Transcript Chapter 9 Uniprocessor Scheduling

Operating Systems:
Internals and Design Principles, 6/E
William Stallings
Chapter 9
Uniprocessor Scheduling
Dave Bremer
Otago Polytechnic, N.Z.
©2008, Prentice Hall
Roadmap
• Types of Processor Scheduling
• Scheduling Algorithms
• Traditional UNIX Scheduling
Scheduling
• An OS must allocate resources amongst
competing processes.
• The resource provided by a processor is
execution time
• The resource is allocated by means of a
scheduling
Overall Aim of Scheduling
• Aim of processor scheduling:
-- assign processes to be executed
by the processor over time,
– in a way that meets system objectives, such as
response time,
throughput, and
processor efficiency.
Scheduling Objectives
• The scheduling function should
– Allocate time fairly among processes
– Prevent starvation of a process
– Use the processor efficiently
– Have low overhead
– Prioritize processes when necessary (e.g. real time
deadlines)
Types of Scheduling
a new process
part of the swapping
function
a process
Two Suspend States
• Remember this diagram from Chapter 3
Scheduling and
Process State Transitions
Nesting of Scheduling Functions
Queuing Diagram (Simplified)
Long-Term Scheduling
• Determines which programs are admitted to the
system for processing
– May be first-come-first-served
– Or based on criteria such as priority, I/O requirements
or expected execution time
• Controls the degree of multiprogramming
• More processes, smaller percentage of time
each process is executed
Medium-Term Scheduling
• Part of the swapping function
• Swapping-in decisions are based on
- the need to manage the degree of multiprogramming
- the resources (memory, I/O) the processes require
• Typically, memory is the limiting resource and the memory
manager acts as the medium-term scheduler
Short-Term Scheduling
• Known as the dispatcher
• Executes most frequently
• Must be very fast
• Invoked when an event occurs
–
–
–
–
Clock interrupts
I/O interrupts
Operating system calls
Signals (e.g., semaphores)
Roadmap
• Types of Processor Scheduling
• Scheduling Algorithms
• Traditional UNIX Scheduling
Aim of Short Term Scheduling
• Main objective is to allocate processor
time to optimize certain aspects of system
behaviour.
• A set of criteria is needed to evaluate the
scheduling policy.
Short-Term Scheduling Criteria:
User vs System
• User-oriented
– Response Time
• Elapsed time between the submission of a request
until there is output.
• System-oriented
– Effective and efficient utilization of the
processor
Short-Term Scheduling Criteria:
Performance vs non-performance
• Performance-related
– Quantitative, easily measured
– e.g. response time and throughput
• Non-performance related
– Qualitative
– Hard to measure
Interdependent Scheduling Criteria
Interdependent Scheduling Criteria (cont.)
Use of Priorities for Scheduling
• Scheduler will always choose a process
of higher priority over one of lower priority
• Have multiple ready queues, one for each
level of priority
Priority Queuing (Simplified)
highest
lowest
Starvation
• Problem with pure priority scheduling:
– Lower-priority may suffer starvation if there is a
steady supply of high priority processes.
• Solution
– Allow a process to change its priority based on its age
or execution history
Alternative Scheduling Policies
FCFS: First-Come-First-Served
RR: Round Robin
SPN: Shortest Process Next
SRT: Shortest Remaining Time
HRRN: Highest Response Ratio Next
Each policy has its own selection function, which determines
the process to be selected next for execution.
Selection Function
• Determines which process is selected next for execution
• May be based on priority, resource requirements, or the
execution characteristics
• Important quantities of execution characteristics:
• w = time spent waiting in system so far
• e = time spent in execution so far
• s = total service time required by the process,
including e (estimated or supplied by the user)
Decision Mode
• Specifies the instants in time at which the
selection function is exercised.
• Two categories: (see next slide)
– Nonpreemptive
– Preemptive
Nonpreemptive vs Premeptive
• Non-preemptive
– Once a process is in the running state, it will continue
until it terminates or blocks itself for I/O
• Preemptive
– Currently running process may be interrupted and
moved to ready state by the OS
– Preemption may occur when new process arrives,
on an interrupt, or periodically.
Alternative Scheduling Policies
The process that has spend the most
time waiting in the system is selected
The process that will take the least
service time waiting is selected
The longer the wait, the higher
the ratio, the sooner to be chosen
This accounts for aging
Focus on the time that has been spent in execution so
far. Penalize processes that has been running longer
Process Scheduling Example
• Example set of
processes, consider
each a batch job
– Service time represents total execution time
First-Come-First-Served
• Each process joins the Ready queue
• When the current process ceases to execute, the
process in the Ready queue the longest is selected
Time of arrival : 0
Arrivals:
A
2
B
4
C
6
D
8
E
3
6
4
5
2
Pr. that finishes:
Finish time:
Turnaround time:
A
3
3
B
9
7
C
13
9
D
18
12
E
20
12
First-Come-First-Served
• A short process may have to wait a very
long time before it can execute
• Favors CPU-bound processes
– FCFS performs better for long processes than short
ones
– I/O processes have to wait until CPU-bound process
completes
– Convoy effect: All other processes wait for the big
process to finish
Round Robin
• Uses preemption based on a clock
– also known as time slicing, because each process is
given a slice of time before being preempted.
Time line:
0
5
10
15
20
The only process
in the system
Pr. that finishes:
Finish time:
Turnaround time
A
4
4
E C B D
15 17 18 20
7 13 16 14
Round Robin
• Clock interrupt is generated at periodic
intervals
• When an interrupt occurs, the currently
running process is placed in the ready
queue
– Next ready job is selected
Effect of Size of
Preemption Time Quantum
Short quantum:
Short processes finish execution relatively quickly
Processing overhead is high
Very short quanta should be avoided
Long quantum:
Favors CPU-bound processes
Very long quanta behaves like FCFS
Useful quide:
Time quantum should be slightly greater than
the time required for a typical interaction
Effect of Size of
Preemption Time Quantum
In this case, most processes will require
at least two time quanta
‘Virtual Round Robin’
• A refinement to the R.R. scheduling
• Avoids the unfairness of R.R.
Get preference over
the main ready queue
When a process is blocked
for I/O, it joins the I/O queue
as usual
Shortest Process Next
• Non-preemptive policy
• Process with shortest expected
processing time is selected next
• Short process jumps ahead of
longer processes
Time line:
Arrivals:
0 1 2 3 4 5 6 7 8 9 10 …
A
B
C
D
E
Finishes:
Finish time:
Turnaround time:
A
3
3
B
9
7
E
11
3
C
15
11
D
20
14
Shortest Process Next
• Advantage: Performance is significantly improved in
terms of response time.
• Disadv: Predictability of longer processes is reduced
• Disadv: Possibility of starvation for longer processes
• Difficulty: need to know / estimate the required
processing time of each process.
• If estimated time for process not correct, the operating
system may abort it
Calculating Program ‘Burst’
• Where:
– Ti = processor execution time for the i-th
instance (turn) of this process
– Si = predicted value for the i-th instance
– S1 = predicted value for first instance;
not calculated
Sn+1 =(1/n) (T1 +T2 + … +Tn)
Sn =(1/(n-1)) (T1 +T2 + … +Tn-1)
Sn+1=(1/n) Tn + ((n-1) / n) Sn
Calculating Program ‘Burst’
• Where:
– Ti = processor execution time for the i-th
instance (turn) of this process
– Si = predicted value for the i-th instance
– S1 = predicted value for first instance;
not calculated
equal-weight averaging
Sn+1 =(1/n) (T1 +T2 + … +Tn)
Sn =(1/(n-1)) (T1 +T2 + … +Tn-1)
Sn+1=(1/n) Tn + ((n-1) / n) Sn
Exponential Averaging
Typically, we would like to give greater weight to more recent
instances
A common technique for predicting a future value on the basis of a
time series of past values is exponential averaging
0<α<1
k
n
Sn+1 = αTn + (1-α)αTn-1 + … + (1-α) αTn-k + … + (1-α) S1
for = 0.8
Sn+1 = 0.8Tn + 0.16Tn-1 + 0.032Tn-2 + 0.0064Tn-3 + …
Exponential Smoothing
Coefficients
Use Of Exponential
Averaging
Use Of
Exponential Averaging
Shortest Remaining Time
• Preemptive version of shortest process next policy
• Must estimate processing time and choose the
shortest
B is preempted at t = 4, due to R(B) = 5
R(C) = 4
the arrival of C, which has
shorter remaining time
R(B) = 5
R(D) = 5
R(E) = 2
R(B) = 5
R(D) = 5
Highest Response Ratio Next
• Choose next process with the greatest ratio
Non-preemptive
R(C) = (5+4) / 4 = 2.25
R(D) = (3+5) / 5 = 1.60
R(E) = (1+2) / 2 = 1.50
R(D) = (7+5) / 5 = 2.40
R(E) = (5+2) / 2 = 3.50
Feedback Scheduling
• If we have no indication of the
relative length of the processes,
then none of SPN, SRT, HRRN
can be used
• Use feedback scheduling to
penalize jobs that have been
running longer
RR (q=1)
RR (q=2)
• Preemptive at time quantum
FCFS
• Favors newer, shorter processes
over older, longer processes
lowerst priority queue
Also called multilevel feedback
Feedback Performance
• Variations exist, simple version pre-empts
periodically, similar to round robin
– But can lead to starvation
Performance Comparison
• Queuing Analysis - Poisson arrivals &
exponential service time assumed
Observation: Any scheduling discipline that chooses
the next item to be served independent of service time
obeys the relationship:
Formulas
Overall Normalized Response Time
Normalized Response
Time for Shorter Process
Normalized Response
Time for Longer Processes
Normalized Turnaround Time
Fair-Share Scheduling
• User’s application runs as a collection of
processes (threads)
• User is concerned about the performance of the
application
• Need to make scheduling decisions based on
process sets, instead of individual processes
Fair-Share Scheduler
Roadmap
• Types of Processor Scheduling
• Scheduling Algorithms
• Traditional UNIX Scheduling
Traditional UNIX Scheduling
• Multilevel feedback using round robin within
each of the priority queues
• If a running process does not block or complete
within 1 second, it is preempted
• Priority is based on process type and execution
history.
Scheduling Formula
Bands
• Priorities are recomputed once per second
• Base priority divides all processes into fixed
bands of priority levels
–
–
–
–
–
Swapper (highest)
Block I/O device control
File manipulation
Character I/O device control
User processes (lowest)
Example of Traditional UNIX Process Scheduling