WO2017208360A1

WO2017208360A1 - Computer system

Info

Publication number: WO2017208360A1
Application number: PCT/JP2016/066056
Authority: WO
Inventors: 雅士高田; 睦細谷; 本村　哲朗
Original assignee: 株式会社日立製作所
Priority date: 2016-05-31
Filing date: 2016-05-31
Publication date: 2017-12-07

Abstract

A computer system comprising a partially configurable programmable logic device, a hypervisor, and a management computer, wherein: the programmable logic device includes a plurality of configurable sectors; the hypervisor or the management computer selects available predetermined sectors from among the plurality of sectors when creating a virtual computer; and the hypervisor allocates the predetermined sectors exclusively to the virtual computer.

Description

Computer system

The present invention generally relates to computer technology, for example, computer virtualization technology.

The use of Field Programmable Gate Array (FPGA), which has a high power performance ratio compared to Central Processing Unit (CPU), in data centers and the like is being studied. The technique disclosed in Patent Document 1 discloses a technique for providing and managing access to an FPGA as a service. Specifically, a system that enables (A) provisioning management for use of a shared FPGA and access control to allow a user's process to access a user-programmed FPGA is a data center server. A user package can be accepted to compile for an FPGA that communicates with. (B) The user package can be image copied to the FPGA along with the attached management payload, and the driver and user key can be used to selectively access the FPGA as a service for the data center virtual machine. Can be made possible. (C) Together, these elements enable the data center to provide an integrated circuit (ICaaS) as a rentable service. It is disclosed that additional services such as billing tracking, provision management, and access control can be provided to the user, enabling lower costs while realizing significant benefits for the data center. Has been.

Special table 2015-507234 gazette

Server resources have increased due to semiconductor miniaturization, and a large number of virtual machines can be executed on one server computer. However, in the virtual machine, there is a high possibility that extra FPGA resources will be left. This is because, unlike CPUs, FPGAs are difficult to share by time division in a short time. Therefore, efficient utilization of FPGA resources is required. Another problem is that when a plurality of virtual machines used by a plurality of users share the resources of one FPGA, information related to the users of each virtual machine is transferred to the other users via the FPGA. It is necessary to prevent it from leaking to the person. An object of the present invention is to efficiently use resources of a programmable logic device such as an FPGA.

A computer system according to an embodiment includes a partially configurable programmable logic device, a hypervisor, and a management computer. The programmable logic device includes a plurality of configurable sectors, and the hypervisor Alternatively, when creating a virtual machine, the management computer selects a predetermined sector that is free from a plurality of sectors, and the hypervisor assigns the predetermined sector exclusively to the virtual machine.

According to the present invention, the resources of the programmable logic device can be used efficiently.

The structural example of the computer system which concerns on this embodiment is shown. The example of the hardware constitutions of a server is shown. An example of the logical configuration of the FPGA is shown. The structural example of a VM management table is shown. The structural example of a FPGA sector management table is shown. It is a flowchart which shows the operation example of the hypervisor in the validation process of FPGA. The example of the management screen for VM creation is shown. It is a flowchart which shows the operation example of the hypervisor in the VM new creation process. It is a flowchart which shows the operation example of the hypervisor in the sector initialization process of FPGA. The example of the management screen for VM change is shown. It is a flowchart which shows the operation example of the hypervisor in the structure change process of VM. The example of the management screen for VM deletion is shown. It is a flowchart which shows the operation example of the hypervisor in the deletion process of VM. It is a flowchart which shows the operation example of the hypervisor in the forced initialization process of a sector. The modification of a computer system is shown.

Hereinafter, an embodiment will be described. In the following description, information may be described using the expression “abc table”, but the information may be expressed in any data structure. That is, the “abc table” can be referred to as “abc information” to indicate that the information does not depend on the data structure.

In the following description, the process may be described using “program” as a subject. However, the program is executed by a processor (for example, a CPU (Central Processing Unit)), so that a predetermined process is appropriately performed. Since at least one of a storage resource (for example, memory) and a communication interface device is used, the subject of processing may be a processor and an apparatus having the processor. Part or all of the processing performed by the processor may be performed by a hardware circuit. The computer program may be installed from a program source. The program source may be a program distribution server or a storage medium (for example, a portable storage medium). The “CPU” may be a so-called core inside the CPU, but for the sake of simplicity of explanation, the “CPU” will be described as being unified. When a CPU and a core are clearly distinguished and called, a core inside the CPU is called a “CPU core”, and when referring to a CPU including a plurality of cores, it is called a “CPU package”. For example, the CPU package may include a CPU core and a programmable logic device.

The management computer also includes an interface device connected to one or more I / O devices including a display device, a storage resource (for example, a memory) that stores information, and a processor connected to the interface device and the storage resource. . The display device may be a display device included in the management computer or a display computer connected to the management computer. The I / O device may be an I / O device (for example, a keyboard and a pointing device or a touch panel) included in the management computer, a display computer connected to the management computer, or another computer. That the management computer “displays the display information” is to display the display information on the display device, and this may be to display the display information on the display device of the management computer. The management computer may transmit display information to the display computer (in the latter case, the display information is displayed by the display computer). The management computer inputting / outputting information may be inputting / outputting information to / from an I / O device of the management computer, or a remote computer connected to the management computer (for example, a display) Information may be input / output to / from the computer. The information output may be a display of information.

In the following description, when the same type of elements are distinguished and described, reference numerals are used such as “sector 120a” and “sector 120b”, and the same types of elements are not distinguished. In some cases, only a common number among the reference symbols is used, such as “sector 120”.

FIG. 1 shows a configuration example of a computer system 1 according to this embodiment.

The computer system 1 includes a server 20, a client 30, a management computer 10, a management client 34, a WEB server 32, and a storage device 36.

The client 30, the WEB server 32, and the server 20 may be connected via a wide area network (WAN) 40, which is an example of a communication network, so that bidirectional communication is possible. The management client 34, the WEB server 32, the management computer 10, and the server 20 may be connected via a local area network (LAN) 42, which is an example of a communication network, so that bidirectional communication is possible. The storage device 36 and the server 20 may be connected via a storage area network (SAN) 44, which is an example of a communication network, so that bidirectional communication is possible.

The server 20 is an example of a computer and includes a hypervisor 200 and an FPGA 23. The hypervisor 200 is a mechanism for realizing a virtual machine (VM) 220 by controlling physical resources (for example, computing resources, storage resources, communication resources, etc.) of the server 20. The FPGA 23 is an example of a programmable logic device, and is a device that can configure a desired logic circuit. The FPGA 23 may have a plurality of sectors, and only a certain sector may be configured (that is, partially).

The storage device (storage system) 36 is a device capable of storing data. The storage device 36 may include a storage device such as a hard disk drive (HDD), a solid state drive (SSD), or a flash memory. The storage apparatus 36 may configure a Redundant Arrays of Inexpensive Disks (RAID) group using a plurality of storage devices.

The VM 220 operating on the server 20 may be able to use a logical storage area (logical volume) provided by the storage device 36 (read and write data).

The client 30 is a computer (terminal) used by the user of the VM 220.

The management computer (or management system) 10 is a computer for managing the server 20 and the storage device 36 included in the computer system 1.

The management client 34 is a computer (terminal) used by the administrator of the computer system 1. An administrator may be able to operate the management computer 10 via the management client 34.

The WEB server 32 may provide a WEB page (Graphical User Interface (GUI)) to the client 30 or the management client 34. For example, the WEB server 32 provides the management client 34 with a WEB page for operating the management computer 10. For example, the WEB server 32 provides the client 30 with a WEB page for operating the VM 220. As an example of the operation, VM creation, VM deletion, or operation or display of a program executed in the VM can be considered.

FIG. 2 shows an example of the hardware configuration of the server 20.

The server 20 may include a CPU 21, a memory 22, a Network Interface Card (NIC) 24, a Host Bus Adapter (HBA) 25, and an FPGA 23 as hardware configuration elements. These elements 21 to 25 may be connected by a bus 26 capable of bidirectional communication.

The CPU 21 is a device for realizing various functions of the server 20 by processing programs and data stored in the memory 22.

The memory 22 is a device for storing programs and data used by the CPU 21 and

other devices

23, 24 and 25. Examples of the memory 22 are Dynamic Random Access Memory (DRAM), Magnetically Responsive Random Access Memory (MRAM), and Ferroelectric Random Access Memory (FeRAM).

The NIC 24 is an interface device for connecting the server 20 to the WAN 40 or the LAN 42. An example of the NIC 24 is an Ethernet (registered trademark) adapter compatible with PCI Express (PCIe).

The HBA 25 is an interface device for connecting the server 20 to the SAN 44. An example of the HBA 25 is a PCIe compatible fiber channel adapter. Note that the HBA 25 may be integrated into the NIC 24.

The FPGA 23 is a kind of programmable logic device and is a device capable of partially configuring an arbitrary logic circuit. An example of the FPGA 23 is a PCIe compatible FPGA card. When viewed as a package, an SRAM memory cell type FPGA has a volatile memory (SRAM) for storing configuration data inside, and a circuit called a logic block simulates the circuit operation by referring to the volatile memory. Yes. Therefore, so-called configuration including loading of configuration data into the volatile memory is required every time the FPGA package is powered on, which is complicated. For this reason, when the FPGA 23 is provided in the form of a card, a nonvolatile memory storing configuration data is mounted on the card. When the FPGA card is powered on, the FPGA package performs configuration (internal configuration) by reading configuration data from the nonvolatile memory. The external configuration includes (1) transmitting configuration data from outside the FPGA 23 to the FPGA 23, and (2) performing configuration of the FPGA with configuration data transmitted from the outside (data from the outside The data may be stored in the nonvolatile memory of the FPGA 23 and read from there, or configuration data from the outside may be received directly). The “configuration” in the following description of the embodiment will be described assuming this external configuration.

FIG. 3 shows an example of the logic (function) configuration of the FPGA 23.

The FPGA 23 may include an endpoint 110, a configuration controller 130, a configurable area 100, and an authentication module 124 as elements of a logical (functional) configuration.

The endpoint 110 provides an endpoint in PCIe to an external element (such as the hypervisor 200). The endpoint 110 corresponds to Single Root I / O Virtualization (SR-IOV), and can provide at least one physical function (PF) 112 and a plurality of virtual functions (VF) 114. The hypervisor 200 exclusively allocates the VF 114 to each VM 220. Note that the PF 112 may be occupied by the hypervisor (that is, the PF 112 is not shown as a function in the PCI Express of the VM 220).

Configurable area 100 is an area where an arbitrary logic circuit can be partially configured. The configurable area 100 may include a plurality of sectors 120 each configurable.

The configuration controller 130 can receive configuration data and configure a logic circuit in at least a part of the configurable area 100. The configuration data is created by processing the source described by, for example, Hardware Description Language (HDL) with logic synthesis software. The endpoint 110, the configuration controller 130, and the authentication module 124 described above may be configured by configuration data, and a circuit may physically exist in the FPGA 23.

When the FPGA 23 is activated, the configuration controller 130 may configure a basic logic circuit in the configurable area 100 based on the basic configuration data. The basic logic circuit may include a plurality of sectors 120, a plurality of access control modules 122, and a bus 132, for example, as shown in FIG. That is, the basic configuration data may include configuration data for a plurality of sectors 120, a plurality of access control modules 122, and a bus 132.

Arbitrary logic circuits can be configured (partial configuration) in the sector 120. That is, the configuration controller 130 can configure a logic circuit in the sector 120 based on the configuration file.

The access control module 122 is a module that controls access from the VF 114 to the sector 120 associated (connected) with itself. In the access control module 122, the VF ID of the VF 114 that is permitted to access the sector 120 may be set. In this case, the access control module 122 permits access from the VF 114 conforming to the VF ID set for itself to the sector 120 associated with itself, and denies access from other VFs. It's okay. The access control module 122 and the sector 120 may be associated on a one-to-one basis, or may be associated on a one-to-N basis (N is an integer of 2 or more). As an exception to the access denial, access from the hypervisor 200 may be permitted. As described above, since the VF 114 is exclusively allocated to a predetermined VM, as a result, the predetermined sector 120 can be exclusively allocated to one or more predetermined VMs 220. The VF ID setting can be changed by accessing from the hypervisor 200.

The authentication module 124 is a module for authenticating the configuration authority for the sector 120 in the configurable area 100. In this embodiment, the configuration data of the sector 120 is described as being transmitted to the PF 112 or VF 114 so that the configuration controller 130 receives the configuration data. However, the configuration data from the path other than the PF 112 or VF 114 is used. Data may be received. The hypervisor 200 preferably sets each sector 120 to be configurable. Accordingly, each authentication module 124 may be set to permit the PF 112 associated with the hypervisor 200 as a standard.

In addition to the hypervisor 200, the VM 220 may also configure the sector 120 assigned to itself. In order to deal with such a case, the authentication module 124 allows the authentication module 124 corresponding to the sector 120 assigned to the predetermined VM 220 to be configured from the VF 114 associated with the predetermined VM, and otherwise. The configuration from the VF 114 (or the VM 220) may be rejected (except for the hypervisor 200). As a result, the sector 120 can be prevented from being configured (for example, tampered) by another VM 220. The authentication module 124 and the sector 120 may be associated with each other one to one, or may be associated with one to N (N is an integer of 2 or more). Note that if it is not assumed that configuration data is received from the VF 114, the authentication module 124 only needs to have a mechanism that limits the functions that can receive configuration data to the PF 112 only.

In the example of FIG. 3, the VF 114a is provided to the VM 220a, and the VF 114a is associated with the

sectors

120a and 120b. That is, the function of the VF 114a is realized by a logic circuit configured in the

sectors

120a and 120b. In this case, the VF ID of the VF 114a may be set in the access control module 122a corresponding to the sector 120a and the access control module 122b corresponding to the sector 120b. A VF 114b is provided to the VM 220b, and the VF 114b is associated with the sector 120c.

In the example of FIG. 3, the sector 120d is an empty sector that is not associated with any VF. In this embodiment, such an empty sector 120 may be already initialized before the VF 114 is associated. This is because if not initialized, the information of the logic circuit configured in this empty sector 120 (an example is information stored in the Block RAM mentioned below) This is because it may be leaked to another VM 220 to be used, which is not preferable in terms of security. Details of the process related to the initialization will be described later.

As examples of VM1 and VM2, the following can be considered.
* VM for server. For example, a part of OLTP processing or OLAP processing using a search engine or a database is left to the FPGA. In the server program, the application program is executed, and the sector dedicated to the VM is configured with configuration data corresponding to the application program.
* Storage VM. Speedup can be expected by causing the FPGA to execute a part of storage processing such as encryption, compression, and deduplication. That is, the storage processing program is executed in the storage VM, and the sector dedicated to the storage VM is configured with configuration data that realizes data compression, data encryption, or data deduplication.

According to the present embodiment, such a flexible operation of creating a plurality of server VMs, creating a plurality of storage VMs alone, or creating them in a mixed manner can also be realized by the server 20 including the FPGA 23. The server VM and the storage VM may be created together by a VM creation operation described later.

Hereinafter, various flows will be described. In the present embodiment, for the sake of simplicity of explanation, assuming that there is one server 20, the hypervisor is in charge of the processing of the sector 120 of the FPGA 23 and the resources allocated to the VM 220, and performs subsequent management. explain. However, when there are a plurality of servers 20, the management computer 10 may perform part or all of the search processing for such resources and sectors 120. For example, when there are a plurality of servers 20, the resources, the free capacity of the sector 120, the used capacity, the load, etc. are monitored for each server 20, and the server 20 that creates the VM 220 is determined. In addition, the hypervisor of the server 20 determines which resource in the server 20 is actually allocated to the VM.

<Operation of enabling FPGA>
When enabling the FPGA 23, the FPGA 23 may execute the following processing. A typical example of the operation will be described later (see FIGS. 6 and 9).

First, the configuration controller 130 initializes the configurable area 100. For example, the configuration controller 130 clears a Block RAM (BRAM) in the configurable area 100.

Next, the configuration controller 130 configures a basic logic circuit in the configurable area 100 based on the basic configuration data. Thereby, as shown in FIG. 3, a plurality of sectors 120, a plurality of access control modules 122, a bus 132, and the like are configured in the configurable area 100. At this time, the configuration controller 130 may associate the sector 120 with the authentication module 124.

Next, the endpoint 110 enables the SR-IOV function. That is, the end point 110 enables a plurality of VFs 114. Thereby, the hypervisor 200 can recognize a plurality of VFs 114.

Each process described above may be executed based on a command issued from the hypervisor 200 to the endpoint 110. The hypervisor 200 may issue these instructions to the endpoint 110 based on the request transmitted from the management computer 10.

Through the above processing, a plurality of sectors 120 and the like are configured in the configurable area 100 of the FPGA 23, and a plurality of VFs 114 are configured in the end point 110.

<Operation of VM creation>
When creating a new VM 220, the management computer 10, the hypervisor 200, and the FPGA 23 may execute the following processing. A typical example of the operation will be described later (see FIG. 9).

First, the management computer 10 determines the sector 120 to be allocated to the new VM 220 based on the FPGA sector management table 300 (see FIG. 5) for managing the correspondence between the VM 220 and the sector 120. The number of sectors assigned to the new VM 220 may be one or more.

Next, the hypervisor 200 determines a VF 114 to be associated with one or more sectors 120 to be allocated to the new VM 220.

Next, the configuration controller 130 initializes the sector 120 associated with the VF 114 determined by the hypervisor 200.

Next, the configuration controller 130 sets the authentication module 124 of the sector 120 associated with the VF 114 to permit the configuration from the VF 114.

Next, the hypervisor 200 sets the VF ID of the VF 114 in the access control module 122 of the sector 120 associated with the VF 114.

Next, the configuration controller 130 configures a logic circuit in the sector 120 assigned to the new VM 220 based on the configuration data.

The hypervisor 200 provides the VF 114 to the new VM 220. The OS 222 running on the new VM 220, the application running on the OS 222, and the like can use the functions configured in the sector 120 corresponding to the VF 114 through a predetermined device driver.

Through the above processing, the new VM 220 can use the function configured in the sector 120 allocated to itself. That is, the resources of one FPGA 23 can be used from a plurality of VMs 220. In some cases, the sector 120 to which the program to be executed by the new VM 220 is assigned is configured. In this case, configuration as an operation when creating a VM may be omitted.

Further, by setting the access control module 122 and the authentication module 124, the VMs 220 other than the new VM 220 cannot access or configure the sector 120 assigned to the new VM 220. That is, the sector 120 allocated to the new VM 220 is not accessed or configured (tampered) by other VMs 220.

Further, by initializing the sector before allocating to the new VM 220, the logic circuit and data configured in the sector 120 by the user of the previous VM 220 are prevented from leaking to the user of the new VM 220. .

Therefore, according to the present embodiment, the resources of the FPGA 23 can be used safely and efficiently by a plurality of VMs 220 and users.

<Operation of deleting VM 220>
The management computer 10 determines a deletion target VM (deletion VM) 220. A typical example of the operation will be described later (see FIG. 13).

The hypervisor 200 deletes the deletion target VM and identifies the VF 114 associated with the deletion VM 220.

The hypervisor 200 deletes the VF ID of the VF 114 from the access control module 122 of the sector 120 corresponding to the specified VF 114.

The configuration controller 130 deletes the configuration permission setting of the VF 114 from the authentication module 124 corresponding to the VF 114.

Next, the configuration controller 130 initializes the sector 120 corresponding to the deleted VM 220. The initialized sector 120 becomes a free sector 120 and can be allocated to another new VM 220.

Through the above processing, the sector 120 allocated to the deleted VM 220 can be released. If the relationship between the VF 114 and the sector 120 remains the same, the deletion of the VD ID related to the access control module 122 and the deletion of the configuration permission setting related to the authentication module 124 may be omitted. However, when it is desired to dynamically change the number of sectors 120 allocated to the VF 114, it is preferable to delete the VF ID related to the access control module 122 and the configuration permission setting related to the authentication module 124.

When the VM 220 is deleted, the sector 120 assigned to the deleted VM 220 is initialized so that the logic circuit and data (for example, data stored in the BRAM) configured in the sector 120 are used next time. Can be prevented from leaking. When this process is executed, the initialization process when allocating the sector 120 to the new VM 220 described above may be omitted. *

FIG. 4 shows a configuration example of the VM management table 350.

The VM management table 350 is a table for managing the VM 220 created in the computer system 1. The VM management table 350 may be held in a storage device of the management computer 10.

Alternatively, the VM management table 350 may be a table for managing the VM 220 created in the server 20. In this case, the VM management table 350 may be held in a storage device of each server 20 (or hypervisor 200).

The VM management table 350 includes, as item values, VM ID 352, server ID 354, OS type 356, CPU number 358, memory capacity 360, NIC number 362, NIC setting 364, storage type 366, storage capacity 368, allocated sector number 370, config. Configuration file ID 374 (abbreviated as Conf file ID) and configuration file ID 374 (abbreviated as Conf file number in the figure). The configuration file is a file that stores the above-described configuration data. However, a data storage format other than a file may be used as long as configuration data can be stored.

The VM ID 352 is an identifier of the VM 220. The VM ID 352 may be a VM name.

The server ID 354 is an identifier of the server 20 having the VM 220 having the VM ID 352. The server ID 354 may be a server name (or host name).

The OS type 356 is a type of the OS 222 that runs on the VM 220 with the VM ID 352. Examples of the OS type 356 are Linux (registered trademark) and Windows (registered trademark).

The CPU number 358 is the number of virtual CPUs of the VM 220 with the VM ID 352.

The memory capacity 360 is a virtual memory capacity of the VM 220 with the VM ID 352.

The NIC number 362 is the number of virtual NICs that the VM 220 with the VM ID 352 has.

The NIC setting 364 is a virtual NIC setting that the VM 220 with the VM ID 352 has. The VM management table 350 may have item values of the NIC setting 364 corresponding to the NIC number 362.

The storage type 366 is a virtual storage type that the VM 220 with the VM ID 352 has. Examples of the storage type 366 are HDD, SSD, flash memory, and the like.

The storage capacity 368 is a virtual storage capacity of the VM 220 with the VM ID 352. The above items 354 to 368 may be changed to other items as long as they indicate the configuration of the VM, and some items may be omitted.

The allocated sector number 370 is the number of sectors 120 of the FPGA 23 allocated to the VM 220 with the VM ID 352.

The configuration file number 372 is the number of configuration files for configuring the logic circuit used by the VM 220 having the VM ID 352.

The configuration file ID 374 is an identifier of a configuration file for configuring a logic circuit used by the VM 220 having the VM ID 352. The configuration file ID 374 may be a configuration file name. The VM management table 350 may have item values of the configuration file ID 374 corresponding to the number of configuration files 372.

These item values may be values input by the administrator from the VM creation management screen (see FIG. 7).

FIG. 5 shows a configuration example of the FPGA sector management table 300.

The FPGA sector management table 300 is a table for managing which sector 120 and VF 114 of the FPGA 23 are allocated to which VM 220 of the FPGA 23 of each server 20. The FPGA sector management table 300 may be held in a storage device of the management computer 10.

Alternatively, the FPGA sector management table 300 may be a table for managing which sector 120 and VF 114 of the FPGA 23 of the server 20 is assigned to which VM 220 created in the server 20. . In this case, the FPGA sector management table 300 may be held in a storage device of each server 20 (or hypervisor 200).

The FPGA sector management table 300 may include an FPGA ID 302, a sector number 304, a VM ID 306, and a VF ID 308 as item values.

The FPGA ID 302 is an identifier of the FPGA 23.

The sector number 304 is a number for identifying the sector 120 included in the FPGA 23 of the FPGA ID 302.

VM ID 306 is an ID of VM 220 to which sector 120 of sector number 304 is assigned. When the sector 120 with the sector number 304 is not assigned to any VM 220 (that is, when it is an empty sector), the VM ID 306 may be “N / A”. When the same VM ID 306 is associated with different sector numbers 304, it may indicate that a sector (that is, a plurality of sectors) with the different sector numbers 304 is assigned to the VM 220 of the VM ID 306. .

The VF ID 308 is the ID of the VF 114 associated with the sector 120 with the sector number 304. When the sector 120 with the sector number 304 is not assigned to any VM 220 (that is, when it is an empty sector), the VF ID 308 may be “N / A”. When the same VF ID 308 is associated with different sector numbers 304, the VF 114 of the VF ID 308 indicates that the sector (that is, a plurality of sectors) 120 with the different sector numbers 304 is associated. It's okay.

FIG. 6 is a flowchart showing an operation example of the hypervisor 200 in the FPGA 23 validation process. In addition, as a timing which performs this operation | movement, although the case where the hypervisor 200 recognizes FPGA23 is considered, you may implement at another timing.

(S100) The management computer 10 transmits an FPGA activation request to the hypervisor 200 of the server 20 that activates the FPGA 23. When the hypervisor 200 receives the FPGA validation request from the management computer 10, the hypervisor 200 proceeds to S102. The FPGA validation request may include a request for validation of the SR-IOV of the FPGA 23.

(S102) The hypervisor 200 instructs the PF 112 of the FPGA 23 to configure a basic logic circuit. As a result, the sector 120, the access control module 122, the bus 132, and the like as shown in FIG. 3 are configured in the configurable area 100 of the FPGA 23. At this time, each sector 120 may be initialized by an initialization process described later (see FIG. 9). Then, the process proceeds to S104.

(S104) The hypervisor 200 commands the PF 112 of the FPGA 23 to enable SR-IOV. As a result, a plurality of VFs 114 are validated at the end point 110 of the FPGA 23, and the hypervisor 200 can recognize these VFs 114. Then, the process proceeds to S106.

(S106) The hypervisor 200 returns a completion response to the FPGA validation request to the management computer 10. Then, this process ends.

Through the above processing, the sector 122, the access control module 122, and the like are configured in the configurable area 100 of the FPGA 23, and the hypervisor 200 can recognize the VF 114 of the FPGA 23.

FIG. 7 shows an example of a management screen 1000 for creating a VM.

The administrator may be able to input the parameters of the VM 220 to be created from the VM creation management screen 1000 shown in FIG.

The parameters that can be input from the management screen 1000 for creating a VM and the item values in the VM management table 350 may be in a correspondence relationship.

When the administrator inputs parameters from the VM creation management screen 1000 and presses the “Done” button 1002, the management computer 10 is notified of this and the VM 220 new creation process shown in FIG. 8 is executed. Good. As described above, in the management screen 1000 described in FIG. 7, the configuration requirements of the VM to be created (for example, Host name to Storage Size in the diagram) and the FPGA configuration requirements to be allocated to the VM (for example, FPGA Sectors to Config. 1) are input. To do.

A template may be used when creating a VM. Specifically, the management computer 10 or the hypervisor 200 stores a template that defines the VM configuration requirements and the FPGA configuration requirements as input in FIG. 7, and when actually creating a VM, the client 30 or In this example, the management client 34 performs a VM creation operation by designating the template. In this case, the screen of FIG. 7 may be a screen for creating the template.

FIG. 8 is a flowchart showing an operation example of the hypervisor 200 in the new creation process of the VM 220.

(S200) The management computer 10 transmits a VM creation request to the hypervisor 200 of the server 20 that creates the VM 220. When receiving the VM creation request from the management computer 10, the hypervisor 200 proceeds to S201. The VM creation request may include parameters input from the VM creation management screen 1000 (each record item value in the VM management table 350; the same applies hereinafter).

(S201) The hypervisor 200 allocates the physical resources of the server 20 to the VM 220 based on the parameters included in the VM creation request. For example, the hypervisor 200 allocates computing resources, storage resources, network resources, and the like to the VM 220. Then, the process proceeds to S202. The resource selection process to be allocated is performed as part of this step. However, the hypervisor 200 may perform the process, or the management computer 10 may perform at least a part of the process.

(S202) The hypervisor 200 determines whether or not the sector 120 needs to be allocated to the VM 220 based on the number of allocated sectors included in the VM creation request. When the determination result is “YES” (when the number of allocated sectors is 1 or more), the hypervisor 200 proceeds to S204, and when “NO” (when the number of allocated sectors is 0), the hypervisor 200 proceeds to S214. The processing may be performed by the hypervisor 200, or at least a part of the processing may be performed by the management computer 10.

(S204) If the determination result in S202 is “YES”, the hypervisor 200 allocates to the VM 220 the empty sectors 120 corresponding to the number of allocated sectors and the VF 114 associated with the empty sectors 120. When the number of allocated sectors is 2 or more, the hypervisor 200 may select a plurality of empty sectors so that the signal transmission distance between the sectors is the shortest. As a result, the operation speed of the logic circuit configured by using a plurality of empty sectors can be improved. Then, the process proceeds to S206.

(S206) The hypervisor 200 initializes the allocated sector 120 (see FIG. 9). Then, the process proceeds to S207.

(S207) The hypervisor 200 sets the authentication module 124 corresponding to the sector 120 allocated to the VM 220 to permit the configuration from the VM 220. Then, the process proceeds to S208.

(S208) The hypervisor 200 sets the VF ID of the VF 114 in the access control module 122 of the sector 120 associated with the VF 114. As a result, the VF 114 can access the sector 120. Then, the process proceeds to S210.

(S210) The hypervisor 200 determines whether or not a configuration file is specified in the VM creation request. The hypervisor 200 proceeds to S212 when the determination result is “YES”, and proceeds to S214 when the determination result is “NO”. The processing may be performed by the hypervisor 200, or at least a part of the processing may be performed by the management computer 10.

(S212) When the determination result in S210 is “YES”, the hypervisor 200 configures a logic circuit based on the designated configuration file in the allocated sector 120 (for example, configuration data from the configuration file). The configuration data is transmitted to the PF 112 while designating the sector, and the configuration controller 130 configures the designated sector using the received configuration data). Then, the process proceeds to S214.

(S214) The hypervisor 200 activates the VM 220. Then, the process proceeds to S216.

(S216) The hypervisor 200 returns a completion response to the VM creation request to the management computer 10. The management computer 10 that has received this completion response reflects the content of the VM creation request (that is, the allocation relationship between the VM (and VF) and the sector) in the FPGA sector management table 300. Then, this process ends.

Through the above processing, the VM 220 can be generated, and the VM 220 can use the function of the logic circuit configured in the sector 120 corresponding to the VF 114.

FIG. 9 is a flowchart showing an operation example of the hypervisor 200 in the sector initialization process of the FPGA 23.

(S120) The hypervisor 200 commands the FPGA 23 to initialize (clear) the sector 120. This instruction may include a sector number to be initialized. The configuration controller 130 of the FPGA 23 that has received this initialization command initializes the sector 120 to be initialized. For example, the configuration controller 130 clears the BRAM of the sector 120. Further, the logic circuit configured in the sector may be cleared. Then, the FPGA 23 returns a completion response to the initialization instruction of the sector 120 to the hypervisor 200.

(S122) When the hypervisor 200 receives the completion response from the FPGA 23, the process proceeds to S124.

(S124) The hypervisor 200 confirms that the sector 120 has been initialized (optional). For example, the hypervisor 200 attempts to read the BRAM of the sector 120 to be initialized, and determines whether or not it has been correctly cleared. Alternatively, a logic circuit configured in the sector 120 may be driven. Then, the process returns to the caller of this process.

The desired sector 120 can be initialized by the above processing. By executing the initialization process of the sector 120 as necessary, for example, when deleting the existing VM 220 and / or allocating the sector 120 to the new VM 220, the logic circuit and data of the sector 120 are changed. , It is possible to prevent leakage to another person's VM 220.

FIG. 10 shows an example of a management screen 1200 for changing VMs.

The administrator may be able to change the configuration (parameter) of the created VM 220 from the VM change management screen 1200 shown in FIG.

The parameters that can be changed from the VM change management screen 1200 and the item values in the VM management table 350 may be in a correspondence relationship.

When the administrator changes the parameter from the VM change management screen 1200 and presses the “Done” button 1202, the management computer 10 is notified to that effect, and the configuration change process of the VM 220 shown in FIG. 11 is executed. Good. Note that a template may be used to change the configuration of the VM as in the VM creation.

FIG. 11 is a flowchart illustrating an operation example of the hypervisor 200 in the configuration change process of the VM 220.

(S300) The management computer 10 transmits a VM change request to the hypervisor 200 of the server 20 including the VM 220 to be changed. When receiving the VM change request from the management computer 10, the hypervisor 200 proceeds to S301. The VM change request may include parameters changed from the VM change management screen 1200.

(S301) The hypervisor 200 stops the VM 220 to be changed. Then, the process proceeds to S302.

(S302) The hypervisor 200 changes the physical resource allocated to the VM 220 based on the parameter included in the VM change request. Then, the process proceeds to S304.

(S304) The hypervisor 200 determines whether the number of assigned sectors after the change included in the VM change request is “increase”, “decrease”, or “match” compared to the original number of assigned sectors. . The hypervisor 200 proceeds to S310 when the number of allocated sectors after the change is “increase”, proceeds to S320 when it is “decreased”, and proceeds to S330 when “match”. Here, the processing is described on the assumption that the sector in which the logic circuit to be continuously used before and after the sector allocation change (including coincidence) is configured is not changed, but the sector change is not limited. . For example, when the sector to be configured is changed, processing for deallocating the sector before the change and assigning the sector for the change destination may be added.

(S310) When it is determined in S304 that the number of allocated sectors after the change is “increasing”, the hypervisor 200 selects a free sector to be newly allocated to the VM 220 according to the increased number. When the number of allocated sectors increases by 2 or more, the hypervisor 200 may select a plurality of empty sectors so that the signal transmission distance between the sectors is as short as possible. Then, the hypervisor 200 sets the configuration permission from the VF 114 provided to the VM 220 and the VF ID in the authentication module 124 and the access control module 122 corresponding to the allocated empty sector 120. As a result, the VF 114 can change and access the configuration of the newly allocated sector 120. Then, the process proceeds to S312.

(S312) The hypervisor 200 determines whether or not a configuration file is specified in the VM change request. If the determination result is “YES”, the hypervisor 200 proceeds to S314, and if “NO”, the hypervisor 200 proceeds to S330.

(S314) If the determination result in S312 is “YES”, the hypervisor 200 configures a logic circuit based on the designated configuration file in the allocated sector 120. A specific example is the same as S212 in FIG. Then, the process proceeds to S330.

(S320) When it is determined in S304 that the number of allocated sectors after the change is “decreasing”, the hypervisor 200 selects one of the sectors 120 allocated to the VM 220 according to the decreased number and the designation of the administrator. The sector 120 to be deallocated is selected. Then, the hypervisor 200 deletes the VF ID and configuration permission setting of the VF 114 provided to the VM 220 from the access control module 122 and the authentication module 124 corresponding to the sector 120 to be deallocated. As a result, the VF 114 cannot access or configure the sector 120 to be deallocated. Then, the process proceeds to S322.

(S322) The hypervisor 200 initializes the sector 120 to be deallocated (see FIG. 9). Then, the process proceeds to S330.

(S330) The hypervisor 200 activates the VM 220 after the change. Then, the process proceeds to S332.

(S332) The hypervisor 200 returns a completion response to the VM change request to the management computer 10. The management computer 10 that has received this completion response reflects the content of the VM change request (that is, the change in the allocation relationship between the VM (and VF) and the sector) in the FPGA sector management table 300. Then, this process ends.

Through the above processing, the configuration of the VM 220 can be changed, and the correspondence relationship between the VM 220 and the sector 120 can also be changed.

FIG. 12 shows an example of a management screen 1300 for deleting a VM.

The administrator can delete the VM 220 already created in the server 20 from the VM deletion management screen 1300 shown in FIG.

When the administrator selects the VM 220 to be deleted from the VM deletion management screen 1300 and presses the “Done” button 1302, the management computer 10 is notified to that effect, and the deletion processing of the VM 220 shown in FIG. 13 is executed. It's okay.

FIG. 13 is a flowchart illustrating an operation example of the hypervisor 200 in the deletion process of the VM 220.

(S400) The management computer 10 transmits a VM deletion request to the hypervisor 200 of the server 20 including the VM 220 to be deleted. When receiving the VM deletion request from the management computer 10, the hypervisor 200 proceeds to S401. The VM deletion request may include the VM ID of the VM 220 selected from the VM deletion management screen 1300.

(S401) The hypervisor 200 stops the VM 220 to be deleted. Then, the process proceeds to S402.

(S402) The hypervisor 200 releases the physical resources allocated to the VM 220. Then, the process proceeds to S404.

(S404) The hypervisor 200 determines whether or not the sector 120 has been allocated to the VM 220. If the determination result is “YES”, the hypervisor 200 proceeds to S406, and if “NO”, the hypervisor 200 proceeds to S412.

(S406) When the determination result in S404 is “YES”, the hypervisor 200 cancels the allocation of the sector 120 to the VM 220. Then, the process proceeds to S408.

(S408) The hypervisor 200 deletes the VF ID and the configuration permission setting of the VF 114 provided to the VM 220 from the access control module 122 and the authentication module 124 corresponding to the sector 120 to deal with the deallocation. As a result, the VF 114 cannot access and configure the sector 120 to be deallocated. Then, the process proceeds to S410.

(S410) The hypervisor 200 performs initialization processing of the sector 120 to be deallocated (see FIG. 9). Then, the process proceeds to S412.

(S412) The hypervisor 200 returns a completion response to the VM deletion request to the management computer 10. The management computer 10 that has received this completion response reflects the contents of the VM deletion request (that is, the deallocation relationship between the VM (and VF) and the sector) in the FPGA sector management table 300. Then, this process ends.

According to the above processing, the sector 120 allocated to the VM 220 to be deleted can be made a free sector. Furthermore, since the empty sector 120 is initialized, it is possible to prevent the logic circuit and data of the sector 120 from leaking to another person's VM 220.

FIG. 14 is a flowchart showing an operation example of the hypervisor 200 in the forced initialization process of the sector 120.

A certain sector may adversely affect the operation of a nearby (adjacent) sector 120. That is, there is a risk of adversely affecting the operation of another person's VM 220. For example, a logic circuit that causes an FPGA thermal runaway in sector 120 is configured and operating. Such a sector 120 may require a forced initialization. Hereinafter, a processing example in such a case will be described.

(S500) The hypervisor 200 requests the temperature of each sector 120 from the FPGA 23, and acquires the temperature of each sector 120. The hypervisor 200 may periodically execute this request. The processing may be performed by the hypervisor 200, or at least a part of the processing may be performed by the management computer 10.

(S502) The hypervisor 200 determines whether or not there is a sector 120 whose temperature is higher than a predetermined threshold. When the determination result is “YES”, the hypervisor 200 proceeds to S504, and when the determination result is “NO”, the hypervisor 200 ends the process. The processing may be performed by the hypervisor 200, or at least a part of the processing may be performed by the management computer 10.

(S504) If the determination result in S502 is “YES”, the hypervisor 200 determines from the access control module 122 and the authentication module 124 corresponding to the sector 120 whose temperature is higher than the predetermined threshold value, the VF ID and the VF ID set therein. Delete configuration permission settings. As a result, the VF 114 with the VF ID cannot access and configure the sector 120 whose temperature is abnormally high. Then, the process proceeds to S506.

(S506) The hypervisor 200 initializes the sector 120 whose temperature is abnormally high (see FIG. 9). Then, this process ends.

By the above processing, the sector 120 having an abnormally high temperature can be forcibly initialized, and adverse effects on the operation of the nearby sector 120 can be prevented. Note that this embodiment can be applied in addition to thermal runaway as long as the predetermined sector adversely affects adjacent sectors because of the configuration data, and the hypervisor has an adverse effect on the state of the predetermined sector. The item for determining may be other than temperature.

FIG. 15 shows a configuration example of a computer system 2 that is a modification of the computer system 1.

The computer system 2 in FIG. 15 is a computer system in which the FPGA 23 existing in each server 20 in FIG.

The FPGA server 38 may include a hypervisor 202 and an FPGA 23. The server 20 may include a hypervisor 201.
The hypervisor 201 of the server 20 and the hypervisor 202 of the FPGA server 38 may cooperate via the LAN 42 so that the VM 220 of the server 20 can use the FPGA 23 of the FPGA server 38 as described above. For example, the hypervisor 201 of the server 20 may transmit the above-described command to the hypervisor 202 of the FPGA server 38 via the LAN 42, so that the computer system 2 can realize the same as the above-described computer system 1. .

According to the configuration of the computer system 2, the resources of the FPGA 23 can be shared by the plurality of servers 20, so that the resources of the FPGA 23 can be used more efficiently.

The embodiment described above is an example for explaining the present invention, and is not intended to limit the scope of the present invention only to the embodiment. Those skilled in the art can implement the present invention in various other modes without departing from the gist of the present invention. For example, in recent years, there is a package configuration including an FPGA (referred to as an FPGA core in this embodiment) in addition to a CPU core in a CPU package. In this case, (A) the non-volatile memory is included in the CPU package, (B) the non-volatile memory is connected to the outside of the CPU package, and (C) the CPU is not provided with the non-volatile memory. There are three patterns in which configuration is performed by transmitting configuration data from outside the FPGA core such as a CPU core when the package is turned on. However, in either case (A) or (C), there is an external configuration. In case (B), the non-volatile memory connected to the CPU package is also considered as a part of the FPGA 23, so that the external configuration It is clear that exists. Therefore, the present embodiment can be applied to such an FPGA. Further, the server 20 may include a VM 220 that does not use the FPGA 23.

1, 2: Computer system 10: Management computer 20: Server 23:

FPGA

200, 201, 202: Hypervisor 220: Virtual computer

Claims

A computer system having a partially configurable programmable logic device, a hypervisor, and a management computer,
The programmable logic device includes a plurality of sectors, each configurable;
The hypervisor or management computer selects a predetermined sector that is free from the plurality of sectors when creating a virtual computer,
The hypervisor is a computer system that exclusively allocates the selected sector to the virtual machine to be created.
The computer system according to claim 1,
The hypervisor transmits a first instruction accompanied by clearing a block RAM (BRAM) corresponding to the selected sector of the programmable logic device before assigning the selected sector to the virtual machine to be created.
The computer system according to claim 2,
The computer system in which the hypervisor instructs the programmable logic device to perform configuration based on configuration data according to the virtual machine to be created after the first instruction.
The computer system according to claim 3,
The hypervisor is a computer system that transmits the first command when the selected sector becomes a free area by deleting another virtual machine.
The computer system according to claim 1,
When the hypervisor or the management computer determines that the state of a certain sector has an adverse effect on an adjacent sector, the computer that transmits a first instruction accompanied by clearing a BRAM corresponding to the certain sector of the programmable logic device system.
The computer system according to claim 1,
The programmable logic device provides a logic circuit configured with the selected sector via a Virtual Function created by Single Root I / O Virtualization (SR-IOV),
The hypervisor is a computer system that exclusively allocates the virtual function to the virtual computer to be created.
A computer system according to claim 6, wherein
The hypervisor or management computer is
Obtain the number of sectors to be dedicated to the virtual machine to be created,
A computer system that assigns the acquired number of sectors to the virtual function to the programmable logic device.
A computer system according to claim 7, wherein
The hypervisor or management computer selects two sectors with the shortest transmission distance between sectors when allocating two or more sectors to the virtual function.
The computer system according to claim 7,
The computer system further includes a web server,
The computer system provides a Graphical User Interface (GUI) capable of inputting an input item including the number of sectors to be allocated as a dedicated component when creating a virtual computer.
The computer system according to claim 7,
The computer system further includes a web server,
The computer system provides a Graphical User Interface (GUI) capable of inputting an identifier of configuration data to be configured in the predetermined sector as a configuration requirement when creating a virtual computer.
A computer system according to claim 9, wherein
The GUI is a computer system that is an input GUI of a template specified when creating a virtual computer.
The computer system according to claim 1,
The selected sector is configured with configuration data that realizes data compression, data encryption, or data deduplication, and the created virtual machine is a storage virtual machine, and other than the created virtual machine A virtual machine is used for a server, and the other sector dedicated to the other virtual machine is configured with configuration data corresponding to an application program executed on the other virtual machine.