WO2016119469A1 - Procédé de gestion de contexte de service, machine principale physique, dispositif pcie et dispositif de gestion de migration - Google Patents

Procédé de gestion de contexte de service, machine principale physique, dispositif pcie et dispositif de gestion de migration Download PDF

Info

Publication number
WO2016119469A1
WO2016119469A1 PCT/CN2015/089815 CN2015089815W WO2016119469A1 WO 2016119469 A1 WO2016119469 A1 WO 2016119469A1 CN 2015089815 W CN2015089815 W CN 2015089815W WO 2016119469 A1 WO2016119469 A1 WO 2016119469A1
Authority
WO
WIPO (PCT)
Prior art keywords
service
memory
pcie device
context
address
Prior art date
Application number
PCT/CN2015/089815
Other languages
English (en)
Chinese (zh)
Inventor
卢胜文
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2016119469A1 publication Critical patent/WO2016119469A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines

Definitions

  • the present invention relates to the field of virtual machines, and in particular, to a service context management method, a physical host, a PCIE device, and a migration management device.
  • a virtual machine manager (English: Virtual Machine Manager, VMM for short) is installed on a physical host, and one or more virtual machines are managed by the VMM.
  • the operating system (English: Operating System, OS for short) and various applications can be run on the machine.
  • the hardware of the general physical host mainly includes: some hardware related to running the virtual machine, such as a central processing unit (English: Central Processing Unit, CPU), memory, hard disk, etc., and some are not directly used to run the virtual machine, but A hardware that performs some specific services and is relatively independent in function, such as a Peripheral Component Interconnection Express (PCIE) device (such as a network card adapter, a graphics card, etc.) in a physical host.
  • PCIE Peripheral Component Interconnection Express
  • Single-Root I/O Virtualization (English: Single-Root I/O Virtualization, SR-IOV for short) is an input/output (English: Input/output, I/O) virtualization technology.
  • a virtual PCIE adapter is virtualized on a physical PCIE adapter, which is called a virtual function (English: Virtual Function, VF for short). Take the NIC adapter that supports SR-IOV as an example.
  • a physical NIC adapter can virtualize multiple virtual NIC adapters. The VF in each virtual machine corresponds to a virtual NIC adapter.
  • the server After the server is virtualized, it can easily implement virtual machine hot migration.
  • the virtual machine hot migration technology refers to the transfer of virtual machines from one physical host to another through hot migration, and the migration process has little impact on business processing.
  • the virtual machine in addition to the need to migrate the data stored in the memory to the target physical host, the virtual machine needs to migrate the state data stored in the register on the PCIE adapter or the internal memory to the target physical host.
  • the registers or memory are typically configured to hold the context of the state offloaded service handled by the PCIE adapter, and the context of the state offload service is used for traffic statistics, control, and business status records.
  • the registers that hold the context of the service are very scattered. If you perform hot migration, the storage and recovery of these registers or memories need to be saved and restored independently, which is very troublesome.
  • the embodiment of the present invention provides a service context management method, a physical host, a PCIE device, and a migration management device.
  • the technical solution is as follows:
  • an embodiment of the present invention provides a service context management method, which is applicable to a virtual machine, where the virtual machine runs on a physical host, and the physical host is connected to a shortcut peripheral component to connect with a standard PCIE device, and the PCIE
  • the device is a hardware device conforming to a single input/output virtualized SR-IOV standard, the PCIE device including at least one physical function PF and at least one virtual function VF, wherein each PF and each VF are configured with unique functions No.
  • the method includes:
  • the virtual machine allocates memory for at least one state offload service carried by the PCIE device;
  • the virtual machine generates a context table in the memory, where the context table is used to record a service context of at least one state offload service carried by the PCIE device;
  • the virtual machine sends the first address of the memory and the function number corresponding to the virtual machine to the PCIE device.
  • the context table is a direct index table or a hash table, where the direct index table includes a correspondence between a keyword KEY and a service context, and the hash table includes a And the at least one state offload service includes a to-be-processed state offload service, where the method further includes:
  • the read service context is sent to the PCIE device.
  • the calculating the hash value C according to the KEY in the request message includes:
  • the size M of the hash bucket is modulo, and the hash value C is obtained.
  • the sending the read service context to the PCIE device includes:
  • the read service context is written into the cache of the PCIE device by means of a cache Cache mechanism.
  • the embodiment of the present invention further provides a service context management method, which is applicable to a PCIE device, where the PCIE device is connected to a physical host, and the PCIE device is a single input/output virtualized SR-IOV standard.
  • the PCIE device includes at least one physical function PF and at least one virtual function VF, wherein each PF and each VF is configured with a unique function number, and the physical host runs at least one virtual machine.
  • the methods include:
  • a context table is saved, where the context table is used to record a service context of a state offload service carried by the PCIE device;
  • the PCIE device acquires a KEY from the service information of the offloaded service in the to-be-processed state;
  • the PCIE device acquires a service context of the to-be-processed offload service according to the first address of the memory corresponding to the unloading service and the KEY according to the to-be-processed state.
  • the context table is a direct index table or a hash table, where the direct index table includes a correspondence between a keyword KEY and a service context, and the hash table includes a Corresponding relationship between the hash value and the service context, the first address of the memory corresponding to the unloading service according to the to-be-processed state, and the service context of the KEY to obtain the unloading service of the to-be-processed state, including:
  • the first address of the memory corresponding to the unloading service in the to-be-processed state is used to search for a corresponding direct index table or a hash table in the memory of the virtual machine;
  • KEY Using the KEY to directly read the service context in the direct index table corresponding to the first address of the found memory, or calculate the hash value C according to the KEY, and use the hash value C to read the found context.
  • the calculating the hash value C according to the KEY includes:
  • the size M of the hash bucket is modulo, and the hash value C is obtained.
  • the first end of the memory corresponding to the unloading service according to the to-be-processed state The address and the KEY acquire the service context of the to-be-processed unloading service, including:
  • the determining, according to the to-be-processed state, service information The function number corresponding to the processing status unloading service, including:
  • the feature is used to indicate a function number corresponding to the to-be-processed state unloading service.
  • the method further includes:
  • the PCIE device receives a first address of a memory of each virtual machine and a function number corresponding to each virtual machine;
  • an embodiment of the present invention further provides a service context management method, where the method includes:
  • the migration management device in the destination physical host acquires the first address of the memory copied from the source physical host memory, and the destination physical host and the source physical host respectively interconnect the standard PCIE with different shortcut peripheral components.
  • the PCIE device is a hardware device supporting a single input and output virtualization SR-IOV standard, the PCIE device includes at least one physical function PF and at least one virtual function VF, wherein each PF and each VF All are configured with a unique function number;
  • mapping a correspondence between the first address of the memory and the function number to a PCIE device of the destination physical host.
  • the embodiment of the present invention further provides a physical host, where the virtual host runs a virtual machine, and the physical host is connected to a fast peripheral component to connect with a standard PCIE device, and the PCIE device supports a single root.
  • Input and output virtualized SR-IOV standard hardware devices the PCIE device including at least one physical function PF and at least one virtual function VF, wherein each PF and each VF is configured with a unique function number, the virtual machine include:
  • An allocating module configured to allocate memory for at least one state offloading service carried by the PCIE device
  • a generating module configured to generate a context table in the memory, where the context table is used to record a service context of at least one state offload service carried by the PCIE device;
  • a sending module configured to send the first address and the function number of the memory to the PCIE device.
  • the context table is a direct index table or a hash table, where the direct index table includes a correspondence between a keyword KEY and a service context, and the at least one state offloading service Including a pending state unloading service, the hash table including a hash value
  • the virtual machine further includes:
  • a receiving module configured to receive a request message sent by the PCIE device to obtain a service context of the to-be-processed state offload service, where the request message includes a first address of the memory and a keyword KEY;
  • a determining module configured to determine, according to the first address of the memory in the request message, a corresponding direct index table or a hash table;
  • a reading module configured to read a service context from the determined direct index table corresponding to the first address of the memory according to the KEY in the request message, or calculate a hash value C according to the KEY in the request message, Using the hash value C to read the service context from the determined hash table corresponding to the first address of the memory;
  • the sending module is further configured to send the read service context to the PCIE device.
  • the reading module is specifically configured to:
  • the KEY in the request message is converted into a total hash value A; and the hash bucket corresponding to the virtual machine in the hash table corresponding to the determined first address of the memory is obtained or the to-be-processed
  • the sending module is specifically configured to: use the read service context to pass the cache The way of the Cache mechanism is written into the cache of the PCIE device.
  • the embodiment of the present invention further provides a PCIE device, where the PCIE device is connected to a physical host, and the PCIE device is a hardware device supporting a single input and output virtualized SR-IOV standard, and the PCIE device is Included in the at least one physical function PF and at least one virtual function VF, wherein each PF and each VF is configured with a unique function number, and the physical host runs at least one virtual machine, and the PCIE device includes:
  • a receiving module configured to receive service information of a to-be-processed unloading service sent by the host side or the network side;
  • a first determining module configured to determine, according to the service information of the offloading service in the to-be-processed state, a function number corresponding to the to-be-processed state unloading service
  • a second determining module configured to uninstall a function corresponding to the service according to the determined to-be-processed state And a first address of the memory corresponding to the unloading service in the to-be-processed state, where the memory stores a context table, where the context table is used to record the bearer of the PCIE device. Status unloads the business context of the business;
  • a first obtaining module configured to obtain a KEY from the service information of the to-be-processed state offloading service
  • the second obtaining module is configured to acquire, according to the to-be-processed state, the first address of the memory corresponding to the unloading service and the KEY to obtain the service context of the to-be-processed state offloading service.
  • the context table is a direct index table or a hash table, where the direct index table includes a correspondence between a keyword KEY and a service context, and the hash table includes a The correspondence between the hash value and the service context, where the second acquiring module is specifically configured to:
  • the first address of the memory corresponding to the unloading service in the to-be-processed state is used to search for a corresponding direct index table or a hash table in the memory of the virtual machine; and the first address corresponding to the memory that is found by using the KEY is directly read. Directly indexing the business context in the table, or calculating the hash value C according to the KEY, and using the hash value C to read the searched service context in the hash table corresponding to the first address of the memory.
  • the second acquiring module is specifically configured to:
  • the KEY in the request message is converted into a total hash value A; and the hash bucket corresponding to the virtual machine in the hash table corresponding to the determined first address of the memory is obtained or the to-be-processed
  • the second acquiring module is specifically configured to:
  • the first determining module is specifically configured to: according to the host side The PCIE interface used in the service information of the to-be-processed state offload service or the service information on the host side to transmit the service information of the to-be-processed offload service determines the to-be-supplied A function number corresponding to the unloading service of the state, the feature is used to indicate a function number corresponding to the unloading service of the to-be-processed state.
  • the receiving module is further configured to receive a memory of each virtual machine a first address and a function number corresponding to each of the virtual machines;
  • the PCIE device further includes: a storage module, configured to save a correspondence between a first address of the memory and the function number.
  • the embodiment of the present invention further provides a migration management device, where the migration management device is disposed in a destination physical host, where the migration management device includes:
  • a first obtaining module configured to acquire a first address of the memory copied from the source physical host memory during the hot migration, where the destination physical host and the source physical host respectively interconnect the standard PCIE device with different shortcut peripheral components Connected
  • the PCIE device is a hardware device supporting a single input/output virtualized SR-IOV standard, the PCIE device including at least one physical function PF and at least one virtual function VF, wherein each PF and each VF Configured with a unique feature number;
  • a second acquiring module configured to obtain the function number corresponding to the migrated virtual machine
  • a configuration module configured to configure a correspondence between the first address of the memory and the function number to a PCIE device of the destination physical host.
  • the embodiment of the present invention further provides a physical host, where the physical host includes: a processor, a memory, an input and output unit, and a bus; the memory is configured to store a computer to execute an instruction, the processor and the The memory is connected by the bus, and when the physical host is running, the processor executes the computer-executed instructions stored by the memory to cause the computer to perform the business context management provided by the foregoing first or third aspect method.
  • the physical host includes: a processor, a memory, an input and output unit, and a bus; the memory is configured to store a computer to execute an instruction, the processor and the The memory is connected by the bus, and when the physical host is running, the processor executes the computer-executed instructions stored by the memory to cause the computer to perform the business context management provided by the foregoing first or third aspect method.
  • the virtual machine allocates memory for at least one state unloading service carried by the PCIE device, and then generates a context table in the memory, and then sends the first address of the memory and the function number corresponding to the virtual machine to the PCIE device, so that the PCIE device can record the first address of the memory.
  • the business context of the state unloading service is obtained from the memory, the operation is convenient, and the speed is fast; when the virtual machine is hot-migrated, the register is avoided.
  • Migration brings The trouble is that the memory is directly migrated to the destination physical machine, and the corresponding relationship between the function number of the virtual machine after the migration and the first address of the virtual machine in the source physical host is configured in the migrated PCIE device.
  • the implementation is convenient and simple, greatly speeding up the hot migration of the virtual machine and shortening the downtime.
  • FIG. 1 is an application scenario diagram provided by an embodiment of the present invention
  • FIG. 2 is a flowchart of a service context management method according to Embodiment 1 of the present invention.
  • FIG. 3 is a flowchart of a service context management method according to Embodiment 2 of the present invention.
  • FIG. 4 is a flowchart of a service context management method according to Embodiment 3 of the present invention.
  • FIG. 5 is a flowchart of a service context management method according to Embodiment 4 of the present invention.
  • FIG. 6 is a structural block diagram of a physical host according to Embodiment 5 of the present invention.
  • FIG. 7 is a structural block diagram of a physical host according to Embodiment 6 of the present invention.
  • FIG. 8 is a structural block diagram of a PCIE device according to Embodiment 7 of the present invention.
  • FIG. 9 is a structural block diagram of a PCIE device according to Embodiment 8 of the present invention.
  • FIG. 10 is a structural block diagram of a migration management device according to Embodiment 9 of the present invention.
  • a physical host runs a virtual machine manager VMM and at least one virtual machine.
  • the VMM can manage one or more virtual machines, and each virtual machine can run an operating system OS and various applications.
  • the physical host can be connected to a PCIE device, which is a hardware device that supports a single input-output virtualized SR-IOV standard.
  • the PCIE device includes at least one physical function (English: Physical Function, PF for short) and at least one virtual function VF, wherein each PF and each VF is configured with a unique function number.
  • the virtual machine performs a specific operation through PF or VF Service, each PF or VF corresponds to a virtual machine, and the correspondence between the function number of the PF or VF and the virtual machine can be saved in the PCIE device.
  • the physical host includes hardware associated with running the virtual machine, such as a CPU and at least one computer readable storage medium (hard disk, memory, etc.).
  • the physical host may further include a communication unit, an input unit, a display unit, and the like.
  • the aforementioned PCIE device may be a PCIE adapter, such as a network card adapter, a graphics card, or the like.
  • the physical host also includes a migration management device.
  • the migration management device on the source physical host migrates the data stored in the memory to the target physical host.
  • the migration management device can be implemented by the aforementioned VMM.
  • An embodiment of the present invention provides a service context management method, which is performed by one of the foregoing at least one virtual machine. Referring to FIG. 2, the method includes:
  • Step 101 The virtual machine allocates memory for at least one state offload service carried by the PCIE device.
  • Step 102 The virtual machine generates a context table in the memory, where the context table is used to record a service context of at least one state offload service carried by the PCIE device.
  • Step 103 The virtual machine sends the first address of the memory and the function number corresponding to the virtual machine to the PCIE device.
  • the PCIE device records the correspondence between the first address and the function number of the memory, and according to the correspondence between the first address of the memory and the function number. , get the business context of the business from the state of the unloaded business.
  • the present invention provides a service context management method, in which a virtual machine allocates memory for at least one state offload service carried by a PCIE device, and then generates a context table in the memory, and then sends the first address of the memory and the function number corresponding to the virtual machine to the PCIE.
  • the device enables the PCIE device to record the correspondence between the first address and the function number of the memory, and obtains the service context of the state offload service from the memory according to the correspondence between the first address and the function number of the memory, which is convenient to operate and fast; When the virtual machine is hot-migrated, it avoids the trouble of register migration.
  • An embodiment of the present invention provides a service context management method, which is performed by the foregoing PCIE device. Referring to FIG. 3, the method includes:
  • Step 201 The PCIE device receives the service information of the offloading service to be processed sent by the host side or the network side.
  • Step 202 The PCIE device determines the function number corresponding to the unloading service to be processed according to the service information of the offloading service in the to-be-processed state.
  • Step 203 The PCIE device saves the first address of the memory corresponding to the unloading service in the to-be-processed state according to the determined function number of the unloading service and the corresponding relationship between the function number and the first address of the memory, and the context table is saved in the memory.
  • the context table is used to record the business context of the state offloading service carried by the PCIE device.
  • Step 204 The PCIE device acquires the KEY from the service information of the offloaded service in the to-be-processed state.
  • Step 205 The PCIE device acquires the service context of the unloaded service in the pending state according to the first address of the memory corresponding to the unloading service and the KEY.
  • the present invention provides a service context management method.
  • the PCIE device determines the function number corresponding to the unloading service to be processed according to the service information of the unloading service to be processed, and the corresponding relationship between the function number and the first address of the memory, and the determined pending status.
  • hot migration the trouble of register migration is avoided, and the memory is directly migrated to the destination physical machine, and the function number corresponding to the virtual machine after migration and the first address of the memory of the virtual machine in the source physical host are The corresponding relationship can be configured into the migrated PCIE device, which is convenient and simple to implement, greatly speeds up the hot migration of the virtual machine, and shortens the downtime.
  • An embodiment of the present invention provides a service context management method, which is performed by the foregoing migration management device. Referring to FIG. 4, the method includes:
  • Step 301 During hot migration, the migration management device in the destination physical host acquires the first address of the memory copied from the source physical host memory, and the destination physical host and the source physical host respectively interconnect the standard PCIE device with different shortcut peripheral components.
  • the PCIE device is a hardware device that supports a single input/output virtualization SR-IOV standard.
  • the PCIE device includes at least one physical function PF. And at least one virtual function VF, wherein each PF and each VF is configured with a unique function number.
  • Step 302 Obtain a function number corresponding to the migrated virtual machine.
  • Step 303 Configure the correspondence between the first address of the memory and the function number to the PCIE device of the destination physical host.
  • the trouble caused by the migration of the register is avoided, and the memory is directly migrated to the destination physical machine, and then the function number corresponding to the virtual machine after the migration and the virtual machine in the source physical host are used.
  • the correspondence between the first address of the memory can be configured into the migrated PCIE device, which is convenient and simple to implement, greatly speeds up the hot migration of the virtual machine, and shortens the downtime.
  • An embodiment of the present invention provides a service context management method. Referring to FIG. 5, the method includes:
  • Step 401 The virtual machine allocates memory for at least one state offload service carried by the PCIE device. This memory is used to record the context of the state offload service.
  • the VM When the VM is started, the VM requests the memory of the physical host for the PCIE device, and then allocates the requested memory to the state offload service. Specifically, after receiving the application of the virtual machine, the PCIE device driver in the physical host allocates memory for the virtual machine, wherein the PCIE device driver in the physical host refers to driving the PCIE device in the hard disk installed in the physical host. Drivers.
  • Step 402 The virtual machine generates a context table in the memory, where the context table is used to record a service context of at least one state offload service carried by the PCIE device.
  • the context table may be a direct index table or a hash table
  • the direct index table includes a correspondence between the keyword KEY and the service context
  • the hash table includes a correspondence between the hash value and the service context.
  • the size M of the bucket is modulo, and a hash value C is obtained; the hash value C is associated with the business context, and the correspondence between the hash value and the business context is obtained.
  • different service packets or service data correspond to different protocols. Therefore, different service packets or service data may adopt different types of KEYs.
  • a service message based on the Remote Direct Memory Access (RDMA) protocol may be used.
  • the queue pair (English: Queue Pair, QP for short) is used as the KEY in the RDMA protocol. It is based on the Transmission Control Protocol (English: Transmission Control Protocol, TCP: Internet Protocol, IP for short).
  • TCP Transmission Control Protocol
  • IP Internet Protocol
  • IP quintuple source IP address, destination IP address, source TCP port, destination TCP port, and protocol number
  • the direct index table is directly indexed by an array, and the KEY with a long number of bits such as the IP quintuple cannot be used as an index of the direct index table, a hash table is needed to implement the index.
  • the state offload service may be a TCP offload service, a small computer system interface (English: Internet Small Computer System Interface, iSCSI) offload service, and an Ethernet Fibre Channel (English: Fibre Channel over Ethernet, FCoE for short) ) Unloading business, etc.
  • Step 403 The virtual machine sends the first address of the memory and the function number corresponding to the virtual machine to the PCIE device.
  • the first address of the memory may be the physical address of the physical host (also referred to as the physical address of the host, English: Host Physical Address, referred to as HPA), or the physical address of the virtual machine (also referred to as the physical address of the client).
  • English: Guest Physical Address (GPA) if it is GPA, when PCIE device accesses this address, it needs to use CPU VT-D (Intel technology) or IOMMU (AMD technology), or other automatic Implement GPA to HPA address translation technology to achieve GPA to HPA conversion.
  • step 403 can also be performed by the VMM.
  • the virtual machine may also save the correspondence between the first address of the memory and the function number for subsequent use.
  • Step 404 The PCIE device receives the first address of the memory sent by each virtual machine and the function number corresponding to each virtual machine, and saves the correspondence between the first place of the memory and the function number.
  • Step 405 The PCIE device receives the service information of the offloading service to be processed sent by the host side or the network side, and the at least one state offloading service includes the unloading service to be processed.
  • the service information of the offloading service in the to-be-processed state includes the service information sent by the host side or the service information sent by the network side.
  • the service information sent by the host side is usually service data
  • the service information sent by the network side is usually a service packet.
  • the host side refers to the system side composed of the CPU, memory, and hard disk in the physical host.
  • the network side refers to the external network side to which the physical host is connected.
  • Step 406 The PCIE device determines, according to the service information of the offloading service in the to-be-processed state, the function number corresponding to the unloading service to be processed.
  • step 406 may include:
  • the function of the offloading service corresponding to the to-be-processed state is determined according to the feature in the service information of the offloading service sent by the host side or the function of receiving the service information of the offloading service of the to-be-processed state.
  • Process ID The function number corresponding to the uninstall service.
  • the feature in the service packet sent by the network side may be a field in the service packet, and different fields may be used as the feature here according to the type of the PCIE device.
  • the PCIE device is a NIC adapter
  • the destination media access control (English: Media Access Control, MAC address) (or virtual LAN number + MAC) can be used as a feature, that is, when receiving a service packet, the NIC adapter
  • the corresponding function number can be determined according to the destination MAC address in the service packet.
  • For the host side to send service data it can be determined according to the PCIE interface used when receiving the host side to transmit service data.
  • Step 407 The PCIE device saves the first address of the memory corresponding to the unloading service in the to-be-processed state according to the determined function number of the unloading service and the corresponding relationship between the function number and the first address of the memory, and the context table is saved in the memory.
  • the context table is used to record the business context of the state offloading service carried by the PCIE device.
  • Step 408 The PCIE device acquires the KEY from the service information of the offloaded service in the to-be-processed state.
  • step 402 the content of the KEY in the service information has been described in step 402, and details are not described herein again.
  • Step 409 The PCIE device acquires the service context of the unloaded service in the pending state according to the first address of the memory corresponding to the unloading service and the KEY.
  • step 409 can be implemented in the following two ways:
  • the PCIE device searches for the corresponding direct index table or hash table in the memory of the virtual machine by using the first address of the memory corresponding to the unloading service in the pending state.
  • the PCIE device directly reads the service context in the direct index table corresponding to the first address of the found memory, or calculates the hash value C according to the KEY, and uses the hash value C to read the first memory found.
  • the first implementation manner may be performed by a direct memory access (English: Direct Memory Access, DMA for short) read/write module in the PCIE device.
  • a direct memory access English: Direct Memory Access, DMA for short
  • DMA Direct Memory Access
  • the hash value C is calculated according to the KEY, including:
  • the hash function (for example, the CRC16 function) is used to convert the KEY to the total hash value A; the size of the hash bucket corresponding to the virtual bucket corresponding to the virtual machine in the hash table or the VF corresponding to the unloading service to be processed is determined; The total hash value A is used to modulo the size M of the hash bucket to obtain a hash value C.
  • the following takes the TCP offload service (the PCIE device as the NIC adapter) as an example.
  • VM1 supports 64K (64*1024) TCP connection
  • VM2 supports 128K (128*1024) TCP connection
  • the network adapter supports the maximum number of TCP connections is 1M (1024*1024).
  • VM1 and VM2 need to perform TCP offload service, and each connection for TCP offload service (a separate offload state task) requires a context of 256 bytes.
  • VM1 is initialized, apply for a hash table that supports 64K entries for the 64K TCP connection. Assume that the hash bucket size of the hash table is also 64K.
  • a hash table supporting 128K entries is allocated. It is assumed that the hash bucket size of the hash table is also 128K, and the maximum size of the NIC adapter is 1M.
  • the NIC adapter When the TCP offload service packet arrives at the NIC adapter, the NIC adapter first obtains the function number of the VF or PF of the exit of the TCP offload service packet according to the VLAN+MAC query MAC address in the TCP offload service packet, and the function number corresponds to Is VM1 and gets a hash bucket size of 64K.
  • the IP quintuple is obtained from the TCP header of the TCP offload service packet.
  • the IP quintuple is used to read the contents of the hash table in the cache of the NIC adapter. If the entry is in the cache of the NIC adapter, the service context is directly obtained.
  • the hash value of the IP quintuple is calculated, which is assumed to be 0x21523; the modulo operation is performed by 0x21523 and 64K, and the hash value is 0x1523, which is indexed by 0x1523, and the corresponding service context is queried in the hash table of VM1 to obtain the service. The context is then loaded into the cache of the NIC adapter for subsequent business processing.
  • the PCIE device sends a request message for acquiring a service context of the to-be-processed offload service to the virtual machine corresponding to the function number of the unloaded service to be processed, and the request message includes the unloading service corresponding to the to-be-processed state.
  • the virtual machine receives a request message sent by the PCIE device to obtain a service context of the to-be-processed offload service.
  • the virtual machine determines a corresponding direct index table or a hash table according to the first address of the memory in the request message.
  • the request message may not include the first address of the memory, and only needs to carry the function number, and the virtual machine obtains the corresponding direct index table according to the function number and the correspondence between the first address of the memory and the function number stored by the virtual machine or Hash table.
  • the virtual machine reads the service context from the direct index table corresponding to the first address of the determined memory according to the KEY in the request message, or calculates the hash value C according to the KEY in the request message, and uses the hash value C to The service context is read in the hash table corresponding to the first address of the determined memory.
  • the manner in which the virtual machine calculates the hash value C according to the KEY in the request message is the same as the manner in which the PCIE device calculates the hash value C in the first implementation manner.
  • the virtual machine sends the read service context to the PCIE device.
  • the fifth step may be implemented in the following manner: the read service context is written into the cache of the PCIE device by using a cache Cache mechanism.
  • the PCIE device receives the service context of the virtual machine to offload the service according to the pending state sent by the request message.
  • Step 410 The PCIE device processes the service information according to the context of the offloaded service according to the to-be-processed state.
  • the processing is performed by the corresponding VF.
  • Step 411 During the hot migration, the migration management device on the source physical host directly migrates the memory block and the context table to the destination virtual machine.
  • Step 412 The migration management device on the destination physical host configures the correspondence between the first address of the virtual machine in the source physical host and the function number corresponding to the virtual machine after the migration to the PCIE device connected to the destination physical host.
  • the migration management device in the destination physical host obtains the first address of the memory copied from the source physical host memory, and the destination physical host and the source physical host respectively have different shortcut peripherals.
  • the component interconnects a standard PCIE device connection.
  • the PCIE device is a hardware device that supports a single input/output virtualization SR-IOV standard.
  • the PCIE device includes at least one physical function PF and at least one virtual function VF, wherein each PF and each The VF is configured with a unique function number; the function number corresponding to the migrated virtual machine is obtained; and the correspondence between the first address of the memory and the function number is configured to the PCIE device of the destination physical host.
  • the migration management device may be a VMM.
  • the present invention provides a service context management method, in which a virtual machine allocates memory for at least one state offload service carried by a PCIE device, and then generates a context table in the memory, and then sends the first address of the memory and the function number corresponding to the virtual machine to the PCIE.
  • the device enables the PCIE device to record the correspondence between the first address and the function number of the memory, and obtains the service context of the state offload service from the memory according to the correspondence between the first address and the function number of the memory, which is convenient to operate and fast; When the virtual machine is hot-migrated, it avoids the trouble of register migration.
  • the embodiment of the present invention provides a physical host, which is applicable to the method provided in Embodiment 1.
  • the physical host 51 runs a virtual machine 510, and the physical host 51 is connected to the fast peripheral component interconnection standard PCIE device 52.
  • the PCIE device 52 is a hardware device supporting a single input/output virtualized SR-IOV standard.
  • the PCIE device 52 includes at least one physical function PF 521 and at least one virtual function VF 522, and each PF 521 and each VF 522
  • the configuration has a unique function number, and the aforementioned virtual machine 510 includes:
  • the allocating module 511 is configured to allocate memory for at least one state offloading service carried by the PCIE device;
  • a generating module 512 configured to generate a context table in the memory, where the context table is used to record a service context of at least one state offload service carried by the PCIE device;
  • the sending module 513 is configured to send the first address and the function number of the memory to the PCIE device, where the PCIE device records the correspondence between the first address and the function number of the memory, and according to the correspondence between the first address of the memory and the function number, the memory is obtained from the memory. Get the business context of the state unloading business.
  • the present invention provides a service context management method, where a virtual machine is carried by a PCIE device.
  • One state unloads the service to allocate memory, and then generates a context table in the memory, and then sends the first address of the memory and the function number corresponding to the virtual machine to the PCIE device, so that the PCIE device can record the correspondence between the first address of the memory and the function number.
  • the business context of the state unloading service is obtained from the memory, and the operation is convenient and fast; in the hot migration of the virtual machine, the trouble caused by the migration of the register is avoided, You need to migrate the memory directly to the destination physical machine, and then configure the mapping between the function number of the virtual machine and the first address of the virtual machine in the source physical host to the migrated PCIE device. This is convenient and simple. It greatly speeds up the hot migration of virtual machines and shortens the downtime.
  • the embodiment of the present invention provides a physical host, which is applicable to the method provided in Embodiment 4.
  • the physical host 61 runs a virtual machine 610, and the physical host 61 is connected to the shortcut peripheral component interconnection standard PCIE device 62.
  • the PCIE device 62 is a hardware device supporting a single input/output virtualized SR-IOV standard.
  • the PCIE device 62 includes at least one physical function PF 621 and at least one virtual function VF 622, and each PF 621 and each VF 622
  • the configuration has a unique function number, and the foregoing virtual machine 610 includes:
  • the allocating module 611 is configured to allocate memory for at least one state offloading service carried by the PCIE device.
  • the VM When the VM is started, the VM requests the memory of the physical host for the PCIE device, and then allocates the requested memory to the state offload service. Specifically, after receiving the application of the virtual machine, the PCIE device driver in the physical host allocates memory for the virtual machine, wherein the PCIE device driver in the physical host refers to driving the PCIE device in the hard disk installed in the physical host. Drivers.
  • the generating module 612 is configured to generate a context table in the memory, where the context table is used to record a service context of the at least one state offload service carried by the PCIE device.
  • the context table may be a direct index table or a hash table
  • the direct index table includes a correspondence between the keyword KEY and the service context
  • the hash table includes a correspondence between the hash value and the service context.
  • the correspondence between the hash value and the service context in the hash table may be established by the virtual machine in the physical host, and may be implemented in the following manner: converting the KEY into the total hash value A; determining the virtual in the hash table The size of the hash bucket corresponding to the VF corresponding to the machine in the hash bucket or the unloading service to be processed; the total hash value A is used to modulate the size M of the hash bucket to obtain a hash value C; The hash value C is associated with the business context, and the correspondence between the hash value and the business context is obtained.
  • different service packets or service data correspond to different protocols. Therefore, different service packets or service data may adopt different types of KEYs.
  • the RDMA protocol-based service packet can use the QP number in the RDMA protocol as the KEY; the TCP/IP-based service packet can use the IP quintuple (source IP address, destination IP address, source TCP port, destination TCP). Port and protocol number) as KEY.
  • the direct index table is directly indexed by an array, and the KEY with a long number of bits such as the IP quintuple cannot be used as an index of the direct index table, a hash table is needed to implement the index.
  • the state offload service may be a TCP offload service, an iSCSI offload service, an FCoE offload service, or the like.
  • the sending module 613 is configured to send the first address and the function number of the memory to the PCIE device, where the PCIE device records the correspondence between the first address and the function number of the memory, and according to the correspondence between the first address of the memory and the function number, the memory is obtained from the memory. Get the business context of the state unloading business.
  • the first address of the memory can be either HPA or GPA. If it is GPA, the PCIE device needs to use the CPU VT-D (Intel technology) or IOMMU (AMD technology) when accessing the address. ), or other technology that automatically implements GPA to HPA address translation, enabling GPA to HPA conversion.
  • the function number of the PF or VF corresponding to the virtual machine may be specified by the virtual machine or may be specified by the VMM.
  • the foregoing at least one state offloading service includes a to-be-processed state offloading service
  • the virtual machine may further include:
  • the receiving module 614 is configured to receive a request message sent by the PCIE device to obtain a service context of the to-be-processed state offload service, where the request message includes a first address of the memory and a keyword KEY;
  • a determining module 615 configured to determine, according to a first address of a memory in the request message, a corresponding direct index table or a hash table;
  • the reading module 616 is configured to read the service context from the direct index table corresponding to the first address of the determined memory according to the KEY in the request message, or calculate the hash value C according to the KEY in the request message, and adopt a hash value C. Reading a business context from a hash table corresponding to the determined first address of the memory;
  • the foregoing sending module 613 is further configured to send the read service context to the PCIE device.
  • the foregoing reading module 616 is specifically configured to:
  • the hash function is used to convert the KEY in the request message to the total hash value A; the size of the hash bucket corresponding to the virtual bucket corresponding to the virtual machine in the hash table or the VF corresponding to the unloading service to be processed in the hash table is determined; The total hash value A is modulo the size M of the hash bucket to obtain a hash value C.
  • TCP offload service (the PCIE device as the network adapter) as an example to describe the working process of the virtual machine 610:
  • VM1 supports 64K (64*1024) TCP connection
  • VM2 supports 128K (128*1024) TCP connection
  • the network adapter supports the maximum number of TCP connections is 1M (1024*1024).
  • VM1 and VM2 need to perform TCP offload service, and each connection for TCP offload service (a separate offload state task) requires a context of 256 bytes.
  • VM1 is initialized, apply for a hash table that supports 64K entries for the 64K TCP connection. Assume that the hash bucket size of the hash table is also 64K.
  • a hash table supporting 128K entries is allocated. It is assumed that the hash bucket size of the hash table is also 128K, and the maximum size of the NIC adapter is 1M.
  • the virtual machine When the virtual machine receives the request message sent by the PCIE device for acquiring the service context of the to-be-processed offloading service, the virtual machine first obtains the hash table according to the first address of the memory. Then, according to the KEY, a hash value C is obtained.
  • the KEY is an IP quintuple, which is assumed to be 0x21523; the modulo operation is performed by 0x21523 and 64K, and the hash value is 0x1523, and the 0x1523 is indexed, and the corresponding business context is queried in the hash table to obtain the business context. Send it to the NIC adapter for later business processing.
  • the foregoing sending module 613 is specifically configured to: write the read service context into the cache of the PCIE device by using a cache Cache mechanism.
  • the present invention provides a service context management method, in which a virtual machine allocates memory for at least one state offload service carried by a PCIE device, and then generates a context table in the memory, and then sends the first address of the memory and the function number corresponding to the virtual machine to the PCIE.
  • the device enables the PCIE device to record the correspondence between the first address and the function number of the memory, and obtains the service context of the state offload service from the memory according to the correspondence between the first address and the function number of the memory, which is convenient to operate and fast; When the virtual machine is hot-migrated, it avoids the trouble of register migration.
  • the embodiment of the present invention provides a PCIE device, which is applicable to the method provided in the second embodiment.
  • the PCIE device 71 is connected to a physical host 72.
  • the physical host 72 runs at least one virtual machine 720 and a PCIE device.
  • the PCIE device 71 includes at least one physical function PF 711 and at least one virtual function VF 712, each PF 711 and each VF 712 being configured with a unique function. No.
  • the PCIE device 71 further includes:
  • the receiving module 713 is configured to receive service information of the offloading service to be processed sent by the host side or the network side;
  • the first determining module 714 is configured to determine, according to the service information of the offloading service in the to-be-processed state, a function number corresponding to the to-be-processed unloading service;
  • the second determining module 715 is configured to save the first address of the memory corresponding to the unloading service to be processed according to the determined function number of the unloading service and the corresponding relationship between the function number and the first address of the memory, and save the memory in the memory.
  • There is a context table where the context table is used to record a service context of at least one state offload service carried by the PCIE device;
  • the first obtaining module 716 is configured to obtain a KEY from the service information of the offloaded service in the to-be-processed state;
  • the second obtaining module 717 is configured to acquire a service context of the to-be-processed unloading service according to the first address of the memory corresponding to the unloading service and the KEY according to the to-be-processed state.
  • the present invention provides a service context management method.
  • the PCIE device determines the function number corresponding to the unloading service to be processed according to the service information of the unloading service to be processed, and the corresponding relationship between the function number and the first address of the memory, and the determined pending status.
  • hot migration the trouble of register migration is avoided, and the memory is directly migrated to the destination physical machine, and the function number corresponding to the virtual machine after migration and the first address of the memory of the virtual machine in the source physical host are The corresponding relationship can be configured into the migrated PCIE device, which is convenient and simple to implement, greatly speeds up the hot migration of the virtual machine, and shortens the downtime.
  • the embodiment of the present invention provides a PCIE device, which is applicable to the method provided in Embodiment 4.
  • the PCIE device 81 is connected to a physical host 82, and the physical host 82 runs to One less virtual machine 820
  • the PCIE device is a hardware device conforming to a single input and output virtualized SR-IOV standard
  • the PCIE device 81 includes at least one physical function PF 811 and at least one virtual function VF 812, each PF 811 and each The VF 812 is configured with a unique function number
  • the PCIE device 81 further includes:
  • the receiving module 813 is configured to receive service information of the offloading service to be processed sent by the host side or the network side.
  • the service information includes service information sent by the host side or service information sent by the network side.
  • the service information sent by the host side is usually service data, and the service information sent by the network side is usually a service packet.
  • the host side refers to the system side composed of CPU, memory, and hard disk in the physical host.
  • the network side refers to the external network side to which the physical host is connected.
  • the first determining module 814 is configured to determine, according to the service information of the offloading service in the to-be-processed state, the function number corresponding to the to-be-processed unloading service.
  • the second determining module 815 is configured to save the first address of the memory corresponding to the unloading service to be processed according to the determined function number of the unloading service and the corresponding relationship between the function number and the first address of the memory, and save the memory in the memory.
  • the context table may be a direct index table or a hash table
  • the direct index table includes a correspondence between the keyword KEY and the service context
  • the hash table includes a correspondence between the hash value and the service context.
  • the first obtaining module 816 is configured to obtain a KEY from the service information of the offloaded service in the to-be-processed state.
  • different service packets or service data correspond to different protocols. Therefore, different service packets or service data may adopt different types of KEYs.
  • the RDMA protocol-based service packet can use the QP number in the RDMA protocol as the KEY; the TCP/IP-based service packet can use the IP quintuple (source IP address, destination IP address, source TCP port, destination TCP). Port and protocol number) as KEY.
  • the second obtaining module 817 is configured to acquire a service context of the to-be-processed unloading service according to the first address of the memory corresponding to the unloading service and the KEY according to the to-be-processed state.
  • the foregoing second obtaining module 817 is specifically configured to:
  • the foregoing second obtaining module 817 can be used to:
  • the hash function is used to convert the KEY in the request message to the total hash value A; the size of the hash bucket corresponding to the virtual bucket corresponding to the virtual machine in the hash table or the VF corresponding to the unloading service to be processed in the hash table is determined; The total hash value A is modulo the size M of the hash bucket to obtain a hash value C.
  • VM1 supports 64K (64*1024) TCP connection
  • VM2 supports 128K (128*1024) TCP connection
  • the network adapter supports the maximum number of TCP connections is 1M (1024*1024).
  • VM1 and VM2 need to perform TCP offload service, and each connection for TCP offload service (a separate offload state task) requires a context of 256 bytes.
  • VM1 is initialized, apply for a hash table that supports 64K entries for the 64K TCP connection. Assume that the hash bucket size of the hash table is also 64K.
  • a hash table supporting 128K entries is allocated. It is assumed that the hash bucket size of the hash table is also 128K, and the maximum size of the NIC adapter is 1M.
  • the NIC adapter When the TCP offload service packet arrives at the NIC adapter, the NIC adapter first obtains the function number of the VF or PF of the exit of the TCP offload service packet according to the VLAN+MAC query MAC address in the TCP offload service packet, and the function number corresponds to Is VM1 and gets a hash bucket size of 64K.
  • the IP quintuple is obtained from the TCP header of the TCP offload service packet.
  • the IP quintuple is used to read the contents of the hash table in the cache of the NIC adapter. If the entry is in the cache of the NIC adapter, the service context is directly obtained.
  • the hash value of the IP quintuple is calculated, which is assumed to be 0x21523; the modulo operation is performed by 0x21523 and 64K, and the hash value is 0x1523, which is indexed by 0x1523, and the corresponding service context is queried in the hash table of VM1 to obtain the service. The context is then loaded into the cache of the NIC adapter for subsequent business processing.
  • the foregoing second obtaining module 817 is specifically configured to:
  • the receiving virtual machine unloads the business context of the service according to the pending state sent by the request message.
  • the foregoing first determining module 814 is specifically configured to: uninstall the feature in the service information of the service according to the to-be-processed state sent by the host side, or receive the host-side transmission to be processed.
  • the PCIE interface used to carry the service information of the service determines the function number corresponding to the unloading service in the pending state, and the feature is used to indicate the function number corresponding to the unloading service in the pending state.
  • the feature in the service packet sent by the network side may be a field in the service packet, and different fields may be used as the feature here according to the type of the PCIE device.
  • the PCIE device is a NIC adapter
  • the destination MAC address (or virtual LAN number + MAC) can be used as a feature.
  • the NIC adapter can determine the corresponding MAC address according to the destination MAC address in the service packet. Function number. For the host side to send service data, it can be determined according to the PCIE interface used when receiving the host side to transmit service data.
  • the foregoing receiving module 813 is further configured to receive a first address of a memory of each virtual machine and a function number corresponding to each virtual machine;
  • the PCIE device may further include: a storage module 818, configured to save a correspondence between a first address of the memory and a function number.
  • the present invention provides a service context management method.
  • the PCIE device determines the function number corresponding to the unloading service to be processed according to the service information of the unloading service to be processed, and the corresponding relationship between the function number and the first address of the memory, and the determined pending status.
  • hot migration the trouble of register migration is avoided, and the memory is directly migrated to the destination physical machine, and the function number corresponding to the virtual machine after migration and the first address of the memory of the virtual machine in the source physical host are The corresponding relationship can be configured into the migrated PCIE device, which is convenient and simple to implement, greatly speeds up the hot migration of the virtual machine, and shortens the downtime.
  • the embodiment of the present invention provides a migration management device, which is disposed in a destination physical host, and is applicable to the method provided in Embodiment 3.
  • the migration management device includes:
  • the first obtaining module 901 is configured to acquire, when hot migration, a first address of the memory copied from the source physical host memory, and the destination physical host and the source physical host respectively connect with different fast peripheral components to interconnect the standard PCIE device, PCIE
  • the device is a hardware device supporting a single input and output virtualization SR-IOV standard.
  • the PCIE device includes at least one physical function PF and at least one virtual function VF, wherein each PF and each VF are configured with a unique function number;
  • the second obtaining module 902 is configured to obtain a function number corresponding to the migrated virtual machine.
  • the configuration module 903 is configured to configure the correspondence between the first address of the memory and the function number to the PCIE device of the destination physical host.
  • the trouble caused by the migration of the register is avoided, and the memory is directly migrated to the destination physical machine, and then the function number corresponding to the virtual machine after the migration and the virtual machine in the source physical host are used.
  • the correspondence between the first address of the memory can be configured into the migrated PCIE device, which is convenient and simple to implement, greatly speeds up the hot migration of the virtual machine, and shortens the downtime.
  • the virtual machine provided by the foregoing embodiment performs virtual machine memory management
  • only the division of each functional module is used as an example.
  • the function distribution may be completed by different functional modules as needed.
  • one of the function assignments is completed by more modules, or multiple of the functions are assigned by the same module, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.
  • the physical host and the service context management method embodiment provided by the foregoing embodiments are in the same concept, and the specific implementation process is described in the method embodiment, and details are not described herein again.
  • a person skilled in the art may understand that all or part of the steps of implementing the above embodiments may be completed by hardware, or may be instructed by a program to execute related hardware, and the program may be stored in a computer readable storage medium.
  • the storage medium mentioned may be a read only memory, a magnetic disk or an optical disk or the like.

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

L'invention concerne un procédé de gestion de contexte de service, une machine principale physique, un dispositif PCIE et un dispositif de gestion de migration, qui appartiennent au domaine des machines virtuelles et sont applicables aux machines virtuelles. Une machine virtuelle est exécutée sur une machine principale physique, la machine principale physique étant connectée à un dispositif d'interconnexion de composant périphériques express (PCIE), le dispositif PCIE étant un dispositif matériel qui satisfait à une norme de virtualisation d'entrée/sortie à racine unique (SR-IOV), et le dispositif PCIE comprenant au moins une fonction physique (PF) et au moins une fonction virtuelle (VF), chacune des PF et chacune des VF étant toutes deux configurées avec un numéro fonctionnel unique. Le procédé comprend les étapes suivantes : une machine virtuelle attribue une mémoire à au moins un service de déchargement d'état supporté par un dispositif PCIE (101) ; la machine virtuelle génère une table de contexte dans la mémoire, la table de contexte étant utilisée pour enregistrer le contexte du ou des services de déchargement d'état supportés par le dispositif PCIE (102) ; et la machine virtuelle envoie une première adresse de la mémoire et un numéro fonctionnel au dispositif PCIE (103).
PCT/CN2015/089815 2015-01-27 2015-09-16 Procédé de gestion de contexte de service, machine principale physique, dispositif pcie et dispositif de gestion de migration WO2016119469A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510041218.6A CN104636185B (zh) 2015-01-27 2015-01-27 业务上下文管理方法、物理主机、pcie设备及迁移管理设备
CN201510041218.6 2015-01-27

Publications (1)

Publication Number Publication Date
WO2016119469A1 true WO2016119469A1 (fr) 2016-08-04

Family

ID=53214977

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/089815 WO2016119469A1 (fr) 2015-01-27 2015-09-16 Procédé de gestion de contexte de service, machine principale physique, dispositif pcie et dispositif de gestion de migration

Country Status (2)

Country Link
CN (1) CN104636185B (fr)
WO (1) WO2016119469A1 (fr)

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104636185B (zh) * 2015-01-27 2018-03-02 华为技术有限公司 业务上下文管理方法、物理主机、pcie设备及迁移管理设备
CN107465626A (zh) * 2016-06-03 2017-12-12 阿里巴巴集团控股有限公司 一种虚拟机热迁移方法及装置
CN108255597B (zh) * 2016-12-28 2022-02-18 北京华耀科技有限公司 基于物理功能的mac地址分配虚拟功能的mac地址方法
CN108063737B (zh) * 2017-11-23 2020-09-08 华中科技大学 一种FCoE存储区域网读请求处理方法及系统
TWI645293B (zh) * 2017-11-30 2018-12-21 財團法人工業技術研究院 在多根輸入輸出虛擬化環境中虛擬機器即時遷移的方法、系統及其計算主機
CN112398817B (zh) * 2018-01-23 2022-02-25 华为技术有限公司 数据发送的方法及设备
CN109753346B (zh) * 2018-12-25 2021-01-01 新华三云计算技术有限公司 一种虚拟机热迁移方法及装置
CN112243046B (zh) 2019-07-19 2021-12-14 华为技术有限公司 通信方法和网卡
CN114253979B (zh) * 2021-12-23 2023-10-03 北京百度网讯科技有限公司 一种报文处理方法、装置及电子设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090313391A1 (en) * 2008-06-11 2009-12-17 Hitachi, Ltd. Computer system, device sharing method, and device sharing program
WO2013066572A2 (fr) * 2011-10-31 2013-05-10 Intel Corporation Migration d'état d'adaptateur d'accès direct à la mémoire à distance dans un environnement virtuel
CN103201721A (zh) * 2012-08-29 2013-07-10 华为技术有限公司 虚拟机热迁移的系统和方法
CN103942087A (zh) * 2014-03-31 2014-07-23 华为技术有限公司 虚拟机热迁移方法及相关装置和集群系统
CN104636185A (zh) * 2015-01-27 2015-05-20 华为技术有限公司 业务上下文管理方法、物理主机、pcie设备及迁移管理设备

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101850318B1 (ko) * 2011-12-09 2018-04-20 삼성전자주식회사 가상 메모리 관리 장치 및 방법
CN102819447B (zh) * 2012-05-29 2015-06-03 中国科学院计算技术研究所 一种用于多根共享系统的直接i/o虚拟化方法和装置
CN102779180B (zh) * 2012-06-29 2015-09-09 华为技术有限公司 数据存储系统的操作处理方法,数据存储系统

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090313391A1 (en) * 2008-06-11 2009-12-17 Hitachi, Ltd. Computer system, device sharing method, and device sharing program
WO2013066572A2 (fr) * 2011-10-31 2013-05-10 Intel Corporation Migration d'état d'adaptateur d'accès direct à la mémoire à distance dans un environnement virtuel
CN103201721A (zh) * 2012-08-29 2013-07-10 华为技术有限公司 虚拟机热迁移的系统和方法
CN103942087A (zh) * 2014-03-31 2014-07-23 华为技术有限公司 虚拟机热迁移方法及相关装置和集群系统
CN104636185A (zh) * 2015-01-27 2015-05-20 华为技术有限公司 业务上下文管理方法、物理主机、pcie设备及迁移管理设备

Also Published As

Publication number Publication date
CN104636185B (zh) 2018-03-02
CN104636185A (zh) 2015-05-20

Similar Documents

Publication Publication Date Title
WO2016119469A1 (fr) Procédé de gestion de contexte de service, machine principale physique, dispositif pcie et dispositif de gestion de migration
US11088949B2 (en) Multicast message filtering in virtual environments
US9529773B2 (en) Systems and methods for enabling access to extensible remote storage over a network as local storage via a logical storage controller
CN111221758B (zh) 处理远程直接内存访问请求的方法和计算机设备
US9294567B2 (en) Systems and methods for enabling access to extensible storage devices over a network as local storage via NVME controller
US7464191B2 (en) System and method for host initialization for an adapter that supports virtualization
US11099872B2 (en) Techniques to copy a virtual machine
EP3206124A1 (fr) Procédé, appareil et système pour accéder à un dispositif de stockage
US10872056B2 (en) Remote memory access using memory mapped addressing among multiple compute nodes
US9092365B2 (en) Splitting direct memory access windows
US9910687B2 (en) Data flow affinity for heterogenous virtual machines
WO2016119468A1 (fr) Procédé de gestion de mémoire de machine virtuelle, machine principale physique, dispositif pcie et son procédé de configuration, et dispositif de gestion de migration
US20130110959A1 (en) Remote direct memory access adapter state migration in a virtual environment
US11757796B2 (en) Zero-copy processing
TW201423422A (zh) 具有PCIe介面的裝置共享系統及方法
US11940933B2 (en) Cross address-space bridging
US9483290B1 (en) Method and system for virtual machine communication
US20230221874A1 (en) Method of efficiently receiving files over a network with a receive file command
US20230224356A1 (en) Zero-copy method for sending key values
US9104600B2 (en) Merging direct memory access windows
US10747594B1 (en) System and methods of zero-copy data path among user level processes
US11979459B1 (en) Configuration of data connections between a host and a shared network adapter
US20230171189A1 (en) Virtual network interfaces for managed layer-2 connectivity at computing service extension locations

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15879667

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 15879667

Country of ref document: EP

Kind code of ref document: A1