AWS:

Five Pillars from Well-Architechted Framework:
https://aws.amazon.com/getting-started/fundamentals-core-concepts/?e=gs2020&p=fundoverview&p=gsrc&c=fo

1. Opertional Excellence
	- IaC (Infrastructure as Code)
		- yaml/json configuration files you write that specify AWS resources (e.g. # of task nodes)
	- Observability
		- log files/data from CloudWatch that monitor AWS resources

2. Security
	- IAM (Identitiy and Access Management)
		- Three objects: principals (who), actions (do), resources (to what)
		- Policies determine who (principal) can do something (actions) to something (resource)
		- Employs principle of least privilege (only allow principals access to what is needed to complete task)
	- Network Security
		- Network-level:
			- "defense in depth" -> security measure at each layer (IP, application, etc.)
			- VPC (virtual private cloud) with subnets (multiple IP addresses) with one internet-facing IP
		- Resource-level:
			- only allow certain communication (e.g. a port) for specific resources (e.g. lambda or EC2)
	- Encryption
		- scramble information upon send-off, de-scramble information at receival (each side needs key)
		- HTTPS endpoints for resources (e.g. database)
		- encryption "at rest" (e.g. within your VPC)

3. Reliability
	- Fault Isolation
		- Resource-level
			- redundancy in resource (e.g. EC2)
			- sharding
		- Availability Zone-level
			- multiple independent (e.g. power) facilities at same region
		- Region
			- possible (but only necessary for large clients) to have resources in multiple regions
	- Limits
		- maximum number of instances (e.g. EC2) for a given AWS account
		- limits blast radius of software misconfiguration or malicious external attacks

4. Performance Efficiency
	- Selection of service
		- Type of Service
			- Compute
				- Virtual Machine (VM)-based
					- EC2 (core product)
					- Elastic Beanstalk
					- Lightsail
				- Container-based
					- ECS (e.g. Docker implementation)
				- Serverless-based (abstracts provisioning)
					- Lambda
					- Fargate (serverless for ECS)
			- Storage
				- Block
					- EBS
				- File-based
					- EFS
				- Object-based
					- S3 (Simple Storage Service)
				- Archival
					- S3 Glacier
			- Database
				- Relational
					- RDS
					- Aurora
				- Non-relational
					- DynamoDB
				- Data warehouse
					- Redshift
			- Network
		- Degree of Management
			- Choosing service within category mainly depends on what level of management the use-case requires
		- Configuration
			- Choosing parameter of service (e.g. size of instance -> t3.small, t3.large, block-storage type -> solid-state, hard drive) affects performance
	- Scaling
		- Vertical
			- increasing instance size (e.g. t3.small -> t3.large)
			- easier to implement but max compute limited by number of instances
		- Horizontal
			- increasing number of instances (e.g. 2 t3.small instances to 4)
			- more difficult to implement (proxy, routing) but no max compute

5. Cost Optimization
	- Pay for use
		- Right-sizing
			- Match resources to requirement (Opex instead of mainframe Capex)
		- Serverless
			- only pay when compute is executed (e.g. Lambda)
		- Reservations
			- discounts if you reserve compute instances
		- Spot instances
			- large discounts if you use spot instances that can be reclaimed by AWS at any time
	- Cost Optimization Lifecycle
		- Review
			- (similar to observability from operational excellence)
			- know and visualize cost and usage across time
		- Track
			- allocate usage and cost to users, applications, cost centers/departments, etc.
			- create usage/cost budgets
		- Optimize
			- change instances/configuration based on budget