Graph-Enabled Network
Automation Solutions with
Neo4j
Dr. Jesús Barrasa - @BarrasaDV
Director Telco&Media - Neo4j
London - 29th October 2019
The #1 Platform for Connected Data
Not this type of graphs!
The Property Graph Model
Optical MUX Filter Card
Are you still NOT using Graphs?
Here are a few reasons why you should
”Graph analysis is possibly the single most effective
competitive differentiator for organisations pursuing
data-driven operations and decisions“
1. Adoption
Adoption Highlights
Top MVPD
2 of the Top 5 US multichannel
video service providers have
chosen Neo4j
Leading OSS Vendors
Half of the leaders in the 2018
Gartner MQ for OSS embed
Neo4j in their products
2 of the 3 world’s largest CSP
use Neo4j in mission critical
solutions
Largest Telcos
IoT OSS/BSS
Governance &
Metadata Mgmnt
IAM &
Fraud Analysis
Common Graph Use Cases in Telco
• Assurance
• Orchestration
• CRM/Support
• Planning &
Optimisation
• Regulatory
compliance
• Data Lineage
• Consent Mgmnt
• Identity &
Access Mgmnt
• Smart Home
• Smart Cities
• Industrial
/Manufacturing
Customer
Experience Mgmt
• Personalisation/Rec
ommendation
• Customer Journey
Analysis
• Customer 360 - KYC
2. Key Business Benefits
• Flexibility: Dynamic Schema
• Developer Productivity: Agility, Declarative query language
Faster Time-To-Market (Solution/Production)
• New data paradigm
• Graph representation of data
Innovation Enabler
• Timeliness
• Accuracy
Better decisions drive better business outcomes
• 10x less CPU with index-free adjacency
• 10x less hardware than other platforms
Hardware efficiency
Neo4j: Graph Platform Benefits
why?
Need to capture complexity
Flexibility
Performance
Need to capture complexity
Flexibility
Performance
Stockholm, March 2019
Need to capture complexity
Flexibility
Performance
Need to capture complexity
Flexibility
Performance
Audience Experiment:
Dependency modelling
Look at this data…
Element Depends On
A B
A C
A D
C H
D J
E F
E G
F J
G L
H I
J N
J M
L M
Element Depends On
A B
A C
A D
C H
D J
E F
E G
F J
G L
H I
J N
J M
L M
Time challenge #1: Does A depend on F ?
?
Look at this data again…
Time challenge #2: Does E depend on M ?
?
M
E
MATCH (a:Element { id: “A”})
MATCH p = (a)-[:DEPENDS_ON*]->(n { id: “N”})
RETURN p
SELECT
d1.ElementId, d2.ElementId, d3.ElementId
FROM dpndncs AS d1
INNER JOIN dpndncs AS d2
ON d1.dependsOnElemId = d2.ElemId
INNER JOIN dpndncs AS d3
ON d2.dependsOnElemId = d3.ElemId
… <arbitrary number of joins>…
WHERE d1.ElementId = “A”
AND d3.ElementId = “N”
Element Depends On
A B
A C
A D
C H
D J
E F
E G
F J
G L
H I
J N
J M
L M
Does X depend on Y ?
Things get more complicated: Route diversity,
SPOF detection…
some example projects
Assurance
IA/RCA
Orchestration
PC
Graph-Based Network Dual Model
Route Oriented Model
Least cost path from A to B
Diverse routes
+
Dependency Oriented Model
No shared underlying resources
IA/RCA
Path Computation
Diverse Routing
?
🏦 :DEPENDS_ON
:DEPENDS_ON
:DEPENDS_ON
IF/AX2431
💥
Customer
Event Correlation
Event Prioritisation
Service Assurance
IA/RCA
MATCH (fe:Link { linkId: $id})<-[:DEPENDS_ON*]-(s:Service)
RETURN max(s.priority) AS severity
{ alarmType: “LOS”,
notifyingEntity: “IF/AX/0/3”, …}
Service Assurance
IA/RCA
GET http://localhost:11001/engine/ia/s1_361_sdh
{"severitySummaryCode": 4,
"severitySummaryDesc": "MAJOR",
"detail": [{"elem":"2217/ol",
"impact" : 1,
Graph Size: ~50M nodes (avg depth: 6)Graph Size: ~1K nodes (avg depth: 5)
Simulation: 128 clients, synchronous requests with1ms wait between requests
AT SCALE
50000x increase in size of dataset -> 1.14x impact in query performance
Graph Native Matters!!!
(Deep) Impact/Root Cause Analysis
Smart Homes
100 requests/day x 11 devices/zone x 1.3 Mill Zones = 1.43 Bill w/req to Neo4j/day
The future…
Customer Journey Analysis
Come meet us at Stand 30 in the Exhibition Area to
see Neo4j in action
Join us at 12:50 for the Data Driven Technologies
roundtable on “How to bring Graph Analytics into
your projects: From Network Operations to Customer
Journey Analysis”
Thank you!

Graph-enabled network automation solutions with Neo4j

  • 1.
    Graph-Enabled Network Automation Solutionswith Neo4j Dr. Jesús Barrasa - @BarrasaDV Director Telco&Media - Neo4j London - 29th October 2019
  • 2.
    The #1 Platformfor Connected Data
  • 4.
    Not this typeof graphs!
  • 5.
    The Property GraphModel Optical MUX Filter Card
  • 7.
    Are you stillNOT using Graphs? Here are a few reasons why you should
  • 8.
    ”Graph analysis ispossibly the single most effective competitive differentiator for organisations pursuing data-driven operations and decisions“
  • 9.
  • 10.
    Adoption Highlights Top MVPD 2of the Top 5 US multichannel video service providers have chosen Neo4j Leading OSS Vendors Half of the leaders in the 2018 Gartner MQ for OSS embed Neo4j in their products 2 of the 3 world’s largest CSP use Neo4j in mission critical solutions Largest Telcos
  • 11.
    IoT OSS/BSS Governance & MetadataMgmnt IAM & Fraud Analysis Common Graph Use Cases in Telco • Assurance • Orchestration • CRM/Support • Planning & Optimisation • Regulatory compliance • Data Lineage • Consent Mgmnt • Identity & Access Mgmnt • Smart Home • Smart Cities • Industrial /Manufacturing Customer Experience Mgmt • Personalisation/Rec ommendation • Customer Journey Analysis • Customer 360 - KYC
  • 12.
  • 13.
    • Flexibility: DynamicSchema • Developer Productivity: Agility, Declarative query language Faster Time-To-Market (Solution/Production) • New data paradigm • Graph representation of data Innovation Enabler • Timeliness • Accuracy Better decisions drive better business outcomes • 10x less CPU with index-free adjacency • 10x less hardware than other platforms Hardware efficiency Neo4j: Graph Platform Benefits
  • 14.
  • 15.
    Need to capturecomplexity Flexibility Performance
  • 16.
    Need to capturecomplexity Flexibility Performance
  • 17.
  • 18.
    Need to capturecomplexity Flexibility Performance
  • 19.
    Need to capturecomplexity Flexibility Performance
  • 20.
  • 21.
    Look at thisdata… Element Depends On A B A C A D C H D J E F E G F J G L H I J N J M L M
  • 22.
    Element Depends On AB A C A D C H D J E F E G F J G L H I J N J M L M Time challenge #1: Does A depend on F ? ?
  • 23.
    Look at thisdata again…
  • 24.
    Time challenge #2:Does E depend on M ? ? M E
  • 25.
    MATCH (a:Element {id: “A”}) MATCH p = (a)-[:DEPENDS_ON*]->(n { id: “N”}) RETURN p SELECT d1.ElementId, d2.ElementId, d3.ElementId FROM dpndncs AS d1 INNER JOIN dpndncs AS d2 ON d1.dependsOnElemId = d2.ElemId INNER JOIN dpndncs AS d3 ON d2.dependsOnElemId = d3.ElemId … <arbitrary number of joins>… WHERE d1.ElementId = “A” AND d3.ElementId = “N” Element Depends On A B A C A D C H D J E F E G F J G L H I J N J M L M Does X depend on Y ?
  • 26.
    Things get morecomplicated: Route diversity, SPOF detection…
  • 27.
  • 28.
  • 29.
    Graph-Based Network DualModel Route Oriented Model Least cost path from A to B Diverse routes + Dependency Oriented Model No shared underlying resources IA/RCA
  • 30.
  • 31.
  • 32.
    MATCH (fe:Link {linkId: $id})<-[:DEPENDS_ON*]-(s:Service) RETURN max(s.priority) AS severity { alarmType: “LOS”, notifyingEntity: “IF/AX/0/3”, …} Service Assurance IA/RCA GET http://localhost:11001/engine/ia/s1_361_sdh {"severitySummaryCode": 4, "severitySummaryDesc": "MAJOR", "detail": [{"elem":"2217/ol", "impact" : 1,
  • 33.
    Graph Size: ~50Mnodes (avg depth: 6)Graph Size: ~1K nodes (avg depth: 5) Simulation: 128 clients, synchronous requests with1ms wait between requests AT SCALE 50000x increase in size of dataset -> 1.14x impact in query performance Graph Native Matters!!! (Deep) Impact/Root Cause Analysis
  • 34.
  • 37.
    100 requests/day x11 devices/zone x 1.3 Mill Zones = 1.43 Bill w/req to Neo4j/day
  • 38.
  • 39.
  • 41.
    Come meet usat Stand 30 in the Exhibition Area to see Neo4j in action Join us at 12:50 for the Data Driven Technologies roundtable on “How to bring Graph Analytics into your projects: From Network Operations to Customer Journey Analysis” Thank you!