Non Preemptive Scheduling Limitations

1
Non Pre-emptive Scheduling Limitations

• The Currently Running process may dominate the
  CPU causing all the Other Processes to starve.
• Will have a profound effect on the Response Times
  of ALL the processes except the very first
  Process i.e. the process which gets to run first.
  
• Except for FCFS and Priority Based Scheduling (
  Priority may be pre-fixed Deadline ) difficult to
  implement.
• The Number of process chunks in the READY Queue
  is NOT fixed and new chunks may arrive
  continuously hence a single READY Queue may
  become too large to manage.

2
Pre-emptive Scheduling Strategies Adopted - 1

• Time share the CPU among the Process Chunks i.e.
  use a Round Robin Scheduling Policy with each
  process chunk allocated a pre-fixed Time Slice
  supported by a Dedicated Interval Timer.
• Stick to pre-assigned Priority based scheduling
  with a fixed number of Pre-Assigned Priorities.
• Use several Ready Queues one for each Priority
  Level.
• CPU Time is distributed in a Round Robin Fashion
  among the Different Priority Levels with a
  Pre-Fixed Time Quantum allotted to each Priority
  Level. Hence each Ready Queue possesses two (2)
  associated attributes , The Priority Level (Pr)
  and the associated Time Slice (T).
• Within one Ready Queue i.e. for the Processes
  having same priority the Process Chunks are
  serviced also in a Round Robin fashion based on
  their Arrival ( normally FCFS).

3
Round Robin Policyfor One Ready Queue

• The Short Term Scheduler maintains a Ready Queue
  of PCBs as well as a List of PCBs of ALL Blocked
  and Swapped Out Processes.
• The PCB of a newly created process is added to
  the end of the Ready Queue (FCFS policy) . The
  PCB of a terminated process is removed from all
  the associated Lists.
• The Scheduler always selects the Process Chunk
  whose PCB is at the Head of the Ready Queue.
• This process is allowed to run for its prefixed
  time slice.
• On completion of this Time slice , a Timer
  Interrupt is generated causing pre-emption of
  that running process and puts its PCB at the
  end of the Ready Queue.
• The Event Handler / Interrupt Handler performs
  the following actions
• (a) Whenever the currently Running
  Process makes any I/O request , or is swapped out
  of memory its PCB is removed from the Ready Queue
  and inserted at the back of the list of Blocked
  and Swapped Out Processes.
• (b) Whenever the I/O operation awaited
  by a process finishes , or the process is swapped
  in , its PCB is removed from the Blocked and
  Swapped Out List and put at the end of the Ready
  Queue.

4
Pre-emptive Priority based Scheduling Multiple
Ready Queues , Assumptions

• As soon as any new process is brought in the
  memory by the Medium Term Scheduler, the
  following events take place in sequence
• a) The currently running process if
  unfinished is brought back at
• the Front of its respective Ready
  Queue.
• b) The Enqueuer is brought into execution
  (running state) after
• a context switch which brings the
  newly arrived process at
• the back of its respective Ready
  Queue.
• c) Next the Dispatcher is invoked in order
  to select the process
• based on the scheduling policy
  illustrated before i.e.
• remaining part of the Time Quantum
  allocated to the
• previously running process is LOST.
• d) However , once ALL process Chunks has
  arrived (rare occurrence) then perfect Round
  Robin Policy is followed among the different
  Priority Levels and minimum Time allotted to
  each process is one Time Quantum as allocated to
  that priority level.

5
Timer generated Interrupt ( an essential feature
of Pre-empted Scheduling Policy / Multitasking - 1

• Implemented by an interrupt timer device The
  code outline
• A. Setting the Time Interval by the user / O.S
• Set Interval ( lt programmable value gt )
• / begin set interval done as a part of
  Context Switch /
• k lt Pre-Fixed Time Quantum gt / Global
  variable /
• Interrupt_count k / Global variable /
• / end set interval /

6
Timer generated Interrupt ( an essential feature
of Pre-empted Scheduling Policy / Multitasking - 2

• Interval-Timer ( ) / will cause the
  Programmable Interrupt Timer Device to issue on
  interrupt once on every k clock ticks /
• / begin repeatedly invoked within the
  running process/
• while (Interrupt_count gt 0)
• / begin while /
• Interrupt_count Interrupt_count
  1
• / end while /
• Int_Request TRUE / Issue
  Interrupt on Reaching 0 Count/
• / end /
• In actual terms one can have a hardware support
  available in a programmable interrupt controller
  (like APSC in Pentium)

7
Pre-emptive Priority Based Scheduling Case
Study- 1

• Consider the following Process mix
• (Lower Number implies Higher Priority )
• --------------------------------------------------
  ----------------------------------
• Process Arrival Time Service time
  Pre-Assigned
• Chunk No. in units in units
  Priority
• --------------------------------------------------
  ----------------------------------
• P0 80 200
  1
• P1 10 400
  3
• P2 130 250
  1
• P3 60 500
  4
• P4 200 300
  2
• --------------------------------------------------
  -----------------------

8
Pre-emptive Priority based Scheduling Case Study
- 2

• Consider the Process mix to be scheduled in order
  of arrival.
• --------------------------------------------------
  ----------------------------------
• Process Arrival Time Service time
  Pre-Assigned
• Chunk No. in units in units
  Priority
• --------------------------------------------------
  ----------------------------------
• P1 10 400
  3
• P3 60 500
  4
• P0 80 200
  1
• P2 130 250
  1
• P4 200 300
  2
• --------------------------------------------------
  -----------------------

9
Pre-emptive Priority based Scheduling Round
Robin Policy Outline

• --------------------------------------------------
  ----------------------------------
• Pre-Assigned Process Process
  Pre-Assigned Time
• Priority Chunks Types
  Slice
• --------------------------------------------------
  ----------------------------------
• 1 P0, P2
  Short, CPU Bound Short, Many
• 2 P4
  Medium Sized Medium , Medium
• 3 P1
  Slightly Higher Size Large, Less
• 4 P3
  Long Largest, Minimum
• --------------------------------------------------
  -----------------------

10
Pre-emptive Priority based Scheduling Round
Robin Policy Outline

• Scheduler takes up processes from each Priority
  Queue in Turn i.e. in a Round Robin Fashion .
• Among all the processes in a Particular Priority
  the Servicing Policy is FCFS.
• No particular Priority should dominate the CPU
  time thereby forcing the other Processes in
  remaining Priority levels to starve.
• Each priority level should be assigned a Fixed
  Percentage Quota of CPU Time.

11
Pre-emptive Priority based Scheduling Case
Study - 3

• Four different Levels of Priority (1..4) with
  lower number implying Higher Priority.
• Servicing Policy CPU goes round robin in the
  following manner
• Within the same queue, round robin based on the
  First Come First Serve (FCFS) basis.
• CPU moves among the different queues with
  prefixed number of Time Slice allocation as
  depicted below
• Priority Level CPU Time(Approx)
  Implication
• 1. 40
  4 Time slices of 10
• 2. 30
  2 Time slices of 20
• 3. 20
  1 Time slice of 30
• 4. 10
  1 Time slice of 40
• N.B Higher Priority Process Chunks gets more
  number of Smaller Sized Chunks .

12
Pre-emptive Priority based Scheduling Policy
Outline

• CPU starts from Priority Level 1 then cycles
  through the remaining Priority Levels.
• If any Priority Level is empty then any remaining
  Time Slice allotted for that level is NOT used up
  and Scheduler goes to the next lower priority
  level.
• If any of the processes at a priority level
  remains unfinished even after ALL the allotted
  Time Slice to that priority level has been used
  up then it may do the two following things
• a) It MUST wait for its turn in
  the NEXT Cycle Non Feedback Queues
• b) All the unfinished processes in
  any Priority Level I is put at the back end of
  the next Priority Level Queue hoping to get any
  unused Time slice in that Priority level
  Feedback Queue .
• 4. If after allocating any time slice to the
  dispatched process, any new process arrives in
  the Ready Queue then that running user process
  gets pre-empted, the left over of the allotted
  time slice to that process is LOST and Enqueuer
  gets invoked.

13
Pre-emptive Priority based Scheduling 5 Non
Feedback Multiple Ready Queues

• Various Queue Structures

P2 250 Units Service P0 200 Units Service
Arrived at 130 Arrived at 80
Priority Level 1 Queue Time slice 10 Units X
4
P4 300 Units Service Arrived at 200
Priority Level 2 Queue Time slice 20 Units X
2
P1 400 Units Service Arrived at 10
Priority Level 3 Queue Time slice 30 Units X
1
P3 500 Units Service Arrived at 60
Priority Level 4 Queue Time slice 40
Units X 1
14
Non Pre Emptive Scheduling Policies Common
Assumptions - 1

• Assume Context switching time 1 unit (fixed).
• Dispatcher Execution time 3 units
  (fixed).
• Enqueuer Execution time 2 units
  (fixed).

15
Pre-emptive Priority based Scheduling 6 All
Types of Multiple Ready Queues

• Dispatcher Enqueuer Block Structures

Context Switch Dispatcher Context Switch 1 Unit
3 Units 1 Unit
Context Switch Enqueuer Context Switch 1 Unit
2 Units 1 Unit

16
Pre-emptive Priority based Scheduling 7 Non
Feedback Multiple Ready Queues

• The GANTT Chart

EnQ P1 (Pr3)
P1(Pr 3) 30 Done 370 Left
EnQ P3 (Pr4)
DiS P1
P1(Pr 3) 7 Done 363 Left
DiS P3
DiS P1
10 13 18
48 53 60
63 68
EnQ P0 (Pr1)
DiS P0
P0(Pr 1) 10 Done 190 Left
DiS P0
P3(Pr 4) 12 Done 488 Left
P0(Pr 1) 10 Done 180 Left
68 80 83
88 98 103
113
17
Pre-emptive Priority based Scheduling 8 Non
Feedback Multiple Ready Queues

• The GANTT Chart Contd.

P0(Pr 1) 30 Done 170 Left
DiS P0
DiS P1
DiS P0
EnQ P2 (Pr1)
P0(Pr 1) 40 Done 160 Left
113 118 128 133
136 146 151
P1(Pr 3) 66 Done 334 Left
P3(Pr 4) 26 Done 474 Left
DiS P3
EnQ P4 (Pr2)
DiS P0
151 181 186
200 203 208
18
Pre-emptive Priority based Scheduling 9 Non
Feedback Multiple Ready Queues

• The GANTT Chart Contd.

P0(Pr 1) 50 Done 150 Left
DiS P2
DiS P0
P2(Pr 1) 10 Done 240 Left
208 218 223
233 238
P2(Pr 1) 20 Done 230 Left
DiS P2
DiS P4
P2(Pr 1) 60 Done 140 Left
238 248 253
263 268
19
Pre-emptive Priority based Scheduling 10 Non
Feedback Multiple Ready Queues

• The GANTT Chart Contd.

DiS P4
DiS P1
P4(Pr 2) 20 Done 280 Left
P4(Pr 2) 40 Done 260 Left
P1(Pr 3) 30 Done 334 Left
268 288 293
313 318
348
DiS P0
DiS P3
P3(Pr 4) 40 Done 437 Left
P0(Pr 1) 70 Done 130 Left
DiS P2
348 353 393
398 408 413
20
Scheduling Policiesincluded in the Syllabus
Self Study

• Scheduling Policy adopted in UNIX Fair Share
  Policy
• Scheduling Policy adopted in LINUX
• Real Time Scheduling
• Non Real Time Scheduling
• 3. Scheduling Policy adopted in WINDOWS XP.