1. Thread Implementation and Scheduling CSE451 Andrew Whitaker

2. Scheduling: A Few Lectures Ago… Scheduler decides when to run runnable threads Programs should make no assumptions about the scheduler  Scheduler is a “black box” scheduler CPU 12345 {4, 2, 3, 5, 1, 3, 4, 5, 2, …}

3. This Lecture: Digging into the Blackbox Previously, we glossed over the choice of which process or thread is chosen to be run next  “some thread from the ready queue” This decision is called scheduling scheduling is policy context switching is mechanism We will focus on CPU scheduling  But, we will touch on network, disk scheduling

4. Scheduling goals Maximize CPU utilization Maximize throughput  Req/sec Minimize latency  Average turnaround time Time from request submission to completion  Average response time Time from request submission till first result Favor some particular class of requests  Priorities Avoid starvation These goals can conflict!

5. Context Goals are different across applications  I/O Bound Web server, database (maybe)  CPU bound Calculating   Interactive system The ideal scheduler works well across a wide range of scenarios

6. Preemption Non-preemptive: Once a thread is given the processor, it has it for as long as it wants  A thread can give up the processor with Thread.yield or a blocking I/O operation Preemptive: A long-running thread can have the processor taken away  Typically, this is triggered by a timer interrupt

7. To Preempt or not to Preempt? Non-preemptive schedulers are subject to starvation Non-preemptive schedulers are easier to implement Non-preemptive scheduling is easier to program  Fewer scheduler interleavings => fewer race conditions Non-preemptive scheduling can perform better  Fewer locks => better performance while (true) { /* do nothing */ }

8. Preemption in Linux User-mode code can always be preempted Linux 2.4 (and before): the kernel was non-preemptive  Any thread/process running in the kernel runs to completion Kernel is “trusted” Linux 2.6: kernel is preemptive  Unless holding a lock

9. Linux 2.6 Preemption Details The task_struct contains a preempt_count variable  Incrementing when grabbing a lock  Decrementing when releasing a lock A thread in the kernel can be preempted iff preempt_count == 0

10. Algorithm #1: FCFS/FIFO First-come first-served / First-in first-out (FCFS/FIFO)  Jobs are scheduled in the order that they arrive  Like “real-world” scheduling of people in lines Supermarkets, bank tellers, McD’s, Starbucks …  Typically non-preemptive no context switching at supermarket!

11. FCFS example Suppose the duration of A is 5, and the durations of B and C are each 1. What is the turnaround time for schedules 1 and 2 (assuming all jobs arrive at roughly the same time) ? Job A B C CB time 1 2 Schedule 1: (5+6+7)/3 = 18/3 = 6 Schedule 2: (1+2+7)/3 = 10/3 = 3.3

12. +No starvation (assuming tasks terminate) -Average response time can be lousy -Small requests wait behind big ones -FCFS may result in poor overlap of CPU and I/O activity  I/O devices are left idle during long CPU bursts Analysis FCFS

13. FCFS at Amazon: Head-of-line Blocking Clients and servers at Amazon communicate over HTTP HTTP is a request/reply protocol which does not allow for re-ordering Problem: Short requests can get stuck behind a long request  Called “head-of-line blocking” Job A B C

14. Algorithm #2: Shortest Job First Choose the job with the smallest service requirement  Variant: shortest processing time first Provably optimal with respect to average waiting time

15. It’s non-preemptive  … but there’s a preemptive version – SRPT (Shortest Remaining Processing Time first) – that accommodates arrivals Starvation  Short jobs continually crowd out long jobs Possible solution: aging How do we know processing times?  In practice, we must estimate SJF drawbacks

16. Longest Job First? Why would a company like Amazon favor Longest Job First?

17. Algorithm #3: Priority Assign priorities to requests  Choose request with highest priority to run next if tie, use another scheduling algorithm to break (e.g., FCFS)  To implement SJF, priority = expected length of CPU burst Abstractly modeled (and usually implemented) as multiple “priority queues”  Put a ready request on the queue associated with its priority

18. Priority drawbacks How are you going to assign priorities? Starvation  if there is an endless supply of high priority jobs, no low- priority job will ever run Solution: “age” threads over time  Increase priority as a function of accumulated wait time  Decrease priority as a function of accumulated processing time  Many ugly heuristics have been explored in this space

19. Priority Inversion A low-priority task acquires a resource (e.g., lock) needed by a high-priority task Can have catastrophic effects (see Mars PathFinder) Solution: priority inheritance  Temporarily “loan” some priority to the lock- holding thread

20. Algorithm #4: RR Round Robin scheduling (RR)  Ready queue is treated as a circular FIFO queue  Each request is given a time slice, called a quantum Request executes for duration of quantum, or until it blocks Advantages:  Simple No need to set priorities, estimate run times, etc.  No starvation Great for timesharing

21. RR Issues What do you set the quantum to be?  No value is “correct” If small, then context switch often, incurring high overhead If large, then response time degrades Turnaround time can be poor  Compared to FCFS or SJF All jobs treated equally  If I run 100 copies of SETI@home, it degrades your service  Need priorities to fix this…

22. Combining algorithms In practice, any real system uses some sort of hybrid approach, with elements of FCFS, SPT, RR, and Priority Example: multi-level queue scheduling  Use a set of queues, corresponding to different priorities  Priority scheduling across queues  Round-robin scheduling within a queue Problems:  How to assign jobs to queues?  How to avoid starvation?  How to keep the user happy?

23. UNIX scheduling: Multi-level Feedback Queues Segregate processes according to their CPU utilization  Highest priority queue has shortest CPU quantum Processes begin in the highest priority queue Priority scheduling across queues, RR within  Process with highest priority always run first  Processes with same priority scheduled RR Processes dynamically change priority  Increases over time if process blocks before end of quantum  Decreases if process uses entire quantum Goals:  Reward interactive behavior over CPU hogs Interactive jobs typically have short bursts of CPU

24. Multi-processor Scheduling Each processor is scheduled independently Two implementation choices  Single, global ready queue  Per-processor run queue Increasingly, per-processor run queues are favored (e.g., Linux 2.6)  Promotes processor affinity (better cache locality)  Decreases demand for (slow) main memory

25. What Happens When a Queue is Empty? Tasks migrate from busy processors to idle processors  Book talks about push vs pull migration Java 1.6 support: thread-safe double- ended queue (java.util.Deque)  Use a bounded buffer per consumer  If nothing in a consumer’s queue, steal work from somebody else