CPU Scheduling Schedulers Structure of a CPU scheduler Criteria - PDF document

CPU Scheduling • Schedulers • Structure of a CPU scheduler • Criteria for scheduling • Scheduling Algorithms – FCFS – SPN – SRT – MLFQ • CPU scheduling in Unix Schedulers start long-term scheduler short-term scheduler suspended ready running ready suspended blocked blocked medium-term scheduler

Short-Term Scheduling • Motivation for multiprogramming: Have multiple processes in memory to keep CPU busy. • Typical execution profile of a process: start terminate wait for I/O wait for I/O wait for I/O CPU burst CPU burst CPU burst CPU burst • CPU scheduler is managing the execution of CPU bursts, represented by processes in ready or running state. Scheduling Decisions • Who is going to use the CPU next? 4 2 ready running ready running 3 1 waiting waiting non-preemptive • Scheduling decision points: – 1 . The running process changes from running to waiting (current CPU burst of that process is over). preemptive – 2 . The running process terminates. – 3 . A waiting process becomes ready (new CPU burst of that process begins). – 4 . The current process switches from running to ready .

Structure of a Scheduler (I) ready queue ready queue PCB scheduler dispatcher CPU PCB scheduler dispatcher CPU select process start new process ? ? What Is a Good Scheduler? Criteria • User oriented: – Turnaround time : time interval from submission of job until its completion – Waiting time : sum of periods spent waiting in ready queue – Response time : time interval from submission of job to first response – Normalized turnaround time : ratio of turnaround time to service time • System oriented: – CPU utilization : percentage of time CPU is busy – Throughput : number of jobs completed per time unit • Any good scheduler should: – maximize CPU utilization and throughput – minimize turnaround time, waiting time, response time • Huh? – maximum/minimum values vs. average values vs. variance

Scheduling Algorithms • FCFS : First-come-first-served • SPN: Shortest Process Next • SRT : Shortest Remaining Time • priority scheduling • RR : Round-robin • Multilevel feedback queue scheduling • Multiprocessor scheduling Structure of a Scheduler (II) (conceptual structure) determine location in queue determine location in queue head head tail tail CPU CPU PCB PCB • Incoming process is put into right location in ready queue. • Dispatcher always picks first element in ready queue.

First-Come-First-Served (FCFS) append at the end of queue append at the end of queue head head tail tail CPU CPU PCB PCB • Advantages: – very simple • Disadvantages: – long average and worst-case waiting times – poor dynamic behavior (convoy effect) Waiting Times for FCFS • Example: P 1 = 24, P 2 = 6, P 3 = 6 W awg = (24+30)/3 = 18 P 1 P 2 P 3 W wc = 30 Different arrival order: W awg = (6+12)/3 = 6 P 2 P 3 P 1 W wc = 12 • Average waiting times is not minimal. • Waiting times may substantially vary over time. • Worst-case waiting times can be very long.

Convoy Effects CPU I/O empty! empty! CPU-bound I/O-bound CPU empty! empty! I/O Convoy Effects CPU I/O idle! idle! CPU-bound I/O-bound

Convoy Effects CPU I/O empty! empty! CPU-bound I/O-bound CPU empty! empty! I/O Shortest Process Next determine location in queue determine location in queue (compare next CPU burst lengths) (compare next CPU burst lengths) CPU CPU long jobs short jobs • Whenever CPU is idle, picks process with shortest next CPU burst . • Advantages: minimizes average waiting times. • Problem: How to determine length of next CPU burst? • Problem: starvation of jobs with long CPU bursts.

SJF Minimizes Average Waiting Time • Provably optimal: Proof: swapping of jobs P long P short P long P short P short P long dW = t short - t long < 0 P short P long • Example: 6 12 8 4 W = 6+18+26 = 50 6 8 12 4 W = 6+14+26 = 46 6 8 4 12 W = 6+14+18 = 38 6 4 8 12 W = 6+10+18 = 34 4 6 8 12 W = 4+10+18 = 32 • Question: How to determine execution time of next CPU burst ?! – wild guess? – code inspection? • Forecasting (i.e. estimation) S n+1 = F(T n , T n-1 , T n-2 , T n-3 , T n-4 , ...) • Simple forecasting function: exponential average: S n+1 = a T n + (1-a) S n • Example: a = 0.8 S n+1 = 0.8T n + 0.16T n-1 + 0.032T n-2 + 0.0064T n-3 + ...

Exponential Averaging: Example 16 14 a = 0.2 12 a = 0.5 10 a = 0.8 8 6 4 2 1 Preemptive SPN: Shortest- Remaining-Time -First • SPN: P 1 and P 3 arrive here P 2 arrives here P 1 P 2 P 3 nil P 3 P 2 P 3 ready queue P 3 • SRT: P 1 and P 3 arrive here P 2 arrives here P 1 P 2 P 1 P 3 nil P 3 P 1 P 3 P 3 P 1 resumes execution P 1 is preempted

Priority Scheduling Selector Selector (compare priorities) (compare priorities) CPU CPU low priority high priority • Whenever CPU is idle, picks process with highest priority . • Priority: – process class, burst length, urgency, pocket depth. • Unbounded blocking: Starvation – Increase priority over time: aging • Conceptually • Priority Queues : priority queue low priority priority low priority q=f(p) (compare priorities) (compare priorities) (compare priorities) (compare priorities) Selector Selector Selector Selector high priority high priority CPU CPU CPU CPU

Round-Robin • FIFO with preemption after time quantum • Method for time sharing • Choice of time quantum: – large: FCFS – small: Processor sharing • Time quantum also defines context-switching FIFO queue overhead end of • Response time smaller time quantum with large time quantums CPU CPU terminate Multilevel Queue Scheduling low priority batch processes (compare priorities) (compare priorities) user processes Selector Selector high-priority user processes kernel processes high priority separate queues, perhaps with different scheduling policies CPU CPU

Multilevel Feedback Queue Scheduling (conceptually) low priority FCFS (quantum = infinity) (compare priorities) (compare priorities) quantum = 16 ms Selector Selector aging quantum = 4ms quantum = 2 ms high priority demotion MFBS: Implementation (Unix System V) • Clock handler generates 60 clock ticks per second. • Each PCB contains a field CPU (“recent CPU usage”), which is incremented on every clock tick while process is running. • Every 60 ticks scheduler is awakened and – ajusts recent CPU usage according to a decay function: decay( CPU ) = CPU /2 – recalculates priorities according to following formula (higher priorities have lower priority values!): priority = CPU /2 + base_priority • Decay rate controls aging. • Priority recalculation controls demotion • Note : This is a simplified view! (For a more detailed description refer to M.J.Bach, The Design of the UNIX Operating System. )

MFBS on UNIX System V: Example • 3 processes, each with base priority 60: Process A Process B Process C time priority CPU count priority CPU count priority CPU count 60 0 60 0 60 0 1 ... 60 1 75 30 60 0 60 0 1 ... 2 67 15 75 30 60 0 1 ... 3 63 7 67 15 75 30 8 ... 67 4 76 33 63 7 67 15 8 ... 67 5 68 16 76 33 63 7