Skip to content

feat(llq): add snip library for 5G xHaul Low-Latency QoS JVD#119

Merged
KB-x merged 3 commits into
mainfrom
update/llq-snips
Jun 2, 2026
Merged

feat(llq): add snip library for 5G xHaul Low-Latency QoS JVD#119
KB-x merged 3 commits into
mainfrom
update/llq-snips

Conversation

@KB-x

@KB-x KB-x commented Jun 1, 2026

Copy link
Copy Markdown
Collaborator

What's New

Snip library for the Low-Latency Queueing (5G xHaul) validated design — 66 files covering the complete QoS + transport + service stack.

  • 36 EVO snips (ACX7509/7100/7024, PTX10001) + 27 Junos snips (MX204/480/304)
  • Headline: priority low-latency hardware scheduler (ACX-only) for sub-10µs eCPRI queuing
  • 7 categories: CoS, firewall, transport, services, policy, interfaces, OAM
  • Full cross-reference headers (Seen-on, Pair-with, Highlights)
  • Topology registry + variables reference + README

Why

LLQ is the most complex CoS design in the JVD repo (8-class model, 5 distinct CoS binding patterns, MAC-based eCPRI MFC filters). Extracting it as a snip library makes the design reusable and auditable.

Details

  • snips/evo/cos/: 17 files — FC, 4 classifiers, 4 rewrites, 3 scheduler variants, scheduler-map, 5 binding patterns
  • snips/junos/cos/: 12 files — same minus ACX-only bindings (IRB, L2 fronthaul, static FC)
  • snips/*/firewall/: MFC IPv4/IPv6 DSCP-to-FC filters + eCPRI MAC-based CCC filter
  • snips/*/transport/: IS-IS (interface + global/SR), MPLS, BGP labeled-unicast, hash-key, load-balance
  • snips/*/services/: EVPN-VPWS-MH, EVPN-ELAN, BGP-VPLS virtual-switch, L3VPN-IRB
  • snips/*/policy/: PPLB, ALLOW_LOOPBACK, SR_NONZERO_LOOPBACKS, next-hop-self
  • snips/evo/interfaces/: LAG + ESI for EVPN multi-homing
  • snips/evo/oam/: CFM maintenance-domain (802.1ag continuity-check)
  • _topology_registry.json: 12 devices mapped to 64 snips by role
  • _variables.md: 40+ template variables documented with examples

Testing

  • All 64 .conf snip bodies verified against source configs in conf/
  • Scheduler split confirmed: ACX uses low-latency, PTX/MX use strict-high
  • Device classification validated via stanza presence scanning

@KB-x KB-x force-pushed the update/llq-snips branch from fcbd679 to c570302 Compare June 2, 2026 00:56
@KB-x KB-x self-assigned this Jun 2, 2026
@KB-x KB-x merged commit d82cfa2 into main Jun 2, 2026
4 checks passed
@KB-x KB-x deleted the update/llq-snips branch June 2, 2026 00:58
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant