Integration
D2 · Resilient
D3 · Performance
~1 phút đọc

Kinesis — Data Streams, Firehose, Analytics & Video Streams

Kinesis family xử lý streaming data real-time. Data Streams cho custom processing, Firehose cho delivery, Analytics cho SQL/Flink trên stream.

kinesis
data-streams
firehose
analytics
streaming
video-streams

Sơ đồ tổng quan

Đang tải sơ đồ…

Tổng quan

Kinesis là platform streaming data của AWS. 4 service: Data Streams (ingest + process), Firehose (delivery), Data Analytics / Managed Flink (SQL/Flink trên stream), Video Streams (video real-time).

Kinesis Data Streams

  • Shard-based: mỗi shard 1 MB/s in, 2 MB/s out, 1000 records/s in.
  • On-demand mode: tự scale shard, không cần capacity planning.
  • Retention: 24h mặc định, tối đa 365 ngày.
  • Enhanced Fan-Out: mỗi consumer có 2 MB/s riêng (push via HTTP/2).
  • KPL (Kinesis Producer Library): batching, aggregation, retry.
  • KCL (Kinesis Client Library): checkpoint, shard balancing, exactly-once processing.

Kinesis Data Firehose

  • Fully managed delivery: buffer (1 MB hoặc 60s) → target.
  • Targets: S3, Redshift (qua S3 COPY), OpenSearch, Splunk, HTTP endpoint, 3rd party.
  • Transform: Lambda function inline, convert JSON → Parquet/ORC.
  • Không lưu data → không replay. Near real-time (không phải real-time).
  • Compression: GZIP, Snappy, Zip cho S3 delivery.

Analytics & Video Streams

  • Managed Service for Apache Flink (trước là Kinesis Data Analytics): chạy Flink job trên stream.
  • Input: Kinesis Data Streams, MSK (Kafka). Output: stream, Firehose, Lambda.
  • Video Streams: ingest video từ camera, xử lý với Rekognition, SageMaker.
  • Use case video: smart home, industrial monitoring, traffic analysis.

Kinesis vs SQS vs MSK

  • Kinesis: ordered per shard, multiple consumers, replay, real-time.
  • SQS: queue, 1 consumer group, delete after consume, no replay.
  • MSK (Managed Kafka): open-source compatible, topic-based, consumer groups, replay.
Info
Đề thi: 'real-time analytics trên clickstream' → Kinesis Data Streams + Flink. 'Deliver logs to S3' → Firehose.
Nội dung liên quan
Flashcards liên quan