Commit 7191c4bd authored by GangCao's avatar GangCao Committed by Jim Harris
Browse files

bdev/qos: add the bandwidth rate limit control 



This patch is to add the core control policy for the
bandwidth rate limit (max bytes per second). Change
the existing functions for a common name and specially
handle the case when IOPS and bandwidth rate limiting
are both enabled.

Change-Id: I9f4565958d472559ef6d8bea52b1fe2a5f3c8969
Signed-off-by: default avatarGangCao <gang.cao@intel.com>
Reviewed-on: https://review.gerrithub.io/413821


Tested-by: default avatarSPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: default avatarDaniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: default avatarJim Harris <james.r.harris@intel.com>
parent 8da7772a

lib/bdev/bdev.c

+64 −16
Original line number Diff line number Diff line
@@ -65,6 +65,7 @@ int __itt_init_ittlib(const char *, __itt_group_id); 
#define SPDK_BDEV_QOS_TIMESLICE_IN_USEC		1000

#define SPDK_BDEV_SEC_TO_USEC			1000000ULL

#define SPDK_BDEV_QOS_MIN_IO_PER_TIMESLICE	1

#define SPDK_BDEV_QOS_MIN_BYTE_PER_TIMESLICE	512

#define SPDK_BDEV_QOS_MIN_IOS_PER_SEC		10000

#define SPDK_BDEV_QOS_MIN_BW_IN_MB_PER_SEC	10
@@ -130,9 +131,16 @@ struct spdk_bdev_qos { 
	 *  only valid for the master channel which manages the outstanding IOs. */

	uint64_t max_ios_per_timeslice;



	/** Maximum allowed bytes to be issued in one timeslice (e.g., 1ms) and

	 *  only valid for the master channel which manages the outstanding IOs. */

	uint64_t max_byte_per_timeslice;



	/** Submitted IO in one timeslice (e.g., 1ms) */

	uint64_t io_submitted_this_timeslice;



	/** Submitted byte in one timeslice (e.g., 1ms) */

	uint64_t byte_submitted_this_timeslice;



	/** Polller that processes queued I/O commands each time slice. */

	struct spdk_poller *poller;

};
@@ -862,6 +870,26 @@ spdk_bdev_put_io(struct spdk_bdev_io *bdev_io) 
	}

}



static uint64_t

_spdk_bdev_get_io_size_in_byte(struct spdk_bdev_io *bdev_io)

{

	struct spdk_bdev	*bdev = bdev_io->bdev;



	switch (bdev_io->type) {

	case SPDK_BDEV_IO_TYPE_NVME_ADMIN:

	case SPDK_BDEV_IO_TYPE_NVME_IO:

	case SPDK_BDEV_IO_TYPE_NVME_IO_MD:

		return bdev_io->u.nvme_passthru.nbytes;

	case SPDK_BDEV_IO_TYPE_READ:

	case SPDK_BDEV_IO_TYPE_WRITE:

	case SPDK_BDEV_IO_TYPE_UNMAP:

	case SPDK_BDEV_IO_TYPE_WRITE_ZEROES:

		return bdev_io->u.bdev.num_blocks * bdev->blocklen;

	default:

		return 0;

	}

}



static void

_spdk_bdev_qos_io_submit(struct spdk_bdev_channel *ch)

{
@@ -871,16 +899,23 @@ _spdk_bdev_qos_io_submit(struct spdk_bdev_channel *ch) 
	struct spdk_bdev_shared_resource *shared_resource = ch->shared_resource;



	while (!TAILQ_EMPTY(&qos->queued)) {

		if (qos->io_submitted_this_timeslice < qos->max_ios_per_timeslice) {

		if (qos->max_ios_per_timeslice > 0 &&

		    qos->io_submitted_this_timeslice >= qos->max_ios_per_timeslice) {

			break;

		}



		if (qos->max_byte_per_timeslice > 0 &&

		    qos->byte_submitted_this_timeslice >= qos->max_byte_per_timeslice) {

			break;

		}



		bdev_io = TAILQ_FIRST(&qos->queued);

		TAILQ_REMOVE(&qos->queued, bdev_io, link);

		qos->io_submitted_this_timeslice++;

		qos->byte_submitted_this_timeslice += _spdk_bdev_get_io_size_in_byte(bdev_io);

		ch->io_outstanding++;

		shared_resource->io_outstanding++;

		bdev->fn_table->submit_request(ch->channel, bdev_io);

		} else {

			break;

		}

	}

}
@@ -1000,16 +1035,25 @@ spdk_bdev_config_json(struct spdk_bdev *bdev, struct spdk_json_write_ctx *w) 
}



static void

spdk_bdev_qos_update_max_ios_per_timeslice(struct spdk_bdev_qos *qos)

spdk_bdev_qos_update_max_quota_per_timeslice(struct spdk_bdev_qos *qos)

{

	uint64_t max_ios_per_timeslice = 0;

	uint64_t max_ios_per_timeslice = 0, max_byte_per_timeslice = 0;



	if (qos->iops_rate_limit > 0) {

		max_ios_per_timeslice = qos->iops_rate_limit * SPDK_BDEV_QOS_TIMESLICE_IN_USEC /

					SPDK_BDEV_SEC_TO_USEC;

		qos->max_ios_per_timeslice = spdk_max(max_ios_per_timeslice,

						      SPDK_BDEV_QOS_MIN_IO_PER_TIMESLICE);

	}



	if (qos->byte_rate_limit > 0) {

		max_byte_per_timeslice = qos->byte_rate_limit * SPDK_BDEV_QOS_TIMESLICE_IN_USEC /

					 SPDK_BDEV_SEC_TO_USEC;

		qos->max_byte_per_timeslice = spdk_max(max_byte_per_timeslice,

						       SPDK_BDEV_QOS_MIN_BYTE_PER_TIMESLICE);

	}

}



static int

spdk_bdev_channel_poll_qos(void *arg)

{
@@ -1017,6 +1061,7 @@ spdk_bdev_channel_poll_qos(void *arg) 


	/* Reset for next round of rate limiting */

	qos->io_submitted_this_timeslice = 0;

	qos->byte_submitted_this_timeslice = 0;



	_spdk_bdev_qos_io_submit(qos->ch);
@@ -1075,8 +1120,9 @@ _spdk_bdev_enable_qos(struct spdk_bdev *bdev, struct spdk_bdev_channel *ch) 
			qos->thread = spdk_io_channel_get_thread(io_ch);



			TAILQ_INIT(&qos->queued);

			spdk_bdev_qos_update_max_ios_per_timeslice(qos);

			spdk_bdev_qos_update_max_quota_per_timeslice(qos);

			qos->io_submitted_this_timeslice = 0;

			qos->byte_submitted_this_timeslice = 0;



			qos->poller = spdk_poller_register(spdk_bdev_channel_poll_qos,

							   qos,
@@ -1266,7 +1312,9 @@ spdk_bdev_qos_destroy(struct spdk_bdev *bdev) 
	new_qos->ch = NULL;

	new_qos->thread = NULL;

	new_qos->max_ios_per_timeslice = 0;

	new_qos->max_byte_per_timeslice = 0;

	new_qos->io_submitted_this_timeslice = 0;

	new_qos->byte_submitted_this_timeslice = 0;

	new_qos->poller = NULL;

	TAILQ_INIT(&new_qos->queued);
@@ -3077,7 +3125,7 @@ _spdk_bdev_update_qos_limit_iops_msg(void *cb_arg) 
	struct spdk_bdev *bdev = ctx->bdev;



	pthread_mutex_lock(&bdev->mutex);

	spdk_bdev_qos_update_max_ios_per_timeslice(bdev->qos);

	spdk_bdev_qos_update_max_quota_per_timeslice(bdev->qos);

	pthread_mutex_unlock(&bdev->mutex);



	_spdk_bdev_set_qos_limit_done(ctx, 0);

test/unit/lib/bdev/mt/bdev.c/bdev_ut.c

+16 −1
Original line number Diff line number Diff line
@@ -629,7 +629,12 @@ basic_qos(void) 
	bdev->qos = calloc(1, sizeof(*bdev->qos));

	SPDK_CU_ASSERT_FATAL(bdev->qos != NULL);

	TAILQ_INIT(&bdev->qos->queued);

	/*

	 * Enable both IOPS and bandwidth rate limits.

	 * In this case, both rate limits will take equal effect.

	 */

	bdev->qos->iops_rate_limit = 2000; /* 2 I/O per millisecond */

	bdev->qos->byte_rate_limit = 8192000; /* 8K byte per millisecond with 4K block size */



	g_get_io_channel = true;
@@ -732,7 +737,12 @@ io_during_qos_queue(void) 
	bdev->qos = calloc(1, sizeof(*bdev->qos));

	SPDK_CU_ASSERT_FATAL(bdev->qos != NULL);

	TAILQ_INIT(&bdev->qos->queued);

	/*

	 * Enable both IOPS and bandwidth rate limits.

	 * In this case, IOPS rate limit will take effect first.

	 */

	bdev->qos->iops_rate_limit = 1000; /* 1000 I/O per second, or 1 per millisecond */

	bdev->qos->byte_rate_limit = 8192000; /* 8K byte per millisecond with 4K block size */



	g_get_io_channel = true;
@@ -815,7 +825,12 @@ io_during_qos_reset(void) 
	bdev->qos = calloc(1, sizeof(*bdev->qos));

	SPDK_CU_ASSERT_FATAL(bdev->qos != NULL);

	TAILQ_INIT(&bdev->qos->queued);

	bdev->qos->iops_rate_limit = 1000; /* 1000 I/O per second, or 1 per millisecond */

	/*

	 * Enable both IOPS and bandwidth rate limits.

	 * In this case, bandwidth rate limit will take effect first.

	 */

	bdev->qos->iops_rate_limit = 2000; /* 2000 I/O per second, or 2 per millisecond */

	bdev->qos->byte_rate_limit = 4096000; /* 4K byte per millisecond with 4K block size */



	g_get_io_channel = true;