introduce bfq i/o scheduler #2

xryoshi · 2012-12-14T02:53:45Z

No description provided.

BFQ uses struct cfq_io_context to store its per-process per-device data, reusing the same code for cic handling of CFQ. The code is not shared ATM to minimize the impact of these patches. This patch introduces a new hlist to each io_context to store all the cic's allocated by BFQ to allow calling the right destructor on module unload; the radix tree used for cic lookup needs to be duplicated because it can contain dead keys inserted by a scheduler and later retrieved by the other one. Update the io_context exit and free paths to take care also of the BFQ cic's. Change the type of cfqq inside struct cfq_io_context to void * to use it also for BFQ per-queue data. A new bfq-specific ioprio_changed field is necessary, too, to avoid clobbering cfq's one, so switch ioprio_changed to a bitmap, with one element per scheduler. Signed-off-by: Paolo Valente <[email protected]> Signed-off-by: Arianna Avanzini <[email protected]>

Add a Kconfig option and do the related Makefile changes to compile the BFQ I/O scheduler. Also let the cgroups subsystem know about the BFQ I/O controller. Signed-off-by: Paolo Valente <[email protected]> Signed-off-by: Arianna Avanzini <[email protected]>

Add the BFQ-v5r1 I/O scheduler to 3.1. The general structure is borrowed from CFQ, as much code. A (bfq_)queue is associated to each task doing I/O on a device, and each time a scheduling decision has to be taken a queue is selected and it is served until it expires. - Slices are given in the service domain: tasks are assigned budgets, measured in number of sectors. Once got the disk, a task must however consume its assigned budget within a configurable maximum time (by default, the maximum possible value of the budgets is automatically computed to comply with this timeout). This allows the desired latency vs "throughput boosting" tradeoff to be set. - Budgets are scheduled according to a variant of WF2Q+, implemented using an augmented rb-tree to take eligibility into account while preserving an O(log N) overall complexity. - A low-latency tunable is provided; if enabled, both interactive and soft real-time applications are guaranteed very low latency. - Latency guarantees are preserved also in presence of NCQ. - High throughput with flash-based devices, while still preserving latency guarantees. - Useful features borrowed from CFQ: cooperating-queues merging (with some additional optimizations with respect to the original CFQ version), static fallback queue for OOM. - BFQ supports full hierarchical scheduling, exporting a cgroups interface. Each node has a full scheduler, so each group can be assigned its own ioprio and an ioprio_class. - If the cgroups interface is used, weights can be explictly assigned, otherwise ioprio values are mapped to weights using the relation weight = IOPRIO_BE_NR - ioprio. - ioprio classes are served in strict priority order, i.e., lower priority queues are not served as long as there are higher priority queues. Among queues in the same class the bandwidth is distributed in proportion to the weights of each queue. A very thin extra bandwidth is however guaranteed to the Idle class, to prevent it from starving. Signed-off-by: Paolo Valente <[email protected]> Signed-off-by: Arianna Avanzini <[email protected]>

ariava added 3 commits December 14, 2012 10:00

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

introduce bfq i/o scheduler #2

introduce bfq i/o scheduler #2

xryoshi commented Dec 14, 2012

introduce bfq i/o scheduler #2

Are you sure you want to change the base?

introduce bfq i/o scheduler #2

Conversation

xryoshi commented Dec 14, 2012