AWS Case Study

Kinesis Data Streams

Use this page as a service-specific case-study starter for architecture, simulation, and optimization scenarios in pinpole.cloud.

Why this service matters

Kinesis Data Streams is the high-throughput ordered stream buffer. Use it when you need ordered processing per key, multiple independent consumers, or near-real-time analytics.

Power:
- Ordered per partition key (within a shard)
- Multiple consumers: shared throughput or Enhanced Fan-Out (EFO)
- Put-to-get delay typically < 1 second

Hard limits / constraints (from the Kinesis Developer Guide):

Important workflows

  • Put Test Record - Publish a sample record to the stream
  • Scale Shards - Increase/decrease shard count
  • Check Hot Shards - Detect skew from partition keys
  • Reshard Plan - Create a split/merge strategy for throughput changes

Configuration sections

  • Why Kinesis (Power + Limits)
  • Stream Mode, Retention & Core Limits
  • Throughput Planning (Hard Numbers)
  • Partitioning (Shard Skew Defense)
  • Producers (PutRecord/PutRecords, KPL)
  • Consumers (Shared vs EFO, Lambda / KCL)
  • Resharding & Scaling Operations
  • Security (Server-side encryption, KMS)

Key configuration points

PointDefault / ValueCategory
EnabledtrueStream Mode, Retention & Core Limits
Capacity Modeon-demand-standardStream Mode, Retention & Core Limits
Shard Count10Stream Mode, Retention & Core Limits
Retention (hours)24Stream Mode, Retention & Core Limits
Max Record Data Blob (bytes)1048576Stream Mode, Retention & Core Limits
Partition Key Max Length (chars)256Stream Mode, Retention & Core Limits
Partition Key Strategyhigh-cardinalityPartitioning (Shard Skew Defense)
Producer LibrarykplProducers (PutRecord/PutRecords, KPL)
Write APIPutRecordsProducers (PutRecord/PutRecords, KPL)
KPL AggregationtrueProducers (PutRecord/PutRecords, KPL)
Max Payload Size (MiB)1Producers (PutRecord/PutRecords, KPL)
Retry Strategybounded-exponentialProducers (PutRecord/PutRecords, KPL)
Back to pinpole.cloud landing page