Distributed File Systems
1
Distributed File Systems
Distributed File Systems (DFS)
• Used by the A&T Unix and Windows
systems.
• Files can be stored on a file server or on
peer systems.
• Reduces the mass storage required to
make large data sets available to large
numbers of users.
• Allows centralized access and update of
files.
• Allows diskless workstations to function.
Goals
• A Distributed File System should look to
a client like a conventional, centralized
file system.
• Enforce access rights and security.
– Prevent viewing or modification as the file
is transmitted over the network.
• Be fault tolerant.
• Be scalable.
Andrew File System (AFS)
• Used at A&T.
• Can run on PCs or Unix systems.
• Requires client computers to have a disk.
• Copies files to the client computer’s hard
drive.
• Accessing a file from a hard drive is faster
than accessing it over a network.
• Network traffic is not necessary during the
use of the file.
AFS Design Assumptions
• Most files are small (<10K).
• Most files are used only by one user.
• Reads occur about 6 times more often
than writes.
• Most writes are by the file owner. Few
files have shared writes.
• Files are referenced in bursts (temporal
locality).
AFS Operation
• A request is sent to the file server when
a file or directory is opened.
• The server copies the file over the
network to the hard drive of the client.
• The user application accesses the file
from the local hard drive.
• When the file is closed, it is copied back
to the server if it has been updated.
Docsity.com
