Unit II – CPU Scheduling

Outline
 Scheduling Objectives
 Levels of Scheduling
 Scheduling Criteria
 Scheduling Algorithms
 FCFS, Shortest Job First, Priority, Round Robin,
Multilevel
 Multiple Processor Scheduling
 Real-time Scheduling
 Algorithm Evaluation
Scheduling Objectives
 Enforcement of fairness
 in allocating resources to processes
 Enforcement of priorities
 Make best use of available system resources
 Give preference to processes holding key
resources.
 Give preference to processes exhibiting good
behavior.
 Degrade gracefully under heavy loads.
Program Behavior Issues

 I/O boundedness
 short burst of CPU before blocking for I/O
 CPU boundedness
 extensive use of CPU before blocking for I/O
 Urgency and Priorities
 Frequency of preemption
 Process execution time
 Time sharing
 amount of execution time process has already received.
Basic Concepts

 Maximum CPU utilization obtained with

multiprogramming.
 CPU-I/O Burst Cycle
 Process execution consists of a cycle of CPU execution
and I/O wait.
CPU Burst Distribution
Levels of Scheduling
 High Level Scheduling or Job Scheduling
 Selects jobs allowed to compete for CPU and other
system resources.
 Intermediate Level Scheduling or Medium
Term Scheduling
 Selects which jobs to temporarily suspend/resume to
smooth fluctuations in system load.
 Low Level (CPU) Scheduling or Dispatching
 Selects the ready process that will be assigned the CPU.
 Ready Queue contains PCBs of processes.
Levels of Scheduling(cont.)
CPU Scheduler
 Selects from among the processes in
memory that are ready to execute, and
allocates the CPU to one of them.
 Non-preemptive Scheduling
 Once CPU has been allocated to a process, the process
keeps the CPU until
 Process exits OR
 Process switches to waiting state
 Preemptive Scheduling
 Process can be interrupted and must release the CPU.
 Need to coordinate access to shared data
CPU Scheduling Decisions

 CPU scheduling decisions may take place

when a process:
 switches from running state to waiting state
 switches from running state to ready state
 switches from waiting to ready
 terminates
 Scheduling under 1 and 4 is non-preemptive.
 All other scheduling is preemptive.
CPU scheduling decisions

new admitted
new exit terminated

interrupt

running
ready

Scheduler
I/O or dispatch
event I/O or
completion event wait

waiting
Dispatcher
 Dispatcher module gives control of the CPU
to the process selected by the short-term
scheduler. This involves:
 switching context
 switching to user mode
 jumping to the proper location in the user program to restart
that program
 Dispatch Latency:
 time it takes for the dispatcher to stop one process and
start another running.
 Dispatcher must be fast.
Scheduling Criteria

 CPU Utilization
 Keep the CPU and other resources as busy as possible
 Throughput
 # of processes that complete their execution per time
unit.
 Turnaround time
 amount of time to execute a particular process from its
entry time.
Scheduling Criteria (cont.)

 Waiting time
 amount of time a process has been waiting in the ready
queue.
 Response Time (in a time-sharing
environment)
 amount of time it takes from when a request was
submitted until the first response is produced, NOT
output.
Optimization Criteria

 Max CPU Utilization

 Max Throughput
 Min Turnaround time
 Min Waiting time
 Min response time
First Come First Serve (FCFS)
Scheduling
 Policy: Process that requests the CPU FIRST
is allocated the CPU FIRST.
 FCFS is a non-preemptive algorithm.
 Implementation - using FIFO queues
 incoming process is added to the tail of the queue.
 Process selected for execution is taken from head of
queue.
 Performance metric - Average waiting time in
queue.
 Gantt Charts are used to visualize schedules.
First-Come, First-Served(FCFS)
Scheduling
 Example  Suppose the arrival
order for the processes
Process Burst Time
P1 24
is
P2 3  P1, P2, P3
P3 3
 Waiting time
 P1 = 0;
Gantt Chart for Schedule
 P2 = 24;
P1 P2 P3  P3 = 27;
0 24 27 30  Average waiting time
 (0+24+27)/3 = 17
FCFS Scheduling (cont.)

 Example  Suppose the arrival order

for the processes is
 P2, P3, P1
Process Burst Time
P1 24  Waiting time
P2 3  P1 = 6; P2 = 0; P3 = 3;
P3 3
 Average waiting time
Gantt Chart for Schedule
 (6+0+3)/3 = 3 , better..
 Convoy Effect:
P2 P3 P1
 short process behind long
0 3 6 process, e.g. 1 CPU bound
30 process, many I/O bound
processes.
Shortest-Job-First(SJF) Scheduling
 Associate with each process the length of its next
CPU burst. Use these lengths to schedule the
process with the shortest time.
 Two Schemes:
 Scheme 1: Non-preemptive
 Once CPU is given to the process it cannot be preempted
until it completes its CPU burst.
 Scheme 2: Preemptive
 If a new CPU process arrives with CPU burst length less
than remaining time of current executing process, preempt.
Also called Shortest-Remaining-Time-First (SRTF).
 SJF is optimal - gives minimum average waiting time for
a given set of processes.
Non-Preemptive SJF Scheduling

 Example
Process Arrival TimeBurst Time
P1 0 7
P2 2 4
P3 4 1
P4 5 4
Gantt Chart for Schedule

P1 P3 P2 P4

0 7 8 12 16
Average waiting time =
(0+6+3+7)/4 = 4
Preemptive SJF Scheduling(SRTF)

 Example
Process Arrival TimeBurst Time
P1 0 7
P2 2 4
P3 4 1
P4 5 4
Gantt Chart for Schedule

P1 P2 P3 P2 P4 P1

0 2 4 5 7 11 16
Average waiting time =
(9+1+0+2)/4 = 3
Determining Length of Next CPU
Burst
 One can only estimate the length of burst.
 Use the length of previous CPU bursts and
perform exponential averaging.
 tn = actual length of nth burst
 n+1 =predicted value for the next CPU burst
  = 0, 0    1
 Define
 n+1 =  tn + (1- ) n
Exponential Averaging(cont.)

 =0
 n+1 = n; Recent history does not count

 = 1
 n+1 = tn; Only the actual last CPU burst counts.

 Similarly, expanding the formula:

 n+1 = tn + (1-) tn-1 + …+ j

(1-)^j tn-j + …
(1-)^(n+1) 0
 Each successive term has less weight than its predecessor.
Priority Scheduling

 A priority value (integer) is associated with

each process. Can be based on
 Cost to user
 Importance to user
 Aging
 %CPU time used in last X hours.
 CPU is allocated to process with the highest
priority.
 Preemptive
 Nonpreemptive
Priority Scheduling (cont.)

 SJN is a priority scheme where the priority is

the predicted next CPU burst time.
 Problem
 Starvation!! - Low priority processes may never execute.
 Solution
 Aging - as time progresses increase the priority of the
process.
Round Robin (RR)
 Each process gets a small unit of CPU time
 Time quantum usually 10-100 milliseconds.
 After this time has elapsed, the process is preempted and
added to the end of the ready queue.
 n processes, time quantum = q
 Each process gets 1/n CPU time in chunks of at most q
time units at a time.
 No process waits more than (n-1)q time units.
 Performance
 Time slice q too large - FIFO behavior
 Time slice q too small - Overhead of context switch is
too expensive.
 Heuristic - 70-80% of jobs block within timeslice
Round Robin Example

 Time Quantum = 20
Process Burst Time
P1 53
P2 17
P3 68
P4 24

Gantt Chart for Schedule

P1 P2 P3 P4 P1 P3 P4 P1 P3 P3

0 20 37 57 77 97 117 121 134 154 162

Typically, higher average turnaround time than SRTF, but better response
Multilevel Queue
 Ready Queue partitioned into separate queues
 Example: system processes, foreground (interactive), background
(batch), student processes….
 Each queue has its own scheduling algorithm
 Example: foreground (RR), background(FCFS)
 Processes assigned to one queue permanently.
 Scheduling must be done between the queues
 Fixed priority - serve all from foreground, then from background.
Possibility of starvation.
 Time slice - Each queue gets some CPU time that it schedules -
e.g. 80% foreground(RR), 20% background (FCFS)
Multilevel Queues
Multilevel Feedback Queue
 Multilevel Queue with priorities
 A process can move between the queues.
 Aging can be implemented this way.
 Parameters for a multilevel feedback queue
scheduler:
 number of queues.
 scheduling algorithm for each queue.
 method used to determine when to upgrade a process.
 method used to determine when to demote a process.
 method used to determine which queue a process will enter
when that process needs service.
Multilevel Feedback Queues

 Example: Three Queues -

 Q0 - time quantum 8 milliseconds (RR)
 Q1 - time quantum 16 milliseconds (RR)
 Q2 - FCFS
 Scheduling
 New job enters Q0 - When it gains CPU, it receives 8
milliseconds. If job does not finish, move it to Q1.
 At Q1, when job gains CPU, it receives 16 more
milliseconds. If job does not complete, it is preempted and
moved to queue Q2.
Multilevel Feedback Queues
Multiple-Processor Scheduling
 CPU scheduling becomes more complex
when multiple CPUs are available.
 Have one ready queue accessed by each CPU.
 Self scheduled - each CPU dispatches a job from ready Q
 Master-Slave - one CPU schedules the other CPUs
 Homogeneous processors within
multiprocessor.
 Permits Load Sharing
 Asymmetric multiprocessing
 only 1 CPU runs kernel, others run user programs
 alleviates need for data sharing
Real-Time Scheduling

 Hard Real-time Computing -

 required to complete a critical task within a guaranteed
amount of time.
 Soft Real-time Computing -
 requires that critical processes receive priority over less
fortunate ones.
 Types of real-time Schedulers
 Periodic Schedulers - Fixed Arrival Rate
 Demand-Driven Schedulers - Variable Arrival Rate
 Deadline Schedulers - Priority determined by deadline
 …..
Issues in Real-time Scheduling
 Dispatch Latency
 Problem - Need to keep dispatch latency small, OS may
enforce process to wait for system call or I/O to complete.
 Solution - Make system calls preemptible, determine safe
criteria such that kernel can be interrupted.
 Priority Inversion and Inheritance
 Problem: Priority Inversion
 Higher Priority Process needs kernel resource currently
being used by another lower priority process..higher
priority process must wait.
 Solution: Priority Inheritance
 Low priority process now inherits high priority until it has
completed use of the resource in question.
Real-time Scheduling - Dispatch
Latency
Algorithm Evaluation
 Deterministic Modeling
 Takes a particular predetermined workload and defines the
performance of each algorithm for that workload. Too specific,
requires exact knowledge to be useful.
 Queuing Models and Queuing Theory
 Use distributions of CPU and I/O bursts. Knowing arrival and
service rates - can compute utilization, average queue length,
average wait time etc…
 Little’s formula - n = W where n is the average queue length, 
is the avg. arrival rate and W is the avg. waiting time in queue.
 Other techniques: Simulations, Implementation