Which features of the AFS design make it more scalable than NFS? What are the limits on its
scalability, assuming that servers can be added as required? Which recent developments offer
greater scalbility?
What will be an ideal response?
The load of RPC calls on AFS servers is much less than NFS servers for the same client workload. This is
achieved by the elimination of all remote calls except those associated with open and close operations, and the
use of the callback mechanism to maintain the consistency of client caches (compared to the use of
getattributes calls by the clients in NFS). The scalability of AFS is limited by the performance of the single
server that holds the most-frequently accessed file volume (e.g. the volume containing /etc/passwd, /etc/hosts,
or some similar system file). Since read-write files cannot be replicated in AFS, there is no way to distribute
the load of access to frequently-used files.
Designs such as xFS and Frangipani offer greater scalability by separating the management and metadata
operations from the data handling, and they reduce network traffic by locating files based on usage patterns.
You might also like to view...
A DSU
a. converts digital signals to analog ones, and vice versa b. converts local digital signal interfaces to voltage levels used on the phone company’s digital circuit c. always includes the CSU function d. is the same as a repeater
Algorithms are written in a format that is specific to a particular programming language.
Answer the following statement true (T) or false (F)