Lecture 7

Question 1

Scheduling Algorithms

Answer 1

• First Come, First Served
• Shortest Job First
• Priority
• Round Robin
• Multilevel Queues

Question 2

First Come, First Served (FCFS)

Answer 2

Processes are executed in the order they arrive

Question 3

Convey Effect

Answer 3

Short processes wait for one long process to get off CPU

Question 4

Shortest Job First (SJF)

Answer 4

Processes with the shortest time (burst time + arrival time)

-Provably optimal

Question 5

Burst Length Prediction

Answer 5

Use exponential average:
Tau_(n+1) = alpha*t_n + (1-alpha)Tau_n
- Tau is the predicted burst time
- t is the actual burst length
Raise (1-alpha) to higher powers to assign smaller weights to older values

Question 6

Shortest Remaining Time First (SRTF)

Answer 6

Accounting for arrival time and burst time, execute the process with the shortest remaining time, sometimes interrupt processes if a shorter process arrives

Question 7

Priority Scheduling

Answer 7

• A priority integer is associated with each process
• CPU is allocated to process with highest priority
• If priority is inverse of CPU burst length, it’s SJF

Question 8

Problem with priority scheduling

Answer 8

• Starvation (low priority processes may never execute)

- Solution is dynamic priority change

Question 9

Aging

Answer 9

• Increasing the priority of a process as it waits

Question 10

Round Robin Scheduling

Answer 10

• Each process gets a small unit of CPU time (time quantum q) usually 10-100 msec
• After q elapses, the process is preempted and added to the end of the ready queue
• Ready queue is a circular queue
• Typically higher average turnaround time than SJF but better response time
• If q is large, it’s basically just FCFS
• If q is small, then the overhead is really high

Question 11

Time Quantum Rule of Thumb

Answer 11

Time quantum should be longer than 80% of CPU bursts

Question 12

Multilevel Queue Scheduling

Answer 12

• Ready queue is partitioned into multiple queues
• Each has its own scheduling algorithm
• Scheduling is needed among queue

Question 13

Multilevel Feedback Queue Scheduling

Answer 13

• A process can move between various queues
• Dynamic priority change
• Process using too much CPU can be moved lower

Question 14

MQS needs to define…

Answer 14

• Number of queues
• Scheduling algorithm for each queue
• Method to move process up/down in queues
• Start queue for each process

Question 15

Multiprocessors

Answer 15

• Multiple CPUs

- Each CPU has many components (caches, data/instructions, decoders, execution units, etc)

Question 16

Multicores

Answer 16

• Multiple execution cores on one processor
• Duplicate some parts of the processor
• Each core has L1 cache, but all cores share L2 cache

Question 17

Multiple-Processor Scheduling

Answer 17

• Divide load among multiple processors (more complex than single CPU scheduling)
• Asymmetric vs Symmetric multiprocessors

Question 18

Asymmetric Multiprocessors

Answer 18

One master processor does the scheduling for others

Question 19

Symmetric Multiprocessors

Answer 19

• Each processor runs its own scheduler

- One ready queue for all processors or separate ready queue for each professor