Chapter 5: Process Scheduling. Outline Basic Concepts CPU-I/O Burst Cycle CPU Scheduler ...

Chapter 5:

Process Scheduling

OutlineBasic Concepts

CPU-I/O Burst Cycle CPU Scheduler Preemptive Scheduling Dispatcher

Scheduling CriteriaScheduling Algorithms

First-Come, First-Serve Scheduling Shortest-Job-First Scheduling Priority Scheduling Round-Robin Scheduling Multilevel Queue Scheduling Multilevel Feedback-Queue Scheduling

Basic ConceptsCPU: One of the primary computer resourceMaximum CPU utilization obtained with

multiprogrammingSeveral processes are kept in memory

CPU–I/O Burst Cycle – Process execution consists of a cycle of CPU execution and I/O wait.

CPU burst distribution determines how to schedule the processes.

Alternating Sequence of CPU And I/O Bursts

Histogram of CPU-burst Times

Normally a process starts with frequent short CPU bursts and then it stabilizes

and shifts to long but infrequent CPU bursts

I/O bound processes usually have many small CPU bursts

CPU bound programs usually have few but long CPU burstsCPU-bound program

CPU SchedulerSelects from among the processes in memory that

are ready to execute, and allocates the CPU to one of them.Is it the same as Short-term scheduler? … Yes

CPU scheduling decisions may take place when a process:1.Switches from running to waiting state.

2.Switches from running to ready state.

3.Switches from waiting to ready.

4.Terminates.(Own will)

Scheduling Scheme given at 1 and 4 is nonpreemptive (Cooperative).

All other scheduling is preemptive.

Preemptive vs Non PreemptiveScheduling

Non Preemptive scheduling If a process is being executed then it keeps the CPU

until one or four conditions of previous slide occur.MS Windows 3.x which was sort of multitasking system

used this methodNot at all efficient.

PreemptiveA process in running state can be brought back to ready

state without any of the conditions of previous slide being met.

Can you guess how it is done?

Problems with preemptive

Although better than non-preemptive still suffers from synchronization problems

E.g.Process P and Q share some data (can be

memory location or a file)P does some calculations changes some of the

data but before it finishes its time expiresQ reads the data but is the data valid?

Dispatcher

Dispatcher module gives control of the CPU to the process selected by the short-term scheduler; this involves:switching contextswitching 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.Should be minimum.

Scheduling CriteriaFollowing criteria is kept under consideration while designing a scheduler

CPU utilization – Keep the CPU as busy as possibleThroughput – # of processes that complete their execution

per unit time (Given Time). Turnaround time – Amount of time to execute a particular

process (Sum of ready, waiting, running, loading times,I/O)Waiting time – Amount of time a process has been waiting

in the ready queueResponse time – Amount of time it takes from when a

request was submitted until the first response is produced, not output (for interactive environment) Time it takes to start responding Not the time it takes to output the response

Optimization Criteria

So the best CPU Scheduling algorithm is which …

Maximizes CPU utilizationMaximizes throughputMinimizes turnaround time Minimizes waiting time Minimizes response time

First-Come, First-Served (FCFS) Scheduling

Process Burst Time

P1 24

P2 3

P3 3

Suppose that the processes arrive in the order: P1 , P2 , P3

The Gantt Chart for the schedule is:

Waiting time for P1 = 0; P2 = 24; P3 = 27

Average waiting time: (0 + 24 + 27)/3 = 17

P1 P2 P3

24 27 300

FCFS Scheduling (Cont.)Suppose that the processes arrive in the order

P2 , P3 , P1 .

The Gantt chart for the schedule is:

Waiting time for P1 = 6; P2 = 0; P3 = 3

Average waiting time: (6 + 0 + 3)/3 = 3Much better than previous case.CPU bound processes will get and hold the CPUConvoy effect short process behind long process (Bad

CPU utilization)

P1P3P2

63 300

Shortest-Job-First (SJR) SchedulingAssociate with each process the length of its next CPU

burst. Use these lengths to schedule the process with the shortest time.

Two schemes: nonpreemptive – once CPU given to the process it cannot be preempted

until it completes its CPU burst. preemptive – if a new process arrives with CPU burst length less than

remaining time of current executing process, preempt. This scheme is know as the Shortest-Remaining-Time-First (SRTF).

SJF is optimal – gives minimum average waiting time for a given set of processes.

Shortest-next-cpu-burst algorithm

Process Burst Time

P1 6

P2 8

P3 7

P4 3

SJF

Average waiting time = (3+16+9+0)/4=7 FCFS=10.25

Example of Non-Preemptive SJF

P4P1 P3

93 240

P2

8

16

Shortest-Job-First (SJR) SchedulingGood for long-term Scheduler in batch systems

User specifies process time during job submissionLower value may mean faster responseToo low a value will cause a time-limit-exceeded error

Difficult to implement at the level of short-term CPU scheduler

Determining Length of Next CPU BurstCan only predict/ estimate the length of next CPU burst.Can be done by using the length of previous CPU

bursts, using exponential averaging.

Examples of Exponential Averaging =0

n+1 = n

Recent history does not count.

=1 n+1 = tn

Only the actual last CPU burst counts.

Since both and (1 - ) are less than or equal to 1, each successive term has less weight than its predecessor.

n+1 =αtn +(1- α) n

tn: length of the nth CPU burst

n+1 : Our predicted Value

α: weight of recent & past history

0<= α<=1

Prediction of the Length of the Next CPU Burst

Initial guess for T(1) = 10 ms Actual 6

Determination of 2 Cycle based on actual (6) and last guess (10)

T(2) = 0.5 x 6 + 0.5 x 10 = 8 ms

Determination of 3 cycle based on actual(4) and last guess (8)

T(3) = 0.5 x 4 + 0.5 x 8 = 6 ms

… Now you can do it

For example: Suppose a process p is given a default expected burst length of 5 time units. When it is run, the actually burst lengths are 10,10,10,1,1,1 (although this information is not known in advance to any algorithm). The prediction of burst times for this process works as follows.

Let e(1) = 5, as a default value.

When process p runs, its first burst actually runs 10 time units, so, a(1) = 10.

e(2) =0.5 * e(1) * 0.5 + a(1) = 0.5 * 5 + 0.5 * 10 = 7.5 This is the prediction for the second cpu burst

The actual second cpu burst is 10. So the prediction for the third cpu burst is:

e(3) = 0.5 * e(2) * 0.5 + a(2) = 0.5 * 7.5 + 0.5 * 10 = 8.75 e(4) = 0.5 * e(3) * 0.5 + a(3) = 0.5 * 8.75 + 0.5 * 10 = 9.38,

So, we predict that the next burst will be close to 10 (9.38) because the recent bursts have been of length 10. At this point, it happens that the process starts having shorter bursts, with a(4) = 1, so the algorithm gradually adjusts its estimated cpu burst (prediction)

e(5) = 0.5 * e(4) * 0.5 + a(4) = 0.5 * 9.38 + 0.5 * 1 = 5.19 e(6) = 0.5 * e(5) * 0.5 + a(5) = 0.5 * 5.19 + 0.5 * 1 = 3.10 e(7) = 0.5 * e(6) * 0.5 + a(6) = 0.5 * 3.10 + 0.5 * 1 = 2.05

Example of Preemptive SJF(Shortest Remaining Time first)

Process Arrival Time Burst Time

P1 0 8

P2 1 4

P3 2 9

P4 3 5SJF (preemptive)

Average waiting time = 6.5Non-Preemptive Average waiting time = 6.5

P1 P2

51 170

P1

10

P4P3

26

Priority SchedulingA priority number (integer) is associated with each processThe CPU is allocated to the process with the highest

priority (smallest integer highest priority). Preemptive Non-preemptive

SJF is a priority scheduling where 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.

If there are n processes in the ready queue and the time quantum is q, then each process gets 1/n of the CPU time in chunks of at most q time units at once. No process waits more than (n-1)q time units.

Performance q large FIFO q small q must be large with respect to context switch,

otherwise overhead is too high.

Example of RR with Time Quantum = 20

Process Burst Time

P1 53

P2 17

P3 68

P4 24

The Gantt chart is:

Typically, higher average turnaround than SJF, but better response.

P1 P2 P3 P4 P1 P3 P4 P1 P3 P3

0 20 37 57 77 97 117 121 134 154 162

Time Quantum and Context Switch Time

Remember: Context switch takes time

If context switching time is approx 10 percent of the time quantum, then about 10 percent of the CPU time will be spent in context switching

Most modern systems have time quanta ranging from 10 to 100 milliseconds

Context switch typically takes less than 10 microseconds

Turnaround Time Varies With The Time Quantum

Average Turnaround time doesn’t necessarily improve as the time-quantum size increasesAverage Turnaround time doesn’t necessarily improve as the time-quantum size increases))

80% of CPU bursts should be shorter than the time quantum

Multilevel QueueReady queue is partitioned into separate queues:

foreground (interactive)background (batch)

Each queue has its own scheduling algorithm, foreground – RRbackground – FCFS

Scheduling must be done between the queues.Fixed priority scheduling; (i.e., serve all from foreground then from

background). Possibility of starvation.Time slice – each queue gets a certain amount of CPU time which it

can schedule amongst its processes; i.e., 80% to foreground in RR20% to background in FCFS

Multilevel Queue Scheduling

Multilevel Feedback Queue

A process can move between the various queues; aging can be implemented this way.

Multilevel-feedback-queue scheduler defined by the following parameters:number of queuesscheduling algorithms for each queuemethod used to determine when to upgrade a processmethod used to determine when to demote a processmethod used to determine which queue a process will

enter when that process needs service

Example of Multilevel Feedback QueueConsider a case if we have three queues:

Q0 – time quantum 8 milliseconds

Q1 – time quantum 16 milliseconds

Q2 – FCFS

SchedulingA new job enters queue Q0 which is served FCFS. When it

gains CPU, job receives 8 milliseconds. If it does not finish in 8 milliseconds, job is moved to tail of queue Q1.

At Q1 job is again served FCFS and receives 16 additional milliseconds. If it still does not complete, it is preempted and moved to tail of queue Q2.

Multilevel Feedback Queues

Outline Basic Concepts


Scheduling Criteria Scheduling Algorithms


Multi-Processor Scheduling

Multiple-Processor SchedulingCPU scheduling more complex when multiple

CPUs are available.Consider Homogeneous processors within a

multiprocessor system.Load sharing Asymmetric multiprocessing – only one

processor accesses the system data structures, alleviating the need for data sharing.

SMP: Each processor is self schedulingMay have separate ready queues

Multiple-Processor Scheduling –Process Affinity

Cache memory?The Data most recently accessed populates it

Process migration to another process Invalidation & Repopulation

Some OS attempt to keep a process running on the same processor –Affinity

Soft Affinity: Attempt to keep a process running on the same processor but not guaranteeing

Hard Affinity: Process cannot migrate to another processor

Multiple-Processor Scheduling –Load Balancing

Utilize the full benefits of multi-processorsLoad balancing: Distribute the workload evenly across all

processors.Necessary where each processors has its own private

queue of eligible processesNot necessary where processors have common run

queueModern OS --private queue or common?Push Migration: move processes from overloaded to idle

processorsPull Migration: Idle processor pulls a waiting process

from a busy processorAny link with Process Affinity?

Multiple-Processor Scheduling Symmetric Multithreading

Multiple logical rather than physical processors to run several threads concurrently (HPT)

Create logical processors on the same physical processor—a view of many Processors to OSEach has its own machine-state registersOwn interrupt handling

Logical CPU

Logical CPU

Physical CPU

Logical CPU

Logical CPU

Physical CPU

System Bus

Multiple-Processor Scheduling Symmetric Multithreading

SMT: a feature provided by hardwareOs needn’t be designed differentlyPerformance gains –If OS is aware.

But why?**Scheduler**

Outline Basic Concepts


Scheduling Criteria Scheduling Algorithms


Multi-Processor Scheduling OS Examples –Self Study Algorithm Evaluation

Algorithm Evaluation How do we select the best algorithm?

Defining a criteria CPU Utilization, response time, throughput Criteria may include several measures

e.g. Maximize CPU utilization |and| Response time is 1 e.g. Maximize Throughput |and| turnaround time

Analytic Evaluation Method The algorithm and some system workload are used to

produce a formula or number which gives the performance of the algorithm for that workload.

Deterministic modeling Queuing models Simulations Implementation

Deterministic Modeling Predetermined workload –defines performance of

each algorithm for that workload. Use of Gantt charts.

Simple and fast Exact numbers for comparison Answers apply only for the cases considered. Performance figures may not be true in general Suitable: If we are running the same program over

and over again. We can easily select an algorithm. Over a set of examples certain trends can be

analyzed, e.g. If all the processes arrive at time 0, the SJF policy will always result in min waiting time If the Arriving time is different the SJF may not always

result in min average waiting time.

Process Arrival Time Burst Time

P1 0.0 7

P2 2.0 4

P3 4.0 1

P4 5.0 4

Gantt chart

Average waiting time = (9 + 1 + 0 +2)/4 = 3

P3P2

42 110

P4

5 7

P2 P1

16

P1

Deterministic Modeling

Queuing Modeling What if the processes that are run, vary from day to day? --

Is deterministic method useful? The distributions of CPU and I/O bursts can be determined/

approximated/ estimated. Mathematical formula: probability of a particular CPU burst Arrival times distribution can also be estimated.

Computer system viewed as a network of queues and servers: ready queue, I/O queue, event queues, CPUs, I/O device controllers, etc. e.g. CPU: ready queue, I/O system :device queue

Input: Arrival and service rates Output: CPU utilization, average queue length, average

waiting time, …

Queuing Modeling

Little’s Formula:n = λ* W

wheren = average queue lengthλ = average arrival rateW = average waiting time in a

queue

Queuing ModelingLet the average job arrival rate be 0.5

Algorithm Average Wait Time

W=tw

Average Queue Length(n)

FCFS 4.6 2.3

SJF 3.6 1.8

SRTF 3.2 1.6

RR (q=1) 7.0 3.5

RR (q=4) 6.0 3.0

Queuing Modeling

Complicated mathematics Distributions (uniform, exponential,

etc) for the arrival and departure rates can be difficult to work with

Assumptions may not be accurate Approximation of the real system

Simulation To get more accurate evaluation Workload generated by assuming some

distribution and a random number generator, or by collecting data from the actual system.

The distributions can be defined mathematically or empirically.

Simulation

Simulation

CharacteristicsExpensive: hours of programming and execution time

May be erroneous because of the assumptions about distributions

Implementation Even simulation is of limited accuracy Best way: Code and implement the scheduling

algorithm and see High cost –coding the algorithm Modify the OS to support it Reaction of the users to changing OS

Environment changes: if short processes are given high priority then user may break their larger processes into many short ones

For example: Design of an automatic Interactive/ non-interactive process classifier Algorithm

Chapter 5: Process Scheduling. Outline Basic Concepts CPU-I/O Burst Cycle CPU Scheduler ...

Documents

Transcript of Chapter 5: Process Scheduling. Outline Basic Concepts CPU-I/O Burst Cycle CPU Scheduler ...