KR101387986B1 - Virtualiztion apparatus - Google Patents

Abstract

The virtualization apparatus according to the present invention comprises a virtual machine including a guest kernel and one or more guest user processes and inserting input / output requests for input and output of one or more physical devices into a shared queue, a hypervisor module for driving the virtual machine, and the physical device. A host kernel comprising one or more device drivers for the guest kernel, wherein the guest kernel includes one or more device drivers and an interprocessor interrupt handler for interacting with the hypervisor module, the hypervisor module in the shared queue. A device emulation thread that checks whether there is an inserted I / O request, handles the polling thread that passes the I / O request to the device emulation thread, and processes the I / O request, and delivers a completion message to the CPU through the interprocessor interrupt handler. It corresponds to a virtualization apparatus including a.

Description

Virtualization Device {VIRTUALIZTION APPARATUS}

The present invention relates to a virtualization device.

Virtualization technology is a concept introduced to virtually divide one mainframe server into several machines. Recently, virtualization technology has been widely used, ranging from small mobile devices to large clustered computing environments that provide cloud services. Virtualization makes devices more efficient to use and has the advantages of portability between different machines, as well as the ability to migrate to another machine while the system is running.

The most basic device virtualization method is the traditional trap-end-emulate method. When the device driver of the guest operating system requests I / O, the switch to the hypervisor occurs because it uses privileged instructions that cannot be executed directly in guest mode. The hypervisor handles the request through device emulation. In addition, when the request is completed, the interrupt is delivered to the guest through the reverse process. As such, I / O in a virtual environment is more complicated than bare-metal environment.

On the other hand, due to the load of device virtualization due to frequent mode switching, I / O-intensive workloads perform very poorly in virtualized environments. Mode switching in each part of the device virtualization process not only costs itself, but also indirect costs, such as polluting the hardware cache, as well as the cost of stopping the entire virtual machine.

Specifically, the mode switching of device virtualization can be largely divided into three processes according to the timing of occurrence. It is the process of requesting I / O in virtualized environment, emulating real device, and delivering completion message after I / O is performed.

During the I / O request process, the guest operating system cannot process I / O commands directly, so special processing is required. In the device virtualization method using the traditional trap-end-emulation method, I / O requests of such guest operating systems generate hardware exceptions. The hypervisor uses these exceptions to determine if I / O has occurred. However, this method involves mode switching between the virtual machine and the hypervisor. In addition, the translation look aside buffer (TLB) is flushed and the hardware cache is polluted when the mode is switched, which completely excludes the advantage of a cache that utilizes locality, thereby degrading system performance. Will not only run the user process that requested I / O, but all processes that have nothing to do with it.

Similar modes of mode switching occur in device emulation and completion message delivery, limiting virtualization performance. At this time, the cost of mode switching can be divided into three categories. First, mode switching uses processor resources to accomplish this. The resulting costs are called direct costs. The second is the indirect cost incurred because the processor executes code that uses different contexts before and after the mode switch. When switching modes, the hardware cache is dirty and a TLB flushes. This has a large impact on cache performance utilizing locality, which causes a lot of load. The third is the synchronization cost incurred by not being able to perform the part that is not directly related to the mode change.

As such, the costs incurred by mode switching are much higher indirect and synchronization costs than direct costs. Therefore, in order to improve I / O performance of virtual machines to bare metal performance, it is necessary to reduce the number of mode switching in the device virtualization process in the virtualization environment.

In this regard, Korean Patent Laid-Open Publication No. 10-2011-0041087 (name of the invention: server virtualization apparatus) performs a host-guest conversion through a virtualization module supported by a CPU of a computer, and controls the control data of a guest machine. A server virtualization apparatus including a hypervisor module is disclosed.

In addition, Japanese Patent Laid-Open No. 2009-288864 (name of the invention: an input / output emulation mechanism of a virtual machine) uses an polling technique as a method for shortening the time required for access according to the operation of a guest-host device, and using an I / O instruction. The process of passing a request to the emulation is disclosed.

The present invention is to solve the above-mentioned problems of the prior art, some embodiments of the present invention is a virtualization eliminating the mode switching occurs when the I / O request, mode switching of the I / O emulation process, mode switching of the interrupt processing process It is an object to provide a device.

As a technical means for achieving the above-described technical problem, the virtualization apparatus according to the first aspect of the present invention includes a guest kernel and one or more guest user processes and inserting input and output requests for input and output of one or more physical devices to the shared queue A host kernel including a virtual machine and a hypervisor module for driving the virtual machine and one or more device drivers for the physical device, wherein the guest kernel includes one or more device drivers and a processor for interacting with the hypervisor module. And an interrupt handler, wherein the hypervisor module checks whether there is an I / O request inserted in the shared queue, processes a polling thread that passes the I / O request to a device emulation thread, and processes the I / O request. Interrupt Through the handler and a device emulation thread passing a completion message to the CPU.

In addition, the virtualization method in the virtualization apparatus according to the second aspect of the present invention comprises the steps of the virtual machine to insert the input / output request to the shared queue for the input and output of the physical device, the input and output request inserted in the shared queue to the hypervisor module Checking by an included polling thread; if the input / output request is in the shared queue, the polling thread forwards the input / output request to a device emulation thread included in the hypervisor module; and directly by the device emulation thread Processing the request and after the device emulation thread processes the input / output request, sending a completion message to the CPU through an interprocessor interrupt handler included in a guest kernel in the virtual machine.

According to one embodiment of the above-described problem solving means of the present invention, it is possible to reduce the cost incurred by the mode switching to achieve the performance close to the bare metal. In other words, the mode change that occurs during I / O request is eliminated by the I / O request technique without switching, and the mode change of the I / O emulation process is removed by the internal device emulation technique of the hypervisor, and the mode change in the interrupt processing is Elimination can be achieved through a switchless I / O completion processing technique. These three techniques reduce the burden of device virtualization by eliminating mode switching in multicore architectures.

1 is a diagram illustrating a virtualization apparatus according to an embodiment of the present invention.
2 is a diagram illustrating a virtualization method in a virtualization apparatus according to an embodiment of the present invention.
3 is a diagram illustrating a procedure of processing an interrupt by a virtualization apparatus according to an exemplary embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, which will be readily apparent to those skilled in the art. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and similar parts are denoted by like reference characters throughout the specification.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "electrically connected" with another part in between . Also, when an element is referred to as "comprising ", it means that it can include other elements as well, without departing from the other elements unless specifically stated otherwise.

1 is a diagram illustrating a virtualization apparatus according to an embodiment of the present invention.

The virtualization apparatus 100 according to the present invention includes a host kernel 110 including a hypervisor module 112 and one or more device drivers 114 and a guest kernel 122 and one or more guest user processes 124. Virtual machine 120. At this time, the hypervisor module 112 includes a polling thread 116 and a device emulation thread 115. Guest kernel 122 also includes an interprocessor interrupt handler 126 and one or more device drivers 125.

1 refers to a hardware component such as software or an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit), and performs predetermined roles .

However, 'components' are not meant to be limited to software or hardware, and each component may be configured to reside on an addressable storage medium and configured to play one or more processors.

Thus, by way of example, an element may comprise components such as software components, object-oriented software components, class components and task components, processes, functions, attributes, procedures, Routines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables.

The components and functions provided within those components may be combined into a smaller number of components or further separated into additional components.

The hypervisor module 112 serves to drive the virtual machine 120. In particular, the device driver 125 for interacting with the hypervisor module 112 is contained within the guest kernel 122 of the virtual machine 120. When the virtual machine 120 inserts an input / output request into the shared queue (not shown) through the device driver 125 for input and output of the physical device 130, the hypervisor module 112 through the polling thread 116. After checking whether there is an I / O request, the I / O request is forwarded to the device emulation thread 115 to process it. Here, the shared queue is located in the hypervisor module 112 and serves to transfer data without additional replication between the hypervisor module 112 and the virtual machine 120.

That is, the virtualization apparatus 100 according to an embodiment of the present invention interacts with the hypervisor module 112 through a shared queue using the device driver 125 using a paravirtualization technique when an input / output request is made. This is called a switchless I / O request technique.

On the other hand, since the I / O request in the guest operating system is not a command that can be directly processed by the virtual machine, it has to be handled through a mode switch to a hypervisor module which is expensive in the past. However, since the virtual device 100 according to an embodiment of the present invention does not generate a mode switch when an input / output request occurs, the virtual device 100 may effectively remove the request from the shared queue and transfer the request to the next process.

The device emulation thread 115 included in the hypervisor module 112 processes the input / output request transmitted by the polling thread 116. That is, the device emulation thread 115 emulates the corresponding physical device 130 through the device driver 114. In this case, since the device emulation thread 115 uses the same context as the hypervisor module 112, no additional mode switching is required. This is called the hypervisor internal device emulation technique.

Meanwhile, in the conventional virtualization apparatus, input / output requests are processed through a rather complicated process. For example, Xen handled it through an independent device domain, and QEMU emulated I / O requests through a user process. This approach increases the device emulation cost because additional switching between domains or between user-kernels is required. On the other hand, the virtualization apparatus 100 according to an embodiment of the present invention may process the input / output request directly in the hypervisor module 112 without the help of other domains or user processes in order to reduce the device response speed.

Experimental results of measuring the sequential read and random read rates by applying the virtualization apparatus 100 according to an embodiment of the present invention are as follows.

The sequential read and random read speeds are improved by 4% and 13%, respectively, using the switchless I / O request method, and the sequential and random read speeds are increased by 2% and 11%, respectively, using only the hypervisor internal device emulation method. That is, compared to the case where the above technique is not applied through the virtualization apparatus 100, the sequential read speed may be improved by 6% and the random read speed by 24%, thereby having a performance close to that of bare metal performance.

After the device emulation thread 115 processes the input / output request as described above, the device emulation thread 115 passes to the CPU through the interprocessor interrupt handler 126 included in the guest kernel 122 in the virtual machine 120. Send a completion message.

At this time, the inter-processor interrupt used in the conventional virtualization apparatus generates two mode switching. The first mode switch occurs when an interrupt is sent to the virtual machine. Interrupts sent to the virtual machine for interrupt handling are passed back to the hypervisor module through mode switching. The hypervisor module determines whether the interrupt should be delivered to the virtual machine, and then passes the interrupt back to the virtual machine for processing. The second mode switch occurs when the LAPIC (Local Advanced Programmable Interrupt Controller) is notified after the interrupt has been processed. Accessing the end-of-interrupt (EOI) registers from the guest operating system is a command that cannot be processed by the virtual machine like I / O requests, so mode switching is required.

The virtualization apparatus 100 according to an embodiment of the present invention removes the mode switching of the interrupt process using the following method.

The first mode switch can be removed by modifying the interrupt descriptor table (IDT) of the guest operating system. That is, the mode switch can be removed by modifying the interrupt descriptor table of the guest operating system included in the virtual machine 120 to directly receive an interrupt indicating that the I / O is completed according to the I / O request.

The second mode switch can be removed using x2APIC. x2APIC is a new version of LAPIC that allows the virtual machine 120 to send an interrupt handling termination message to the CPU without going through the hypervisor module 112 using machine special registers (MSRs) instead of traditional port I / O or memory I / O methods. . Through X2APIC, EOI can be sent an interrupt processing termination message without changing mode. The method of eliminating the two mode switching is referred to as an input / output completion processing technique without switching.

Although the above-described switching-free input / output completion processing technique can be applied separately, it can be applied simultaneously with the above-described two techniques (switch-less input / output request technique and hypervisor internal device emulation technique). When the input / output completion processing technique without switching is applied in the present invention, the performance equivalent to bare metal can be obtained by further improving the sequential read speed and the random read speed independently or by interacting with the above two techniques.

2 is a diagram illustrating a virtualization method in a virtualization apparatus according to an embodiment of the present invention.

First, when the virtual machine 120 makes an input / output request for input / output of the physical device 130 (S210), the device driver 125 included in the virtual machine 120 inserts an input / output request into the shared queue (S220). . In other words, an I / O request is like enqueuing an I / O request to a shared queue.

Next, the polling thread 116 included in the hypervisor module 112 checks whether the input / output request is in the shared queue, and then transmits the input / output request to the device emulation thread 115 (S230). The device emulation thread 115 directly processes the input / output request transmitted from the polling thread 116 (S240).

Next, the device emulation thread 115 transmits a completion message to the CPU running the guest operating system through the interrupt interprocessor handler (IPI) 126 (S250).

3 is a diagram illustrating a procedure of processing an interrupt by a virtualization apparatus according to an exemplary embodiment of the present invention.

The physical device 130 generates an interrupt when the input / output according to the input / output request of the virtual machine 120 is completed (S310).

Next, the virtual machine modifies the interrupt descriptor table of the virtual machine 120 to process the interrupt (S320). Conventional virtual machines do not handle interrupts, so additional mode switching occurs. On the other hand, the virtualization apparatus 100 according to an embodiment of the present invention corrects the interrupt descriptor table to directly receive an interrupt indicating that the I / O is completed according to the I / O request, thereby directly processing the interrupt and thus eliminating the mode change. can do.

Next, after processing the interrupt, the virtual machine 120 directly accesses the EOI register using the x2APIC to transfer an interrupt processing end message to the CPU without mode switching (S330). In this case, the x2APIC may transmit an interrupt processing end message to the CPU without the help of the hypervisor module 112 using the machine special register.

One embodiment of the present invention may also be embodied in the form of a recording medium including instructions executable by a computer, such as program modules, being executed by a computer. Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. In addition, the computer-readable medium may include both computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Communication media typically includes any information delivery media, including computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave, or other transport mechanism.

While the methods and systems of the present invention have been described in connection with specific embodiments, some or all of those elements or operations may be implemented using a computer system having a general purpose hardware architecture.

The foregoing description of the present invention is intended for illustration, and it will be understood by those skilled in the art that the present invention may be easily modified in other specific forms without changing the technical spirit or essential features of the present invention. will be. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

The scope of the present invention is defined by the appended claims rather than the detailed description and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

100: virtualization device 110: host kernel
112: hypervisor module 114: device driver
115: device emulation thread 116: polling thread
120: virtual machine 122: guest kernel
124: Guest User Process 125: Device Driver
126: Interprocessor Interrupt Handler

Claims (6)

In the virtualization device,
A virtual machine including a guest kernel and one or more guest user processes and inserting input and output requests for input and output of one or more physical devices to a shared queue;
A host kernel including a hypervisor module for driving the virtual machine and one or more device drivers for the physical device,
The guest kernel includes one or more device drivers and an interprocessor interrupt handler for interacting with the hypervisor module,
The hypervisor module
The shared queue,
A polling thread that checks whether there is an input / output request inserted in the shared queue, and passes the input / output request to a device emulation thread;
And a device emulation thread that processes the input / output request and delivers a completion message to a CPU through the interprocessor interrupt handler.
The method according to claim 1,
The virtual machine has modified the interrupt descriptor table of the guest operating system included in the virtual machine to directly receive the interrupt that completed the input and output according to the input and output request.
The method according to claim 1,
After the interrupt processing, the virtual machine transfers the interrupt processing end message to the CPU using a machine special register (MSR) without passing through the hypervisor module.
In the virtualization method in the virtualization device,
The virtual machine inserting an I / O request into the shared queue included in the hypervisor module for the I / O of the physical device,
Checking, by the polling thread included in the hypervisor module, the I / O request inserted into the shared queue;
If the input / output request is in the shared queue, the polling thread forwarding the input / output request to a device emulation thread included in the hypervisor module;
The device emulation thread directly processing the input / output request; and
And after the device emulation thread processes the input / output request, sending a completion message to a CPU through an interprocessor interrupt handler included in a guest kernel in the virtual machine.
5. The method of claim 4,
And the virtual machine modifies the interrupt descriptor table of the guest operating system included in the virtual machine to directly receive an interrupt indicating that the I / O is completed according to the I / O request.
5. The method of claim 4,
After the interrupt processing, the virtual machine transfers the interrupt processing end message to the CPU using a machine special register (MSR) without passing through the hypervisor module.

Images