doc: add VBSK overhead analysis doc

Add a new developer guide describing VBSK overhead analysis.

Signed-off-by: David B. Kinder <david.b.kinder@intel.com>

doc/developer-guides/VBSK-analysis.rst

@@ -0,0 +1,147 @@
+.. _vbsk-overhead:
+
+VBS-K Framework Virtualization Overhead Analysis
+################################################
+
+Introduction
+************
+
+The ACRN Hypervisor follows the Virtual I/O Device (virtio)
+specification to realize I/O virtualization for many
+performance-critical devices supported in the ACRN project. The
+hypervisor provides the virtio backend service (VBS) APIs, that make it
+very straightforward to implement a virtio device in the hypervisor. We
+can evaluate the virtio backend service in kernel-land (VBS-K) framework
+overhead through a test virtual device called virtio-echo. The total
+overhead of a frontend-backend application based on VBS-K consists
+of VBS-K framework overhead and application-specific overhead. The
+application-specific overhead depends on the specific frontend-backend
+design, from microseconds to seconds. In our HW case, the overall VBS-K
+framework overhead is on the microsecond level, sufficient to meet the
+needs of most applications.
+
+Architecture of VIRTIO-ECHO
+***************************
+
+virtio-echo is a virtual device based on virtio, and designed for
+testing ACRN virtio backend services in the kernel (VBS-K) framework. It
+includes a virtio-echo frontend driver, a virtio-echo driver in ACRN
+device model (DM) for initialization, and a virtio-echo driver based on
+VBS-K for data reception and transmission. For more virtualization
+background introduction, please refer to:
+
+* :ref:`introduction`
+* :ref:`virtio-hld`
+
+virtio-echo is implemented as a virtio legacy device in the ACRN device
+model (DM), and  is registered as a PCI virtio device to the guest OS
+(UOS). The virtio-echo software  has three parts:
+
+-  **virtio-echo Frontend Driver**: This driver runs in the UOS. It prepares
+   the RXQ and notifies the backend for receiving incoming data when the
+   UOS starts. Second, it copies the received data from the RXQ to TXQ
+   and sends them to the backend. After receiving the message that the
+   transmission is completed, it starts again another round of reception
+   and transmission, and keeps running until a specified number of cycle
+   is reached.
+-  **virtio-echo Driver in DM**: This driver is used for initialization
+   configuration. It simulates a virtual PCI device for the frontend
+   driver use, and sets necessary information such as the device
+   configuration and virtqueue information to the VBS-K. After
+   initialization, all data exchange are taken over by the VBS-K
+   vbs-echo driver.
+-  **vbs-echo Backend Driver**: This driver sets all frontend RX buffers to
+   be a specific value and sends the data to the frontend driver. After
+   receiving the data in RXQ, the fronted driver copies the data to the
+   TXQ, and then sends them back to the backend. The backend driver then
+   notifies the frontend driver that the data in the TXQ has
+   been successfully received. In virtio-echo, the backend driver
+   doesn't process or use the received data.
+
+:numref:`vbsk-virtio-echo-arch` shows the whole architecture of virtio-echo.
+
+.. figure:: images/vbsk-image2.png
+   :width: 900px
+   :align: center
+   :name: vbsk-virtio-echo-arch
+
+   virtio-echo Architecture
+
+Virtualization Overhead Analysis
+********************************
+
+Let's analyze the overhead of the VBS-K framework. As we know, the VBS-K
+handles notifications in the SOS kernel instead of in the SOS user space
+DM. This can avoid overhead from switching between kernel space and user
+space. Virtqueues are allocated by UOS, and virtqueue information is
+configured to VBS-K backend by the virtio-echo driver in DM, thus
+virtqueues can be shared between UOS and SOS. There is no copy overhead
+in this sense. The overhead of VBS-K framework mainly contains two
+parts: kick overhead and notify overhead.
+
+-  **Kick Overhead**: The UOS gets trapped when it executes sensitive
+   instructions that notify the hypervisor first. The notification is
+   assembled into an IOREQ, saved in a shared IO page, and then
+   forwarded to the VHM module by the hypervisor. The VHM notifies its
+   client for this IOREQ, in this case, the client is the vbs-echo
+   backend driver. Kick overhead is defined as the interval from the
+   beginning of UOS trap to a specific VBS-K driver e.g. when
+   virtio-echo gets notified.
+-  **Notify Overhead**: After the data in virtqueue being processed by the
+   backend driver, vbs-echo calls the VHM module to inject an interrupt
+   into the frontend. The VHM then uses the hypercall provided by the
+   hypervisor, which causes a UOS VMEXIT. The hypervisor finally injects
+   an interrupt into the vLAPIC of the UOS and resumes it. The UOS
+   therefore receives the interrupt notification.  Notify overhead is
+   defined as the interval from the beginning of the interrupt injection
+   to when the UOS starts interrupt processing.
+
+The overhead of a specific application based on VBS-K includes two
+parts: VBS-K framework overhead and application-specific overhead.
+
+-  **VBS-K Framework Overhead**: As defined above, VBS-K framework overhead
+   refers to kick overhead and notify overhead.
+-  **Application-Specific Overhead**: A specific virtual device has its own
+   frontend driver and backend driver. The application-specific overhead
+   depends on its own design.
+
+:numref:`vbsk-virtio-echo-e2e` shows the overhead of one end-to-end
+operation in virtio-echo.  Overhead of steps marked as red are caused by
+the virtualization scheme based on VBS-K framework. Costs of one "kick"
+operation and one "notify" operation are both on a microsecond level.
+Overhead of steps marked as blue depend on specific frontend and backend
+virtual device drivers. For virtio-echo, the whole end-to-end process
+(from step1 to step 9) costs about 4 dozens of microsecond. That's
+because virtio-echo does little things in its frontend and backend
+driver which is just for testing and there is very little process
+overhead.
+
+.. figure:: images/vbsk-image1.png
+   :width: 600px
+   :align: center
+   :name: vbsk-virtio-echo-e2e
+
+   End to End Overhead of virtio-echo
+
+:numref:`vbsk-virtio-echo-path` details the path of kick and notify
+operation shown in :numref:`vbsk-virtio-echo-e2e`. The VBS-K framework
+overhead is caused by operations through these paths. As we can see, all
+these operations are processed in kernel mode which avoids extra
+overhead of passing IOREQ to userspace processing.
+
+.. figure:: images/vbsk-image3.png
+   :width: 900px
+   :align: center
+   :name: vbsk-virtio-echo-path
+
+   Path of VBS-K Framework Overhead
+
+Conclusion
+**********
+
+Unlike VBS-U processing in user mode, VBS-K moves things into the kernel
+mode and can be used to accelerate processing. A virtual device
+virtio-echo based on VBS-K framework is used to evaluate the VBS-K
+framework overhead. In our test, the VBS-K framework overhead (one kick
+operation and one notify operation) is on the microsecond level which
+can meet the needs of most applications.

doc/developer-guides/index.rst

@@ -12,6 +12,7 @@ Developer Guides
    GVT-g-kernel-options
    trusty
    l1tf
+   VBSK-analysis
    modularity
    ../api/index
    ../reference/kconfig/index