AI Workload Routing
aiworkloadrouting.com
§1 — Definition
Network Compute Fabric
An architectural term for relating network infrastructure, distributed compute, interconnect, optical transport, workload movement, and operational coordination.
§2 — Relationships
Closest related capability
Difference
Relationship
Closest comparison and adjacent concepts.
Network Compute Fabric is related to AI Workload Routing, but each addresses a distinct architectural question.
What separates them
Network Compute Fabric addresses a distinct architectural boundary from AI Workload Routing.
How they work together
The concepts can be evaluated together when their respective infrastructure roles are relevant.
See also
§3 — Standards and AuthorityWork in progress ↗ Work in progress ↗
Where the terminology comes from.
Network Compute Fabric is a package-level architectural synthesis rather than a term governed by a single standards body. Its constituent capabilities carry different levels of institutional support. "Artificial Intelligence and Communication" appears directly in Recommendation ITU-R M.2160-0. Data Center Interconnect is directly represented in published OIF interoperability work. Multimodal Data Fusion appears directly in active IEEE standards development. Other namespaces either map to established underlying architectural functions or remain LJP-defined compound terms for which no institutional body currently defines the complete concept. Each capability page identifies its own authority basis, standards-development status, and evidentiary boundary.
Benchmarking Terminology for AI Network Fabrics
Internet Engineering Task Force Datatracker · draft-calabria-bmwg-ai-fabric-terminology-02
Active individual Internet-Draft describing terminology for Ethernet-based AI network fabrics, including RDMA, RoCEv2, UET, congestion, collectives, and topology.
Work in progress with no formal IETF standing or endorsement; supports underlying technical convergence only.
Benchmarking Methodology for AI Training Network Fabrics
Internet Engineering Task Force Datatracker · draft-calabria-bmwg-ai-fabric-training-bench-02
Active individual Internet-Draft describing benchmarking methods and KPIs for Ethernet-based AI training fabrics and their workload behavior.
Work in progress with no formal IETF standing or endorsement; it does not define the LJP compound term.
§4 — Evaluation
Apply the distinction to the decision at hand.
Helps an enterprise compare the relevant architectural boundary and dependencies without treating this namespace as a deployment recommendation.
§5 — LJP Foundation
How this capability fits the package.
Network Compute Fabric distinguishes its architectural role from adjacent package capabilities and from implementation-specific choices.
A peer capability within the Intelligent Compute Fabric architecture story.
§6 — Machine-Readable Resources
Public identity and discovery resources.
- Namespace manifesthttps://networkcomputefabric.com/namespace.json
- Public ontologyhttps://networkcomputefabric.com/ontology.jsonld
- LLM summaryhttps://networkcomputefabric.com/llms.txt
- Robots policyhttps://networkcomputefabric.com/robots.txt
- Sitemaphttps://networkcomputefabric.com/sitemap.xml
§7 — Credibility Boundary
What this reference does not claim.
This namespace does not prescribe an architecture, vendor, standard interpretation, configuration, routing policy, optical design, workload policy, service commitment, or implementation method.
This namespace is an LJP editorial construct. It claims no standards ownership or external endorsement and selects no vendor or implementation; protected methods and transaction materials are not disclosed.