Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

The year of the graph: do you really need a graph database? How do you choose one?

443 views

Published on

Graph databases have been around for more than 15 years, but it was AWS and Microsoft getting in the domain that attracted widespread interest. If they are into this, there must be a reason.

Everyone wants to know more, few can really keep up and provide answers. And as this hitherto niche domain is in the mainstream now, the dynamics are changing dramatically. Besides new entries, existing players keep evolving.

I’ve done the hard work of evaluating solutions, so you don’t have to. An overview of the domain and selection methodology, as presented in Big Data Spain 2018

Published in: Technology

The year of the graph: do you really need a graph database? How do you choose one?

  1. 1. THEYEAR OFTHE GRAPH: DOYOU REALLY NEED A GRAPH DATABASE? HOW DOYOU CHOOSE ONE? The year of the graph: do you really need a graph database?The year of the graph: do you really need a graph database?
  2. 2. ABOUT ME  Working with data since 1992  Graph DBs since 2005  Databases  Modeling  Research  Analysis  Consulting  Entrepreneurship  Journalism
  3. 3. THE PROBLEMWITH RELATIONAL DBS  Not good at relations!  * Unintuitive model  * Hard to query  * Does not scale
  4. 4. GRAPH USE CASES: OPERATIONAL APPS  Image: Neo4j
  5. 5. GRAPH USE CASES: OPERATIONAL APPS  Smart Home - IoT  * LeadingTelco in the Nordics  * 1,5 Million Homes  * Real-time processing
  6. 6. GRAPH USE CASES: ANALYTICS  Image: Stanley Wang
  7. 7. GRAPH USE CASES: ANALYTICS  Drug Discovery  * Leading Pharma  * Data on genes, proteins, etc  * Identification of causal relationships
  8. 8. GRAPH USE CASES: DATA INTEGRATION  Image: Ontotext
  9. 9. GRAPH USE CASES: DATA INTEGRATION  Knowledge Graph for Search  * Leading Retailer in DACH  * 200Million+ MAU, 300K+ search requests  * Improve coverage, response time, bottomline
  10. 10. GRAPH USE CASES: MACHINE LEARNING  Image: Oracle
  11. 11. GRAPH USE CASES: MACHINE LEARNING  Anti-Fraud in real-time  * LeadingTelco in China  * 600 Million Users  * Compliance, trust
  12. 12. KNOWLEDGE GRAPHS EVERYWHERE, GRAPH DATABASES ASTHE FOUNDATION
  13. 13. GRAPH DATABASES ARE BOOMING.. SO HOW DOYOU CHOOSE ONE?  Existing research is:  * Outdated  * Shallow  * Expensive  * Marketing oriented
  14. 14. EVALUATING GRAPH DATABASES: HTTP://YEAROFTHEGRAPH.XYZ  Premises:  * Always up to date  * Holistic evaluation  *Value for money  * Hands-on  Why me:  * Hands-on since 2005  *Top-tier analyst since 2013  * Independent  Free Newsletter!
  15. 15. WHAT GRAPH DATABASES ARE NOT: ANALYTICAL &VIZ FRAMEWORKS,THIN GRAPH LAYERS
  16. 16. REAL GRAPH DATABASES GO ALLTHEWAY  Operational vs. Analytical  * Fully-fledged graph API  * Operations & Analytics  * Future-proof, integrated  Native vs. Non-native  *Designed as a graph database  * Storing data in a native format  * Optimized for graph
  17. 17. GRAPH DATABASETYPES: LPG (LABELED PROPERTY GRAPH)  * Non-standard format, query  * Poor schema support  * Interoperability  * Flexible, generic  * Fast traversals  * Scalability
  18. 18. LPG GRAPH DATABASE USE CASES  Operational  applications  Graph  Traversals  Graph  Algorithms
  19. 19. GRAPH DATABASETYPES: RDF (RESOURCE DESCRIPTION FRAMEWORK)  * W3C: Interoperability, Maturity  * Rich & Flexible Schema  * Semantics, Rules, Reasoning  * Performance  * Scalability  * Complexity
  20. 20. RDF GRAPH DATABASE USE CASES  Data Integration  Knowledge Graph  AI
  21. 21. GRAPH DATABASETYPES: MULTI-MODEL (DOCUMENT, KEY-VALUE)  * Flexibility  *Tooling  * May not be optimized for graph  * May have to move data around  * Lock-in (cloud vendors)
  22. 22. EVALUATION CRITERIA: HOLISTIC, DATA-DRIVEN  Application Development  * Engine & API  * Data Model  * Query Language  DevOps  * Interoperability  * Deployment & Configuration  Advanced Analytics  *Advanced Graph  *Visualization  Vendor Credentials & Support  *Vendor Credentials  * Support & Community
  23. 23. EVALUATION CRITERIA: PERFORMANCE & COST  * Not all vendors participate in benchmarks  * Most benchmarks are not done by 3rd party  * Complex, demanding exercise  * Hard to compare LPG vs RDF  * Not all vendors have public pricing & license info  * Considered, not included as data points  * Informed judgement needs experience, hands-on
  24. 24. TAKEAWAY POINTS  * Graph DBs are here to stay  * Different Graph DB models  LPG  RDF  Multi-model  * Each best suited to different use cases  * Evaluation is hard!
  25. 25. EVALUATING GRAPH DATABASES IS HARD.. BUT SOMEBODY HASTO DO IT Impressive work. I’m not aware of another source that is as comprehensive as this one. JONATHAN LACEFIELD, SENIOR DIRECTOR OF PRODUCT MANAGEMENT, DATASTAX ENTERPRISE SERVER We did not have the time, resources, or expertise to evaluate all options properly. If we did, our choices would have been different. APPLICATIONARCHITECT
  26. 26. EVALUATING GRAPH DATABASES: JUST DOTHE MATH  * 30+ options  * Costs time and money  * Requires expertise  * Lack of proper evaluation ->  Sub-optimal decisions ->  Report cost:  1 Day ofTop-tier consultant  Access to updates, consulting

×