Data Store Concepts

Notes regarding data store concepts and terminology

Data persistence

• Persistent.

Data is durable and survive restarts or power cycles. Glacier, RDS

• Transient.

Data is just temporarily stored and passed along to another service or persistent store. SQS, SNS

• Ephemeral.

Data is lost after service stop, restart, power cycles. EC2 instance store, Memcached

IOPS vs Throughput

Transfer speed and data volume that can be transferred

IOPS

Input/Output Operations per second (IOPS). Measure of how fast we can read write to a device

Throughput

Measure of how much data can be moved at a time

Consistency models (ACID & BASE)

ACID

Atomic - transactions are “all or nothing”

Consistent - transactions must be valid

Isolated - transactions can’t mess with one another.

Durable - completed transaction must stick around

BASE

Basic Availability - values availability event if stale

Soft-state - might not be instantly consistent across stores

Eventual consistency - will achieve consistency at some point