Tutorial 4
Scheduling
Why do we need scheduling?
• To manage processes according to
requirements of a system, like:
– User responsiveness or
– Throughput
• Performance of a scheduler is determined
mainly by:
– Context switch time
– Scheduling policy
1, 2, 3, 4, 5, 6, Context Switch!
• Switching from one process running on the
CPU to another process
• Saves all the registers of outgoing process
(to memory), then loads all the registers of
incoming process (from memory)
• Can be time-costly; mostly hardwaredependent
Scheduling
• The mechanism that determines when the
CPU will be allocated to processes, and in
what order
• Two classes of scheduling strategies:
– Nonpreemptive (aka Batch)
– Preemptive
Non-preemptive policies
• Allow any process to run to completion
once it has been allocated to the CPU.
Current process does not get interrupted.
• Some examples:
– First Come First Serve (FCFS)
– Shortest Job Next (SJN)
– Priority scheduling
– Deadline scheduling
Preemptive policies
• Allow another process to interrupt current
process if:
– It has a higher priority
– The time quantum has elapsed
• Some examples:
– Round Robin
– Multiple-level Queues
Scheduling examples
• Given three threads, their execution times
and I/O needs, apply scheduling policies
• Threads are placed on ready queue in
order: T1, T2 then T3
* Specific to Round Robin:
– Time Quantum of 3ms
– Context switch time considered negligible in
this example
First Come First Serve
Thread
CPU
1st I/O
2nd I/O
T1
10 ms
n/a
n/a
T2
15 ms at 2ms for 5ms
T3
12 ms at 4ms for 2ms at 8ms for 2ms
1
CPU:
0
I/O:
5
2
10
3
2
15
2
n/a
20
3
3
25
30
3
35
3
40
Shortest Job First (Nonpreemptive)
Thread
CPU
1st I/O
2nd I/O
T1
10 ms
n/a
n/a
T2
15 ms at 2ms for 5ms
T3
12 ms at 4ms for 2ms at 8ms for 2ms
1
CPU:
0
I/O:
5
3
10
2
3
15
3
2
20
2
n/a
25
3
3
30
35
40
Shortest Job First (Preemptive)
Thread
CPU
1st I/O
2nd I/O
T1
10 ms
n/a
n/a
T2
15 ms at 2ms for 5ms
T3
12 ms at 4ms for 2ms at 8ms for 2ms
1
CPU:
0
I/O:
5
3
10
2
3
15
3
2
20
2
3
25
3
n/a
2
30
35
40
Priority (Preemptive)
Thread Prio
CPU
1st I/O
2nd I/O
n/a
n/a
T1
2
10 ms
T2
3
15 ms at 2ms for 5ms
T3
1
12 ms at 4ms for 2ms at 8ms for 2ms
3
CPU:
0
I/O:
1
5
3
3
1
10
3
15
3
1
20
2
n/a
2
25
30
2
35
40
Round Robin
Thread
CPU
1st I/O
2nd I/O
T1
10 ms
n/a
n/a
T2
15 ms at 2ms for 5ms
T3
12 ms at 4ms for 2ms at 8ms for 2ms
3
1
CPU:
0
I/O:
2
3
1
5
13
2
10
2
n/a
1
15
3
3
20
2
32
2
25
3
30
3
2
35
40
Recapping Scheduling in Java
Java 1.2.1 and prior (“green”) threads:
• Threads are managed by Java VM
• One thread runs at a time
• Thread is only taken off the CPU when:
– it yields
– it blocks (on a resource)
– it exits
– a higher priority thread becomes runnable
(not always)
New-Style Scheduling in Java
Post-1.2.1 versions of Java:
• Threads are managed by the host OS
• Allows multiple threads to run
“concurrently”
• Thread is taken off the CPU as in previous
versions, but also
– when its time quantum expires (if supported
by OS)