DistOS 2019F 2019-11-04

From Soma-notes
Jump to navigation Jump to search

Readings

GFS & Chubby

Discussion Questions

GFS

  • What was the key task GFS was built for?
  • What were the unusual design decisions made for GFS (i.e., how did they change semantics from UNIX-like filesystems)
  • Why did those design decisions make sense given the task GFS was designed for?
  • To what extent is GFS distributed? How is it centralized?
  • How does GFS achieve fault tolerance and high performance?
  • What are the different kinds of servers in GFS? What are their roles?

Chubby

  • To what extent is Chubby like a UNIX filesystem? How does its semantics differ?
  • Why does Chubby run on so few computers? Is it scalable?
  • What is Paxos? What is it for? Why is it important to Chubby?
  • What are the use cases for Chubby? How was it used with GFS?

Other

  • Do you think Chubby and GFS are sufficient for Google today? Has anything changed in their use cases?
  • Other thoughts or questions?