Common Workflows

There is an existing pyNTM training repository on GitHub that extensively covers common workflows with pyNTM.

pyNTM Training Modules repository module 1 covers what network modeling is, the problem it solves for, and the common use cases

Please see the pyNTM Training Modules repository module 2 for info and walk-through exercises for

  • Directions on how to get started using pyNTM

  • Setting up a practice/demo environment

  • Finding Shortest Path(s)

  • Failing/Unfailing Interfaces

  • Finding traffic demands egressing a given interface

  • Finding all ECMP paths for a specific demand

  • Simple visualization exercise

Please see the pyNTM Training Modules repository module 3 for info and walk-through exercises for

  • Adding a new Node

  • Adding a new link

  • Adding traffic to the traffic matrix

  • Changing Interface/Circuit capacity

  • Changing an Interface metric

  • Working with RSVP LSPs

Please see the pyNTM Training Modules repository module 4 for info and walk-through exercises for

  • RSVP LSP model data files

  • RSVP types and behaviors

  • Auto bandwidth

  • Fixed bandwidth

  • LSPs and Demands

  • Getting an LSP path

  • Seeing demands on an LSP

  • Demand path when demand is on LSP

  • Shared Risk Link Groups (SRLGs)

  • Adding an SRLG

  • Failing an SRLG

Please see the pyNTM Training Modules repository module 5 for info and walk-through exercises for

  • How to create a visualization using the WeatherMap

  • WeatherMap visual components overview

Checking Network Health

There are a some results to watch for in your simulations that will indicate a network augment or re-architecture of your existing or planned network may be helpful.

IGP routing is deterministic and much simpler to interpret; one obvious warning sign is over-utilized links.

It gets a bit more difficult with RSVP, especially with auto-bandwidth enabled, to determine if the network is under stress. RSVP auto-bandwidth behavior can be non-deterministic, meaning that there may be multiple different end-states the network will converge to, depending on the order in which the LSPs signal and how long each layer 3 node takes to compute the paths for its LSPs and a host of other factors.

With this being the case, there are a few behavior in the model to watch for when running RSVP that may indicate a network augment or re-architecture may be helpful:

  • Large quantities of LSPs not on the shortest path

  • LSPs reserving less bandwidth than they are carrying

  • Some LSPs not being able to signal due to lack of available setup bandwidth in the path