Please enable JavaScript.

Coggle requires JavaScript to display documents.

System Design Basics, SDI, TinyURL - Coggle Diagram

- - - - Least Connection Method
      - Least Response Time Method
      - Least Bandwidth Method
      - Round Robin Method
      - Weighted Round Robin Method
      - IP Hash
- - - - service layer
      - memory layer
      - local disk layer, faster than going to network storage
  - - - global caches
      - distributed caches
  - - - database change, invalidate cache
    - - Write-through cache
        
        Pros
        
        data consistency
        
        cons
        
        high latency
      - Write-around cache
        
        Pros
        
        avoid flood cache by data that will not be re-requested
        
        cons
        
        cache miss
      - Write-back cache
        
        pros
        
        low latency
        
        hig throughput for write intensive applications
        
        cons
        
        risk of data loss when cache crashes
- - - - Pros
        
        uniform allocation of data
      - cons
        
        fixes the total number of DB servers
        
        has to redistribution of data and has downtime for adding new servers
        
        solution: Consistent Hashing
  - - - alias: Data sharding
      - exampe: zip code range into different table
      - cons: possible unbalanced servers
      - definition: row range based
    - - definition: feature based
      - cons: possible additional server when specific feature grows
      - Example: photo, users, friends list in different server
      - Pros: easy to implement, less impact on app
    - - definition: create a look up service that knows the partitioning schema, query the directory
      - Pros: loosely coupled, no impact on the application
  - - - problem: difficult when tables are in different servers
      - Work arounds: denormalize the database
    - - Problem: difficult to enforce data integrity constrain such as foreign keys are difficult
      - workarounds: denormalize the database
    - - reasons
        
        not uniformed data distribution: eg, zip code
        
        a lot of load on a partition: eg, too many request on photo partition
      - challenges: difficult to deal with downtime
      - solution to downtime: directory based partitioning
        
        cons: complexity & single point failure
- - - - store in rows-columns
      - examples: MySQL, Oracle, MS SQL server, SQLites, Postgres...
  - - - Key-value
        
        definition
      - Document Databases
        
        definition
        
        data in document
        
        examples
        
        CouchDB, MongoDB
      - Wide-column Database
        
        definition
        
        example
        
        Cassandra & HBase
        
        senarios
        
        large datasets
      - Graph Database
        
        definition
        
        data in nodes, properties and lines
        
        example
        
        Neo4J, InfiniteGraph
  - - - SQL
        
        vertically & expensive
      - NoSQL
        
        horizontally & cost-effective
    - - SQL
        
        structured query language
      - NoSQL
        
        unstructured query language
    - - SQL
        
        fixed upfront
      - NoSQL
        
        dynamic
    - - SQL
        
        Atomicity, Consistency, Isolation, Durability compliant
      - NoSQL
        
        sacrifice ACID for performance and scalibility
    - - SQL -
        
        row & columns
      - NoSQL
        
        key-value, document, graph, and columnar
  - - - need ACID
      - structured data and slow growth
    - - cloud computing and storage, store in multiple servers
      - rapid development, no prepare ahead of time
      - no or little structured data
- - - - add/remove headers
      - encrypting/decrypting
      - compressing a resource
  - - - anonymous proxy
        
        hide the initial IP
      - transparent proxy
        
        first IP address can be viewed, cache the website
    - - retrieves resources on behalf of the clients, appearing as if they originated from the proxy server