Lecture 25 - Distributed Systems

Content locked

The catalog is public. Reading this content requires approved access.

What is a Distributed System?

Why Go Distributed?

New Challenges

Pipes vs Network Sockets

Reliable Messages: Layering Strategy

Technique #1: ACK

Technique #2: Timeout

Lost Ack Issues

Technique #3: Receiver Remembers Messages

TCP

Distributed File Systems

Goals

NFS — Network File System

Architecture

Mount Tree

General Strategy: Export FS

NFS Network API: Strategy 1 — Wrap UNIX Calls

Problem: Server Crash

Strategy 2: Stateless Protocol

Strategy 3: Inode Requests

Strategy 4: File Handles

NFS Protocol Walkthrough

Can NFS Include Append?

Idempotent Operations

pwrite is idempotent

append is NOT idempotent

What’s Idempotent?

Strategy 5: Client Logic on Top of Idempotent Requests

Replica Suppression in TCP is Stateful

Write Buffering

Server Write Buffer Lost (Problem)

Solutions

NFS Summary

Source