Skip to content

Latest commit

 

History

History
157 lines (107 loc) · 4.98 KB

glossary.md

File metadata and controls

157 lines (107 loc) · 4.98 KB
Raw

Basic Considerations

  1. Architectural pieces that can be used
  2. How each architectural piece works with the other pieces
  3. Optimal way to use the pieces -- tradeoffs

Distribured Systems

Scalability

  • Ability for system to grow as demand increases.
  • Horizontal scaling: add more servers
  • Vertical scaling: add more power to your server (CPU, RAM, Storage, etc.)

Reliability

  • The probability that a system will fail in a given period.
  • Distributed systems are reliable if they can keep delivering services even though one or multiple hardware / software components fail.

Efficiency

Serviceability or Manageability

Load Balancing

Caching

Overview

  • Caching enables you to maximize your resources
  • Locality of reference principle: Recently requested data is likely to be requested again
  • Cache is like short-term memory
  • Caches are faster than the original data source
  • Caches contain the most recently accessed items

Application Server Cache

  • You can place a cache directly on a request layer node
  • This enables the local storage of response data
  • Each time a request is made to the service, the node will quickly return local cached data if it exists

Content Distribution Network (CDN)

  • Kind of cache used for sites leveraging large amounts of static media
  • CDNs hold static assets
  • When requesting an asset from the CDN, CDN is checked locally first; if the asset is not found locally, the CDN checks the backend servers.
  • After checking the backend servers, CDN will cache the asset locally.
  • Usually CDNs are only needed if the system you are building is quite large and calls for it.

Cache Invalidation

  • Need to keep the cache valid
  • If data that is in the cache is modified in the DB, it needs to be invalidated in the cache
  • Write-through cache: write directly to the cache and the db
  • Write-around cache: skip the cache and write directly to the db
  • Write-back cache: write directly to the cache, and update the main db under specified conditions

Cache Eviction Policies

  • First in First Out (FIFO): Cache evicts the first block accessed
  • Last in First Out (LIFO): Cache evicts the last block accessed
  • Least Recently Used (LRU): Disgards the least recently used item first
  • Most Recently Used (MRU): Disgards the most recently used item first
  • Least Frequently Used (LFU): Discards the least frequently used items
  • Random Replacement (RR): Randomly discards a candidate

HTTP Caching

Source: HTTP Caching

  • Every browser ships with and implementation of an HTTP cache
  • You just need to ensure that each server response provides the correct HTTP header directives
  • These directives instruct the broswer on when and for how long the browser can cache the response
    • ETag Header: validation token
    • Cache-Control: who can cache the reponse and under which conditions
  • Locally cached responses are used until the resource expires
  • Each application needs to define its own cache hierarchy for optimal performance

Browser Cache

Source: What is a browser cache, and why is it important?

  • Assets stored by browser cache
    • Images
    • HTML
    • CSS
    • JavaScript
  • Browsers typically cache static assets (assets that do not change from visit to visit)
  • What to cache and for how long is determined by the website
  • The main pitfall of local caching is that if an asset is swapped, you may have to clear your cache before being able to see the new asset.

Data Partitioning

Indexes

Proxies

Redundancy and Replication

SQL vs NoSQL

CAP Theorem

Consistent Hashing

Long-Polling vs WebSockets vs Server-Sent Events

HTTP Request

  1. Client sends a request to a server
  2. Server calculates the request
  3. Server sends a response to the client

AJAX Polling

  • Client repeatedly polls (or requests) a server for data
  • Requests are sent at regular intervals (e.g. 0.5 seconds)
  • Server calculates the response and sends it back

HTTP Long-Polling

  • Aka "Hanging Get"
  • Instead of sending back an empty response, the request "hangs" until the server has something to send back to the client
  • Once the client recieves the request, typically another request is sent out

WebSockets

  • Creates a bi-drectional (full duplex) communication channel over a single TCP collection
  • TCP: Transmission Control Prorocol
  • Enables communication between client and server with lower overheads
  • Facilitates real-time data transfer from the server
  • Standardizes the conversation between client and server
  • Open conversation

Server-Sent Events (SSEs)

  • Client establishes a persistent long-term connection to the server
  • Server uses this connection to send data to the client
  • This is a one-directional connection
  • Server can send to client, but client cannot send to server