Cache

Location of Caching

Client (Browser) Caching

Cache data in browser.

You can do it manually, store things like user info such as profile pic or email. You don’t want to request that very time. Or some data for offline working or faster rendering.

There are libraries doing this automatically, such as Apollo GraphQL Client. It can cache GraphQL queries automatically, so that repetitive requests aren’t made.

• Caching in Apollo Client
• Configuring the Apollo Client cache

Server-Side Caching

Reverse Proxy can cache data.

CDN Caching

Application Caching

There are applications like Redis that cache in memory. Cache Invalidation is required. There are algorithms like LRU.

Database Caching

Databases support caching, to save time from queries.

• MySQL Query Cache

• Postgresql Cache

• Query Level Caching

  • Use query hash as key to cache
  • Gets complicated when query is complex, changing a single cell may affect many cache.

• Object Level Caching primer

Types of Caching

Cache-Aside

graph LR;
    Client-->Cache
    Client-->Storage

Write-Through

sequenceDiagram
    participant Client
    participant Cache
    participant DB

    Client->>Cache: 1. Write to Cache
    Cache->>DB: 2. Store in DB
    Cache->>Client: 3. Return to user

Write-Begind (write-back)

Write data to database asynchronously. Good for situations that have super high frequency database write.

sequenceDiagram
    participant Client
    participant Cache
    participant Queue
    participant Event Processor
    participant DB

    Client->>Cache: 1. Write to Cache
    Cache->>Event Processor: 2. Add to queue (as publisher)
    Cache->>Client: 3. Return to user
    Event Processor->>DB: 4. Async: select and execute event (as Consumer)

Refresh-Ahead

graph LR;
    Cache-->Client
    Database-->Cache

Database automatically refresh recently accessed cache prior to expiration.

• Reduced Latency
• Drawback: Have to accurately predict which items are going to be queried.

Reference

• Cache