Lecture 7: CPU Scheduling 1

Question 1

Process States: New

Answer 1

Process is being created

Question 2

Process States: Running

Answer 2

Instructions are being executed

Question 3

Process States: Waiting

Answer 3

Process is waiting for some event to occur

Question 4

Process States: Ready

Answer 4

Process is waiting to be assigned the CPU

Question 5

Process States: Terminated

Answer 5

Process has finished execution

Question 6

One processor → one process running at a time

Answer 6

• All other processers wait for CPU to be free

* Often running process does need CPU (e.g., I/O requests)

Question 7

Multi programming goal: avoid CPU idle times → some

process is using the CPU all the time

Answer 7

• Keep multiple processes in memory at one time
• When one process has to wait the CPU is allocated to another process
• Scheduling algorithms: decide which process goes into the CPU

Question 8

CPU (Short-Term) Scheduler:

Answer 8

• first example of resource manager(manages sharing of CPU)
• applicable to either multiprocessing or multi-threading (we’ll assume former)
• What? A kernel process that is always active (daemon)
• responsible for choosing process to run from ready queue

Question 9

When Does the Scheduler Get Invoked?

Answer 9

1. When there’s an interrupt
   • from timer: moves process from running to ready state
   • upon I/O interrupt: moves process from wait to ready state
2. When running process terminates
3. When running process issues an I/O request (block) and moves to wait state

Question 10

Non-Preemptive:

Answer 10

Processes only give up CPU voluntarily
• simpler to implement (no timer interrupts)
• greedy or buggy process can starve others

Question 11

Preemptive:

Answer 11

Processes also may be preempted by an interrupt
• adds complexity
• adds protection, better at ensuring fairness, as well as
doing better in other scheduling metrics

Question 12

Throughput:

Answer 12

Measured in # processes completed / time unit

Number of processes that complete execution per time unit (Good = High)

Question 13

Turnaround Time:

Answer 13

Average (time of completion –time of creation)
- Related metric: Waiting Time (average time on ready queue)
- Waiting vs Turnaround: no penalty for processes w/ long processing times
(Turnaround Time = Processing /Blocked Time + Waiting Time)

Question 14

Response Time:

Answer 14

Time of 1st response is produced–time of creation

Question 15

Overhead:

Answer 15

Time spent related to scheduling (e.g., context switch time)

Question 16

Fairness:

Answer 16

How much variation in waiting time

 a) minimal
 b) proportional (to processing time)
 c) never infinite (aka:  no starvation)

Question 17

CPU Utilization

Answer 17

What percentage of time is the CPU doing useful work?

Question 18

Dispatch time:

Answer 18

Time it takes to choose next running process including schedule time +
context switch time. If lengthy, effects effectiveness of scheduler

Question 19

First-Come First-Serve (FCFS)

Answer 19

Non-Preemptive Policy

+: Easy to implement (ready queue = scheduling queue)
No starvation
–: Convoy Effect: Order of arrival determines performance
e.g., many processes might end up waiting for a big one to get off the CPU
(Food for thought: maybe shortest process should go first instead?)

Question 20

Shortest-Job-First (SJF)

Answer 20

Non-Preemptive Policy
Idea: Rank processes by CPU time requests.
Optimizes (minimizes) average waiting time

Preemptive Version of SJF
Idea: Currently running process can be preempted if new process
arrives with shorter remaining CPU burst