RDS Link to heading

  • Supports relational: SQL Server, Oracle, Mysql, Aurora, PostgreSQL, MariaDB
  • Supports nosql: dynamodb
  • Data warehousing: redshift (OLAP)
  • Encryption is supported (only applies if the db is being created and encryption option is enabled)
  • All RDS are OLTP

Backups Link to heading

  • two types: automated backups and database snapshots
  • automated backups:
    • aws takes a full daily snapshots with a retention period of 35 days.
    • enabled by default
    • stored in s3
    • during the backup process, the db i/o may be suspended or increased latency may be seen
  • database snapshots:
    • user initiazed
    • triggered when a db is deleted
    • the restore process creates a new RDS
    • snapshots can be encrypted

Multi-AZ Link to heading

  • Databases are replicated to different AZs (subnets). DR only, not performance.

Read Replica Link to heading

  • a way to scale a database cluster, the replicas are asynchronous copies of the main db.
  • ready-heave workloads
  • doesn’t work with: SQL Server and Oracle
  • for scaling, not DR
  • automatic backups needed enabled
  • limit of a 5 read replicas per db
  • each read-replica have its own DNS endpoint
  • read-replicas can have multi-az and/or different region
  • read-replicas of multi-az source databases
  • read-replicas can be promoted to its own database (write/read)

Engine: Aurora Link to heading

  • mysql-compatible
  • x5 times better performance than mysql ?
  • starts with 10Gb, scales in 10Gb increments to 65Tb
  • scales up to 32vcpus and 244Gb of mem
  • 2 copies of data is maintained on each AZ, minimum of 3 AZs. 6 copies of the data.
  • disks are automatically scaned and repaired
  • replicas:
    • aurora replicas: 15 (failovers)
    • mysql read replicas: 5
  • endpoints for both lead and replicas have the same endpoint (failover purposes)

Engine: SQL Server Link to heading

  • For RDS IOPS provisioned storage, the maximum volume can be 32TB

Limitations Link to heading

  • Don’t use it for large binary objects (AWS S3 is an option)
  • It doesn’t scale automatically (dynamodb does)
  • When data is not well structured or is unpredictable (use dynamodb)
  • It doesn’t support DB2 or HANA DB
  • Not complete access to the underlying EC2 instances