SAA

• Shared 2 MB/sec, all shards shared 2MB/s
• Enhanced: Per shard 2MB/sec

• Max retiention 365 days
• Can replay data
• same partition goes to same shard

• Serverless
• Near real time

• Each record 1MB max
• S3 backup
• Destionations: S3, Redshift, ElasticSearch

• Stream can be replayed
• Stream is realtime
• Firehose can do transformation using Lambda

• For SQS, use GroupId in order to group message together

• Same patition key goes to same shard

Kinesis: max number of consumer = shard number

SISO: 100 GroupID = 100 consumers

• STOMP, AMQP, MQTT protocols
• Migrating application to cloud
• Has Failover

• Using EFS as storage

SAP

• Kinesis data is automatically replicated synchronously to 3 AZ

• Order at Shard level
• You can merge shard / reshard

• Producers
  • KPL for C++, Java
  • Kinesis Agent: can write to Data Stream and Kinesis Data Firehose
• Consumers
  • KCL: Checkpoint, coordinated reads write to DynamoDB

• You can put Source Records into S3
• You can put transformation failures job into S3
• You can put Delivery Failures job into S3

• FIrehose is near real time because it has buffer
• Max size for buffer is 32 MB, once reach that will flush buffer
• Max time is 1 min
• Buffer sizse can be automaticlly increase to increase throughput
• If you need real-time flush, use Data streams with Lambda