1. CISC3595 CPU Scheduling

2. Goal: improve CPU utilization
Basic Concepts
Multiprogramming
Load multiple program into memory, and run them simultaneously
A process (or kernel-level thread) is executed until it must wait (for I/O completion, child exiting…), OS then switches to run another process (or thread)
Goal: improve CPU utilization

3. Short-term scheduling (CPU scheduling)
Process Scheduling
Short-term scheduling (CPU scheduling)
Select one process from ready queue to be executed
Medium-term scheduling, Long-term scheduling
What is ready queue ?
Contains PCB for all processes that lines up waiting for CPU
Not necessarily FIFO queue

4. Process Control Block (PCB): review
Information associated with each process
Process state
Program counter:
address of next instruction to be executed
CPU registers:
Contents of CPU registers
CPU scheduling information
Process priority, …
Memory-management information
Value of base and limit registers, page/segmentation tables (to be studied later)
Accounting information
Amount of CPU time used …
I/O status information
I/O device allocated, list of open files… 4

5. Process Control Block (PCB): review5

6. Context Switch: review
P0, P1 are running concurrently, i.e.,
their executions are interleaved. 6

7. Process Scheduling Queues: review
OS maintain multiple process queues for scheduling purpose:
Job queue – set of all processes in the system
Ready queue – set of all processes residing in main memory, ready and waiting to execute
Device queues – set of processes waiting for an I/O device
Processes migrate among the various queues 7

8. Ready Queue And I/O Device Queues: review8

9. CPU and I/O Bursts
Process execution consists of a cycle of CPU execution and I/O wait
CPU burst:
Data processing, calculation
E.g. sorting, searching
I/O burst:
Read from I/O device
File system access
What about the myCopy program?

10. Histogram of CPU-burst Times
A large number
short CPU burst
Small number of
long CPU burst

11. CPU Scheduler
Selects one process from ready queue, and allocates CPU to execute it
When to make CPU scheduling :
1. a process switches from running to waiting state
2. a process switches from running to ready state
3. a Process switches from waiting to ready
4. a process terminates
Nonpreemptive scheme: only run CPU scheduling under 1 and 4
i.e., current running process gives up CPU
Otherwise, preemptive scheduling scheme

12. Dispatcher module Scheduler => Dispatcher
Dispatcher module: gives control of CPU to the process selected by short-term scheduler
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
Including context switching latency

13. CPU Scheduling Algorithms
First-come, First-Served Scheduling
Shortest-Job-First Scheduling
Priority Scheduling
Round-Robin Scheduling
Multilevel Queue Scheduling
Multilevel Feedback Queue Scheduling

14. CPU Scheduling Criteria
CPU utilization – percentage of time CPU is busy (not idle)
Throughput – # of processes that complete their execution per time unit
Turnaround time – amount of time to execute a particular process
From process submission to time of completion
Waiting time – amount of time a process has been waiting in ready queue
Response time – amount of time it takes from when a request was submitted until first response is produced (for time-sharing environment)

15. Optimization Criteria
Maximize CPU utilization
Maximize throughput
Minimize turnaround time
Minimize waiting time
Minimize response time
Achieved performance measure is random variable depending on the statistics of workload
Usually, optimize average measure (average case performance)
Sometimes, optimize minimum or maximum value (worst case performance)
Or sometimes, variance of the measure

16. First-Come First-Served Scheduling
Serve processes in the order of arrivals, non- preemptive
Process Burst Time
P1 24
P2 3
P3 3
Suppose that processes arrive in the order: P1 , P2 , P Gantt Chart for FCFS schedule is:
Waiting time for P1 = 0; P2 = 24; P3 = 27
Average waiting time: ( )/3 = 17 P1 P2 P3 24 27 30

17. FCFS Scheduling (Cont.)
Process Burst Time
P1 24
P2 3
P3 3
Suppose that processes arrive in the order P2 , P3 , P1
The Gantt chart for FCFS schedule is:
Waiting time for P1 = 6; P2 = 0; P3 = 3
Average waiting time: ( )/3 = 3
Much better than previous case
Short process behind long process P1 P3 P2 6 3 30

18. Process Burst Time P1 10 P2 4 P3 6 P4 5 Exercise
Draw Gantt chart for FCFS schedule for the following set of processes arrived in the order of P1, P2, P3, P4.
Process Burst Time
P1 10
P2 4
P3 6
P4 5
Average waiting time
What order of arrival of these processes will give smallest average waiting time ?

19. FCFS Scheduling A Convoy effect happens when
What will happen if the system has
one CPU-bound process
many I/O-bound processes
A Convoy effect happens when
a set of processes need to use a resource for a short time
one process holds the resource for a long time, blocking all of the other processes
Result: poor utilization of the other resources in the system.
Other problems of FCFS ?
Ill-suited for time-sharing systems (interactive environment)

20. Shortest-Job-First (SJF) Scheduling
Schedule process with shortest-next-CPU-burst (i.e., the length/duration of next CPU burst)
Two schemes:
nonpreemptive – once CPU given to the process it cannot be preempted until completes its CPU burst
preemptive – if a new process arrives with CPU burst length less than remaining time of current executing process, preempt. Also called Shortest-Remaining-Time-First (SRTF)

21. Shortest-Job-First (SJF) Scheduling
Among non-preemptive scheduling algorithms, SJF achieves minimum average waiting time for a given set of processes that arrives at the same time.
Suppose the duration of CPU burst of n processes is
l1,l2,…ln , li<lj for i<j, (it’s sorted in ascending order
Assume the schedule that achieves minimum average waiting time is not SJF, it means that there exists at least a pair of processes, j and k, such that lj<lk, but process j is scheduled to run after process k
The schedule: … lk … lj …
We can switch lk and lj to decrease avg. waiting time.
Assumption is wrong, and SJF achieves minimum avg. waiting time. 21 21

22. Example of Non-Preemptive SJF
Process Arrival Time Burst Time P P P P
SJF (non-preemptive)
Average waiting time = ( )/4 = 4 P1 P3 P2 7 3 16 P4 8 12

23. Example of Preemptive SJF
Process Arrival Time Burst Time P P P P
SJF (preemptive)
Average waiting time = ( )/4 = 3 P1 P3 P2 4 2 11 P4 5 7 16

24. Length of Next CPU Burst ?
Predict next CPU burst length
Using length of previous CPU bursts, using exponential moving averaging

25. Exponential Average Prediction

26. Examples of Exponential Averaging
 =0
n+1 = n,, recent history does not count
 =1
n+1 =  tn, only previous CPU burst counts
Expand formula:
n+1 =  tn+(1 - ) tn -1 + …
+(1 -  )j  tn -j + …
+(1 -  )n +1 0
each successive term has less weight than its predecessor
Both  and (1 - ) are less than or equal to 1

27. Priority Scheduling Can be Preemptive or Nonpreemptive
CPU is allocated to process with the highest priority
A priority number (integer) is associated with each process
smallest integer  highest priority
Can be Preemptive or Nonpreemptive
Preemptive: newly arrived higher priority process preempt current running process
Priority can be internal or external
Internal: decided based on some quantities of process
resource requirement, ratio of I/O burst and avg burst, next CPU bursts length
External: importance, funds paid,…

28. Priority Scheduling Problem of priority scheduling Solution
Indefinite blocking, starvation: low priority processes may never execute
Solution
Aging – as time progresses increase the priority of the waiting process
E.g., increase the priority of a waiting process by 1 every 15 minutes… 28 28

29. Round Robin (RR) So far, we see FCFS, SJF, Priority Scheduling
How they perform for time-sharing system ?
Each process gets a small unit of CPU time (time quantum), usually 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 ready queue and time quantum is q, then each process gets 1/n of CPU time in chunks of at most q time units at once. No process waits more than (n- 1)q time units.

30. 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 20 37 57 77 97
117
121
134
154
162

31. Round Robin (RR) Downside of having very small q ?
Choosing q, time quantum
q large  FIFO
q small  processor sharing
Downside of having very small q ?
q must be large with respect to context switch, otherwise overhead is too high 31 31

32. Quantum and Context Switch

33. Turnaround Time Varies vs Time Quantum

34. Multilevel Queue Ready queue is partitioned into separate queues
Realtime process
foreground (interactive) process
background (batch) process
Each queue has its own scheduling algorithm
Realtime – Earliest Deadline First
foreground – RR
background – FCFS
Scheduling must be done between queues
Fixed priority scheduling; (i.e., serve all from realtime, then foreground, then from background).
Low priority process might experience starvation
Time slice – each queue gets a certain amount of CPU time
i.e., 80% to foreground in RR, 20% to background in FCFS

35. Multilevel Queue Scheduling
How do we assign process to appropriate queue ?

36. Multilevel Feedback Queue Scheduling
Compared to multilevel queue scheduling
A process can move between various queues; aging can be implemented this way
To favor shorter jobs, or process with short CPU burst
Penalize process that have been run longer by moving it to lower priority queue …
Feed the process back to a different queue

37. Example of Multilevel Feedback Queue
I/O, terminate
Process becomes
ready
Timer goes off
I/O, terminate
I/O, terminate
Q0 – RR with quantum 8 milliseconds
Q1 – RR quantum 16 milliseconds
Q2 – FCFS
Strict priority among queues Q1>Q2>Q3

38. Multilevel Feedback Queue
Effect of multilevel feedback queue
I/O bound and interactive processes stay in highest priority queue
CPU bound process stays in low priority queue
Multilevel-feedback-queue scheduler defined by the following parameters:
number of queues
scheduling algorithms 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

39. Multiple-Processor Scheduling
CPU scheduling more complex when multiple CPUs are available
Homogeneous processors within a multiprocessor
Two appraches
Symmetric multiprocessing: each processor is self-scheduling
Share a common queue, or private queue
Asymmetric multiprocessing: one processor (master server) handles scheduling decision, I/O Processing, and other system activities
– only one processor accesses the system data structures, alleviating the need for data sharing

40. Multiple-Processor Scheduling: considerations
Recall L1 cache ?
To avoid cache invalidation and repopulation, scheduling taking into account processor affinity: a process has an affinity to the processor that it is currently running on
Avoid migrating a process to other processor
Load balancing: keep workload balanced among processors
Needed if each processor has its private queue
Push migration and pull migration

41. Real-Time Scheduling
Hard real-time systems – 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

42. Thread Scheduling
Local Scheduling – How the threads library decides which thread to put onto an available LWP
Global Scheduling – How the kernel decides which kernel thread to run next

43. Case studies: Windows XP and Linux
Homework: comparing CPU scheduling used by Windows XP and Linux
Details to come
Evaluation of different CPU scheduling algorithms
By hand
Through mathematical model
Through simulation studies
Actual Implementation and measurement