Drop-In, Not Rip-Out: The Migration Path [3/6]
Paper 2 showed that access-deployable coherent modules now exist. This paper addresses the operational question: how does an operator migrate from a 10G SFF installed base to 100G coherent — without a forklift, a hard cutover, or a procurement commitment it may later regret.
The Migration Question
The access network upgrade debate is too often framed as binary: keep running an overloaded 10G infrastructure or commit to a forklift replacement. Both framings miss the point. A QSFP28-form-factor 100G ZR module drops into the same slots currently housing 10G SFF optics — no chassis change, no hard cutover. But form factor compatibility is only part of the picture. Two questions determine whether the migration is genuinely smooth: how does traffic move from 10G to 100G without service risk, and what happens to the transceiver’s management interface when host equipment is eventually modernized? The second carries financial consequences most procurement models miss.
|
7 yrs Standard transceiver depreciation lifecycle |
3 Migration phases — no hard cutover required |
1 Part number — SFF ↔ CMIS via firmware (single-SKU architectures |
Form Factor Compatibility — One Underappreciated Architectural Reality
One underappreciated architectural reality: 100G ZR QSFP28 modules are electrically and mechanically compatible with the QSFP28 slots in the vast majority of access aggregation and backhaul equipment. An operator upgrading from 10G SFF DWDM does not need a new chassis, new line cards, or a maintenance window — the upgrade is a module substitution, not a system replacement. Compatibility is not unconditional: operators should verify chassis firmware support for coherent pluggables and validate thermal management at the target power envelope before wide deployment.
This changes the migration risk calculus. A forklift forces a hard cutover — all services at a node moved simultaneously, no coexistence possible. Module substitution allows one link at a time, one service at a time, with the legacy 10G SFF module in the adjacent slot as live rollback until the new link is confirmed stable. No all-or-nothing moment. One honest friction point: NMS platforms may have limited coherent pluggable diagnostic visibility until CMIS support is enabled, affecting MTTR in Phase 1 until the management stack catches up.
The optical layer is also compatible: 100G ZR operates over existing ITU-T DWDM infrastructure — same fiber, same grid, no external amplifiers or dispersion compensators needed for metro and regional access distances.
| “Coherent QSFP28 offers the industry the opportunity to seamlessly upgrade 10G DWDM networks with 100G transceivers without the need to re-engineer the network and allows reuse of existing 100G router and switch ports.”
— Vladimir Kozlov, CEO, LightCounting LLC, September 2024 |
Incremental Migration Eliminates the Hard Cutover — Capex Follows Traffic
One principle is worth stating before the mechanics: in access networks, the dominant cost driver is rarely hardware price — it is misalignment between upgrade cycles across layers. A migration architecture that keeps those cycles independent outperforms one that forces them to align.
A forklift requires operators to spend before need is proven at each node. The incremental module-based path allows capex to follow actual congestion — link by link, site by site — rather than precede it on a fixed program schedule.
A practical phased migration architecture has three stages, each with defined entry criteria and full rollback capability:
Phase 1 — Pilot validation. Select the three to five highest-congestion links. Deploy 100G ZR QSFP28 modules in SFF-8636 mode on existing hosts. Validate reach, power, and basic I2C monitoring. Measure actual TCO. De-risk the decision before wider commitment.
Phase 2 — Congestion-priority rollout. Expand to all saturated links, migrating traffic service by service. Legacy 10G SFF modules remain in adjacent slots as live rollback until each link is confirmed stable. Capex follows congestion, not a fixed timetable.
Phase 3 — Systematic base refresh. As host equipment is modernized to CMIS-capable platforms on its own cycle, a software command switches each deployed module from SFF-8636 to CMIS — unlocking full telemetry and zero-touch provisioning. No new transceivers required. The hardware already in the field becomes more capable as the surrounding infrastructure evolves.
Figure 1: Phased migration architecture — 10G SFF to 100G ZR QSFP28. No hard cutover, no forklift, full rollback at every stage. Source: Arycs Technologies.
Lifecycle Coupling Between Layers — The Hidden Cost Driver Most Operators Miss
The most expensive part of a network upgrade is rarely the hardware itself. It is the moment you are forced to replace hardware before its lifecycle ends because a dependency in an adjacent layer has changed. That pattern — lifecycle coupling — is what makes the SFF-8636 to CMIS transition financially consequential in ways that most procurement models miss.
Standard transceiver depreciation runs approximately seven years. The critical question: what happens when the host equipment is replaced with a CMIS-only platform partway through that lifecycle — at year 3, 4, or 5?
The two-part-number problem. Some vendors offer two separate SKUs: one for SFF-8636 hosts, one for CMIS hosts. Under a host upgrade mid-lifecycle, the SFF-8636 module becomes incompatible with the new platform. The operator faces a write-off on unused lifecycle value plus a full replacement purchase. The graphic below models this across three upgrade timing scenarios, normalized to 100% = one module at Year 0 list price. The replacement module cost reflects 5% compound annual price erosion — a standard industry assumption as yields improves and competition matures. Even with that erosion reducing the replacement cost to 86–77%, the two-SKU total runs 177–186% vs. 100% for the single-SKU path.
Figure 2: Total transceiver cost — two-SKU vs. single-SKU architecture across three host upgrade scenarios. Normalized to 100% = one module at Year 0 list price. Replacement module cost reflects 5% compound annual price erosion. Source: Arycs Technologies analysis
An architecture that allows the management interface to evolve independently of the hardware eliminates this coupling entirely. A firmware-switchable single SKU — where SFF-8636 and CMIS are selectable via software — carries no compatibility risk across its full seven-year lifecycle. When the host is upgraded, a software command switches the module’s management mode. The transceiver depreciates on its original schedule. No write-off. No new purchase. Not all 100G ZR implementations offer this; operators should treat it as a procurement criterion.
The spare parts benefit follows directly: one part number, one sparing pool, no risk of deploying the wrong variant in the field. For operations teams managing hundreds of sites, the simplification is immediate and ongoing.
| “CMIS advertising enables hosts to write generic software for managing CMIS-compliant modules, significantly reducing integration time and accelerating the deployment of new capabilities in end-user networks.”
— OIF, CMIS: Path to Plug and Play, 2024 |
The Migration Path Is Tractable — If the Architecture Is Right
The migration from 10G SFF to 100G ZR coherent does not require a forklift, a hard cutover, or capital committed before capacity is needed. Each link upgrades on its own timeline, with full rollback and no service disruption. One practical friction point: operations teams accustomed to 10G direct detect often find mixed management environments disorienting during the transition — Phase 1 pilots should budget for that learning curve alongside the technical validation.
The management interface question is where procurement decisions become lifecycle decisions. A two-SKU strategy creates depreciation risk invisible at purchase and real at host refresh. An architecture that decouples the transceiver’s lifecycle from the host platform’s upgrade schedule eliminates that risk entirely. Simpler procurement, simpler sparing, and the peace of mind that the hardware investment survives whatever modernization timeline the network requires. This reflects a broader pattern: in access networks, architectural coupling between layers — not component cost — drives long-term TCO. Paper 4 builds the complete financial case.
#NetworkMigration #QSFP28 #PluggableOptics #FutureProofNetworks #BuiltToScale
About Arycs Technologies
Arycs delivers power-efficient, coherent-class optical connectivity based on silicon photonics, coherent DSP, and advanced optical architectures. Our solutions provide industry-leading bandwidth per watt, deterministic performance, and flexible network evolution for AI, cloud, telecom, and edge infrastructure. Designed for real-world deployment, Arycs Technologies enables networks to scale with growing AI demand without disruptive redesign or hardware replacement.
arycs-tech.com | LinkedIn: Arycs Technologies
Drop-In, Not Rip-Out: The Migration Path [3/6]
