Basics

Basic Component

1. DNS(domain network system): domain -> ip address.
2. web server: process query and return response.
3. database: separate data tier and computation tier.
4. load balancer: split the traffic to different server.
5. cache: between web server and DB to improve the data access in memory.
6. CDN: content delivery network, deliver the static content to save the web server capability and improve the response time.
7. data center: each has its own web servers, Db, caches. Share NoSQL.
8. message queue: seperate the producer/consumers logic, to make web servers/specific service more scalable.
9. logging: aggregate the errors message and monitor message together.
10. metrics: host level metrics, aggregated level metrics(DB, cache performance), DAU, revenue.
11. monitoring: monitor the health status of the system.
12. automation: like CICD.

Items

1. Vertical & horizontal scaling: single machine boost vs multi-machine boost. For the horizontal scaling, resharding data(consistent hashing), celebrity problem(hot key), join and de-normalization(join the normalized data to one).
2. Data replication: replicate the database to improve the stability/availibility of DB.
3. Stateless web tier(stateless architecture): all servers share storage NoSQL.
4. Stateful architecture: each server is responsible for specific part of users. Hard to handle the failure cases.

How to scale

1. Stateless architecture: keep web tier stateless
2. Replication: build redundancy at every tier
3. Cache: as much you can
4. Multiple data center: support
5. Sharding: scale the DB
6. Message queue: separate the producer/consumers, split tiers into individual services
7. Monitor/logging/automation/metrics

System Design Steps

1. Understand and clarify the scope
2. Propose high level design
3. Design deep dive
4. Wrap: further discussion.Monitor/Rollback/Error logs/Scal up