When using AIO, submit all requests which exists in the vring in a single
io_submit instead of one io_submit for each descriptor.
Benchmarks:
Short version: 15%+ increase in IOPS, small increase in BW.
Read IOPS:
Before:
vda: ios=291792/0, merge=0/0, ticks=35229/0, in_queue=31025, util=61.30%
After:
vda: ios=333114/0, merge=0/0, ticks=47983/0, in_queue=46630, util=62.22%
Write IOPS:
Before:
vda: ios=0/271716, merge=0/0, ticks=0/33367, in_queue=26531, util=59.96%
After:
vda: ios=0/327485, merge=0/0, ticks=0/23789, in_queue=22475, util=55.74%
Read BW:
Before:
READ: io=7526.0MB, aggrb=1246.3MB/s, minb=1275.1MB/s, maxb=1275.1MB/s, mint=6040msec, maxt=6040msec
After:
READ: io=7526.0MB, aggrb=1315.5MB/s, minb=1346.7MB/s, maxb=1346.7MB/s, mint=5723msec, maxt=5723msec
Write BW:
Before:
WRITE: io=7526.0MB, aggrb=1110.2MB/s, minb=1136.9MB/s, maxb=1136.9MB/s, mint=6779msec, maxt=6779msec
After:
WRITE: io=7526.0MB, aggrb=1113.5MB/s, minb=1140.3MB/s, maxb=1140.3MB/s, mint=6759msec, maxt=6759msec
Signed-off-by: Sasha Levin <levinsasha928@xxxxxxxxx>
---
tools/kvm/disk/core.c | 2 -
tools/kvm/disk/raw.c | 38 +++++++++++++++++++++++++++++------
tools/kvm/include/kvm/disk-image.h | 8 +++++-
tools/kvm/util/read-write.c | 8 +-----
tools/kvm/virtio/blk.c | 3 ++
5 files changed, 42 insertions(+), 17 deletions(-)
diff --git a/tools/kvm/disk/core.c b/tools/kvm/disk/core.c
index 4915efd..b18014e 100644
--- a/tools/kvm/disk/core.c
+++ b/tools/kvm/disk/core.c
@@ -5,8 +5,6 @@
#include <sys/eventfd.h>
#include <sys/poll.h>
-#define AIO_MAX 32
-
int debug_iodelay;
#ifdef CONFIG_HAS_AIO
diff --git a/tools/kvm/disk/raw.c b/tools/kvm/disk/raw.c
index caa023c..1162fb7 100644
--- a/tools/kvm/disk/raw.c
+++ b/tools/kvm/disk/raw.c
@@ -10,9 +10,9 @@ ssize_t raw_image__read_sector(struct disk_image *disk, u64 sector, const struct
u64 offset = sector << SECTOR_SHIFT;
#ifdef CONFIG_HAS_AIO
- struct iocb iocb;
+ struct iocb *iocb = &disk->iocb[disk->count++];
- return aio_preadv(disk->ctx, &iocb, disk->fd, iov, iovcount, offset,
+ return aio_preadv(disk->ctx, iocb, disk->fd, iov, iovcount, offset,
disk->evt, param);
#else
return preadv_in_full(disk->fd, iov, iovcount, offset);
@@ -25,9 +25,9 @@ ssize_t raw_image__write_sector(struct disk_image *disk, u64 sector, const struc
u64 offset = sector << SECTOR_SHIFT;
#ifdef CONFIG_HAS_AIO
- struct iocb iocb;
+ struct iocb *iocb = &disk->iocb[disk->count++];
- return aio_pwritev(disk->ctx, &iocb, disk->fd, iov, iovcount, offset,
+ return aio_pwritev(disk->ctx, iocb, disk->fd, iov, iovcount, offset,
disk->evt, param);
#else
return pwritev_in_full(disk->fd, iov, iovcount, offset);
@@ -79,19 +79,33 @@ int raw_image__close(struct disk_image *disk)
close(disk->evt);
-#ifdef CONFIG_HAS_VIRTIO
+#ifdef CONFIG_HAS_AIO
io_destroy(disk->ctx);
#endif
return ret;
}
+static int raw_image__aio_submit(struct disk_image *disk)
+{
+ int ret;
+
+ ret = io_submit(disk->ctx, disk->count, disk->iocb_ptrs);
+ if (ret > 0)
+ disk->count = 0;
+
+ return ret;
+}
+
/*
* multiple buffer based disk image operations
*/
static struct disk_image_operations raw_image_regular_ops = {
.read_sector = raw_image__read_sector,
.write_sector = raw_image__write_sector,
+#ifdef CONFIG_HAS_AIO
+ .submit = raw_image__aio_submit,
+#endif
};
struct disk_image_operations ro_ops = {
@@ -120,8 +134,13 @@ struct disk_image *raw_image__probe(int fd, struct stat *st, bool readonly)
disk = disk_image__new(fd, st->st_size, &ro_ops_nowrite, DISK_IMAGE_REGULAR);
#ifdef CONFIG_HAS_AIO
- if (disk)
+ if (disk) {
+ int i;
+
disk->async = 1;
+ for (i = 0; i < AIO_MAX; i++)
+ disk->iocb_ptrs[i] = &disk->iocb[i];
+ }
#endif
}
@@ -132,8 +151,13 @@ struct disk_image *raw_image__probe(int fd, struct stat *st, bool readonly)
*/
disk = disk_image__new(fd, st->st_size, &raw_image_regular_ops, DISK_IMAGE_REGULAR);
#ifdef CONFIG_HAS_AIO
- if (disk)
+ if (disk) {
+ int i;
+
disk->async = 1;
+ for (i = 0; i < AIO_MAX; i++)
+ disk->iocb_ptrs[i] = &disk->iocb[i];
+ }
#endif
return disk;
}
diff --git a/tools/kvm/include/kvm/disk-image.h b/tools/kvm/include/kvm/disk-image.h
index 56c08da..6eba950 100644
--- a/tools/kvm/include/kvm/disk-image.h
+++ b/tools/kvm/include/kvm/disk-image.h
@@ -20,14 +20,14 @@
#define SECTOR_SHIFT 9
#define SECTOR_SIZE (1UL << SECTOR_SHIFT)
+#define AIO_MAX 256
+#define MAX_DISK_IMAGES 4
enum {
DISK_IMAGE_REGULAR,
DISK_IMAGE_MMAP,
};
-#define MAX_DISK_IMAGES 4
-
struct disk_image;
struct disk_image_operations {
@@ -37,6 +37,7 @@ struct disk_image_operations {
int iovcount, void *param);
int (*flush)(struct disk_image *disk);
int (*close)(struct disk_image *disk);
+ int (*submit)(struct disk_image *disk);
};
struct disk_image {
@@ -50,6 +51,9 @@ struct disk_image {
int evt;
#ifdef CONFIG_HAS_AIO
io_context_t ctx;
+ struct iocb iocb[AIO_MAX];
+ struct iocb *iocb_ptrs[AIO_MAX];
+ u32 count;
#endif
};
diff --git a/tools/kvm/util/read-write.c b/tools/kvm/util/read-write.c
index 55473ba..4ccd554 100644
--- a/tools/kvm/util/read-write.c
+++ b/tools/kvm/util/read-write.c
@@ -321,24 +321,20 @@ ssize_t pwritev_in_full(int fd, const struct iovec *iov, int iovcnt, off_t offse
int aio_pwritev(io_context_t ctx, struct iocb *iocb, int fd, const struct iovec *iov, int iovcnt,
off_t offset, int ev, void *param)
{
- struct iocb *ios[1] = { iocb };
-
io_prep_pwritev(iocb, fd, iov, iovcnt, offset);
io_set_eventfd(iocb, ev);
iocb->data = param;
- return io_submit(ctx, 1, ios);
+ return 0;
}
int aio_preadv(io_context_t ctx, struct iocb *iocb, int fd, const struct iovec *iov, int iovcnt,
off_t offset, int ev, void *param)
{
- struct iocb *ios[1] = { iocb };
-
io_prep_preadv(iocb, fd, iov, iovcnt, offset);
io_set_eventfd(iocb, ev);
iocb->data = param;
- return io_submit(ctx, 1, ios);
+ return 0;
}
#endif
\ No newline at end of file
diff --git a/tools/kvm/virtio/blk.c b/tools/kvm/virtio/blk.c
index d1a0197..153b11b 100644
--- a/tools/kvm/virtio/blk.c
+++ b/tools/kvm/virtio/blk.c
@@ -128,6 +128,9 @@ static void virtio_blk_do_io(struct kvm *kvm, struct virt_queue *vq, struct blk_
virtio_blk_do_io_request(kvm, req);
}
+
+ if (bdev->disk->ops->submit)
+ bdev->disk->ops->submit(bdev->disk);
}
static void set_config(struct kvm *kvm, void *dev, u8 data, u32 offset)
--
1.7.8
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
