CS439: Principles of Computer Systems
Homework 10, Part 1
Due in Section on Friday, April 22, 2016 Part 1 of the
homeworks must be completed before section and brought to section. Please refer to the
homework turnin instructions.
Suppose a computer
system and all of its applications are completely bug free. Suppose
further that everyone in the world is completely honest and
trustworthy. In other words, we do not need to consider fault isolation.
- How should the operating system allocate the processor? Should
it give all of the processor to each application until it no longer
needs it? If there are multiple tasks ready to execute at the same
time, should it schedule the task with the least amount of work to do
or the most? Justify your answer.
You may assume a uniprocessor.
- How should the operating system allocate physical memory between
applications? What should happen if the set of applications does not
all fit in memory at the same time?
Note: These are thought questions---you are not expected to
determine how operating systems currently solve these problems.
Instead, you should think about how you would solve the problem.
Now suppose the
computer system needs to support fault isolation. What hardware
and/or operating support do you think would be needed to protect an
application's data structure in memory from being corrupted by other
applications? Assume more than one application may reside in memory
at a time.
(From Anderson and Dahlin, Ch1 question 2.)
Again, a thought question.
In two-phase commit, suppose the master fails after writing
``GLOBAL_COMMIT'' to disk and sending ``GLOBAL_COMMIT'' to all
participating nodes but before any of the nodes receive this
message. Suppose that the ``GLOBAL_COMMIT'' message reaches i of
the n participants and that the participants implement a protocol
in which they communicate with one another when they suspect the
primary has died. What is the minimum number of participants that
must receive the ``GLOBAL_COMMIT'' message to allow the
participants to complete the transaction without waiting for the
master to recover?
Sun's network file system (NFS) protocol provides reliability via:
- at-most-once semantics
- at-least-once semantics
- two-phase commit