gistfile1.txt

80 pages in pre-production length

Preface

Part I. Actor Application Types
1. Domain-Driven
    Domain-driven messages are “facts”
2. Work Distribution
    Routers and routees
      Random
      Round Robin
      Smallest Mailbox
      Broadcast
      ScatterGatherFirstCompletedOf
      Consistent Hash Routing
    BalancingDispatcher will be deprecated soon!
    Work distribution messages are “commands”

Part II. Patterns of Actor Usage
3. The Extra Pattern 
    The problem
    Avoiding ask
    Capturing context
    Sending yourself a timeout message
    Why not use a Promise?
    How to test this logic
4. The Cameo Pattern
    The Cameo Companion Object Factory Method
    How to test this logic

Part III. Best Practices
5. Actors Should Do Only One Thing
    Single Responsibility Principle
    Create specific supervisors
    Keep the error kernel simple
    Failure zones
6. Avoid Blocking
    Futures delegation example
      Java futures are blocking!
    Pre-defining parallel futures
    Parallel futures with the zip method
    Sequential futures
    Callbacks versus monadic handling
    Futures and ExecutionContext
    Push, don’t pull
    When you must block
    Managed Blocking in Scala
7. Avoid Premature Optimization
    Start simple
      Deterministic
      Declarative
      Immutable
      Functional
    Layer in complexity via indeterminism
    Optimize with mutability
      Add mutability with Compare and Swap (CAS)
      Compose CAS with Software Transactional Memory
      Add locks as a last resort
    Prepare for race conditions
      Avoid coding in a way where timing matters
      Tell, don’t ask
8. Be Explicit
    Name Actors and ActorSystem instances Create specialized messages
    Create specialized exceptions
    Beware the “thundering herd”
      Dampen message overload
      Use circuit breakers for failure overload
9. Don’t Expose Actors
    Avoid using “this”
      The Companion Object Factory Method
    Never use direct references
    Don’t close over variables
    Use immutable messages with immutable data
10. Help Yourself in Production
    Make debugging easier
    Add Metrics
    Externalize business logic
    Use semantically-useful logging
    Aggregate your logs with a tool like Flume
    Use unique IDs for messages
11. Tune Akka Applications with the Typesafe Console
    Fixing Starvation
      How do you fix it?
      Does PinnedDispatcher help? 
      Failure zones
      Routers
    Sizing dispatchers
    The Parallelism-Factor setting 
    Actor Mailbox Size
    Throughput Setting
    Edge Cases