Express Scripts: Driving Digital Transformation from Mainframe to Microservices
1 like2,377 views
The document presents a session on digital transformation strategies at Express Scripts, focusing on transitioning from a mainframe to a microservices architecture. Key topics include the implementation of a data sync platform, Kafka libraries, and achieving transactional consistency while improving application reliability and performance. The presentation emphasizes the importance of building reusable tools and frameworks to enhance project delivery and decrease overall costs.
Express Scripts: Driving Digital Transformation from Mainframe to Microservices
- 1. Express Scripts: Driving Digital Transformation from Mainframe to Microservices Ankur Kaneria, Principal Architect, Kevin Petrie, Senior Director, Alan Hsia, Group Manager, Product Marketing, Confluent
- 2. 2 Speakers Kevin Petrie Senior Director, Attunity Kevin is a senior director and technology evangelist at Attunity (www.attunity.com), a Burlington, MA-based provider of Big Data integration and automation software. He has 20 years of experience in technology, with a focus on Big Data, storage and the Cloud. Kevin has held strategy, services and marketing leadership roles with EMC and Symantec. Prior to receiving his MBA from Haas School of Business at UC Berkeley, Kevin was a financial writer with TheStreet.com, analyzing the data networking and telecommunications industries. Kevin has a bachelor of arts degree from Bowdoin College. He is an outdoor fitness nut, bookworm, husband and father of three boys. Ankur Kaneria Principal Architect, Express Scripts Ankur is a principal architect at Express Scripts, based out of Austin, Texas. His areas of expertise include telecom, e-commerce and health care verticals. Ankur’s 15+ years of experience spans software architecture and design, high availability real-time software systems, high performance system design, cloud computing and virtualization. His main focus is on driving technology transformation across the company, helping engineering teams migrate from legacy tech stack to a modernized environment. Being a software engineer at heart, Ankur likes to take practical approach while building software system design and architecture with usability, reliability and fault tolerance in mind. Alan Hsia Group Manager, Product Marketing, Confluent Alan is a group manager of product marketing at Confluent where he leads all product marketing efforts for our solutions areas. Prior to joining Confluent, he led product at VMware for their hybrid cloud platform, VMware Cloud Foundation. Alan also spent 6 years at Parakletos Ventures, a boutique venture capital fund focusing on tech investments across mobile, semiconductors, networking in the US and Asia.
- 3. 3 Session Overview ● This session will be one hour in length ● The last 10-15 minutes will be reserved for Q&A ● Submit questions by entering them into the GoToWebinar panel ● The slides and recording will be available after the event
- 4. 4 Confluent Platform The Event Streaming Platform Built by the Original Creators of Apache Kafka® Apache Kafka® Development & Stream Processing Confluent Platform Operations and Security Support&Services Mission-critical Reliability Complete Event Streaming Platform Freedom of Choice
- 5. 5 Complete Portfolio of Products and Services Built around Kafka Professional Services Enterprise Support Kafka Training Confluent Cloud Confluent Platform *Software, services and support across the entire adoption lifecycle
- 6. Driving Digital Transformation from Mainframe to Microservices March 12, 2019 Ankur Kaneria - Express Scripts.
- 7. 7 © 2018 Express Scripts. All Rights Reserved. Agenda ✓ Who we are ✓ Architecture & Design ✓ Existing Architecture ✓ Entity Architecture ✓ Data Sync Platform ✓ Current Pipeline Footprint ✓ Achieving Transactional Consistency ✓ Kafka Libraries and connectivity options – and our approach ✓ Supporting Ecosystem & Other Considerations ✓ Data Pipeline Health ✓ Initial Load of Data ✓ What’s Next
- 8. 8 © 2018 Express Scripts. All Rights Reserved. What does Express Scripts do? ✓ Part of Cigna Corporation ✓ Just like your medical plan covers visits to your doctor, your Express Scripts prescription plan covers the medicine your doctor prescribes ✓ We’re one of the largest pharmacy benefits managers in the United States and one of the country’s largest pharmacies ✓ Serve more than 100 million people ✓ Claims Volume ✓ Daily ~6 million ✓ Annual ~2 Billion
- 9. 9 © 2018 Express Scripts. All Rights Reserved. Executive Summary • In the past, we rebuilt the same functionality for every big project • Entity Platform strategy focuses on building once and reusing • This increases up-front costs, but decreases capital spend over time & increases speed of delivery WHY WE’RE MOVING TO ENTITIES Pharmacy As A Platform Rebates 20/20 Support Clients Manage Products Orders Member Prescription Orders Claim Benefit Contract Claim Benefit Contract Member Benefit Pharmacy As A Platform Rebates 20/20 Support Clients Manage Products Orders Prescription Benefit Member Claim Contract Data Sync
- 10. 10 © 2018 Express Scripts. All Rights Reserved. Walking the Interaction and Business Tier Stack User Interaction Channels (Applications) Business Platforms User Interface Logic Human Workflow UI Frameworks Business platforms consists of entities with REST-based APIs that completely control access to a data domain. They are first class citizens.RequiredRequired Web Sites Mobile Apps …and more... Entity Platform Entity Platform Entity Platform Data Data Data Person Membership Referral Order Shipment Drug Price Quote and more... Interaction APIs Aggregations UX-Specific Views Optional User-specific order status Cross-entity items, not aligned with a UI that don’t map to an entity. Optional Business Process & Logic
- 11. 11 © 2018 Express Scripts. All Rights Reserved. Claims History Lookup Member Site (Applications) Business Platforms Member Website Used by our members to view claims history, orders etc. Require d Required User Accounts Person Entity Claims Entity Data Data Data 1. Member login-in. Authenticated through User Accounts 2. A resource id that uniquely identifies the member is retrieved 3. Using the unique member key, appropriate member specific claims history data is accessed and displayed to the member 1 2 3
- 12. 12 © 2018 Express Scripts. All Rights Reserved. Target Sync Platform Data Sync Platform Architecture & Design Source Service Platform Source DB Source Service Target Service Platform Retry Topic Retry Stream Processor Target DB Target Service CDC Topic Update Topic CDC Stream Processor Update Stream Processor on error ✓ Change Data Capture ✓ Kafka ✓ Target Sync Platform Error Topic
- 13. 13 © 2018 Express Scripts. All Rights Reserved. Pipeline Footprint at a Glance Attunity Kafka • Replicate • Version 6.0 • 4 VMs • Each 8 Core, 32 GB RAM, Linux • Active/Standby • Brokers • 8 node cluster • 8 x 24Cores, 256GB RAM, 22TB HD • 4TB SSD for OS • Enterprise Manager (AEM) • Version 6.1 • 1 VM • 8 Core, 32 GB RAM, Windows • ZooKeeper • 5 nodes ensemble • 5 x 36Cores, 256GB RAM, 7TB HD • 4TB SSD for OS and (500 GB for ZK data) • 1 Schema Registry, 1 Control Center
- 14. 14 © 2018 Express Scripts. All Rights Reserved. Why Confluent Platform ✓ Enterprise Support ✓ Access to Kafka subject matter experts ✓ Ability to reach out for support with operational/production issues ✓ Help & guidance ✓ Operations & Monitoring ✓ Confluent control center ✓ Features & Capabilities ✓ KSQL ✓ Confluent Hub/Connector Ecosystem ✓ Replicator for active/active topology ✓ Better documentation
- 15. 15 © 2018 Express Scripts. All Rights Reserved. Topic.T1 Topic.T2 Topic.T3 CDC Stream Join Table1 Table2 Table3 Tran1 Tran2 Tran3 Tran4 Tran5 Tran6 Transactional Consistency and CDC Merged Stream
- 16. 16 © 2018 Express Scripts. All Rights Reserved. Transactional Consistency and CDC – Contd… ✓ Transactions can impact multiple tables ✓ Attunity bulk writes all events making up a transaction ✓ Natural order of events guaranteed only at partition level ✓ Kafka streams applications need to aggregate events ✓ Aggregated events need to be applied as a single unit
- 17. 17 © 2018 Express Scripts. All Rights Reserved. Grouping Related Events ✓ Aggregate events in the same window ✓ Event emitted for each aggregation ✓ How to know all events arrived? https://kafka.apache.org/11/documentation/streams/d eveloper-guide/dsl-api.html#session-windows
- 18. 18 © 2018 Express Scripts. All Rights Reserved. Possible Solution ✓ Related events should be together ✓ https://www.confluent.io/blog/put-several-event-types-kafka-topic/ ✓ Handle deserializing different event types ✓ Confluent Schema Registry support for multiple Avro types on same topic? vs.
- 19. 19 © 2018 Express Scripts. All Rights Reserved. Partitioning Strategies ✓ Transaction ✓ Potential out of order events partition 0 partition 1 partition 0 partition 1 ✓ Entity key
- 20. 20 © 2018 Express Scripts. All Rights Reserved. CDC Stream Processor Stream Time (ms) 0 50 100 150 200 250 300 350 400 450 500 550 600 650 Update Topic Partition ✓ Leverages Kafka Processor API and state store ✓ Punctuation handles inactivity gap State Store CDC Topic Partition
- 21. 21 © 2018 Express Scripts. All Rights Reserved. Update Processor ✓ Receives aggregated events from CDC processor ✓ Merges new state with current ✓ Create ✓ Update ✓ Integrity checks ✓ Error handling
- 22. 22 © 2018 Express Scripts. All Rights Reserved. Pipeline Data Integrity ✓ Retry/Error processor ✓ One way flow of data 1 2 3 Source DB Retry Topic CDC Topic Entity Update Topic Target Sync Platform Retry Stream Processor CDC Stream Processor Update Stream Processor Data Services REST API
- 23. 23 © 2018 Express Scripts. All Rights Reserved. Which Kafka Library to use? ✓ Several options ! ✓ Producer & consumer APIs ✓ Streams API ✓ Connectors ecosystem (Confluent hub) & connector API ✓ Spring’s wrapper (spring-kafka) ✓ No silver bullet ! ✓ Every use-case is different and unique ✓ Do your own due diligence
- 24. 24 © 2018 Express Scripts. All Rights Reserved. General Guidelines ✓ Producer & consumer APIs ✓ Applications directly producing data streams (e.g. clickstreams, logs, IoT…) ✓ Read a stream and perform real-time actions on it (e.g. send email, generate alarm, send notification…) ✓ Streams API ✓ Applications wanting to consume from Kafka and produce back into Kafka, also called stream processing ✓ Very powerful & offers advanced features for complex processing logic ✓ Kstream, Ktable, stateless/stateful processing, state store, aggregations, punctuation…
- 25. 25 © 2018 Express Scripts. All Rights Reserved. General Guidelines (continued) ✓ Connectors ecosystem (Confluent hub) & connector API ✓ Interactions with data stores ( e.g. S3, PostgreSQL, MongoDB, HDFS…) ✓ Spring’s wrapper (spring-kafka) ✓ Java based applications ✓ Spring native out of the box capabilities ✓ Low key approach to get you rolling in short time ✓ Hybrid Approach ✓ Streams API for advanced/complex logic ✓ Spring-kafka for applications with generic streams processing and handling
- 26. 26 © 2018 Express Scripts. All Rights Reserved. Reconciliation / Data Integrity ✓ High level existence check ✓ Data Compare Source vs. Target Missing on Target Missing on Source +
- 27. 27 © 2018 Express Scripts. All Rights Reserved. Performance & Health of the Pipeline ✓ Latency monitoring ✓ Throughput monitoring
- 28. 28 © 2018 Express Scripts. All Rights Reserved. Prerequisites ✓ Initial load is needed for stateful data ✓ Initial load strategies ✓ CDC provided functionality ✓ Sqoop/Spark, Kafka connectors ✓ Micro-batched extraction
- 29. 29 © 2018 Express Scripts. All Rights Reserved. What’s Next ✓ Data Lake for analytics ✓ Incorporate Elasticsearch for improved searching ✓ Microservices to act on data events in real time ✓ Stream Registry
- 30. MARCH 2019KEVIN PETRIE CDC STREAMING WEBINAR WITH EXPRESS SCRIPTS AND CONFLUENT
- 31. 31© 2017 Attunity 31© 2017 Attunity Leader in change data capture • Most comprehensive platform supportability matrix • Modern architecture & UI • End-to-end offering captures changed data and delivers analytics ready data structures Leader in cloud integration and cloud migration solutions • Relied on by the top cloud computing providers • Migrated and integrated more than 100,000 databases and mainframes to AWS, Azure and Google • Complete cloud data delivery capabilities, including the last mile Leader in ease of use and automation • Automated target table creation and data instantiation • Automated source and target mappings • Quickly create and deploy analytics ready structures The right architecture for efficiently capturing large volumes of changed data from heterogeneous source systems and delivering it in real-time to Streaming & Cloud Platforms, Data Warehouses and Data Lakes ATTUNITY – Modernize and Automate Data Integration
- 32. 32© 2017 Attunity 32© 2017 Attunity CDC STREAMING WITH ATTUNITY CONVERT DATABASE TO LIVE FEED WITH NO CODING • Reduce labor, impact and latency of publishing DB transactions to record stream • Automatically configure publication of database transactions to Kafka, etc. • Define Kafka broker as target, select topics, execute and monitor tasks enterprise-wide • Rename schemas/tables, add/drop columns, filter records to be published • Fully support and automate all topic, partition configuration options in Kafka
- 33. 33© 2017 Attunity 33© 2017 Attunity CDC STREAMING WITH ATTUNITY Generate real-time events; no scripting Auto-provision one DB producer for many consumers Schema evolution; easy schema registry integration DATABASE AS A STREAM ONE TO MANY STREAM METADATA Auto-map data types for simpler consumption DATA FORMAT CONVERSION Fully support all topic, partition options CONFIGURATION FLEXIBILITY AMAZON S3 ADLS
- 34. 34© 2017 Attunity 34© 2017 Attunity • Ensure transactions are committed to Kafka in correct order • Notify consumer applications when transactions are ready for processing • Restart/recover tasks from beginning of transaction TRANSACTIONAL CONSISTENCY WITH ATTUNITY
- 36. 36 Q & A
- 37. 37 Resources and Next Steps See more customer stories https://confluent.io/customers Try Confluent Platform https://cnfl.io/download Sign up for Confluent Community https://cnfl.io/slack Download Replicate Whitepaper https://discover.attunity.com/why -attunity-replicate-is-the-better-alt ernative-lp1724.html
- 38. 38 Thank you for joining us!