Experiences with Modeling Network Topologies at Multiple Levels of Abstraction
summary
A Multi-Abstraction-Layer Topology called MALT represents desired and actual topology across design, deployment, config, operation, measurement, and analysis, cutting N^2 integration to roughly N+M. It uses an entity–relationship model with about 250 entity kinds and ~20 relationship kinds, explicit relationships, and principles of orthogonality. A replicated storage service Mulch (Spanner-backed) holds thousands of shards with millions of entities, immutable versions, a domain query language, and generators that translate high-level intent to L3/L1 designs; the system is production and actively evolving.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:05Network Lifecycle Management
- 2:31Automation
- 4:36Standard Representation
- 5:36Network Design
- 7:07Entity Relationships
- 8:16Example
- 8:45Humans vs Computers
- 9:06Shorthand notation
- 9:22Attributes
- 9:46Software Ecosystem
- 10:15Query Language
- 11:16Mulchup
- 12:46Misconceptions
- 14:11Evolution
- 14:56Entity Relationship Model
- 15:25Data Flow Style
- 16:02Questions
related talks
21mChange Management in Physical Network Lifecycle Automation
20mPhysical Network Design at Scale
28mMagma Abstractions and Interfaces A Systems Approach to Access Networks
16mMatryoshka: Realizing Hyperscale Data Center Network Design for the AI Era
10m10x Network Design Engineer
22mUnderstanding Lifecycle Management Complexity of Datacenter Topologies
10m10x Network Design Engineer
10mLightning Talk: NSM, Network Service Mesh at Scale for Telco Networking