicable IEEE standard for the rate you are running.

The takeaway from the numbers is plain. Multimode is a short-link technology that gets shorter every speed step, and single-mode is the one fiber whose reach does not move when the rate does.

The optics: SR, LR, and what the transceiver costs

The fiber type does not work alone. It is matched to a transceiver, and the optic is where most of the cost and the reach actually live. Multimode runs on short-reach optics, the SR family, built around the 850 nm VCSEL, which is cheap to make and drive. Single-mode runs on the longer-reach families, LR, FR, DR, and the rest, built around precise 1310 or 1550 nm lasers, which cost more because launching light into a 9 micron core demands tighter tolerances.

The optic and the fiber have to match end to end. An SR multimode module needs multimode fiber of the right grade. An LR single-mode module needs single-mode. Plug a single-mode optic onto multimode fiber, or the reverse, and the link does not work or works far below spec. The transceiver naming carries the reach and the fiber type: SR for short-reach multimode, LR and FR and DR for single-mode, with the number after it usually telling you the fiber or lane count.

The cost story has flipped over the last few cycles. Single-mode optics, once several times the price of multimode, have dropped toward a small premium as silicon photonics matured and volume climbed. That single change is why single-mode now shows up on links that used to be automatic multimode. The structured cabling work covers how the optics ride in the spaces. The point here is to size the fiber to the optic you will actually plug in.

The connectors: LC and MPO/MTP

Two connector families cover almost all data center fiber. The duplex LC is the small two-fiber connector on most switch optics, one fiber to transmit and one to receive, and it is what an SR or LR module in a switch port usually takes. The MPO, often sold under the MTP brand, is a multi-fiber connector that carries 8, 12, 16, or 24 fibers in one ferrule, and it is what the high-speed parallel links and the high-density trunks depend on.

The connector you need follows the optic. A duplex optic, two fibers, takes LC. A parallel optic that drives many fibers at once takes MPO. The polarity and the fiber count on an MPO have to be right end to end, because a wrong-polarity MPO link reads as continuous on a quick check and refuses to pass traffic, which is exactly the failure the structured cabling work walks through in detail. Get the MPO fiber count, gender, and polarity off and the whole link has to be re-cabled.

On single-mode, the connector polish matters too. Reflection-sensitive single-mode links use angle-polished APC connectors, the green ones, to throw reflections into the cladding instead of back at the laser. The fiber certification work covers reflectance and why APC exists.

Parallel optics and MPO breakout

Parallel optics are why MPO connectors took over the high-speed links. Instead of one pair of fibers carrying everything through wavelength tricks, a parallel optic drives several fibers at once, each carrying a lane of the total rate, through one MPO ferrule. A 400G DR4 runs eight single-mode fibers in parallel, four transmit and four receive, through an MPO. A 400G SR8 runs sixteen multimode fibers. The lanes combine to make the rate.

Parallel optics also give you breakout. One high-speed port on an MPO can fan out into several lower-speed ports, an 800G port broken into two 400G or eight 100G links through breakout cabling and cassettes, because the lanes are already separate fibers. That is how a spine switch with high-rate ports feeds a row of lower-rate server links from one trunk.

The design consequence is that the fiber count and the MPO format have to be planned around the breakout the network team will run, not just the link count. Order the trunk fiber count, the gender, and the polarity to match the cassette and the breakout plan. This is where the high-fiber-count migration the structured cabling work describes lives, and it is the most expensive thing to get wrong, because a trunk built to the wrong format is a re-pull.

The data center link classes and their media

Inside a data center the links break into a few distance classes, and each one has a media that fits. The shortest, server to the top-of-rack switch inside one rack, is often not fiber at all but a direct attach copper cable, DAC, a fixed-length twinax assembly with the transceivers built onto the ends, cheap and low-power for a few meters. In-row and row-to-row links, up to tens or low hundreds of meters, have been the multimode zone, OM4 on SR optics, because the reach fit and the optics were cheap.

The longer links, across the hall, between halls, and out to other buildings, are single-mode, OS2 on LR or DR or FR optics, because nothing else reaches. That much has been stable for years. What is moving is the middle. As single-mode optics dropped in price, more of the in-row and cross-hall links that used to be automatic multimode are going single-mode, especially on new high-speed builds.

Map the link distances first, then pick the media per class. The intra-rack stays copper DAC for the very short hops. The open question is where the multimode-to-single-mode line falls for everything longer, and that line keeps sliding toward single-mode.

Copper DAC vs fiber for the shortest hops

For the very shortest links, copper still wins on cost and power. A direct attach copper cable, the twinax DAC, connects a server to the top-of-rack switch over a few meters with the transceivers molded onto the ends, no separate optic to buy and far less power per port than any fiber link. Inside a single rack, DAC is usually the right answer, and active optical cables, AOC, extend the same idea a bit farther on fiber when the run outgrows copper.

The line is short. DAC reach is a few meters, and at the higher rates even that shrinks, so the moment a link leaves the rack or stretches past a handful of meters, fiber takes over. The Cat6A copper access and PoE side is its own world the structured cabling work covers. The point for fiber selection is just to know that the shortest hops are not fiber's job, and forcing fiber onto a one-meter intra-rack link wastes money and power for nothing.

Which fiber is best for a data center?

The selection comes down to the link distance, the rate, the cost, and how long the plant has to last. For short links inside the hall, multimode on OM4 has been the cost choice, because the SR optics were cheap and the reach fit. For the long links, single-mode on OS2, because it is the only thing that reaches. That is the textbook split, and it still describes the physics.

The reason the textbook answer is shifting is future-proofing. Multimode reach shrinks every time the rate steps up, so a multimode link that fits today can fall short at the next refresh, and a re-pull in a live room is the most expensive cable you will ever install. Single-mode reach does not move with the rate, and the optic premium that used to make single-mode expensive has mostly collapsed. So on a new build, especially a high-speed one, more designers pull single-mode even on links short enough for multimode, to buy a plant that does not need re-cabling when 800G and beyond land.

The honest version: if the link is short, cheap optics matter most, and the rate is settled, multimode still makes sense. If the plant has to live fifteen to twenty-five years through rate steps you cannot predict, single-mode is the lower-risk pull. Decide per project, with the refresh horizon in front of you, not on shop habit.

The move to single-mode as speeds rise

The shift toward single-mode in the data center is the timely consideration in 2026, and it is driven by two curves crossing. Multimode reach keeps falling as rates climb, while single-mode optic cost keeps falling as silicon photonics scale. The point where single-mode became viable for links as short as 100 m has already arrived for high-speed builds, and a base-8 single-mode design is now a common default where base-12 multimode used to be automatic.

This does not retire multimode overnight. Plenty of existing OM4 plant has years of service left at the rates it was built for, and short in-rack links still favor cheap optics. But for new high-speed fiber, the design center has moved. The question on a fresh build is less whether single-mode is worth the premium and more whether multimode is worth the reach risk.

AI and high-speed fiber

AI training clusters are pushing fiber selection harder than anything before them, because the back-end fabric that connects GPU to GPU carries enormous bandwidth between every node, and almost all of it is fiber at 400 and 800G. The fiber count per rack in a GPU pod dwarfs a conventional server row, thousands of links where a normal hall had hundreds, almost entirely on MPO parallel optics.

The type choice in an AI fabric leans on single-mode and short-reach multimode together. The high-rate links that have to reach across the fabric run single-mode OS2 on DR and FR optics with parallel MPO, because the reach and the rate push past what multimode holds economically at that scale, and because single-mode does not need re-pulling at the next rate step. Short multimode SR still serves some in-rack and very short hops where the optic cost wins.

The sheer volume of identical MPO trunks is the practical problem. The fiber type and grade per link, the MPO format, and the labeling all have to be planned for that density before the first trunk lands, which is exactly the AI and GPU cluster cabling the structured cabling work details. Get the type and format right at the order, because field-fixing a fabric of thousands of trunks is not a thing you do.

Loss budget and testing, in brief

Every link has a loss budget, the total light loss it is allowed across the fiber, the connectors, and any splices, and the budget depends on the fiber type and the link length. Single-mode runs very low attenuation, well under 0.5 dB/km, so the connectors dominate its budget on a short link. Multimode runs higher attenuation, around 3 dB/km at 850 nm, so the fiber itself eats more of the budget as the run grows. The optic also has a power budget, the loss it can tolerate and still link, and the cabling loss has to fit inside it.

You prove the link by test, not by assumption. Tier 1 insertion-loss certification with an optical loss test set gives the end-to-end loss against the budget, and Tier 2 with an OTDR maps where the loss lives. The how, the bi-directional averaging, the per-event caps, and the acceptance limits, is the fiber OTDR and certification work, and that is where to go for the testing detail. The selection point is only this: pick the fiber type and grade so the link loss fits inside the optic power budget at the distance you are running, then test to confirm it.

Bend radius and handling

Fiber is glass, and it does not forgive a tight bend or a hard pull. Bend a fiber tighter than its minimum bend radius and it leaks light, a macrobend that shows up as loss on a trace, and bend it hard enough or often enough and the glass cracks. The minimum radius is set by the cable, commonly given as a multiple of the cable diameter, larger under pulling tension than at rest, so hold the radius at every turn and especially where slack stacks up behind a patch panel.

Bend-insensitive fiber has made this more forgiving, tolerating tighter bends with less loss, but it is a margin, not a license to kink the cable. The loss from a tight bend is also wavelength-dependent, worse at the longer wavelength, which is why a bend can pass at 1310 nm and fail at 1550, a trap the certification work covers.

Handling is the rest of it. Keep end faces clean and capped, because a dirty connector is the leading cause of fiber loss. Dust on a core drives up loss and reflection and grinds into the mating face. Inspect before you mate, every time. Do not exceed the cable pull tension, use proper lubricant and sweeps, and never cinch a fiber bundle with a tight zip tie, which crushes the glass into a loss point.

Pre-terminated trunks vs field termination

Fiber comes to the job two ways, pre-terminated or field-terminated, and the data center has moved hard toward pre-terminated. A pre-terminated trunk is built and tested in the factory, a length of fiber with MPO or LC connectors already on both ends, often with breakout cassettes to match, so the install is plug-and-play and the connector quality is factory-controlled. Field termination means the installer puts the connectors on in place, by fusion splicing pigtails or by mechanical connectors, which takes skill, time, and clean conditions.

For the high-fiber-count, high-density plant a modern hall needs, pre-terminated wins on speed and on consistency. You cannot field-terminate thousands of MPO trunks in a GPU pod on schedule, and factory terminations test better and more uniformly than field work under a raised floor. Field termination still has its place for odd lengths, repairs, and runs where a trunk cannot be ordered to length.

The tradeoff moves to the order. Pre-terminated means the lengths, the connector types, the polarity, and the fiber counts all have to be right when you order, because a trunk built to the wrong spec is scrap. Field termination keeps the flexibility at the cost of labor and variability. Most large fiber plants now lean pre-terminated and put the care into the trunk schedule.

Labeling and management of the fiber type

A fiber plant you cannot identify is a fiber plant you cannot run, and the fiber type and grade are part of what the labeling has to carry. Every cable gets labeled at both ends to a consistent scheme, the TIA-606 administration the structured cabling work lays out, and the record should capture not just the port-to-port assignment but the media, the OM4 or OS2 grade, and the MPO polarity, so a tech at the rack knows what is in the tray without guessing.

The reason this matters more for fiber type than people expect is mixing. A jumper of the wrong type or grade dropped into a link, an OM3 patch on an OM4 trunk or a single-mode jumper on a multimode link, reads fine on a quick look and fails on loss or fails to link. The label and the record are how you keep the types from getting crossed over the life of the room.

Set the scheme before the first trunk lands, and keep the as-built honest. The day the operations team trusts the record over a physical trace is the day a high-density fiber plant becomes maintainable.

The cost tradeoff and where it crosses over

The cost of a fiber link is the fiber plus the optics plus the labor, and the type changes where the money sits. Multimode glass and single-mode glass cost about the same per foot now. The difference was always the optic. A short multimode link on cheap SR optics has been the low-cost choice, while single-mode carried a per-optic premium that added up across thousands of ports.

That premium has shrunk. Single-mode optics that used to run several times the price of multimode are now closer to a small premium, sometimes near parity at the high rates, as silicon photonics matured and volume climbed. So the system cost crossover has moved. At low rates and short reach, multimode still costs less. At high rates, long counts, and a long service life, single-mode's optic premium is small enough that the reach insurance is worth it, and the re-pull you avoid dwarfs the optic difference.

Cost the whole link over the refresh horizon, not the optic on day one. The cheapest fiber to install is sometimes the most expensive to own, because the multimode link that fit at 100G is the one you re-pull at 800G. Run the numbers per project, because the crossover keeps sliding toward single-mode as rates rise.

Field example: an OM4 row that linked at 100G and failed at 400G

A row of switches in a new hall went in on OM4 with 100G SR optics, certified clean, links up, everyone happy. Eighteen months later the refresh dropped 400G modules into those ports, and a stretch of the longest in-row links would not establish. The optics were good. The fiber was clean and passed loss. The links simply would not come up at the higher rate.

The cause was reach, not damage. Those runs were near 120 m, fine for 100G SR on OM4, but past the roughly 100 m a 400G SR4.2 multimode optic supports. Nothing failed. The same glass that carried 100G to that distance could not carry 400G that far, because the higher rate spends the modal-dispersion budget faster. The fix was a re-pull of those runs in single-mode, in a live row, at the worst possible time and cost.

The lesson the team kept: certify-clean does not mean future-proof. A multimode link is only as good as the reach at the rate you will eventually run, and on the long in-row runs they should have pulled single-mode the first time. They cost the OM4 savings against one re-pull and never won that trade again.

What to document

A fiber plant whose type and grade nobody recorded is a plant the next person has to trace by hand. Capture the fiber type and grade per link, single-mode OS2 or multimode OM4 and the rest, the connector type and MPO polarity and fiber count, the link length, the optic and rate the link was built for, the loss certification result, and the labeling that ties it all to the field.

Two records carry the most weight later. The media-by-link record, so a refresh knows what reach each run actually has before the new optics arrive, and the certification result, so the plant performance is proven. A package that says only fiber, without the type, grade, polarity, and reach, leaves the next team guessing at the one thing that decides whether the link survives the next rate step.

Common mistakes

• Pulling multimode to a distance or a future rate the optic cannot hold, then re-pulling in a live room.
• Mixing fiber types or grades on one link, an OM3 jumper on OM4 or a single-mode patch on multimode.
• Plugging a single-mode optic onto multimode fiber, or the reverse, so the link fails or runs far below spec.
• Ignoring where the rate is heading, so a multimode link that fits today falls short at the next refresh.
• Getting the connector type, MPO polarity, or fiber count wrong, so a link reads continuous and never passes traffic.
• Exceeding the optic loss budget with too many connectors or too long a run for the grade.
• Reading jacket color as the whole answer instead of confirming the grade on the cable print.
• Bending fiber past its minimum radius or cinching it with tight zip ties, leaking light at a macrobend.

Field checklist

Standards and references

The fiber grades and the cabling framework come from the TIA and ISO/IEC standards. ANSI/TIA-568, the .3 part, covers optical fiber components and performance, including the OM and OS grade definitions and the test requirements. TIA-942 is the data center telecommunications infrastructure standard that frames the spaces and the media. ISO/IEC 11801 is the international cabling standard that parallels TIA-568 and carries its own grade and class designations. The fiber-type physics, the core sizes and the categories, also trace to the IEC 60793 fiber standards and the ITU-T G.652 single-mode categories that OS1 and OS2 map to.

The reach by rate is an IEEE matter. The Ethernet physical-layer specifications, the 100GBASE, 400GBASE, and 800GBASE families, set the supported distance for each optic over each fiber grade, which is why the reach numbers belong to the optic and the IEEE spec, not to the fiber alone. Confirm the supported distance against the specific transceiver datasheet and the applicable IEEE standard for the rate.

Edition letters and the specific values, the bandwidth figures, the attenuation caps, the reach distances, move between cycles, so verify the OM and OS grade specs, the loss values, and the supported reach against the adopted standard editions, the transceiver specifications, and the project documents before citing them. The standards give the framework. The optic datasheet and the project specification control the link.

Units, terms, and acronyms

Fiber type carries vocabulary from the glass, the optic, and the connector, and the same idea reads differently across a cut sheet, a drawing, and a spec. The terms below travel across the whole fiber-selection scope.

Single-mode (SM)

Fiber with a ~9 micron core carrying one mode, low loss, long reach, laser source, graded OS1/OS2

Multimode (MM)

Fiber with a ~50 micron core carrying many modes, shorter reach from modal dispersion, VCSEL source, graded OM1 to OM5

Modal dispersion

Pulse spreading from many modes arriving at different times, the limit on multimode distance

OM3 / OM4 / OM5

Laser-optimized multimode grades; OM3 and OM4 aqua at 850 nm, OM5 lime green wideband for SWDM

OS1 / OS2

Single-mode grades; OS2 is low-water-peak glass for long and outside-plant runs

VCSEL

Vertical-cavity surface-emitting laser, the low-cost 850 nm source for multimode SR optics

SR / LR / FR / DR

Transceiver reach families; SR short-reach multimode, LR/FR/DR single-mode

MPO / MTP

Multi-fiber connector carrying 8 to 24 fibers in one ferrule; MTP is a brand of MPO, the basis of parallel optics

DAC

Direct attach copper, a short twinax assembly with built-on transceivers for intra-rack links

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is data center structured cabling?
• What is the difference between Tier 1 and Tier 2 fiber testing?
• What is a rack unit (U)?
• What kind of batteries do data centers use?
• How do you size a data center generator?