This preview shows pages 1–4. Sign up to view the full content.
CSC 7103
Midterm Review Page 1 of 13
Midterm Review
A couple of problems based on Online Load Balancing/scheduling
algorithms
Know the
Greedy Algorithm on Identical Machines
and its competitive
level.
Slowfit Algorithm should read l
j
(k).
Right above it make all l
i
(j) read l
j
(i).
Know key points in the proof – we are showing it is not possible for all
machines to be overloaded, but the calculations show there is at least one
machine that is underloaded.
We run the slow fit but guessing a value of A and doubling each time it
fails to achieve the right value for A.
This approach does not hurt us
because the optimal algorithm is almost failing as much as you are.
Using
this observation you should be able to bound slow fit with the optimal and
show that it is still somewhat competitive.
Think about a formal way to
prove this with a parameterized algorithm and a constant C.
Questions on the exam
45 problems max on the exam
First part will be definitions – definitions of concepts and design issues
surrounding those concepts.
This will be ~3040% of the total of the exam.
Multicasting sequence type problem (similar to first sample problem)
Leader election is important (12 problems).
Don’t forget Dr. Bush’s talk.
Synchronous and Asynchronous algorithms are important.
Distributed mutual exclusion
Raymond’s Algorithm (or the advanced version – path compression)
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document CSC 7103
Midterm Review Page 2 of 13
Scheduling problem – We may be asked to competitive factor given
machines and jobs.
Stone’s algorithm
Questions like what is the difference between leader election and mutual
exclusion.
Look at sample problems!!!
Misc
Slowfit is not the same as greedy.
Slowfit gets a list of candidates and we
choose the lowest index.
Greedy just chooses the smallest sum of load.
Provide answer to A,B,C of Sample problem 3 by 10/20/09 for 5 bonus
points.
Chapter 1
Cooperative autonomous systems
(CAS) emphasize the design of distributed
applications for open system environments.
Degree of coupling
is measure of how centralized/decentralized a system is.
In terms of hardware a system would be tightly coupled if interprocessor
communication overhead is relatively close to its intraprocessor
communication time.
Example a multiprocessor PC is more tightly
coupled than a LAN.
DOSs present a centralized logical view of the software system that runs
under a loosely coupled multiple computer system.
The goal is to provided
a single computer view of a multiple computer system.
Figure 1.3 – Basic model of an operating system
web browser, email client
Application level
CSC 7103
Midterm Review Page 3 of 13
database, etc.
Subservice level
compiler, shells
Utilities
file mgmt, process mgmt, memory
mgmt
Service level
drivers, interrupts, cpu, memory
Kernel
The key distinction between a network operating system and a DOS is the
concept of transparency – the underlying system is transparent to the user.
Problems with DOSs:
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
This is the end of the preview. Sign up
to
access the rest of the document.
This note was uploaded on 01/25/2010 for the course CSC 7103 taught by Professor Kannan,r during the Fall '08 term at LSU.
 Fall '08
 Kannan,R
 Operating Systems

Click to edit the document details