Streaming at Scale

Distributed Stream Processing with Apache Flink

Introduction: The Real-Time Imperative

1. What You Will Learn
2. Prerequisites
3. How This Book Is Structured
4. Version Notes

Chapter 1: The Era of Streaming Data

1. From Batch to Streaming: Why Latency Matters
2. The Lambda Architecture and Its Pain Points
3. From Lambda to Kappa: Unified Processing Models
4. Core Challenges of Distributed Stream Processing
5. Where Flink Fits in the Streaming Landscape
6. Summary

Chapter 2: Stream Processing Fundamentals

1. Data Streams, Events, and Event Streams
2. Stream vs Batch: Execution Models and Mindset
3. Ordering, Latency, and Throughput in Distributed Systems
4. Partitioning and Shuffling in Stream Processing
5. Fault Tolerance Models: At-Most-Once, At-Least-Once, Exactly-Once
6. The Role of Time in Stream Processing
7. Summary

Chapter 3: Flink Architecture Overview

1. The Flink Runtime: Components and Responsibilities
2. Job Submission: From Application Code to Running Job
3. The Job Graph: Nodes, Edges, and Stream Transformations
4. From Job Graph to Execution Graph
5. Task Slots, Parallelism, and Resource Allocation
6. The Life Cycle of a Flink Job
7. Summary

Chapter 4: Building Your First Flink Application

1. Development Environment and Dependencies
2. Project Structure for a Flink Application
3. Writing a WordCount Streaming Job
4. Running Locally and Inspecting Results
5. Understanding the Generated Job Graph
6. Packaging and Submitting to a Cluster
7. Summary

Chapter 5: The DataStream API

1. Sources and Sinks: The Boundaries of Your Pipeline
2. Core Transformations: Map, FlatMap, Filter, KeyBy
3. Reduction and Aggregation Operators
4. Process Functions and Fine-Grained Control
5. Operator Chaining and Pipelining
6. Side Outputs and Branching Streams
7. Summary

Chapter 6: Time, Timestamps, and Watermarks

1. Processing Time, Event Time, and Ingestion Time
2. Why Event Time Matters for Correct Results
3. Assigning Timestamps to Streams
4. Watermarks: The Heart of Event-Time Processing
5. Watermark Strategies and Generators
6. Handling Out-of-Order Events
7. Summary

Chapter 7: Windowing

1. The Concept of Windows
2. Time Windows: Tumbling, Sliding, and Session
3. Count and Global Windows
4. Triggers: Controlling When Windows Fire
5. Evictors and Window Functions
6. Late Data and Allowed Lateness
7. Summary

Chapter 8: State Management Fundamentals

1. Why State Changes Everything
2. Keyed State vs Operator State
3. State Types: Value, List, Map, Reducing, Aggregating
4. Accessing and Updating State
5. State Backends: Memory, FileSystem, and RocksDB
6. Choosing the Right State Strategy
7. Summary

Chapter 9: Fault Tolerance: Checkpoints and Savepoints

1. The Distributed Snapshot Problem
2. Checkpointing in Flink: The Barrier Flow Algorithm
3. Configuration and Tuning Checkpoints
4. Fault Recovery and State Restoration
5. Savepoints: Managed Checkpoints for Operations
6. Exactly-Once, At-Least-Once, and At-Most-Once Guarantees
7. Summary

Chapter 10: Kafka Integration

1. Why Kafka and Flink Are a Natural Pair
2. The Flink Kafka Consumer
3. The Flink Kafka Producer
4. Offset Management and Checkpointing
5. Partitioning Semantics and Ordering
6. Kafka Integration Patterns and Best Practices
7. Summary

Chapter 11: Connectors and External Systems

1. Connector Architecture and Interfaces
2. Database Connectors: JDBC, Elasticsearch, Redis
3. File System Connectors: HDFS, S3, Local
4. Change Data Capture: Debezium and Flink CDC
5. Building Custom Sources and Sinks
6. Choosing Connectors for Your Use Case
7. Summary

Chapter 12: The Flink Table API and SQL

1. The Table API: Declarative Stream Processing
2. Table Environments and Execution Mode
3. Defining Tables and Schemas
4. SQL for Streaming: Queries and Semantics
5. Bridging DataStream and Table APIs
6. When to Use SQL vs DataStream
7. Summary

Chapter 13: Stream Joins and Enrichment

1. Joining Event Streams: The Challenge
2. Stream-Stream Joins: Interval and Window Joins
3. Stream-Table Joins: Enriching with Reference Data
4. Temporal Table Joins and Schema Evolution
5. Broadcast State for Lookup and Enrichment
6. Handling Asynchrony and Skewed Data
7. Summary

Chapter 14: Advanced Data Processing Patterns

1. Deduplication and Idempotency
2. Anomaly Detection on Streams
3. Sessionization and User Activity Tracking
4. Multi-Dimensional Aggregations
5. Multi-Stage Pipelines and Topology Design
6. Real-Time Feature Computation for ML
7. Summary

Chapter 15: Serialization and Data Formats

1. Flink’s Type System
2. Serialization: Kryo, Java, Generic Types
3. Optimizing Serialization Performance
4. Avro, Protobuf, and Schema Management
5. Schema Evolution in Streaming
6. Binary Formats and Network Efficiency
7. Summary

Chapter 16: Parallelism, Backpressure, and Scaling

1. Parallelism Model and Task Allocation
2. Understanding and Detecting Backpressure
3. Rescaling Jobs Without Losing State
4. Data Skew and Its Impact
5. Slot Sharing and Co-Localization
6. Elastic Scaling Patterns
7. Summary

Chapter 17: Memory Management and Resource Tuning

1. Flink’s Memory Layout
2. Task, Network, and Managed Memory
3. Tuning TaskManager and JobManager Resources
4. JVM Garbage Collection and Flink
5. Monitoring Memory Pressure
6. Memory-Related Failure Modes
7. Summary

Chapter 18: Performance Tuning and Optimization

1. Latency vs Throughput Trade-Offs
2. Optimizing State Access and Size
3. Checkpoint Optimization Strategies
4. Network Shuffle Optimization
5. Operator-Level Optimization Techniques
6. Profiling and Diagnosing Performance Problems
7. Summary

Chapter 19: Deployment and Operations

1. Deployment Modes: Standalone, YARN, Kubernetes
2. Cluster Modes: Session, Application, Per-Job
3. High Availability Configuration
4. Resource Management and Cluster Sizing
5. Configuration Management
6. Production Deployment Checklist
7. Summary

Chapter 20: Observability, Monitoring, and Debugging

1. The Flink Web Dashboard
2. Built-In Metrics and Custom Metrics
3. Integrating with Prometheus and Grafana
4. Logging Strategies for Distributed Jobs
5. Debugging Techniques and Tools
6. Incident Response and Troubleshooting
7. Summary

Chapter 21: Security, Testing, and Reliability

1. Authentication and Authorization
2. Encryption in Transit and at Rest
3. Network Security and Kerberos
4. Testing Flink Applications
5. CI/CD Pipelines for Flink Jobs
6. Production Reliability Patterns
7. Summary

Chapter 22: Production Architecture Patterns and Anti-Patterns

1. End-to-End Real-Time Analytics Architecture
2. Fraud Detection Architecture
3. CDC-Based Event-Driven Architecture
4. Multi-Tenant Flink Platforms
5. Common Anti-Patterns and How to Avoid Them
6. Version Upgrades and Long-Term Maintenance
7. Summary

Conclusion: Principles for Streaming at Scale

References