numam-spdk/doc/virtio.md
Dariusz Stojaczyk 71ea826507 bdev_virtio: added doc page
Change-Id: Ia88ae52117068ac395dad9ad3d7ac818e41077fb
Signed-off-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/380956
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
2017-10-20 19:44:23 -04:00

42 lines
1.8 KiB
Markdown

# Virtio SCSI driver {#virtio}
# Introduction {#virtio_intro}
Virtio SCSI driver is an initiator for SPDK @ref vhost application. The
driver allows any SPDK app to connect to another SPDK instance exposing
a vhost-scsi device. The driver will enumerate targets on the device (which acts
as a SCSI controller) and create *virtual* bdevs usable by any SPDK application.
Sending an I/O request to the Virtio SCSI bdev will put the request data into
a Virtio queue that is processed by the host SPDK app exposing the
controller. The host, after sending I/O to the real drive, will put the response
back into the Virtio queue. Then, the response is received by the Virtio SCSI
driver.
The driver, just like the SPDK @ref vhost, is using pollers instead of standard
interrupts to check for an I/O response. It bypasses kernel interrupt and context
switching overhead of QEMU and guest kernel, significantly boosting the overall
I/O performance.
Virtio SCSI driver supports two different usage models:
* PCI - This is the standard mode of operation when used in a guest virtual
machine, where QEMU has presented the virtio-scsi controller as a virtual
PCI device.
* User vhost - Can be used to connect to a vhost-scsi socket directly on the
same host.
# Multiqueue {#virtio_multiqueue}
The Virtio SCSI controller will automatically manage virtio queue distribution.
Currently each thread doing an I/O on a single bdev will get an exclusive queue.
Multi-threaded I/O on bdevs from a single Virtio-SCSI controller is not supported.
# Limitations {#virtio_limitations}
The Virtio SCSI driver is still experimental. Current implementation has many
limitations:
* supports only up to 8 hugepages (implies only 1GB sized pages are practical)
* single LUN per target
* only SPDK vhost-scsi controllers supported
* no RPC
* no multi-threaded I/O for single-queue virtio devices