Distributed File System

Design Distributed File System
1. GFS (Google File System)
  1. Client -> Web Server -> Database Server -> GFS
2. HDFS (Hadoop Distributed File System)
Scenario
1. Read / Write Files.
  1. For example 1000T
2. Store on multiple machines
  1. 100K machines
Service
1. Client (Web Server / Database Server)
2. Server
  1. Peer to Peer
    Client talks to one server. All servers are equal.
    1. Pro: Fault tolerance.
    2. Cons: Synchronization.
  2. Master and Slave
    Client talks to the master. Master manage slaves. Master doesn’t store data. Slaves partition data.
    1. Pro: Synchronization.
    2. Cons: Fault tolerance.

levendlee