Delta Lake OSS: Create reliable and performant Data Lake by Quentin Ambard
Denunciar
Compartilhar
•7 gostaram•1,260 visualizações
Delta Lake OSS: Create reliable and performant Data Lake by Quentin Ambard
•7 gostaram•1,260 visualizações
Denunciar
Compartilhar
Baixar para ler offline
Delta Lake is an open source framework living on top of parquet in your data lake to provide Reliability and performances. It has been open-sourced by Databricks this year and is gaining traction to become the defacto delta lake format. We’ll see all the goods Delta Lake can do to your data with ACID transactions, DDL operations, Schema enforcement, batch and stream support etc !
Mais conteúdo relacionado
Mais procurados
![[DSC Europe 22] Lakehouse architecture with Delta Lake and Databricks - Draga...](https://cdn.slidesharecdn.com/ss_thumbnails/draganberic-lakehousearchitecturewithdeltalakeanddatabricks-221130080712-6e817e95-thumbnail.jpg?width=384&fit=bounds)
![[DSC Europe 22] Overview of the Databricks Platform - Petar Zecevic](https://cdn.slidesharecdn.com/ss_thumbnails/petarzecevic-databricksoverview-221130080703-c60d93de-thumbnail.jpg?width=384&fit=bounds)
Mais procurados(20)
![[DSC Europe 22] Lakehouse architecture with Delta Lake and Databricks - Draga...](https://cdn.slidesharecdn.com/ss_thumbnails/draganberic-lakehousearchitecturewithdeltalakeanddatabricks-221130080712-6e817e95-thumbnail.jpg?width=768&fit=bounds)
![[DSC Europe 22] Overview of the Databricks Platform - Petar Zecevic](https://cdn.slidesharecdn.com/ss_thumbnails/petarzecevic-databricksoverview-221130080703-c60d93de-thumbnail.jpg?width=768&fit=bounds)
Similar a Delta Lake OSS: Create reliable and performant Data Lake by Quentin Ambard
Similar a Delta Lake OSS: Create reliable and performant Data Lake by Quentin Ambard(20)
Mais de Paris Data Engineers !
Mais de Paris Data Engineers !(11)
Último
Último(20)
Delta Lake OSS: Create reliable and performant Data Lake by Quentin Ambard
- 2. Databricks Workspace Collaborative Notebooks, production jobs & business insights Managed platform Cloud Native Databricks: Unified Data Analytics Platform ML Runtime For your Big data and Machine Learning Lifecycle ...
- 3. ● A typical Data Lake Architecture ● The Delta Architecture ● Inside Delta Lake ● Demo The Delta Agenda
- 4. Enterprises have been spending millions of dollars getting data into data lakes Data Lake
- 5. The aspiration is to do data science and ML on all that data using Apache Spark! Data Lake Data Science & ML • Recommendation Engines • Risk, Fraud Detection • IoT & Predictive Maintenance • Genomics & DNA Sequencing
- 6. Data Lake But the data is not ready for data science & ML The majority of these projects are failing due to Complex pipeline and unreliable data! Data Science & ML • Recommendation Engines • Risk, Fraud Detection • IoT & Predictive Maintenance • Genomics & DNA Sequencing
- 7. What does a typical data lake project look like?
- 8. Evolution of a Cutting-Edge Data Lake Events ? AI & Reporting Streaming Analytics Data Lake
- 9. Evolution of a Cutting-Edge Data Lake Events AI & Reporting Streaming Analytics Data Lake
- 10. Challenge #1: Historical Queries? Data Lake λ-arch λ-arch Streaming Analytics AI & Reporting Events λ-arch1 1 1
- 11. Challenge #2: Messy Data? Data Lake λ-arch λ-arch Streaming Analytics AI & Reporting Events Validation λ-arch Validation 1 21 1 2
- 12. Reprocessing Challenge #3: Mistakes and Failures? Data Lake λ-arch λ-arch Streaming Analytics AI & Reporting Events Validation λ-arch Validation Reprocessing Partitioned 1 2 3 1 1 3 2
- 13. Challenge #4: Updates? Data Lake λ-arch λ-arch Streaming Analytics AI & Reporting Events Validation λ-arch Validation Reprocessing Updates GDPR... Partitioned UPDATE & MERGE Scheduled to Avoid Modifications 1 2 3 1 1 3 4 4 4 2 Reprocessing
- 14. Challenge #5: Stability at scale? Data Lake λ-arch λ-arch Streaming Analytics AI & Reporting Events Validation λ-arch Validation Reprocessing Updates GDPR... Small filesPartitioned UPDATE & MERGE Scheduled to Avoid Modifications 1 2 3 1 1 3 4 4 4 2 5 5 Reprocessing
- 15. Data reliability challenges with data lakes No atomicity: failed jobs leaves data in corrupt state requiring tedious recovery✗ No quality enforcement: creates inconsistent and low quality data Lack of consistency / isolation: makes it almost impossible to mix delete, appends and reads, batch and streaming
- 16. Let’s try it instead with
- 17. ● Open Format Based on Parquet ● By the creator of Apache Spark ● With Transactions ● Using Spark API’s A New Standard for Building Data Lakes
- 18. Is there a better architecture? Data Lake λ-arch λ-arch Streaming Analytics AI & Reporting Events Validation λ-arch Validation Reprocessing Updates GDPR... Small filesPartitioned UPDATE & MERGE Scheduled to Avoid Modifications 1 2 3 1 1 3 4 4 4 2 5 5 Reprocessing
- 19. Data Lake AI & Reporting Streaming Analytics Business-level Aggregates Filtered, Cleaned Augmented Raw Ingestion The Bronze Silver Gold CSV, JSON, TXT… Kinesis
- 20. Data Lake AI & Reporting Streaming Analytics Business-level Aggregates Filtered, Cleaned Augmented Raw Ingestion The Bronze Silver Gold CSV, JSON, TXT… Kinesis Quality Delta Lake allows you to improve the quality of your data until it is ready for consumption.
- 21. Data Lake AI & Reporting Streaming Analytics Business-level Aggregates Filtered, Cleaned Augmented Raw Ingestion The Bronze Silver Gold CSV, JSON, TXT… Kinesis Raw data with minimal parsing Supports long retention (years)
- 22. Data Lake AI & Reporting Streaming Analytics Business-level Aggregates Filtered, Cleaned Augmented Raw Ingestion The Bronze Silver Gold CSV, JSON, TXT… Kinesis Intermediate data with some cleanup applied. Schema enforcement/evolution, data expectation Queryable for easy debugging!
- 23. Data Lake AI & Reporting Streaming Analytics Business-level Aggregates Filtered, Cleaned Augmented Raw Ingestion The Bronze Silver Gold CSV, JSON, TXT… Kinesis Clean data, ready for consumption. Read with Spark, Presto, Glue* *Coming Soon
- 24. Data Lake AI & Reporting Streaming Analytics Business-level Aggregates Filtered, Cleaned Augmented Raw Ingestion The Bronze Silver Gold CSV, JSON, TXT… Kinesis • Full ACID Transactions • Open Source (Apache License) • Powered by
- 25. Data Lake AI & Reporting Streaming Analytics Business-level Aggregates Filtered, Cleaned Augmented Raw Ingestion The Bronze Silver CSV, JSON, TXT… Kinesis Streams move data through the Delta Lake •Low-latency or manually triggered •Eliminates management of schedules and jobs Gold
- 26. Data Lake AI & Reporting Streaming Analytics Business-level Aggregates Filtered, Cleaned Augmented Raw Ingestion The Bronze Silver CSV, JSON, TXT… Kinesis Delta Lake also supports batch jobs and standard DML while streams run UPDATE DELETE MERGE OVERWRITE • Retention • Corrections • GDPR INSERT Gold
- 27. Data Lake AI & Reporting Streaming Analytics Business-level Aggregates Filtered, Cleaned Augmented Raw Ingestion The Bronze Silver CSV, JSON, TXT… Kinesis Easy to recompute when business logic changes: • Clear tables • Restart streams DELETE DELETE Gold
- 28. How do I use ?
- 29. dataframe .write .format("delta") .save("/data") Get Started with Delta using Spark APIs dataframe .write .format("parquet") .save("/data") Instead of parquet... … simply say delta Add Spark Package pyspark --packages io.delta:delta-core_2.12:0.1.0 bin/spark-shell --packages io.delta:delta-core_2.12:0.1.0 <dependency> <groupId>io.delta</groupId> <artifactId>delta-core_2.12</artifactId> <version>0.1.0</version> </dependency> Maven
- 30. How does work?
- 31. Delta On Disk my_table/ _delta_log/ 00000.json 00001.json date=2019-01-01/ file-1.parquet Transaction Log Table Versions (Optional) Partition Directories Data Files
- 32. Log Structured Storage Changes to the table are stored as ordered, atomic units called commits Add 1.parquet Add 2.parquet Remove 1.parquet Remove 2.parquet Add 3.parquet 000000.json 000001.json …
- 33. Handling Massive Metadata Large tables can have millions of files in them! How do we scale the metadata? Use Spark for scaling! Add 1.parquet Add 2.parquet Remove 1.parquet Remove 2.parquet Add 3.parquet Checkpoint … 0009.json 0010.json checkpoint-1.parquet 0011.json … Transaction log
- 34. Transactional Log Parquet Files Delta Lake ensures data reliability Streaming ● ACID Transactions / full DML ● Data quality ● Unified Batch & Streaming ● Time Travel/Data Snapshots Key Features High Quality & Reliable Data always ready for analytics Batch Updates/Deletes
- 35. Support concurrent operation Notebook/User 1: SELECT * FROM customers WHERE firstname='xxx' Notebook/User 2: INSERT INTO customers (firstname, …) VALUES ('marc', …) Notebook/User 3: DELETE FROM customers WHERE firstname='quentin'
- 36. Support concurrent operation Isolation level: WriteSerializable Delta solves conflict optimistically Concurrent modifications on a table triggers a rollback
- 37. Upsert/Merge: Fine-grained Updates MERGE INTO customers -- Delta table USING updates ON customers.customerId = source.customerId WHEN MATCHED THEN UPDATE SET address = updates.address WHEN NOT MATCHED THEN INSERT (customerId, address) VALUES (updates.customerId, updates.address)
- 38. Ensure Data Quality* Enforce metadata, schema, and quality declaratively. Inserts will fail if data doesn’t respect schema or quality table("warehouse") .location(…) // Location on DBFS .schema(my_schema) // Optional strict schema checking .metastoreName(…) // Registration in Hive Metastore .description(…) // Human readable description for users *Coming Soon .expect("validTimestamp", // Expectations on data quality* "timestamp > 2012-01-01 AND …", "fail / alert / quarantine")
- 39. Unified batch and streaming Concurrent stream/batch with exactly-once processing guarantee Data Lake AI & Reporting Streaming Analytics Join stream with table/stream Bronze Silver CSV, JSON, TXT… Kinesis DELETE DELETE Gold
- 40. SELECT count(*) FROM events TIMESTAMP AS OF timestamp SELECT count(*) FROM events VERSION AS OF version Time Travel spark.read.format(" delta").option("timestampAsOf", timestamp_string).load("/events/") INSERT INTO my_table SELECT * FROM my_table TIMESTAMP AS OF date_sub( current_date(), 1) Reproduce experiments & reports Rollback accidental bad writes
- 41. Demo time !
- 42. Workshop Delta & MLFlow Jeudi 7 Novembre 9h-12h30 https://dbricks.co/workshop-databricks