1. Intro to Multiprocessing
CS/COE 0449 (term 2184)
Jarrett Billingsley

2. Class announcements
don't forget… project 3 is due next Wednesday night
OMET completion is at 20%
15 responses out of 56 needed for cookies
step up your game
CS449 (2184)

3. Multiprocessing
CS449 (2184)

4. Where'd it come from? computers used to be huge like.. physically
and therefore expensive
like.. millions of dollars
so many people had to be able to share one computer at once
like.. thoth
multiprocessing is: running multiple processes (programs) at the same time on one computer
CS449 (2184)

5. And then they got small
personal computers (PCs) weren't A Thing until the 1970s, and weren't really popular until well into the 1980s
limited and basic at first, but…
CS449 (2184)

6. Scheduling
CS449 (2184)

7. AAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
All my children
at any given time, there are dozens or hundreds of processes running
Code
Memory
Process 1
my turn! my turn!!!
I'm the ONLY process! that means I'm the BEST!
Code
Memory
Process 173
Kernel
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
Code
Memory
Process 9
Code
Memory
Process 92
I want all the CPU!!
no I want all the CPU!!
CS449 (2184)

8. we call these CPU-bound we call these IO-bound
Who's who
some processes use LOTS OF CPU
other just sit around most of the time, waiting for input or output (I/O)
enter your choice: rock
the computer chose rock. tie!
enter your choice: paper
the computer chose rock. you win!
you: 1, computer: 0
enter your choice: _
g++ somefile.cpp
Loading…
we call these CPU-bound
we call these IO-bound
CS449 (2184)

9. I'm turning this car around
you've been playing for 3 hours…
lemme just beat this boss okay???
if we give the CPU-bound processes priority…
…then the IO-bound ones never get a turn.
mommy has work to do, Billy
if we give the IO-bound processes priority…
cow!!!!!!
…then the CPU-bound ones run really slowly.
CS449 (2184)

10. Scheduling
scheduling is how we make the processes take turns on the CPU
this is an entire unit in CS1550
all the many ways of doing scheduling do the same basic thing: P1 P2 P5 P9 P1 P7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
time
we're slicing up time and giving each slice to a different process.
scheduling answers the question: "which process goes next?"
CS449 (2184)

11. Collaborative scheduling
in this method, each process has to say "I'm done with my turn" before another process gets to run
this is simple to implement and doesn't require any fancy OS or hardware features!
…but we all know the problem with taking turns
CS449 (2184)

12. What a hog while(true) {}
the problem with collaborative scheduling is…
if a process never gives up the CPU, nothing else can run.
not even the kernel.
so it's unfortunately really, really easy to make a collaboratively scheduled OS freeze…
while(true) {}
aaaand now you have to restart the computer.
CS449 (2184)

13. Preemptive scheduling
in this method, the OS can interrupt a process that's taking too long
k, I'm done
doo dee doo
but.. but I wasn't…
haha suckerrrr P1 P2 P5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
P2, your turn
ok P2 that's enough… P5, go!
Kernel
CS449 (2184)

14. Interrupted by an interrupt
to make this work, we need special hardware in the CPU
wheee!!!
hey timer, wake me up in 100ms
CPU P2
Kernel
P2, gtfo
RING RING!!!!
the timer causes an interrupt – like a signal, but lower level
only the kernel is allowed to respond to interrupts
the quantum is how long a process has before it's preempted
CS449 (2184)

15. The process lifecycle
CS449 (2184)

16. The directed graph of life
processes can be in one of many states
Running
normal exit
scheduled
fork()!
preempted
Created
Ready (sleeping)
Dead
(or yielded)
blocking syscall
abnormal termination – crashed or killed
OS detects syscall can complete
Blocked
CS449 (2184)

17. Scheduling points
whenever a process gives up the CPU to the kernel, the kernel has a chance to schedule another process ("who's next?")
there are a few ways for it to give up the CPU
it could get preempted
it used up its whole quantum and the OS interrupted it
it could yield
it voluntarily gives up control – it's playing nice!!
it could use a blocking syscall
such as reading data from a file
it could exit
either on purpose or by crashing.
CS449 (2184)

18. and then it can schedule a process from the ready set
Ready, set, blocked
internally, the OS has these two (or more) sets of processes
Blocked
Ready P1 P4 P5 P9
P12 P7
yay!!
when the kernel takes control, it can see if any of the blocked processes can be moved to the ready set
and then it can schedule a process from the ready set
CS449 (2184)

19. But context switches are haaaardddd
processes are pretty big things with many parts
Stack
Heap
Globals
Code
open files
internal OS data structures the kernel uses to keep tabs on the process 34 35
process 14 has been running for 8 years and shows no signs
of stopping. what the hell is it doing
all this crap
sure would be nice to get some of the benefits of processes without all the overhead…
CS449 (2184)