FabricEval

Traffic engineering evaluation framework built for hyperscale networks.

Get Started

What is FabricEval?

FabricEval is an evaluation framework that enables users to study the behavior and performance of traffic engineering (TE) systems. It models all major components of a TE system, including network topology, traffic demands, TE formulation and algorithm, as well as router configuration. FabricEval is also shipped with production-like network configurations based on Google's Jupiter fabrics.

FabricEval diagram

Flexible and modular

Thanks to Bazel, each functional component of FabricEval is packaged as an individual module. Users have the ultimate flexibility to modify/replace any module and observe the effects on TE.

Efficient and scalable

TE can be efficiently evaluated in networks with thousands of routers and millions of links. FabricEval supports building large-scale hierarchical topologies, and visualized graph interaction in Neo4j.

Production-like configurations

Google has made open source a collection of production-like network configs, ranging from large-scale topologies featuring heterogeneous router hardware to traffic demands following the Gravity model, as well as the same TE algorithm powering the Orion SDN controller.