NVIDIA Spectrum SN2000: three missing 25/100GbE switches, and the fabric/console/power facts SN2410 and SN2700 were missing #1595

Merged
jaylouisw merged 1 commit from catalogue/1116-0325 into main 2026-11-16 06:30:50 +00:00
Owner

What

One vendor family: NVIDIA (Mellanox) Spectrum SN2000, the 25G/100G open-Ethernet line. Three new specs and two corrections, all from NVIDIA/Mellanox's own product briefs and the SN2000 1U hardware manual.

Added (3)

Model Ports Capacity Power
SN2010 18x SFP28 1/10/25G + 4x QSFP28 40/100G 1.7 Tb/s, 1.26 Bpps 57 W ATIS, half-width
SN2100 16x QSFP28 40/56/100G 3.2 Tb/s, 2.38 Bpps 94.3 W ATIS, half-width
SN2201 48x 1G RJ45 + 4x QSFP28 40/100G 448 Gb/s, 667 Mpps 98 W ATIS, 4 fans

Each carries fabric capacity, mgmt port, console port, dimensions/weight, airflow/PSU and the breakout rules, every field with its source URL.

Corrected (2)

  • SN2410 had ports and a source URL but no fabric capacity, no management port, no console, no power/weight provenance. Now 4,000,000 Mbps (4 Tb/s, 2.97 Bpps), 165 W ATIS, 43.8 x 438 x 436 mm, 8.52 kg, RJ45 MGT + USB console.
  • MSN2700-CS2RC had no source URL at all, was catalogued as a plain switch while the brief lists BGP4/OSPFv2/VRRP and 64 VRFs, and carried 32 QSFP28 with no mgmt port, console or capacity. Now l3-switch, 6,400,000 Mbps, 150 W ATIS, 11.1 kg, MGT + USB console.

Discrepancies, stated rather than smoothed

  1. SN2201 capacity. NVIDIA publishes 448 Gb/s. That is 48 x 1G + 4 x 100G counted once; the full-duplex sum the family uses elsewhere (3.2 Tb/s = 16 x 100G x 2) would be 896 Gb/s. The catalogue carries the published 448 Gb/s and the arithmetic is written into fabric.note - no reconciled number was invented.
  2. SN2010 1GbE. The brief headline says "18 ports of 10/25GbE"; the family hardware manual marks the SFP28 cages 1/10/25GbE with 1GbE needing a 1G SFP. The catalogue follows the manual, so 1G is a speed capability of the cage, not a separate port.
  3. Console connector. The three data-centre models expose only a USB console; the SN2201 has a front RJ45 console. That distinction is in notes, not a connector field, because audit-port-types correctly refuses a console port that resolves to a data cage - declaring connector: usb there made the audit report 5 contradictions on the first attempt, which is the rule working as intended.
  4. Vendor split, not touched. The family sits across devices/mellanox/ (SN2010, SN2100, SN2201, SN2410) and devices/nvidia/ (SN2700). Consolidating a vendor key is its own decision and would break references, so it is flagged here rather than done.

Validation

  • node scripts/import-devices.js --apply: Added: 3 duplicates: 1057 rejected: 7 refused duplicate spellings: 2 (the 7 rejects and 2 refusals are pre-existing entries in old research/incoming/ files, untouched).
  • node scripts/syntax-check.js: 333 files, all OK.
  • node scripts/audit-port-types.js: No port contradicts its declared type.
  • node scripts/audit-devices.js: No hard defects.
  • node scripts/regen-registry-doc.js: L3 switch 175 -> 179; docs/DEVICE_REGISTRY.md regenerated; README headline 2,109 -> 2,112 models (252 vendors unchanged).
  • Targeted suites, all green: import/fuzz-importer/importer-roundtrip (69 assertions), fabric + fabric-agreement (58), catalogue-hygiene (25), port-count (25), wired-port-has-media (36), missing-devices (66), port-roundtrip (45), doc-catalog-counts (24).

Sources

Unverifiable / open

  • The SN2201 RJ45 console is stated in the family manual's management-interfaces table and in the SN2201 user manual, but the SN2201's own product brief 404s at the pb-sn2201.pdf path, so no second vendor document corroborates it.
  • The SN2410 brief's physical table gives width 438 mm and depth 436 mm (a 2 mm difference); carried as printed.
  • 56GbE is advertised for QSFP28 cages but the engine has no 56G line rate, so it is not declared - same treatment as the rest of the catalogue.
## What One vendor family: **NVIDIA (Mellanox) Spectrum SN2000**, the 25G/100G open-Ethernet line. Three new specs and two corrections, all from NVIDIA/Mellanox's own product briefs and the SN2000 1U hardware manual. ### Added (3) | Model | Ports | Capacity | Power | |---|---|---|---| | SN2010 | 18x SFP28 1/10/25G + 4x QSFP28 40/100G | 1.7 Tb/s, 1.26 Bpps | 57 W ATIS, half-width | | SN2100 | 16x QSFP28 40/56/100G | 3.2 Tb/s, 2.38 Bpps | 94.3 W ATIS, half-width | | SN2201 | 48x 1G RJ45 + 4x QSFP28 40/100G | 448 Gb/s, 667 Mpps | 98 W ATIS, 4 fans | Each carries fabric capacity, mgmt port, console port, dimensions/weight, airflow/PSU and the breakout rules, every field with its source URL. ### Corrected (2) - **SN2410** had ports and a source URL but **no fabric capacity, no management port, no console, no power/weight provenance**. Now 4,000,000 Mbps (4 Tb/s, 2.97 Bpps), 165 W ATIS, 43.8 x 438 x 436 mm, 8.52 kg, RJ45 MGT + USB console. - **MSN2700-CS2RC** had **no source URL at all**, was catalogued as a plain `switch` while the brief lists BGP4/OSPFv2/VRRP and 64 VRFs, and carried 32 QSFP28 with no mgmt port, console or capacity. Now `l3-switch`, 6,400,000 Mbps, 150 W ATIS, 11.1 kg, MGT + USB console. ## Discrepancies, stated rather than smoothed 1. **SN2201 capacity.** NVIDIA publishes **448 Gb/s**. That is 48 x 1G + 4 x 100G counted once; the full-duplex sum the family uses elsewhere (3.2 Tb/s = 16 x 100G x 2) would be **896 Gb/s**. The catalogue carries the published 448 Gb/s and the arithmetic is written into `fabric.note` - no reconciled number was invented. 2. **SN2010 1GbE.** The brief headline says "18 ports of 10/25GbE"; the family hardware manual marks the SFP28 cages `1/10/25GbE` with 1GbE needing a 1G SFP. The catalogue follows the manual, so 1G is a speed capability of the cage, not a separate port. 3. **Console connector.** The three data-centre models expose only a **USB** console; the SN2201 has a **front RJ45 console**. That distinction is in `notes`, not a `connector` field, because `audit-port-types` correctly refuses a console port that resolves to a data cage - declaring `connector: usb` there made the audit report 5 contradictions on the first attempt, which is the rule working as intended. 4. **Vendor split, not touched.** The family sits across `devices/mellanox/` (SN2010, SN2100, SN2201, SN2410) and `devices/nvidia/` (SN2700). Consolidating a vendor key is its own decision and would break references, so it is flagged here rather than done. ## Validation - `node scripts/import-devices.js --apply`: `Added: 3 duplicates: 1057 rejected: 7 refused duplicate spellings: 2` (the 7 rejects and 2 refusals are pre-existing entries in old `research/incoming/` files, untouched). - `node scripts/syntax-check.js`: **333 files, all OK**. - `node scripts/audit-port-types.js`: **No port contradicts its declared type.** - `node scripts/audit-devices.js`: **No hard defects.** - `node scripts/regen-registry-doc.js`: L3 switch 175 -> 179; `docs/DEVICE_REGISTRY.md` regenerated; README headline 2,109 -> 2,112 models (252 vendors unchanged). - Targeted suites, all green: import/fuzz-importer/importer-roundtrip (69 assertions), fabric + fabric-agreement (58), catalogue-hygiene (25), port-count (25), wired-port-has-media (36), missing-devices (66), port-roundtrip (45), doc-catalog-counts (24). ## Sources - https://network.nvidia.com/files/doc-2020/pb-sn2010.pdf - https://network.nvidia.com/files/doc-2020/pb-sn2100.pdf - https://network.nvidia.com/files/doc-2020/pb-sn2410.pdf - https://network.nvidia.com/files/doc-2020/pb-sn2700.pdf - https://networking-docs.nvidia.com/sn2201hw/specifications - https://docs.nvidia.com/networking/display/nvidia-spectrum-sn2000-1u-switch-systems-hardware-user-manual.pdf - https://www.nvidia.com/en-us/networking/ethernet-switching/ ## Unverifiable / open - The SN2201 **RJ45 console** is stated in the family manual's management-interfaces table and in the SN2201 user manual, but the SN2201's own product brief 404s at the `pb-sn2201.pdf` path, so no second vendor document corroborates it. - The SN2410 brief's physical table gives width 438 mm and depth 436 mm (a 2 mm difference); carried as printed. - 56GbE is advertised for QSFP28 cages but the engine has no 56G line rate, so it is not declared - same treatment as the rest of the catalogue.
NVIDIA Spectrum SN2000: the 25/100GbE family's three missing switches, and the fabric/console facts the SN2410 and SN2700 entries were carrying without
Some checks failed
PR rebase / Rebase onto main and arm auto-merge (pull_request) Failing after 10m54s
CI / Classify the change (pull_request) Failing after 11m3s
CI / Test (Node 18) (pull_request) Has been cancelled
CI / Test (Node 20) (pull_request) Has been cancelled
CI / Test (Node 22) (pull_request) Has been cancelled
CI / Server smoke test (pull_request) Has been cancelled
CI / Docker build (pull_request) Has been cancelled
5edc0ed807
Three new specs from NVIDIA's own product briefs and the SN2000 1U hardware
manual: SN2010 (18x SFP28 + 4x QSFP28, 1.7 Tb/s, 57 W half-width), SN2100
(16x QSFP28, 3.2 Tb/s, 94.3 W) and SN2201 (48x 1G RJ45 + 4x QSFP28, 448 Gb/s,
98 W - the only box in the family with an RJ45 console port).

Two existing entries corrected against the same sources:
- SN2410 had ports and a source URL but no fabric capacity, no management or
  console port and no power/weight provenance. Now 4,000,000 Mbps (4 Tb/s,
  2.97 Bpps), 165 W ATIS, 43.8 x 438 x 436 mm, 8.52 kg, mgmt + USB console.
- MSN2700-CS2RC had no source URL, was a plain "switch" although the brief
  lists BGP4/OSPFv2/VRRP and 64 VRFs, and carried 32 QSFP28 with no mgmt port,
  console or capacity. Now l3-switch, 6,400,000 Mbps, 150 W, 11.1 kg.

Discrepancies are stated in the specs rather than smoothed: NVIDIA publishes
448 Gb/s for the SN2201, which is not the full-duplex sum of its cages
(896 Gb/s), and the SN2010 brief's "18 ports of 10/25GbE" turns into "1G needs
a 1G SFP" in the family manual, so 1G stays a speed capability, not a port.

Vendor key is left as-is: the family is split across devices/mellanox/ (four
models) and devices/nvidia/ (SN2700), which is upstream of this run.
jaylouisw force-pushed catalogue/1116-0325 from 5edc0ed807
Some checks failed
PR rebase / Rebase onto main and arm auto-merge (pull_request) Failing after 10m54s
CI / Classify the change (pull_request) Failing after 11m3s
CI / Test (Node 18) (pull_request) Has been cancelled
CI / Test (Node 20) (pull_request) Has been cancelled
CI / Test (Node 22) (pull_request) Has been cancelled
CI / Server smoke test (pull_request) Has been cancelled
CI / Docker build (pull_request) Has been cancelled
to cdfa1acdf0
Some checks failed
CI / Classify the change (pull_request) Waiting to run
PR rebase / Rebase onto main and arm auto-merge (pull_request) Waiting to run
CI / Test (Node 18) (pull_request) Has been cancelled
CI / Test (Node 20) (pull_request) Has been cancelled
CI / Test (Node 22) (pull_request) Has been cancelled
CI / Server smoke test (pull_request) Has been cancelled
CI / Docker build (pull_request) Has been cancelled
2026-11-16 05:08:38 +00:00
Compare
jaylouisw force-pushed catalogue/1116-0325 from cdfa1acdf0
Some checks failed
CI / Classify the change (pull_request) Waiting to run
PR rebase / Rebase onto main and arm auto-merge (pull_request) Waiting to run
CI / Test (Node 18) (pull_request) Has been cancelled
CI / Test (Node 20) (pull_request) Has been cancelled
CI / Test (Node 22) (pull_request) Has been cancelled
CI / Server smoke test (pull_request) Has been cancelled
CI / Docker build (pull_request) Has been cancelled
to fc7fed87d1
Some checks failed
PR rebase / Rebase onto main and arm auto-merge (pull_request) Waiting to run
CI / Classify the change (pull_request) Successful in 34s
CI / Test (Node 18) (pull_request) Has started running
CI / Test (Node 20) (pull_request) Has started running
CI / Test (Node 22) (pull_request) Failing after 1m41s
CI / Server smoke test (pull_request) Has started running
CI / Docker build (pull_request) Has started running
2026-11-16 05:10:55 +00:00
Compare
jaylouisw force-pushed catalogue/1116-0325 from fc7fed87d1
Some checks failed
PR rebase / Rebase onto main and arm auto-merge (pull_request) Waiting to run
CI / Classify the change (pull_request) Successful in 34s
CI / Test (Node 18) (pull_request) Has started running
CI / Test (Node 20) (pull_request) Has started running
CI / Test (Node 22) (pull_request) Failing after 1m41s
CI / Server smoke test (pull_request) Has started running
CI / Docker build (pull_request) Has started running
to 51516dbb37
All checks were successful
CI / Classify the change (pull_request) Successful in 17s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 20s
CI / Server smoke test (pull_request) Has been skipped
CI / Docker build (pull_request) Has been skipped
CI / Test (Node 20) (pull_request) Successful in 1m43s
CI / Test (Node 18) (pull_request) Successful in 1m51s
CI / Test (Node 22) (pull_request) Successful in 2m3s
2026-11-16 06:26:43 +00:00
Compare
Author
Owner

This branch conflicts with main, so it was not rebased:

README.md devices/mellanox/sn2010.json devices/mellanox/sn2100.json devices/mellanox/sn2201.json devices/mellanox/sn2410.json devices/nvidia/msn2700-cs2rc.json docs/DEVICE_REGISTRY.md research/incoming/devices-nvidia-spectrum-sn2000.json 

A conflict is a decision about behaviour rather than a mechanical fix, so this is left for a human. Rebasing by hand and force-pushing will let the automation pick it up again.

This branch conflicts with `main`, so it was not rebased: ``` README.md devices/mellanox/sn2010.json devices/mellanox/sn2100.json devices/mellanox/sn2201.json devices/mellanox/sn2410.json devices/nvidia/msn2700-cs2rc.json docs/DEVICE_REGISTRY.md research/incoming/devices-nvidia-spectrum-sn2000.json ``` A conflict is a decision about behaviour rather than a mechanical fix, so this is left for a human. Rebasing by hand and force-pushing will let the automation pick it up again.
jaylouisw force-pushed catalogue/1116-0325 from 51516dbb37
All checks were successful
CI / Classify the change (pull_request) Successful in 17s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 20s
CI / Server smoke test (pull_request) Has been skipped
CI / Docker build (pull_request) Has been skipped
CI / Test (Node 20) (pull_request) Successful in 1m43s
CI / Test (Node 18) (pull_request) Successful in 1m51s
CI / Test (Node 22) (pull_request) Successful in 2m3s
to b43820b308
Some checks failed
CI / Classify the change (pull_request) Successful in 13s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 16s
CI / Server smoke test (pull_request) Has been skipped
CI / Docker build (pull_request) Has been skipped
CI / Test (Node 20) (pull_request) Failing after 1m30s
CI / Test (Node 18) (pull_request) Failing after 1m38s
CI / Test (Node 22) (pull_request) Failing after 1m36s
2026-11-16 06:29:32 +00:00
Compare
jaylouisw scheduled this pull request to auto merge when all checks succeed 2026-11-16 06:30:43 +00:00
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jaylouisw/packetforge!1595
No description provided.