The concern is documenting main components of kinesis

1. Stream


  • there is a immutable unified log — the only allowed operation is APPEND
    • no insert or removal
    • the amount of data is so high that just a touch could distabilize the whole stream
    • seems like apt for FUNCTIONAL PROGRAMMING
  • each record has a sequence number

2. Shard

  • the unit of data sequencing within the stream
  • each shard has its own log


3. Writing

  • how does it know where to write
  • record has
    • key
    • data
  • stream has a key range associated with it
  • every shard has a key range associated w/ it
  • records with same key will certainly end up in the same shard


4. Storage

  • 24 hours, usually
  • There is a retention defined

5. Reading

  • reading is not removing
  • log can be read by multiple consumers
  • 1 consumer can read from 1 or more shards simultaneously