SCHEDULING

1
SCHEDULING

• Kshama Desai
• Bijal Shah
• Kishore Putta
• Kashyap Sheth

2
SCHEDULING

• Scheduling manages CPU allocation among a group
  of ready processes and threads.
• Concept Of Scheduling
• Based on mechanism of context switching.
• Requirement of shared resources by processes
• and threads.

3
Multiprogramming

• Allows many processes to load and
  time-multiplexing of the threads.
• Per CPU, a single thread executes at any given
  time.
• Need for threads to perform concurrent I/O
  operations.
• High CPU multiplexing rate gives the impression
  of concurrent execution of processes / threads.

4
Scheduling Policies and Mechanism

• Scheduling Policy determines the thread to which
  the CPU should be allocated and its time of
  execution.
• Scheduling mechanisms determine the way process
  manager multiplexes the CPU and the state of the
  thread.
• Thread Scheduling
• Scheduler determines the transition of thread
  from the running state to the ready state.

5

• Steps in Thread Scheduling
• Wait in Ready list for CPU allocation.
• On CPU allocation, state of the thread changes
  from ready to running state.
• During execution, thread waits in resource
  managers pool on subsequent request for an
  unavailable resource (if needed).
• After complete execution, it leaves the CPU, or
  returns to ready state.

6

• Simpler Processor Scheduling Model

Preemption / Yield
Done
New Thread
Ready List
Scheduler
CPU
Job
Job
Ready
Job
Running
Resource Manager
Request
Allocate
Job
Job
Blocked
Resources
7

• Running Thread Ceases CPU use for following
  reasons
• Thread completes execution and leaves the system.
• Thread requests an unavailable resource. State is
  changed to Blocked. On availability of resource,
  state is changed to Ready state.
• Thread voluntarily decides to release CPU
• System Preempts the thread by changing its state
  from Running to Ready

8
Scheduling Mechanisms

• Scheduler is implemented in the hardware as well
  as in
• the OS software.
• Scheduler implements 3 mechanisms
• a) Enqueuer b) Context Switching c) Dispatch
  element .
• Sequence of Schedule Mechanism
• Enqueuer mechanism places the process into Ready
  state and decides its priority.
• Context switch element helps to remove a process
  from CPU and bring in a new process.
• Dispatch element allocates CPU to the new
  incoming process.

9
CPU SHARING

• Two types of Schedulers
• Voluntary CPU sharing (Non-preemptive scheduler)
• Involuntary CPU sharing (Preemptive scheduler)
• Voluntary CPU Sharing
• Hardware includes special yield machine
  instruction.
• Address of the next instruction is saved in a
  designated memory location.

10

• Disadvantage of yield instruction
• Failure of periodic access of a process to yield
• instruction blocks other processes from
• executing CPU, until the process exits or
• requests a resource.
• Solution of this problem is if the system gives
  self-interrupt.

11

• Involuntary CPU sharing
• Incorporates interval timer device.
• Interval of time is decided by the system
  programmer.
• Internal timer invokes the scheduler periodically
• Advantage
• Process executing in an infinite loop cannot
  block
• other processes from running (executing ) the
• CPU.

12
Performance

• Scheduler affects the performance of a
  multiprogrammed CPU to a great extent.
• Imbalanced process selection for CPU execution by
  a scheduler will lead to Starvation. This leads
  to the need for strategy selection of schedulers

13
Strategy Selection

• Criteria to select a scheduling strategy depends
  on the goals of the OS.
• Scheduling algorithms for modern OS use internal
  priorities.
• Involuntary CPU sharing have time quantum /
  timeslice length. Optimal schedule can be
  computed, provided there is no new entry in the
  ready list while prior present processes are
  being served.

14
Model For Scheduling

• Performance metrics to compare scheduling
  strategies
• Service Time
• Wait time
• Turnaround time.
• Process model and the metrics are used to compare
  the performance characteristics of each
  algorithm.
• The general model must fit each specific class of
  the OS
• environment.
• Turnaround time is the most critical performance
  metric in batch multiprogrammed system.
• Time sharing systems focus on single phase of
  thread execution.

15

• Simpler Processor Scheduling Model

Preemption / Yield
Done
New Thread
Ready List
Scheduler
CPU
Job
Job
Ready
Job
Running
Resource Manager
Request
Allocate
Job
Job
Blocked
Resources
16
Types of Scheduling Methods

• Non- Pre-emptive
• Pre-emptive

17
Non-Preemptive

• Non-Pre-emptive
• Once a process enters the running state, it is
  not removed from the processor until it has
  completed its service time.
• There are 4 Non-Pre-emptive strategies
• First Come First Serve (FCFS)
• Shortest Job Next (SJN)
• Priority Scheduling
• Deadline Scheduling

18
Pre-emptive

• Based on prioritize computation.
• Process with highest priority should always be
  the one using the CPU.
• If a process is currently using the CPU and a new
  processor with higher priority enters the ready
  list, the process on the processor should be
  removed and returned to the ready list until it
  is once again the highest priority process in the
  system.

19
First Come First Serve (FCFS)

• Processes are assigned the CPU in the order they
  request it.
• If a running process blocks, the first process on
  the queue is run the next. When this blocked
  process becomes ready, like a newly arrived job,
  it is put on the end of the queue.

20
FCFS Example

• Example load

21
FCFS Example

• Turnaround Time
• Average of finishing times
• TTRnd (P0) t(P0) 350
• TTRnd (P1) t(P1) TTRnd (P0) 125 350
  475
• TTRnd (P2) t(P2) TTRnd (P1) 475 475
  950
• TTRnd (P3) t(P3) TTRnd (P2) 250 950
  1200
• TTRnd (P4) t(P4) TTRnd (P3) 75 950
  1275
• Average turnaround time
• TTRnd (350 475 950 1200 1275)/5
• 4250/5
• 850

22
FCFS Example

• Wait Time
• Average time in wait before first run
• From the Gantt chart
• W(P0) 0
• W(P1) TTRnd (P0) 350
• W(P2) TTRnd (P1) 475
• W(P3) TTRnd (P2) 950
• W(P4) TTRnd (P3) 1200
• Average wait time
• W (0 350 475 950 1200)/5 2975/5
  595

23
Advantages and Disadvantages of FCFS

• Advantages
• Easy to understand and easy to program
• It is fair
• Requires only single linked list to keep track of
  all ready processes
• Disadvantages
• Does not perform well in real systems
• Ignores the service time request and all other
  criteria that may influence the performance with
  respect to the turnaround or waiting time.

24
Shortest Job Next (SJN)

• Here each process is associated with its length.
• The ready queue is maintained in the order of
  increasing job lengths.
• When current process is done, pick the one at the
  head of the queue and run it.

25
SJN Example

• Example load

Average turnaround time TTRnd (800 200
1275 450 75)/5 2800/5 560 Average wait
time W (450 74 800 200 0)/5 1525/5
305
26
Advantages and Disadvantages of SJN

• Advantages
• Minimizes wait time
• Disadvantages
• Long running threads may starve
• NOTE
• It requires prior knowledge of service time
• Optimal only when all the jobs are available
  simultaneously.

27
Priority Scheduling

• Each process is assigned a priority and the
  runnable process with highest priority is allowed
  to run.
• Priorities can be assigned statically or
  dynamically
• With static priority, starvation is possible
• Dynamic (internal) priority solves the problem of
  starvation

28
Priority Scheduling

• Example load

Average turnaround time TTRnd (1275 375
850 250 925)/5 3675/5 735 Average wait
time W (925 250 375 0 850)/5 2400/5
480
29
Deadline Scheduling

• For hard real-time systems, process must finish
  by a certain time
• Turnaround time and wait times are irrelevant
• We need to know maximum service time for each
  process
• Deadline must be met for each period in a
  processs life

30
Deadline Scheduling Example

• Example load

0 75 200
550
1025 1275
P4
P0
P2
P3
P1
31
Round Robin

• Most widely used scheduling algorithm
• It tries to be fair by equally distributing the
  processing time among all the processes
• When the process uses up its quantum, it is put
  on the end of the ready list

32
How to choose the length of the quantum?

• Setting the quantum length too short causes many
  process switches and lowers the CPU efficiency
• Setting the quantum length too long may cause
  poor response to short interactive requests
• Solution
• Around 20-50 msec is a reasonable compromise.

33
Multiple-Level Queues

• It is an extension of priority scheduling
• The ready queue is partitioned into separate
  queues
• Foreground
• Background
• Scheduling must be done between the queues
• It uses 2 strategies of scheduling
• One to select the queue
• Another to select the process in the queue such
  as Round Robin

34
LINUX Scheduling Mechanism

• Linux threads are kernel threads hence scheduling
  is based on threads.
• It is based on time-sharing techniques.
• Each thread has a scheduling priority
• Default value is 20, but can be altered using the
  nice (value) system call to a value of
  20-value
• Value must be in the range -20 to 19, hence the
  range of priority is between 1 and 40
• Quality of service is proportional to priority.
• The scheduler keeps track of what processes are
  doing and adjust the priorities periodically,
  i.e., the priorities are dynamic

35
Windows NT Scheduling Mechanism

• Windows NT is a pre-emptive multithreading OS.
• Even here the unit of scheduling is thread.
• It uses 32 numerical thread priorities from 1 to
  31 (0 being reserved for system use).
• 16 to 31 for use by time critical operations
• 1 to 15 (dynamic priorities) for program threads
  of typical applications.

36
References

• Nutt, Gary. Operating Systems. Third Edition,
  Pearson Education Inc, 2004.
• Tanenbaum, Andrew. Modern Operating Systems,
  Prentice-Hall Of India Pvt. Ltd.
• http//www.windowsitpro.com/Article/ArticleID/302/
  302.html