Database access is through database CNAME which points to primary instance
All reads and writes happen from primary instance
Backups & Snapshots happen from StandBy
When failover happens, database CNAME points to standby. Now it becomes Primary.This takes 60 to 120 seconds. Hence this switchover can have some downtime
There is only 1 standby replica
MultiAZ mode can be only within the same region
MultiAZ Cluster Deployment
Can replicate to one writer & 2 readers. The readers are in different AZs. This mode can have only 2 Readers while Aurora can have more
Replication between Writers & Readers is Synchronous
Data written to a Writer is confirmed to be committed when atleast 1 reader confirms replication
Each instance has its own storage
The end points available are
Cluster Endpoint which points to writer. It can be used for Reads, Writes & Administrative functions
Reader End Point points to the reader instances
Instance End Points point to each of the instances within the cluster
This runs on much faster hardware. Any writes are first written on to local NVME SSD Storage & then flushed to EBS
Replication is done through Transaction Logs
Failover can happen in upto 35s. Additional time is required to apply transaction logs to reader nodes