Class 9
CS 170
27 April 2020

On the board
------------

1. Last time
2. Scheduling intro
3. Scheduling disciplines

---------------------------------------------------------------------------

1. Last time

    Therac-25 discussion

    Example of what can go wrong in concurrent programming (and, more
    generally, with poor programming practice). 

    Another kind of end-to-end test: radiation phantom.

2. Scheduling intro

    High-level problem: operating system has to decide which process (or
    thread) to run.

    A. When scheduling decisions happen:

        [draw process transition diagram]

                                                   exit (iv)
                                                   |-------->[terminated]
              admitted         (ii) interrupt      |
	[new] -->   [ready]   <-------------- [running]
	              ^        --------------->      |
       I/O or event    \       scheduler dispatch    |
     completion (iii)   \                 ___________|
                         \               /
			      [waiting] v  I/O or event wait (i)


	scheduling decisions take place when a process:

	    (i) Switches from running to waiting state 
	    (ii) Switches from running to ready state 
	    (iii) Switches from waiting to ready 
	    (iv) Exits 

	preemptive scheduling: at all four points

	nonpreemptive scheduling: at points (i), (iv), only (this is the
	definition of nonpreemptive scheduling)



    B. what are the metrics and criteria?

	--turnaround time
	    time for each process to complete

        --waiting/response/output time. variations on this definition,
        but they all capture the time spent waiting for something to
        happen rather than the time from launch to exit.

            --time between when job enters system and starts executing
            following your text, we will call this "response time"
            [some texts call this "waiting time"]

            --time from request to first response (e.g., key press to
            character echo)
            we will call this "output time" 
            [some texts call this "response time"]


	--system throughput
	    # of processes that complete per unit time

	--fairness

	    different possible definitions:

		--freedom from starvation

		--all users get equal time on CPU

		--highest priority jobs get most of CPU

		--etc.

	    [often conflicts with efficiency. true in life as well.]

    C. context switching has a cost

        --CPU time in kernel

	    --save and restore registers

	    --switch address spaces 

	--indirect costs

	    --TLB shootdowns, processor cache, OS caches (e.g., buffer
	    caches)

	--result: more frequent context switches will lead to worse
	throughput (higher overhead)


3. Scheduling disciplines

    Assume first that processes/threads do not do I/O.

        (Completely unrealistic but lets us build intuition, as a
        warmup.)

    A. FCFS/FIFO

	--run each job until it's done

	--P1 needs 24 seconds

	--P2 needs 3 seconds

	--P3 needs 3 seconds
	
	--[ P1         P2  P3 ] 

	--throughput: 3 jobs / 30 seconds = .1 jobs/sec

	--average turnaround time?
	    1/3(24 + 27 + 30) = 27

	--observe: scheduling P2,P3,P1 would drastically reduce average
	turnaround time

	--advantages to FCFS:
	    --simple
	    --no starvation
	    --few context switches

	--disadvantage:
	    --short jobs get stuck behind long ones


    B. SJF and STCF

        --SJF
            Schedule the job whose next CPU burst is the shortest

        --STCF
            preemptive version of SJF: if job arrives that has a shorter
            time to completion than the remaining time on the current job,
            immediately preempt CPU to give to new job

        --idea:
	    --get short jobs out of the system
	    --big (positive) effect on short jobs, small
	    (negative) effect on large jobs

	--example:

	    process     arrival time         burst time
	    P1		0			7
	    P2		2			4
	    P3		4			1
	    P4		5			4

	    preemptive:
	    0  1  2  3  4  5  6  7  8  9  10 11 12 13 14 15
	    P1 P1 P2 P2 P3 P2 P2 P4 P4 P4 P4 P1 P1 P1 P1 P1

        --defer advantages and disadvantages to after we introduce
        relaxation of "no I/O" assumption
    
    let's decide we also care about response time (as defined by our
    book: duration between when a job arrives and when it is first
    scheduled.)

    C. Round-robin (RR)

	--add timer

	--preempt CPU from long-running jobs. per time slice or quantum

	--after time slice, go to the back of the ready queue

	--most OSes do something of this flavor (see MLFQ below)

	--advantages:

	    --fair allocation of CPU across jobs
	    --low average response time when job lengths vary
	    --good for output time if small number of jobs

	--disadvantages:

	    --what if jobs are same length?
	    --example: 2 jobs of 50 time units each, quantum of 1
	    --average turnaround time: 100 (vs. 75 for FCFS)

	--how to choose the quantum size?
	    
	    --want much larger than context switch cost

	    --majority of bursts should be less than quantum

	    --pick too small, and spend too much time context switching

	    --pick too large, and response time suffers (extreme case:
	    system reverts to FCFS)

	    --typical time slice is between 10-100 milliseconds. context
	    switches are usually microseconds or tens of microseconds
	    (maybe hundreds)

    D. Incorporating I/O (relax assumption from before)

        motivating example:

       	    3 jobs
	    A, B: both CPU bound, run for a week
	    C: I/O bound, loop
		1 ms of CPU
		10 ms of disk I/O

	    by itself, C uses 90% of disk
	    By itself, A or B uses 100% of CPU

	    what happens if we use FIFO?
		--if A or B gets in, keeps CPU for 2 weeks

	    what about RR with 100msec time slice?
		--only get 5% disk utilization

	    what about RR with 1msec time slice?
		--get nearly 90% disk utilization
		--but lots of preemptions

        what about STCF? (applied to remaining burst time)
                --would give us good disk utilization ... 

        --advantages to STCF
            --disk utilization (see above)
            --optimal (minimum) waiting/response/output time (can prove this)
            --low overhead (no needless preemptions)

        --disadvantages to STCF
            --long-running jobs can get starved
	    --does not optimize turnaround time (only
	    waiting/response/output time)
            --requires predicting the future

    
	so useful as a yardstick for measuring other policies (good way
	to do CS design and development: figure out what the absolute
	best answer is, then figure out how to approximate it)

	however, can attempt to estimate future based on past (another
	thing that people do when designing systems):

	    --Exponentially weighted average a good idea 

	    --t_n: length of proc's nth CPU burst

	    --\tao_{n+1}: estimate for n+1 burst

	    --choose \alpha, 0 < \alpha <= 1

	    --set \tao_{n+1} = \alpha * t_n + (1-\alpha)*\tao_n

	    --this is called an exponential weighted moving average
	    (EWMA)

	    --reacts to changes, but smoothly

	upshot: favor jobs that have been using CPU the least amount of
	time; that ought to approximate STCF


    E. Priority

	--priority scheme: give every process a number (set by
	administrator), and give the CPU to the process with the highest
	priority (which might be the lowest or highest number, depending
	on the scheme)

	    --can be done preemptively or non-preemptively

	--generally a bad idea to use strict priority because of
	starvation of low priority tasks.
	
	    (--note: SJF is priority scheduling where priority is the
	    predicted next CPU burst time )

	--solution to this starvation is to increase a process's
	priority as it waits

    F. Multi-level feedback queue (MLFQ)

        three ideas:

	    --*multiple queues, each with different priority*. 32
	    queues, for example.
	    
	    --round-robin within each queue (flush a priority level
	    before moving onto the next one). 

	    --feedback: process's priority changes, for example based on
	    how little or much it's used the CPU

        advantages:
	    --approximates SRTCF

	disadvantages:
	    --can't donate priority
	    --not very flexible
	    --not good for real-time and multimedia
	    --can be gameable: user puts in meaningless I/O to keep job's
	    priority high

	see textbook (chap 8) for details

    G. Lottery and stride scheduling

	[citations:
	
	C. A. Waldsburger and W. E. Weihl. Lottery
	Scheduling: Flexible Proportional-Share Resource Management.
	Proc. Usenix Symposium on Operating Systems Design and
	Implementation, November, 1994.
	http://www.usenix.org/publications/library/proceedings/osdi/full_papers/waldspurger.pdf]

	C. A. Waldsburger and W. E. Weihl.  Stride Scheduling:
	Deterministic Proportional-Share Resource Management.
	Technical Memorandum MIT/LCS/TM-528, MIT Laboratory for
	Computer Science, June 1995. 
	http://www.psg.lcs.mit.edu/papers/stride-tm528.ps

	Carl A. Waldspurger. Lottery and Stride Scheduling:
	Flexible Proportional-Share Resource Management, Ph.D.
	dissertation, Massachusetts Institute of Technology,
	September 1995. Also appears as Technical Report
	MIT/LCS/TR-667.
	http://waldspurger.org/carl/papers/phd-mit-tr667.pdf
	]

	Issue lottery tickets to processes 
	-- Let p_i have t_i tickets 
	-- Let T be total # of tickets, T = \sum t_i
	-- Chance of winning next quantum is t_i / T
	-- Note lottery tickets not used up, more like season tickets

	controls long-term average proportion of CPU for each
	process

	can also group processes hierarchically for control
	    --subdivide lottery tickets
	    --can model as currency, so there can be an exchange
	    rate between real currencies (money) and lottery tickets

	--lots of nice features

	    --deals with starvation (have one ticket --> will make
	    progress)

	    --don't have to worry that adding one high priority job will
	    starve all others

	    --adding/deleting jobs affects all jobs proportionally (T
	    gets bigger)

	    --can transfer tickets between processes: highly useful if a
	    client is waiting for a server. then client can donate
	    tickets to server so it can run.

		--note difference between donating tix and donating
		priority. with donating tix, recipient amasses enough
		until it runs. with donating priority, no difference
		between one process donating and 1000 processes donating

	--many other details

	    --ticket inflation for processes that don't use their whole
	    quantum

	    --use fraction f of quantum; inflate tix by 1/f until it
	    next gets CPU

	--disadvantages

	    --latency unpredictable

	    --expected error somewhat high 

		--for those comfortable with probability: this winds up
		being a binomial distribution. variance n*p*(1-p) -->
		standard deviation \proportional \sqrt(n),
		    --where:
			p is fraction of tickets owned
			n is number of quanta

	--in reaction to these disadvantages, Waldspurger and Weihl
	proposed *Stride Scheduling*
 
	    --basically, a deterministic version of lottery scheduling.
	    less randomness --> less expected error.

	    --see textbook (chap 9) for details
	
    H. What Linux does

	--the current Linux scheduler (post 2.6.23), called CFS
	(Completely Fair Scheduler), roughly reinvented the ideas of
	Stride Scheduling