CPU scheduling algorithms

multilevel feedback queue is defined by the following characteristics

-# of queues -scheduling alg. -alg. for moving processes b/w queues -alg. to decide queue a process starts in

when higher queue has relative priority over lower queue

-each queue gets a time slice, which it can schedule as smaller slices to its processes -avoids starvation, but higher processes more likely to be disrupted by lower-priority ones

multilevel queue scheduling algorithms

-higher queue has absolute priority over lower queues -higher queue has relative priority -combination

in round robin, the time slice(q) needs to be

-large enough with respect to context switch time -small enough for system to be continually responsive

first-come first-served

-processes all allocated to CPU in order they arrive -simplest -processes run until completing -not preemptive

weaknesses of priority

completely dependent on the priority being set reasonably -a process might get one so bad it never runs

round robin

define a time quantum/slice, set timer interrupt to stop running process if burst is longer than slice -ready processes added to back of queue

priority(scheduling algorithm)

each process has a priority that determines how soon it is scheduled compared to other processes -priority can be # or func

processes in a multilevel queue

each process is permanently assigned a given queue based on input or an inherent property

round robin is like

first come first served but redesigned to add preemption

strengths of priority

flexible

strengths of shortest job first

guaranteed to produce an optimal schedule in terms of waiting time

when higher queues have absolute priority over lower queues

higher-importance processes less likely to be stalled by lower-importance ones, but this might lead to starvation for some processes

multilevel feedback queue

multilevel queue where processes can move between queues

multilevel queue not useful for

nonpreemptive systems

weakness of shortest job first

not possible, longer processes might never get run

average waiting time for round robin

often long

shortest job first

processes are scheduled to CPU with priority to process with next-shortest burst -can be preemptive or non

soft real-time scheduling just requires that

processes run with absolute priority over other processes

multilevel queue

ready queue partitioned into multiple separate queues -queues are given a priority order

strengths of first-come first-served

really simple, easy to write and understand

combination of absolute/relative priority in multilevel queues

schedule top 2 queues(system, quasi-realtime) have abs. priority over rest -priority w/in those 2 is relative

absolute priority leads to

starvation, almost always a bad idea

the round robin is the basis of

the scheduling algorithm for most interactive time-sharing systems

round robin designed for

time-sharing systems

in round robin, each process is guaranteed

to have to wait at most (n-1)q time units until it gets the processor back

weaknesses of first-come first-served

unpredictable wait times and performance - nonpreemptive nature makes it unsuitable for modern OSs

real-time scheduling

used for processes that need to respond to events as quickly as possible

hard real-time scheduling requires that

we either service process's latency requirements or refuse the process

starvation

when a process has such a bad priority that some other process always has a better one and it never runs