Skip to content

Latest commit

 

History

History
323 lines (276 loc) · 7.89 KB

study-plan.md

File metadata and controls

323 lines (276 loc) · 7.89 KB
Raw

Draft - Learning Practical Software Design

This is a map of the territory that we'll be eventually trying to cover in the podcast. It will change as we learn more (links to the updated version will be updated and announced)

Introduction

Block 1: Walking over important concepts (talks and drawings)

Strategic Viewpoint: Concepts and decisions of DDD

  • Important factors
  • Good design
    • Practical stories
    • What is good design
    • Why is it imports
  • Factors that affect design
    • Natural Boundaries
      • Languages
      • Teams
      • Organizations
    • Enabling Resources
      • Teams
      • Technologies
      • Knowledge
    • Constraints
      • Budget
      • Time
      • Scope and Quality
      • Political aspects
  • Strategic concepts related to design
    • Domain Model
    • Ubiquitous Language
    • Bounded Context
    • Context Map
    • Project deliverables
  • Processes

Tactical Viewpoint: Concepts and factors inside BC

  • Concepts
    • Long Running Process
    • Behavior
      • Aggregate
      • Aggregate Root
      • Identity
      • Value Object
      • use case
    • Integration
      • anti-corruption
    • Task-based UI
      • inductive vs. deductive
      • eventual consistency
    • "big data" processing
    • Continuous delivery

Blueprint Viewpoint: Consistent patterns and practices

  • TODO based on block 3

Practical Viewpoint: Devil is in the details

  • TODO based on block 4

Block 2: Practical Foundation (with isolated samples)

Module: Messaging

Module: Event Sourcing

Module: Distributed Architecture Elements

  • CQS
  • CQRS perspective
    • Elements: Client - Application Server - Projection Host
    • Contracts: Command - Event - View
    • Relation to Bounded contexts and UL
  • Modeling distributed world
    • Temporal decoupling
    • Eventual consistency
    • Volatility
    • Transactional scopes
  • Team scalability patterns
    • Outsourcing CQRS elements
    • Developing CQRS elements in parallel
    • Developing features in parallel
    • Risk mitigation
  • Performance scalability patterns
    • Application Server - Partitioning and load balancing
    • Client - Load balancing
    • Projection Host - multiplexing and partitioning

Module: Deriving state with Projections, Queries and Indexes

  • Definition
  • Simple implementation
    • Document storage interfaces
    • Projection Code
    • Performance optimizations and tricks
    • Building automated rebuilder
    • Tracking changes in code
  • Eventual consistency
  • Concurrent changes
  • Scaling and replication

Module: Capturing the Domain

  • Domain modeling exercises

Module: Dealing with Time and Uncertainty

  • TODO

Module: Applied Machine Learning

  • Core principles of "learning"
    • deterministic
    • non-deterministic
  • Supervised vs. unsupervised learning
  • Validation and experiments
  • Dealing with data
  • Classical approaches
    • Regressions
    • Classification
  • "Brute-force" approaches
    • Neural Networks
    • Genetic Algorithms
  • Hybrid approaches
    • Evolutionary Neural networks
    • Self-recovering distributed systems
  • Distributing CPU-intensive machine learning tasks

Block 3: Bringing it all together (diving into large sample)

Module: Development Process

  • Source code versioning
  • Unit testing
  • Continuous integration
  • Lowering development friction
  • Rapid iterations
  • Keeping things simple and separate
  • Distributing work and outsourcing

Module: Breaking Down Common System Archetypes

  • Application Server
  • Thin client
    • API
    • Web
    • Native
      • Mobile
      • Desktop
  • Hybrid systems
    • Smart "Fat" client
    • Occasionally connected system

Module: Application Server Blocks

  • Application Service
  • Aggregate
  • View Projection
  • Long Running Process (Task)
  • Event Port
  • Domain Service
  • Bounded Context

Module: Application Server Infrastructure

  • Atomic document storage
    • Atomicity
    • Retrial and conflict resolution
    • Local files
    • Cloud / PaaS
  • Append-only storage
    • Local files
    • Cloud / PaaS
    • SQL
  • Streaming blob storage
    • Local files
    • Cloud / PaaS
  • Message dispatcher
    • Quarantine
    • Deduplication manager
    • Message overflows
    • Debugging and replays
  • Task runner
  • Startup tasks

Module: Big data systems

  • Partitioned storage
  • Distributed processing
  • Message and record streaming
  • Real-time aspects
  • Performance improvements
  • Data Visualization

Module: Practical Machine Learning

  • Basics
    • Supervised vs unsupervised learning
    • Error
    • Data quality and validation
  • Deterministic heuristics
  • Non-deterministic brute-force
    • Neural networks
    • Genetic algorithms
  • Scaling out CPU-intensive learning
  • Practical applications
  • Deployment, maintenance and life hacks

Block 4: Keeping System Alive and Growing (case studies)

Module: Project Case Studies and Lessons Learned

  • Lokad Salescast
  • Lokad Hub
  • Lokad Watchtower
  • Lokad Salescast2
  • Event Day
  • Human Record
  • TODO: ask for more external cases

Module: Deployment Routines

  • Options and costs
    • With downtime
    • With reduced SLAs
    • With partial functionality
    • Without downtime
  • Managing versions
  • Incremental deployments
  • Rolling back on failures

Module: Managing Failures

  • TODO: gather from Lokad experience

Module: Reducing Development Friction and Lifehacks

  • TODO: gather from Lokad experience

Module: Specialised technology overview

  • Persistence
    • Relational
      • MS SQL
      • mySQL
      • Postgres
      • Oracle
    • NoSQL
      • Redis
      • CouchDb
      • MongoDB
      • Azure
      • Amazon
    • NoDB
      • Files / Blobs
      • Memory caching
  • Messaging
    • Broker-based
    • Broker-less
  • Push and pull infrastructures
  • Hardware factors
    • Rust drives
    • SSD
    • RAM
    • Multi-core processing (including low-power ARM scenarios)
    • Dedicated hardware