e55c0ad55f
Adds 15 Sonnet-quality blog articles for fo-blog-v1 fine-tuning: tutorials, comparisons, tech deep-dives covering 400G/800G topics. Also adds seed-blog-training-data.py script for learning_corpus import.
23 lines
7.4 KiB
Markdown
23 lines
7.4 KiB
Markdown
---
|
|
title: "Fiber Plant Audit Before a 100G Upgrade: What to Check and Why"
|
|
type: tutorial
|
|
target_audience: technical
|
|
score: 9/10
|
|
---
|
|
|
|
Network teams that migrate from 10G to 100G and experience a wave of link instability within the first month almost uniformly skipped the fiber plant audit. The instability they experience is not caused by the 100G modules being defective or incompatible — it is caused by fiber infrastructure that was adequate for 10G's generous margin but is inadequate for 100G's tighter power budget, now exposed by the migration. The audit takes one to three days depending on the scale of the deployment. The post-migration firefighting it prevents takes weeks of engineer time and generates the kind of escalation heat that terminates migration projects and ends careers. The audit is not optional.
|
|
|
|
The OTDR testing methodology for a pre-migration audit differs from installation verification testing in important ways. Installation OTDR tests are typically single-direction, single-wavelength measurements done immediately after connector installation when connectors are new and clean. Pre-migration OTDR tests should be bidirectional — measuring loss from both ends and averaging the results to eliminate directional asymmetry from angled polish connectors — and should be performed at the wavelength of the target 100G technology. For 100GBASE-SR4, that is 850 nm. For 100GBASE-LR4, that is 1310 nm. For 100GBASE-CWDM4, it is 1271 to 1331 nm. Using an OTDR at 1310 nm to test a path that will carry 100GBASE-SR4 at 850 nm gives you results that do not map to the actual link budget because multimode fiber attenuation at 850 nm (typically 3.0 to 3.5 dB/km on OM4) is significantly different from attenuation at 1310 nm (approximately 1.0 dB/km). Always test at the operating wavelength.
|
|
|
|
Interpreting OTDR results against 100G power budget specifications requires understanding which events are measurement artifacts and which represent real optical loss. An OTDR event that shows 0.02 dB of reflectance gain — where the fiber appears to gain optical power rather than lose it — is a Fresnel reflection artifact at a connector that has an air gap, not a measurement of real gain. These ghost reflections can appear upstream of real events and create a false picture of the fiber path topology. Every event marker in an OTDR trace above 0.15 dB should be verified as a real connector pair location by cross-referencing against the fiber path documentation. An unmapped 0.25 dB event on a path that was supposed to have only three connector pairs is either a damaged splice or an undocumented connector that will consume budget headroom at 100G.
|
|
|
|
Fiber type compatibility is the most common and most expensive surprise in 100G migrations from legacy infrastructure. OM1 fiber, which was widely deployed in campus and enterprise buildings through the mid-2000s, has a 50-micron or 62.5-micron core and a minimum modal bandwidth of approximately 200 MHz·km at 850 nm for the 62.5-micron variant. The IEEE 802.3ba standard for 100GBASE-SR4 requires OM3 with a minimum modal bandwidth of 2000 MHz·km or OM4 with 4700 MHz·km. OM1 at 850 nm supports a maximum 100GBASE-SR4 distance of approximately 33 meters for 62.5-micron core, which means almost any OM1 run longer than the patch cord connecting a server to a top-of-rack switch will fail at 100G SR4. Teams that deployed OM1 in horizontal cable runs with 20 to 50 meter lengths between equipment rooms and server racks face complete fiber replacement for those segments, regardless of how well-maintained the connectors are.
|
|
|
|
OM2 is slightly better but not by much. The OM2 specification at 850 nm gives a maximum 100GBASE-SR4 reach of approximately 26 to 30 meters, depending on the specific OM2 fiber product. As with OM1, runs longer than that distance are not upgradeable to 100G SR4 without fiber replacement. OM3 supports 100GBASE-SR4 to 70 meters, which covers most intra-building horizontal runs, though it does not leave significant margin for longer runs in large facilities. OM4 is the minimum fiber type that makes 100G SR4 deployable without distance anxiety for runs up to 100 meters, and OM5 extends this further through wideband multimode operation. An infrastructure audit that characterizes all fiber paths by type — OM1, OM2, OM3, OM4, OS2 — and maps them against the path length data is the essential first step before any budget is allocated to 100G module procurement.
|
|
|
|
Connector degradation over time is the second category of audit findings that the migration team needs to quantify before deploying 100G. Connectors installed in the late 2000s and early 2010s, now 12 to 15 years into service life, have accumulated years of dust, mating cycles, and physical wear. Published data on MPO connector insertion loss degradation in operational environments shows that connectors cleaned once per year at annual maintenance see median insertion loss increases of 0.08 dB per mated pair per year. A connector pair that measured 0.3 dB at installation in 2010 may be at 1.2 to 1.5 dB by 2026. At 10GBASE-SR with a channel budget of 7.5 dB on OM3, this degradation is absorbed easily. At 100GBASE-SR4 with a channel budget of 1.9 dB on OM3, a single mated pair at 1.5 dB consumes 79 percent of the entire budget before accounting for any other loss in the path.
|
|
|
|
The audit checklist that prevents post-migration firefighting structures the work into three phases. The pre-test phase gathers all existing fiber plant documentation — installation records, previous OTDR trace files, fiber type certifications, and connector installation dates. These documents are frequently incomplete or absent for infrastructure installed more than eight years ago, in which case the physical test data becomes the sole basis for decisions. The test phase executes bidirectional OTDR traces at operating wavelength for every fiber path that will carry 100G traffic, supplemented by insertion loss measurement with an OLTS test set for paths with events that are marginal or ambiguous in the OTDR data. The analysis phase compares measured insertion loss against the 100G budget specification for the relevant technology type, applies a three-year aging factor to each connector pair measurement, and classifies each path as pass, marginal, or fail.
|
|
|
|
Remediation decisions for marginal and failing paths follow a cost-effectiveness filter. For a path where the only issue is connector contamination — measured insertion loss above 0.5 dB per mated pair on what should be a clean connector — wet-then-dry cleaning plus re-test brings most of those connections into compliance at negligible cost. For paths where insertion loss is elevated due to fiber bends or physical damage to the fiber, remediation requires either re-routing the cable to eliminate the bend or replacing the affected segment. For OM1 paths that are too short-reach for SR4 regardless of connector condition, the only practical option is fiber replacement. A decision rule that routes OM1 paths shorter than 20 meters to "accept as SR4 compatible," paths of 20 to 33 meters to "test with SR4 module before committing," and paths over 33 meters to "replace fiber or use single-mode LR4" correctly classifies most OM1 scenarios without requiring individual engineering judgment on each circuit. The economics of remediation versus replacement versus technology change should be calculated at the path level rather than applied uniformly, because a uniform policy will over-invest in some paths and under-invest in others.
|