阅读论文 GFS(2003),可以思考以下问题:
- Why is atomic record append at-least-once, rather than exactly once?
- How does an application know what sections of a chunk consist of padding and duplicate records?
- How can clients find their data given that atomic record append writes it at an unpredictable offset in the file?
- The paper mentions reference counts -- what are they?
- If an application uses the standard POSIX file APIs, would it need to be modified in order to use GFS?
- How does GFS determine the location of the nearest replica?
- Does Google still use GFS?
- Won't the master be a performance bottleneck?
- How acceptable is it that GFS trades correctness for performance and simplicity?
- What if the master fails?
问题的答案在这里
在阅读讲义请先思考以下问题:
- Why are we reading this paper?
- What is consistency?
- "Ideal" consistency model
- Challenges to achieving ideal consistency
- GFS goals:
- High-level design / Reads / Writes
- Record append
- Housekeeping
- Failures
- Does GFS achieve "ideal" consistency?
- Authors claims weak consistency is not a big problems for apps
- Performance (Figure 3)
[Lab 02 的上机说明](6.824 Lab 2: Raft.html)