Messaging

Use Cases for a Message Bus

Notifications
- pipeline steps
  - single implementation (for most steps)
    - usage pattern: push-pull pipeline
    - exchange type: direct
    - queues: single queue for each pipeline step, with multiple consumers per queue
  - multiple implementations (e.g., user target resolvers)
    - usage pattern: push-pull pipeline
    - exchange type: direct
    - queues: different queue for each "type" of user-target resolver, with multiple consumers per queue
- delivery channels
  - usage pattern: pub/sub or push-pull pipeline, depending on channel implementations
  - exchange type: fanout ( ? )
Cross-process event handling
- usage pattern: pub/sub
- exchange type: direct and/or topic
Course publish event (CMS → LMS)
- usage pattern: request-reply RPC or pub/sub
- exchange type: default or direct

message delivery guarantees: at-most-once, at-least-once, exactly-once
message durability: persistent or transient
message routing: automatic filtering
competing or non-competing consumers
usage patterns
- request-reply RPC (pattern: task distribution, topology: service bus)
- publish-subscribe (pattern: data distribution, topology: data distribution tree)
- push-pull pipeline (pattern: parallel task distribution and collection, topology: parallelized pipeline)
exchanges (per AMQP 0-9-1)
- types:
  - default (automatic routing to named queue)
  - direct (route based on message routing key) - usually for unicast in task distribution patterns
  - fanout (broadcast to multiple queues) - usually for broadcast routing in pub/sub patterns
  - topic (route based on match between routing key and queue binding pattern) - usually for multicast in pub/sub patterns
  - headers (route based on any/all match of header values) - "direct on steroids"
- attributes: name, durability (survive broker restart), auto-delete (deleted when all queues have finished using it)
queues
- attributes: name, durable, exclusive, auto-delete
consumers
- push or pull APIs
- message acks, rejects, bulk rejects
- message prefetching
messages
- serialization format: JSON, thrift, protocol buffers, messagepack
- durability: persistent or ephemeral

Bazaar - time-based activities with intelligent prompting and synchronous chat for learners
Discussions
- forum activity - including notifying moderators and forum participants
- social awareness
Course Dates
Course Announcements
User Progress in course
Analytics Use cases:
- Instructors reaching out to learners
- Automatically notifying/supporting struggling learners
- Verified upgrade reminders (+ Jasper Ho Deen Abdul-Hathi (Deactivated))