Amazon RDS :: Aurora provisioned
- optimal read : upto ram 1024 GB, 40,000 mbps
- burstable : 2 Gb ram , 4 cpu unit
- Memory optimized -- SSD, eni-speed-high
- single, multi-az mode
Aurora RDS :: serverless v2
A. Intro
- engine :
Postgres
3x and MySQL
5x
- serverless v2 improvement : instant auto-scaling with no warm-up time.
- forwarding writes from secondary to primary region. developer dont nned to write logic.
- no capacity planning : configure min/max ACU (1acu === 2 gb ram , equ n/w and cpu)
- scale to 128TB per db instance

- OLTP | rdbms
- cluster design, fault tolerance by design:
- cluster spans multiple AZ in region.
- 1 primary + ( 5-16 )read replica(assign priority, can change at any time.)
- if no reader, the primary itself re-created.
B. DR
- RPO :
1 sec
| RTP < 1 min

- aurora has no stand-by instance thing

- has read replicas only in Aurora cluster, only replica can promote as primary, during DR.
- read replica has 2 purpose : reader endpoint + availability
- self-healing,
- continuous s3 backup,
- PITR
C. Advantages
1 Global Aurora database
- cross
region
replicas in less than a sec
. | single Database spans over multiple region.
- 1 Primary Region (read / write)
- Up to 5 secondary (read-only) in each region,
replication
lag is less than 1 second

- Up to 16 Read Replicas per secondary region
- RPO: less than second
- RTO: less than a minute

- provides more comprehensive failover capabilities

- Managed planned failover
- Unplanned failover ("detach and promote")
2 integration ML service
SageMaker
and Comprehend
- fraud detection, ads targeting, sentiment analysis, product recommendations
3 Auto-scaling (storage and compute are separate)
storage
scaling :: EBS volume - 10 GB to 128 TB
. /64TB?
compute
Instance :: type( eg: d.r3.large,etc), --RAM++, --cpu++.
Read replicas
: (built-in, dont need to create CW + ASG, auto happens bts)
- can add, CW metric --> triggers --> auto up/down read replicas
- AWS cloud optimized and claim
3x
Performance improvement (on Postgres)
- master +
6-15
Read Replica, with fast replication
5 Availability (cluster arch)
6 copies
for data access 3 AZ : cluster
( with reader and writer endpoint)
- instant fail-over (<30s) +
self healing
from peer2peer replication.


6 backtracking 
- rewind the DB cluster to any time you specify
- faster compared to restoring a DB cluster via PITR or snapshot
7 cloning
- usecase : create test env from prod.
- faster than backup
- restore uses
copy-on-write
- use same volume
- for new changes additional storage allocated and data copied to it.
8 more
- Isolation and
security
Industry compliance
- Push-button
scaling
- Advanced Monitoring
- Routine
Maintenance
+ Automated Patching
with Zero Downtime
- Backtrack:
restore
data at any point of time without using backups. (earliest :5 mim ago
)
B pricing
C. demo
- select engine
- select versions (so many available)
- template - prod (allow to configure everything)
- admis + password
- max i/o or standard
- ec2 instance or serverless
- choose : avialability - replica,etc
- vpc, subnet, Ibv4
- public access
- sg
- port
- authentication
- db name
- ...
=== READY ====
- add more read replica
- add cross az replica
- add region (global database)
- horizontal scaling policy (trigger : metric -CPU usage)
- max 15 and min 1