Shaping the next BNP CIB infrastructure with Elastic

BNP Paribas Corporate and Institutional Banking (CIB), the global investment banking subsidiary of the BNP Group is working on the unification of all infrastructure assets to give the best and simplest experience to their users. The Elastic Stack was chosen to be the critical platform to serve data to BNP CIB technical and business teams at scale. In this presentation Alex Afflerbach and Emmanuel Salzard will present the architecture that was designed to serve a multi-tenant logging platform, go through the challenges that came up in their project such as resource segregation or dealing with Petabytes scale topology.

Alex Afflerbach | Head of IT Operations Analytics Platform | BNP CIB
Emmanuel Salzard | Global Domain head of E.M Infrastructures and H.P.C. | BNP CIB

Transcript

1. Emmanuel Salzard, Global Domain Head of Infrastructure V2 Alex Afflerbach, Head of IT Operations Analytics Platform, IV2 Shaping the next BNP Paribas CIB Infrastructure with The Elastic Stack BNP Paribas, Corporate and Institutional Banking November 9, 2016
2. 2 BNP Corporate & Investment Banking •  Global provider of financial solutions to corporate and institutional clients •  Approximately 30,000 people in 57 countries across EMEA, Asia-Pacific and the Americas •  Tailored solutions for clients across capital markets, securities services, financing, treasury and financial advisory •  Global leader across multiple disciplines; expertise include structured financing, derivatives and risk management, to name a few
3. Infrastructure V2 (IV2) •  TARDIS Digital Platform §  Customer-oriented o  IT Store, IT Wall, IT Talk, IT Lab §  Unified §  Product driven §  With intelligence everywhere 3
4. 4 Infrastructure V2 (IV2) •  Aramchek – IT Operation Analytics –  Root cause analysis / Bad behavioral analytics –  Forecasting –  Cruise control •  Log indexation as a service –  Provide in-house managed Elastic environment for business specific data (non-IV2)
5. Monitoring IV2 with Kibana 5
6. Why we engaged with Elastic? •  Elastic Stack §  Open source technology §  Large & active community §  Partnering opportunities 6 •  Specific project requirements §  Complete solution - from ingest to visualization §  Multi-tenant security §  Dedicated support
7. Functional Implementation Storage Platform Visualization & Exploration Data Processing Alerting, Monitoring & Anomaly Detection Enterprise Services Referentials Identity Management Authentication Rights Management Alerting Frameworks Other Outputs, APIs Telemetry Data Collection Agents Plugins/Outputs System Logs Application Logs SNMP File Metadata Net flows CPU Memory Storage NIC Any numerical Event, Etc. Etc. Semi-Structured Metrics Application
8. Functional Implementation System Logs Application Logs SNMP File Metadata Net flows Memory Storage NIC Any numerical Event, Etc. Metrics Semi-Structured Application CPU Etc. Front-End REST API VISUALIZATION INGEST
9. Functional Implementation 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Size in Terabytes Billions of Documents Linear (Size in Terabytes) Elas1c Performance Workshop Compression with Deflate 2.X Upgrade Security Project Kick-off
10. Multi-Tenant Ingestion Kibana Dashboard 10
11. 11 Scaling Clusters for Petabytes of Data •  Multi-tenancy cluster topology –  Shared vs. Dedicated •  Managing very high volumes in Elasticsearch –  Annual ingest rate of 1 PB –  10 year archiving requirements •  Sharing resources between tenants/regions –  Tribe Node
12. 12 Lessons Learned •  Enhanced Index Management (segment level) –  data retention, sharding and physical & logical partitioning •  Index Management (at a functional level) –  proper “type” separation by index and indices •  Take Security Seriously •  Multi-Tenant Management –  Cluster scaling and monitoring •  Respect the JVM –  JVM Heap (32 GB) vs. Data managed •  Disaster Recovery Strategy –  Datacenter isolation •  Ingestion Buffer (Kafka)
13. Future projects with Elastic •  Move away from threshold monitoring to predictive analysis and enable anomaly detection with Prelert 13 •  Make the IV2 Digital Platform smarter §  Self-healing §  Building a knowledge base §  Taking advantage of Elasticsearch scoring capabilities
14. www.elastic.co 14
15. Except where otherwise noted, this work is licensed under hNp://creaQvecommons.org/licenses/by-nd/4.0/ CreaQve Commons and the double C in a circle are registered trademarks of CreaQve Commons in the United States and other countries. Third party marks and brands are the property of their respecQve holders.