FlamingButter

This project visualise ISIS L2 topology using Neo4j database and help planning MPLS LSP bandwidth demands.

Installation requirements

Project should work fine on Linux and OSx with the following requirements.

• Neo4j server version 2.3.7 installed locally
  • disable authentication dbms.security.auth_enabled=false in configuration file neo4j-server.properties
• python 2.7
• python modules
  • lxml==3.6.0
  • py2neo==2.0.8
  • requests==2.9.1

On Windows use Vagrant + VirtualBox. From the root of the project type vagrant up When VM is up open in Neo4j browser http://localhost:7474/browser/

Files

isisparser.py - import ISIS topological data in Neo4j graph database it require file with output from command show isis database extensive | display xml

place_demands.py - uses Neo4j API to calculate a path between Source and Destination ISIS Nodes and place bandwidth per path.

get_demands.py - (optional) download and save LSP demands from InfluxDB.

resolve_demands.py - (optional) helper script to convert CSV file with Hostname,Loopback IP,LSP bandwidth to a CSV file with Source,Destination,LSP bandwidth

settings.py - project settings in Python file

spf.py - implementation of REST API query to Neo4j DB.

Neo4j Database schema

What is connected and how?

// What is related, and how
MATCH (a)-[r]->(b)
WHERE labels(a) <> [] AND labels(b) <> []
RETURN DISTINCT head(labels(a)) AS This, type(r) as To, head(labels(b)) AS That
LIMIT 10

This	To	That
Router	LINK	Router
Router	SimNN	Router
Prefix	ISIS	Router

Relationship LINK and ISIS have property called metric. metric represent ISIS cost.

CypherQueries

The following two queries delete all information in neo4j database. delete connected nodes and links - match (n)-[r]-() delete n,r delete not connected nodes - match (n) delete n

list of subnets connected to more that 2 routers (not p2p)

match (a:Prefix)-[r]->(:Router) with a, count(*) as n where n > 2 match (a)-->(z:Router) return a,z

Smimilar as previous query - find all ISIS DIS nodes (LAN segmens)

match (a:Router)-[r]->(z:Router) where NOT (a.name =~ '.+.00$') return a,r,z

Example Dijkstra ECMP query

match (a:Router{name:'at-vie05b-ri2-re0.00'}), (z:Router{name:'HU-MON-MONR-RA2.00'}), 
p = allshortestPaths((a)-[r:LINK*]->(z))  
WITH reduce(cost = 0, rel in rels(p) | cost + rel.metric) as sumcost, p
return sumcost,extract(n IN nodes(p)| n.name)