Kinesis is an alternative to Kafka. It is great for processing, analyzing and transforming real-time streaming data.
Kinesis Streams
- Good for general real-time streaming data
- Processed in shards/partitions
- Size limit: 1MB
- Retention limits: 1 to 365 days
Kinesis Analytics
- Good for both real-time streaming and batched data
- Use case: ETL, metric generation, responsive analytic
-
ML algorithms:
- RANDOM_CUT_FOREST for online anomaly detection
- HOTSPOTS for finding dense areas in the feature space
Kinesis Firehose
- Data aggregation and exportation
- Batch writes to target destinations (S3(compression supported), Redshift, ElasticSearch, Splunk)
- Near real-time output because of the batching
- Automatic scaling
Kinesis Video Streams
- Streaming data from cameras, DeepLens etc.
- One device, one stream
- Retention: 1 hour to 10 years