operating systems

goals of an OS

resource allocator, make system easier for users to use, manage CPU time, memory space, etc

tasks of an OS: memory management

allocate/deallocate memory space, keep track of used memory blocks, which process is using them, decide which processes to move in/out of memory

shortest job first scheduling (how)

append each process with the length of the next CPU burst. Schedule jobs with the shortest time

pros/cons of SJF

optimal, but hard to know future CPU burst lengths

shared memory cons

potentially vulnerable to conflicts. race conditions may appear

message passing pros/cons

useful when no conflict resolution is wanted but is slower (each read/write needs kernel operations)

when is the message passing model NEEDED

to communicate between different systems that don’t share memory

message passing model

processes use a message-passing protocol to send/receive info between cooperating processes

shared memory model

processes agree on a region of memory to share among cooperating processes. they read/write there to exchange info

independent vs cooperating processes

independent - doesn’t share any data while executing. cooperating - can affect/be affected by other processes

zombie process

child process that has been terminated but whose parents haven’t called wait() yet. kept in the system to return status to parent eventually

allocating system resources to child processes

either obtain resources directly from the OS. or share a subset of resources from the parent

shortest remaining time

if we allow SJF to be preemptive, we can interrupt a currently running process if it would run longer than some new process

I/O vs CPU bound

I/O - most time spent waiting for memory
CPU - most time spent in execution

process states

new, running, waiting, ready, terminated

starvation in scheduling

low priority processes may not execute

layered OS kernel

highest layer is UI, lowest is hardware. layers can only call functions from the layer below

convoy effect

short jobs held waiting by long jobs in FCFS

monolithic kernel

all kernel functions are in a single static binary file that runs a single address space