The HA-NFS and AFS papers are the main readings:

- The HA-NFS paper is simple and well written.  You should read it
  in its entirety and be familiar with every aspect of its design.

- The AFS paper contains many experimental results and analyses, 
  but the basic issues behind it are simple.  Here are a few questions
  that you should focus on while reading this paper:

  - What were the goals of the AFS designers?

  - Describe the initial AFS system design, with regards to
    - Client caching policies
    - Namespace management
    - Client-server communication and process management

 - What sort of experimental evaluation methodology did they choose?

 - What were the results of the experimental evaluation of the
   original prototype and how did it influence their subsequent
   revised version of AFS?

 - What were the effects of their changes for performance?

 - How does AFS performance compare to NFS for a large number of
   clients?   Which factors contribute to this performance
   difference?

 - Briefly, what were the changes to AFS to improve system 
   manageability?


The other two papers are complimentary (but mandatory) readings.


- On the Hartung paper, read pages 383 to 389 (up to but not 
  including the Performance section), focusing on Sections

  - "Basic Operation" (p. 386), 
  - "Virtualization" (p. 387), and
  - "Continuous Availability" (p. 388).  

  Think about the common themes (if any) between this system and 
  AutoRAID or HA-NFS.


- The Hitz paper describes the high-level design of the Network
  Appliance file system design (WAFL, "write anywhere file layout").
  Skim through this paper and look for answers to the following 
  questions:

  - What is a snapshot?

  - How does WAFL solve the RAID5 small write problem?  (RAID4, which
    is mentioned in this paper is similar to RAID5 except that it 
    uses a dedicated disk to store parity blocks).

  - What are the benefits of logging NFS requests to NVRAM?