Process Management

• a process manager manages processes, so what's a process?
• it is not quite the same as a program or an executable file
• consider this life-cyle for a process and how each transition comes about
• the fact that programs need different resources at different times allows for multiprogramming
• multiprogramming will lead to a lot of jumping around in the flow of control from the hardware's perspective
• the process manager portion of the OS needs to keep up with all the processes
• so there is a process table with an entry for each process
• consider how this entry helps when context-switching and what needs to be saved/restored
• in addition to keeping up with each process' information, the process manager must also track order in a queue

Process Scheduling

• schedulers can be non-preemptive or preemptive; preemptive requires an interval timer
• how do we decide who get the CPU at any given time? priority - either implicitly or explicitly
• most scheduling systems can also change a process' priority as needed
• a good scheduler will prevent starvation
• some non-preemptive scheduling algorithms:
  • first come first serve (FCFS)
  • shortest job first (SJF) (also called shortest job next or SJN)
  • assigned priority
  • deadline
• measuring effectiveness with service time, wait time, turnaround time (or its inverse, throughput rate)
• consider how these measures apply to a fast-food drive-through
• compute these measures for the following list of processes for FCFS and SJF
  
  

  Process Number (Arrival Order)	Service Time
  0	350
  1	125
  2	475
  3	250
  4	75

• generally the optimal schedule given known service times is SJF
• however, in the real world, most of the time we really don't know the service time until the program finishes running
• another unrealistic aspect of this exercise is we haven't taken into account any I/O
• furthermore, scheduling normally takes place in concert with context-switching and that means it is preemptive
• in a preemptive scenario, service time is broken up into pieces called quantums
• some preemptive scheduling algorithms:
  • round-robin (RR)
  • multiple-level queues (also called multilevel feedback queues)
• consider how RR would work with the earlier process list