[计算机软件及应用]Linux 25 and 26 scheduling.pptVIP

Group Scheduling If A has 1 process and B has 3 processes, A should obtain 25% time of CPU and B should obtain 75%? Seemingly, it’s sensible, but it’s unfair! So…. * Michael Owen At HIT * I/O Scheduling Before Linux 2.6 Deadline Anticipatory Complete Fair Queuing and Noop * Michael Owen At HIT * I/O scheduling perform two actions to minimize seeks: * Michael Owen At HIT * merging sorting Try to keep the disk head moving in a straight line, like: Elevator Before Linux 2.6 * Michael Owen At HIT * Elevator scheduling Elevator scheduling Maintains a single queue for disk read and write requests Keeps list of requests sorted by block number Drive moves in a single direction to satisfy each request * Michael Owen At HIT * Starvation and read versus write problems may occur * Michael Owen At HIT * I/O Scheduling Deadline Anticipatory Complete Fair Queuing and Noop Deadline scheduling * Michael Owen At HIT * Deadline scheduling Deadline scheduling Uses three queues Incoming requests Read requests go to the tail of a FIFO queue Write requests go to the tail of a FIFO queue Each request has an expiration time(write: 5s, read: 0.5s) * Michael Owen At HIT * Overcome the starvation problem and also read versus write problem * Michael Owen At HIT * I/O Scheduling Deadline Anticipatory Complete Fair Queuing and Noop Anticipatory I/O scheduling Anticipatory I/O scheduling Delay a short period of time (default time: 6ms) after satisfying a read request to see if a new nearby request can be made because of the principle of locality * Michael Owen At HIT * Performance improvement * Michael Owen At HIT * I/O Scheduling Deadline Anticipatory Complete Fair Queuing and Noop * Michael Owen At HIT * CFQ scheduling (Linux 2.6.18) There is one queue for each process submitting I/O Service the queues round robin Default scheduling in Linux now, it performs in nearly all workload without any pathological corner cases Noop scheduling Not perform sorting, but perform