Outline for 3/4/98

• Last time: File protection / access control
• Administrative
  • Optional problems will be on the Homework webpage later today. Solutions will be posted eventually (before studying for finals begins)
  • Questions?
• Objective:
  • Introduce distributed systems
  • Distributed file systems

• Economic argument - lots of small machines are more cost effective than a single supercomputer-class machine.
• Sharing - some physical devices (printers, scanners) and logical resources (files, databases)
• Geographically distributed applications (banking, collaboration)
• Parallel processing - collections of machines cooperating on a single problem.

Some Issues in Distributed Systems

• Transparency - degree to which the location and boundaries between nodes are visible.
• Performance - latency, bandwidth
• Scalability - behavior as the size of the system grows.
• Reliability and fault tolerance - Parts go down
  • How to detect? What to do with the surviving parts?
• Security

• Various technologies (multiaccess bus - twisted pair, fiber optics; wireless - IR, radio; switched; store-and-forward)
• Usual distinctions
  • LANs (household ethernet); WANs (internet)
  • Packet-switching; Circuit-switching
• Trends - network technology is undergoing significant performance improvements - big impact on future system design

• Naming
• Location transparency/ independence
• Caching
• Consistency
• Replication
• Availability and updates

• \\His\d\pictures\castle.jpg
  • Not location transparent - both machine and drive embedded in name.
• NFS mounting
  • Remote directory mounted
    over local directory in local naming hierarching.
  • /usr/m_pt/A
  • No global view

Hints

• A valuable distributed systems design technique that can be illustrated in naming.
• Definition: information that is not guaranteed to be correct. If it is, it can improve performance. If not, things will still work OK. Must be able to validate information.
• Example: Sprite prefix tables

• Location of cache on client - disk or memory
• Update policy
  • write through
  • delayed writeback
  • write-on-close
• Consistency
  • Client does validity check, contacting server
  • Server call-backs

• Server crashes
  • State (if any kept) lost, reconstruct upon recovery (dialog with clients?)
  • Stateless server - all requests from clients are self-contained
• Network partitions
  • Client response - optimistic (continue to use what's in cache) or pessimistic (conservative)

• File name maps to set of replicas, one of which will be used to satisfy request
• Goal: availability
• Update strategy
  • Atomic updates - all or none
  • Primary copy approach
  • Voting schemes
  • Optimistic, then detection of conflicts