CN110071900A - The method and apparatus that data are sent - Google Patents

The method and apparatus that data are sent Download PDF

Info

Publication number
CN110071900A
CN110071900A CN201810064378.6A CN201810064378A CN110071900A CN 110071900 A CN110071900 A CN 110071900A CN 201810064378 A CN201810064378 A CN 201810064378A CN 110071900 A CN110071900 A CN 110071900A
Authority
CN
China
Prior art keywords
virtual
port
virtual machine
address
mark
Prior art date
Legal status (The legal status 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 status listed.)
Granted
Application number
CN201810064378.6A
Other languages
Chinese (zh)
Other versions
CN110071900B (en
Inventor
付斌章
苏金钊
谭焜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huawei Technologies Co Ltd
Original Assignee
Huawei Technologies Co Ltd
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 Huawei Technologies Co Ltd filed Critical Huawei Technologies Co Ltd
Priority to CN201810064378.6A priority Critical patent/CN110071900B/en
Priority to CN202011203280.8A priority patent/CN112398817B/en
Publication of CN110071900A publication Critical patent/CN110071900A/en
Application granted granted Critical
Publication of CN110071900B publication Critical patent/CN110071900B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L63/00Network architectures or network communication protocols for network security
    • H04L63/02Network architectures or network communication protocols for network security for separating internal from external traffic, e.g. firewalls
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L61/00Network arrangements, protocols or services for addressing or naming
    • H04L61/09Mapping addresses
    • H04L61/10Mapping addresses of different types
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/01Protocols
    • H04L67/08Protocols specially adapted for terminal emulation, e.g. Telnet
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L69/00Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
    • H04L69/30Definitions, standards or architectural aspects of layered protocol stacks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • G06F2009/45587Isolation or security of virtual machine instances

Abstract

The embodiment of the invention provides the method and apparatus that a kind of data are sent, it is related to field of communication technology, method includes when the first virtual machine and the second virtual machine establish data connection, first virtual machine is according to the virtual address of the first virtual machine, determine the global identity GID of the corresponding RDMA network of the first virtual machine, first virtual machine passes through the corresponding first port of GID, to host transmit queue to QP attribute setting information, QP attribute setting information includes the first instruction and the second instruction, host deletes the first instruction, it is third instruction by the second instruction modification, the attribute of QP queue is arranged according to modified QP attribute setting information for host, so that HCA card sends data to the second virtual machine based on the QP queue after setting.It is isolated using the virtual network that virtual machine can be thus achieved in the present invention with physical network, guarantees the safety of virtual machine, and RoCE agreement can be compatible with.

Description

The method and apparatus that data are sent
Technical field
This application involves the method and apparatus that field of communication technology more particularly to data are sent.
Background technique
Currently, the remote direct data based on fusion Ethernet accesses (Remote Direct Memory Access Over Converged Ethernet, RoCE) agreement is widely used in cloud computing, using RoCE agreement, on the one hand, can To solve traditional transmission control protocol/Internet Protocol (Transmission Control Protocol/ Internet Protocol, TCP/IP) agreement sending and receiving data when, network delay is big, central processing unit (Central Processing Unit, CPU) high problem is loaded, another aspect can be compatible with Ethernet protocol, meet low investment, low fortune Seek the requirement of cost.Multiple virtual machines are usually provided in server in cloud computing, meanwhile, it is equipped with and supports RoCE agreement Host channel adapter (Host Channel Adapter, HCA), which includes multiple physical ports, for each physics Port, the physical port can be abstracted as the Ethernet interface and a support remote direct data of a support Ethernet protocol The RDMA interface of (Remote Direct Memory Access, RDMA) agreement is accessed, administrator can be in the Ethernet interface The corresponding physical IP address of upper configuration virtual machine.
In the prior art, it when the virtual machine in different server needs to carry out data communication, needs to open on HCA card Open hardware virtualization function, such as SR-IOV.In this way, virtual machine can be built according to the physical IP address of administrator configurations with other side Vertical connection establishes queue to (Queue Pairs, QP) on corresponding RDMA interface and the source address of the QP queue is arranged as this The physical IP address of itself, destination address are the physical IP address of other side.During subsequent data transmission, HCA card is directly read The data in the corresponding memory of application program are taken, the data are packaged according to source address in QP queue and destination address, are obtained The data message is transmitted to data message, and by physical network.
As it can be seen that in the prior art, can not can not be mentioned by the network virtualization of software realization RoCE agreement, the i.e. prior art It is isolated for a kind of virtual network that virtual machine can be thus achieved with physical network, guarantees the safety of virtual machine, and can be compatible with The data transmission method for uplink of RoCE agreement.
Summary of the invention
The embodiment of the invention provides the method and apparatus that a kind of data are sent, and virtual network data isolation may be implemented, And reduce the security risk of virtual machine.The technical solution is as follows:
In a first aspect, providing a kind of method that data are sent, the method is applied to equipment, and the equipment includes first Virtual machine, host and host channel adapter HCA card, wherein first virtual machine includes the port of virtual RDMA network interface card, institute The method of stating includes: virtual address of first virtual machine according to first virtual machine, determines that first virtual machine is corresponding RDMA network global identity GID;First virtual machine is by the corresponding first port of the GID, to the master For machine transmit queue to QP attribute setting information, the first port is the end of the virtual RDMA network interface card of first virtual machine Mouthful, the QP attribute setting information includes the first instruction and the second instruction, and first instruction is used to indicate the host for QP Source address in queue is set as the virtual address of first virtual machine, and second instruction is used to indicate the host for institute The virtual address that the destination address in QP queue is set as second virtual machine is stated, the QP queue is HCA card correspondence QP queue in, first virtual machine communicated with second virtual machine used in QP queue;Described in the host is deleted Second instruction modification is third instruction by the first instruction, and the third instruction is used to indicate the host will be in QP queue Destination address be set as the physical address of second virtual machine;The host is set according to modified QP attribute setting information The attribute of the QP queue is set, so that the HCA card sends number to second virtual machine based on the QP queue after setting According to.
The embodiment of the present invention, when host receives the QP attribute setting information of the first virtual machine transmission, QP attribute setting Information includes the first instruction and the second instruction, and the first instruction is used to indicate the host and sets for the source address in QP queue The virtual address of one virtual machine, the second instruction are used to indicate host and set the second virtual machine for the destination address in QP queue Virtual address, host delete the first instruction, are third instruction by the second instruction modification, and third instruction is used to indicate host for QP Destination address in queue is set as the physical address of the second virtual machine;Then it is arranged according to modified QP attribute setting information The attribute of QP queue, so that HCA card sends data to the second virtual machine based on the QP queue after setting.Based on the present invention To realize that the virtual network of virtual machine is isolated with physical network, guarantee the safety of virtual machine, and RoCE agreement can be compatible with.
In one possible implementation, it is described by second instruction modification be third instruct, comprising: the host The virtual address of first mapping table according to the pre-stored data and second virtual machine in second instruction, determines described the The physical address of two virtual machines, first mapping table include the corresponding relationship of virtual address and physical address;The host will The virtual address of second virtual machine in second instruction is revised as the physical address of second virtual machine, to obtain Third instruction.
In one possible implementation, first virtual machine further includes the port of virtual ethernet network interface card, described Method further include: first virtual machine obtains the mark of the preset virtual RDMA network interface card and the mark of the first port Know;First virtual machine is according to pre- in the mark, the mark of the first port and the host of the virtual RDMA network interface card The second mapping table first stored, determines the mark of corresponding second port, and second mapping table includes virtual RDMA network interface card Mark, virtual RDMA network interface card port mark and virtual ethernet network interface card port mark three's corresponding relationship, it is described Second port is the port of the virtual ethernet network interface card of first virtual machine;First virtual machine is according to preset The corresponding virtual address of the mark of GID generating algorithm and the second port, determines the first GID of the first port;It is described The address of the first port is set the first GID by first virtual machine.
In one possible implementation, first virtual machine is according to the mark of the virtual RDMA network interface card, described Pre-stored second mapping table, determines the mark of corresponding second port in the mark of first port and the host, comprising: First virtual machine sends inquiry request information to the host, carries the virtual RDMA network interface card in the inquiry request Mark and the first port mark;The host is according to the inquiry request information and pre-stored second mapping Table determines the mark of corresponding second port;The host sends inquiry response information, the inquiry to first virtual machine The mark of the second port is carried in response message.
In one possible implementation, the method also includes: when first virtual machine detected carrying When stating the ethernet address modification information of the mark of second port, according to the mark of the second port and pre-stored third Mapping table, determines the mark of corresponding virtual RDMA network interface card and the mark of the first port, and the third mapping table includes void The three couple of the mark of the port of the mark of port of quasi- ethernet nic, the mark of virtual RDMA network interface card and virtual RDMA network interface card It should be related to;First virtual machine is according to the event type carried in the ethernet address modification information, to the first end The address of mouth is updated.
In one possible implementation, first virtual machine is carried according in the ethernet address modification information Event type, the address of the first port is updated, comprising: when the event type be network interface card start when, it is described First virtual machine obtains the second port and identifies corresponding virtual address;First virtual machine according to the virtual address and Preset GID generating algorithm determines the 2nd GID of the first port, and sets second for the address of the first port GID。
In one possible implementation, first virtual machine is carried according in the ethernet address modification information Event type, the address of the first port is updated, comprising: when the event type be network interface card close when, it is described First virtual machine deletes the address of the first port.
Second aspect, provides a kind of equipment, and the equipment includes the first virtual machine, host and host channel adapter HCA card, first virtual machine include virtual ethernet network interface card and virtual RDMA network interface card, and the virtual ethernet network interface card includes void Quasi- ethernet port, the virtual RDMA network interface card includes the port of virtual RDMA port virtual RDMA network interface card;Wherein: described first Virtual machine, for according to the virtual address of first virtual machine, determining when needing to establish data connection with the second virtual machine The global identity GID of the corresponding RDMA network of first virtual machine;First virtual machine, is also used to by described The corresponding first port of GID, Xiang Suoshu host send QP attribute setting information, and the first port is first virtual machine The port of the virtual RDMA network interface card, the QP attribute setting information include the first instruction and the second instruction, first instruction It is used to indicate the virtual address that the host sets the source address in QP queue to first virtual machine, second instruction It is used to indicate the virtual address that the host sets the destination address in the QP queue to second virtual machine, the QP Queue is in the corresponding QP queue of the HCA card, and first virtual machine communicates used QP team with second virtual machine Column;Second instruction modification is third instruction for deleting first instruction by the host, and the third instruction is used The destination address in QP queue is set in the instruction host physical address of second virtual machine;The master Machine, for the corresponding QP queue of first virtual machine to be arranged in the HCA card according to modified QP attribute setting information Attribute so that the HCA card based on the QP queue after setting to second virtual machine send data.
The embodiment of the present invention, when host receives the QP attribute setting information of the first virtual machine transmission, QP attribute setting Information includes the first instruction and the second instruction, and the first instruction is used to indicate the host and sets for the source address in QP queue The virtual address of one virtual machine, the second instruction are used to indicate host and set the second virtual machine for the destination address in QP queue Virtual address, host delete the first instruction, are third instruction by the second instruction modification, and third instruction is used to indicate host for QP Destination address in queue is set as the physical address of the second virtual machine;Then it is arranged according to modified QP attribute setting information The attribute of QP queue, so that HCA card sends data to the second virtual machine based on the QP queue after setting.Based on the present invention To realize that the virtual network of virtual machine is isolated with physical network, guarantee the safety of virtual machine, and RoCE agreement can be compatible with.
In one possible implementation, the host, is specifically used for: the first mapping table according to the pre-stored data and institute The virtual address for stating second virtual machine in the second instruction, determines the physical address of second virtual machine, described first Mapping table includes the corresponding relationship of virtual address and physical address;By the virtual of second virtual machine in second instruction The physical address of second virtual machine is revised as in address, to obtain third instruction.
In one possible implementation, first virtual machine further includes the port of virtual ethernet network interface card, described First virtual machine, is also used to: obtaining the mark of the preset virtual RDMA network interface card and the mark of the first port;According to institute It states the mark of virtual RDMA network interface card, pre-stored second mapping table in the mark of the first port and the host, determines The mark of corresponding second port, second mapping table include the port of the mark of virtual RDMA network interface card, virtual RDMA network interface card Mark and virtual ethernet network interface card port mark three's corresponding relationship, the second port be first virtual machine The virtual ethernet network interface card port;According to the corresponding void of mark of preset GID generating algorithm and the second port Quasi- address, determines the first GID of the first port;The first GID is set by the address of the first port.
In one possible implementation, first virtual machine, is specifically used for: Xiang Suoshu host sends inquiry request Information carries the mark of the virtual RDMA network interface card and the mark of the first port in the inquiry request;The host, It is specifically used for: according to the inquiry request information and pre-stored second mapping table, determines the mark of corresponding second port; Inquiry response information is sent to first virtual machine, the mark of the second port is carried in the inquiry response information.
In one possible implementation, first virtual machine, is also used to: carrying the second end when detecting Mouthful mark ethernet address modification information when, according to the mark of the second port and pre-stored third mapping table, Determine the mark of corresponding virtual RDMA network interface card and the mark of the first port, the third mapping table includes virtual ethernet Three's corresponding relationship of the mark of the port of the mark of the port of network interface card, the mark of virtual RDMA network interface card and virtual RDMA network interface card; According to the event type carried in the ethernet address modification information, the address of the first port is updated.
In one possible implementation, first virtual machine, is specifically used for: when the event type opens for network interface card When dynamic, obtain the second port and identify corresponding virtual address;According to the virtual address and preset GID generating algorithm, It determines the 2nd GID of the first port, and sets the 2nd GID for the address of the first port.
In one possible implementation, first virtual machine, is specifically used for: when the event type is network interface card pass When closing, the address of the first port is deleted.
The third aspect, provides a kind of equipment, and the equipment includes: that the equipment includes: memory, processing circuit and master Machine channel adapter HCA card;Wherein, the memory is used for store code;The processing circuit, for according to the storage Store code runs the first virtual machine and host in device;The processing circuit is also used to through first virtual machine and host Realize method described in first aspect;The HCA card, for sending data to second virtual machine.
Fourth aspect provides a kind of computer readable storage medium, including instruction, when described instruction is transported on computers When row, the computer is made to execute method described in first aspect.
5th aspect, provides a kind of computer program product comprising instruction, when run on a computer, so that Computer executes method described in first aspect.
The embodiment of the present invention, when host receives the QP attribute setting information of the first virtual machine transmission, QP attribute setting Information includes the first instruction and the second instruction, and the first instruction is used to indicate the host and sets for the source address in QP queue The virtual address of one virtual machine, the second instruction are used to indicate host and set the second virtual machine for the destination address in QP queue Virtual address, host delete the first instruction, are third instruction by the second instruction modification, and third instruction is used to indicate host for QP Destination address in queue is set as the physical address of the second virtual machine;Then it is arranged according to modified QP attribute setting information The attribute of QP queue, so that HCA card sends data to the second virtual machine based on the QP queue after setting.Based on the present invention To realize that the virtual network of virtual machine is isolated with physical network, guarantee the safety of virtual machine, and RoCE agreement can be compatible with.
Detailed description of the invention
Fig. 1 is the flow chart of TCP/IP data transmission in network provided in an embodiment of the present invention;
Fig. 2 is the flow chart of RDMA data transmission in network provided in an embodiment of the present invention;
Fig. 3 is the architecture diagram of RoCE agreement provided in an embodiment of the present invention;
Fig. 4 is a kind of structural schematic diagram of equipment provided in an embodiment of the present invention;
Fig. 5 is a kind of structural schematic diagram of equipment provided in an embodiment of the present invention;
Fig. 6 is a kind of port of virtual ethernet network interface card provided in an embodiment of the present invention and the port of virtual RDMA network interface card The flow chart of correlating method;
Fig. 7 is the affiliated party of a kind of port of virtual ethernet network interface card provided by the invention and the port of virtual RDMA network interface card The exemplary flow chart of method;
Fig. 8 is a kind of flow chart of QP attribute setting information management method provided in an embodiment of the present invention;
Fig. 9 is a kind of structural schematic diagram of equipment provided in an embodiment of the present invention.
Specific embodiment
Fig. 1 is the flow chart of TCP/IP data transmission in network provided in an embodiment of the present invention, as shown in Figure 1, in tradition TCP/IP network in, firstly, first server needs to copy data to ICP/IP protocol stack from the buffer area of application program Buffer area;Then, data are packaged into data packet according to ICP/IP protocol, and data packet is copied to the buffering of trawl performance Area;Finally, data packet is transmitted to second server by network interface card.Correspondingly, in traditional TCP/IP network, firstly, the Data packet in the buffer area of trawl performance is copied to the buffer area of ICP/IP protocol stack by two servers, then, by data packet It is decapsulated according to ICP/IP protocol, obtains data, finally, copying data to application from the buffer area of ICP/IP protocol stack The buffer area of program.
In this way, problem of both will lead to, on the one hand, the number of data copy is more, causes communication delay larger.Separately On the one hand, protocol stack is packaged reconciliation encapsulation process to data and is required to central processing unit (Central Processing Unit, CPU) it is operated, cause the load of CPU higher.For example, inquiring can produce each time in search for application Several inquiry requests to back-end server are given birth to, and these inquiry requests have deadline requirement.So only cutting The inquiry response of search for application is only returned to before the time, result can just be considered, and be more than the inquiry of deadline Response can be searched application program discarding.Therefore, when the delay in TCP/IP network is larger, the inquiry response that is dropped Ratio will dramatically increase.On the one hand, it will lead to the search quality decline of search for application;On the other hand, a large amount of invalid Inquiry response transmits in TCP/IP network and further increases network delay.In another example cpu load is higher will to will lead to server The computing capability that can be provided reduces.In cloud computing, CPU is kernel scheduling resource, and cpu load is higher, and will result directly in can be with The decline of tenant's quantity of service, to reduce the profitability of total system.
In order to solve to postpone in TCP/IP network larger, the higher problem of cpu load, rdma protocol is applied and is given birth to, RDMA Agreement has relative to ICP/IP protocol and postpones the low and small advantage of cpu load.Fig. 2 is RDMA provided in an embodiment of the present invention The flow chart of data transmission in network, as shown in Fig. 2, in RDMA network, firstly, application program before carrying out data transmission Registers memory in the operating system of server, and the address of registers memory is sent to the HCA card of server, then, server HCA card can be with the registers memory of direct read/write application program.In this way, data send and receive no longer need buffer area it Between repeatedly copied, delay can substantially reduce.Meanwhile HCA card also achieves the function of rdma protocol stack, i.e., by HCA card Encapsulation process is conciliate to complete the encapsulation of data packet.In this way, release cpu resource, the load of CPU is reduced, server can be with There are more cpu resources to distribute to tenant's use.
Rdma protocol is realized based on infinite bandwidth (Infiniband, IB) network architecture.However, in cloud computing scene Under, ICP/IP protocol is still most important network protocol.Therefore, in order to be compatible with ICP/IP protocol, industry proposes RoCE association View.Currently, RoCE technology mainly includes two versions RoCEv1 and RoCEv2.Fig. 3 is RoCE provided in an embodiment of the present invention association The architecture diagram of view, as shown in figure 3, the link layer association of IB agreement is substituted in the link layer protocol of ICP/IP protocol in RoCEv1 version View.In RoCEv2 version, User Datagram Protocol (User in the network layer protocol and transport layer protocol of ICP/IP protocol Datagram Protocol, UDP) the network layer protocol that IB agreement is substituted.Specifically, the network layer of ICP/IP protocol is assisted View is used for route learning and transmission, and the udp protocol of ICP/IP protocol then can be used for equivalent route (Equal- CostMultipathRouting, ECMP).Currently, generalling use RoCEv2 version in system for cloud computing.
In the prior art, it when the virtual machine in different server needs to carry out data communication, needs to open on HCA card Hardware virtualization function, such as SR-IOV.In this way, virtual machine can be established according to the physical IP address of administrator configurations and other side Connection, established on corresponding RDMA interface queue to (Queue Pairs, QP) and be arranged the QP queue source address be should be from The physical IP address of body, destination address are the physical IP address of other side.During subsequent data transmission, HCA card is directly read Data in the corresponding registers memory of application program are packaged the data according to source address in QP queue and destination address, Data message is obtained, and the data message is transmitted by physical network.
In conventional TCP/IP agreement, administrator guarantees to realize that the virtual network of virtual machine is isolated with physical network The safety of virtual machine, it will usually which virtual ip address is set in virtual machine.However, being based on the prior art, virtual machine setting is virtual After IP address, when establishing connection with opposite end virtual machine, which can set the source address in QP queue to the virtual machine The virtual ip address of the virtual machine, destination address are set as the virtual ip address of opposite end virtual machine, and subsequent HCA card is according to the QP team When column are packaged the data of the virtual machine, the source address of the data message is the virtual ip address of the virtual machine, destination Location is the virtual ip address of opposite end virtual machine.Due to there is no the related road of virtual ip address in the network equipment of physical network By list item, so can not be transmitted to the data message.
The method that data that the embodiment of the invention provides a kind of in cloud computing scene are sent, this method are applied to equipment. In this scenario, as shown in figure 4, virtual machine 410, host 420 can be set in the equipment and supported RoCE or other classes Like the HCA card 430 of agreement, wherein virtual machine 410 is connected with host 420 and HCA card 430, and host 420 is connected with HCA card 430, Host 420 can be with management function virtual machine or monitor of virtual machine (Virtual Machine Monitor, VMM) Or a variety of implementations such as Hypervisor, the present invention is without limitation.Specifically, as shown in figure 5, virtual machine 410 includes virtual Ethernet nic 411, virtual RDMA network interface card 412, virtual ethernet management module 413 and virtual RDMA management module 414.The master Machine 420 includes virtual RDMA control module 421, can also include virtual ethernet control module 422 and controller 423.
Wherein, HCA card 430 includes multiple physical ports, and administrator can virtualize HCA card 430, for HCA The physical port can virtually be turned to one or more groups of virtual RoCE network interface cards by each physical port on card 430, administrator, often The virtual RoCE network interface card of group may include a virtual ethernet network interface card 411 and a virtual RDMA network interface card 412.The present invention is with one Port virtual is introduced for turning to one group of virtual RoCE network interface card, other situations are similar therewith.Wherein, virtual ethernet network interface card 411 include the port of multiple virtual ethernet network interface cards, and virtual RDMA network interface card 412 includes the port of multiple virtual RDMA network interface cards.
Virtual machine 410 is connected with virtual ethernet network interface card 411 and virtual RDMA network interface card 412 respectively, virtual ethernet network interface card 411 are connected by virtual ethernet control module 422 with the port of corresponding HCA card 430, and virtual RDMA network interface card 412 passes through void Quasi- RDMA control module 421 is connected with the port of corresponding HCA card 430, meanwhile, virtual RDMA control module 421 and controller 423 are connected.
Virtual RDMA management module 414, for registering and managing virtual RDMA network interface card 412, virtual ethernet management module 413 for registering and managing virtual ethernet network interface card 411, virtual ethernet control module 422, for sending to virtual machine 410 Ethernet protocol message controlled, virtual RDMA control module 421, for being sent to virtual machine 410 QP attribute setting Information is controlled, controller 423, for store the mark of virtual RDMA network interface card, the port of virtual RDMA network interface card mark and The corresponding relationship of the three of the mark of the port of virtual ethernet network interface card and the virtual address and physical address of opposite end virtual machine Corresponding relationship.The virtual address of virtual machine can be virtual ip address, be also possible to virtual media access control (Media Access Control, MAC) address, similarly, physical address can be physical IP address, it is also possible to physical MAC address, this It is introduced for invention virtual ip address and physical IP address, other situations are similar therewith.
In the prior art, from the perspective of operating system, a HCA network interface card is a HCA equipment, a HCA equipment Including an ethernet device and a RDMA equipment.For a HCA equipment, HCA driver is first in an operating system Create an infrastructure device.And an ethernet device and a RDMA equipment are taken out in the infrastructure device.It is logical in RDMA During letter, IP address is provided on ethernet device, for establishing connection, RDMA equipment is used for the read-write operation of data.By Only specify IP address when carrying out RDMA operation in application program, thus operating system must provide it is a kind of true by IP address Determine the method for RDMA equipment.
In HCA card, since ethernet device and the equal carry of RDMA equipment are under identical infrastructure device, so HCA drives It is dynamic it is corresponding to be automatically determined out according to the IP address and pre-stored GID generating algorithm of ethernet device for RDMA equipment GID, and the GID is set by the address of RDMA equipment.Meanwhile when the IP address of ethernet device changes, HCA driving The GID of RDMA equipment can be automatically updated according to the IP address of ethernet device.In this way, application program can be according to itself IP Address and pre-stored GID generating algorithm, determine corresponding RDMA equipment, to realize the transmission of data.However, for void Quasi- ethernet nic and virtual RDMA network interface card, the two are independent equipment in an operating system, without any association.Therefore, For ease of description, a kind of port for providing virtual ethernet network interface card corresponding to virtual machine first of the embodiment of the present invention and void The correlating method of the port of quasi- RDMA network interface card is introduced, and as shown in Figure 5, Figure 6, specific steps can be such that
Step 601, the first virtual machine 410 obtains the mark of preset virtual RDMA network interface card 412 and the mark of first port.
In an implementation, the mark of virtual RDMA network interface card 412 can be stored in advance in the virtual RDMA of virtual machine by administrator In management module 414, it is managed with will pass through virtual RDMA management module 414 to virtual RDMA network interface card 412.As virtual RDMA After management module 414 detects that certain virtual RDMA network interface card 412 starts, virtual RDMA management module 414 is available to be stored in advance The virtual RDMA network interface card 412 mark, then to RDMA trawl performance module (such as ib-core module) send registration report Text carries the mark of the virtual RDMA network interface card 412 in the logon message, to register to the virtual RDMA network interface card 412. Meanwhile link channel type can also be carried in the logon message, which can be RDMA, RDMA trawl performance mould After block receives the logon message, the pipe of the GID of the port to virtual RDMA network interface card each on the virtual RDMA network interface card 412 can be stopped Reason.For example, virtual RDMA management module 414 calls ib_register_device () interface, it should to ib_core Module registers Virtual RDMA network interface card 412.Wherein, virtual RDMA management module 414 is by the parameter in ib_register_device () interface The mode that link layer type is appointed as " IB " stops ib_core to the first port of the virtual RDMA network interface card 412 The management of gib_cache.
Virtual RDMA management module 414 is after the registration for completing the virtual RDMA network interface card 412, virtual RDMA management module 414 can be scanned in the bus, inquire the mark of the first port on the virtual RDMA network interface card 412, wherein first port Mark can be first port port numbers.
Step 602, the first virtual machine 410 is according to the mark, the mark of first port and host of virtual RDMA network interface card 412 Pre-stored second mapping table, determines the mark of corresponding second port in 420.
Wherein, the second mapping table include the mark of virtual RDMA network interface card 412, virtual RDMA network interface card port mark and void Three's corresponding relationship of the mark of the port of quasi- ethernet nic, the second port are the described virtual of first virtual machine The port of ethernet nic.
In an implementation, virtual RDMA control module 421 can be set in host 420, the virtual RDMA control module 421 In can be previously stored with the second mapping table.Second mapping table can be preset by administrator, when administrator is by the object of HCA When reason port virtual turns to virtual ethernet network interface card 411 and virtual RDMA network interface card 412, administrator can be by the virtual RDMA network interface card The mark of the port of each virtual RDMA network interface card and corresponding each virtual ether in 412 mark, the virtual RDMA network interface card 412 Three's corresponding relationship of the mark of the port of net network interface card is stored in second mapping table.Wherein, the virtual ethernet network interface card The mark of port can be the media access control (Media Access Control, MAC) of the port of virtual ethernet network interface card Address.
Virtual RDMA management module 414, can be with after obtaining the mark of mark and first port of virtual RDMA network interface card 412 The mark of the corresponding second port of the first port is inquired by virtual RDMA control module 421.Specific steps can be such that
Step 1, the first virtual machine 410 send inquiry request information to host 420.
Wherein, the mark of virtual RDMA network interface card 412 and the mark of first port can be carried in inquiry request.
Step 2, host 420 determine corresponding second end according to inquiry request information and pre-stored second mapping table The mark of mouth.
In an implementation, after virtual RDMA control module 421 receives the inquiry request information, to the inquiry request information into Row parsing, obtains the mark of the virtual RDMA network interface card 412 carried in the inquiry request information and the mark of first port, then root According to the mark of virtual RDMA network interface card 412 and the mark of first port, the mark of corresponding second port is inquired in the second mapping table Know (i.e. the mark of the corresponding second port of first port).
Step 3, host 420 send inquiry response information to the first virtual machine 410.
In an implementation, virtual RDMA control module 421 is after the mark for inquiring the corresponding second port of the first port, Inquiry response information can will be sent to virtual RDMA management module 414, carry the second port in the inquiry response information Mark.
It should be noted that in order to improve search efficiency, administrator can only be cached in virtual RDMA control module 421 Complete second mapping table is stored in the controller 423 of host 420 by the list item for the second mapping table inquired.It is empty Quasi- RDMA control module 421 receive inquiry request information that virtual RDMA management module 414 sends (for the ease of distinguishing, with It is known as the first inquiry request information down), the mark of virtual RDMA network interface card 412 and the mark of first port are obtained, then first at this Ground inquiry whether there is the mark of the corresponding second port of the first port.If it is, to the virtual RDMA management module 414 Send inquiry response information, carry in the inquiry response information mark of the second port (for the ease of distinguishing, hereinafter referred to as For the first inquiry response information);Otherwise, virtual RDMA control module 421 to controller 423 send inquiry request information (in order to Convenient for distinguishing, hereinafter referred to as the second inquiry request information).Controller 423 receives what the virtual RDMA control module 421 was sent Second inquiry request information obtains the mark of virtual RDMA network interface card 412 and the mark of first port, then inquires pre-stored The mark that whether there is the corresponding second port of the first port in second mapping table, if it is, will be controlled to virtual RDMA Module 421 sends inquiry response information, carry in the inquiry response information second port mark (for the ease of distinguishing, Hereinafter referred to as the second inquiry response information);Otherwise, query failure message is returned to virtual RDMA control module 421.When virtual When RDMA control module 421 receives the second inquiry response information, the mark of the second port is obtained, and manage to virtual RDMA It manages module 414 and sends the first inquiry response information.Meanwhile virtual RDMA management module 414 is by the mark of the virtual RDMA network interface card 412 Three's corresponding relationship of the mark of knowledge, the mark of the first port and second port is stored in local, provides quickly so as to subsequent Inquiry.When virtual RDMA control module 421 receives query failure message, returns and inquire to virtual RDMA management module 414 Failure information can pass through display or short message etc. after the virtual RDMA management module 414 receives the query failure message Mode notifies administrator, so that administrator in time modifies to mistake.
Step 603, the first virtual machine 410 is corresponding virtual according to the mark of preset GID generating algorithm and second port IP address determines the first GID of first port.
In an implementation, GID generating algorithm is previously stored in virtual RDMA management module 414.The GID generating algorithm can be with Using the GID conversion method of professional standard.For example, the address size of GID is 128, if being GID by IPv4 address conversion, Since the address size of IPv4 is 32, can be using the address IPv4 as low 32 of GID, high 96 of GID all mend 0;If being GID by IPv6 address conversion, since the address size of IPv6 is 128, directly the address IPv6 can be made For GID.
After virtual RDMA management module 414 receives inquiry response information, which is parsed, is obtained The mark of the corresponding second port of the first port.Virtual RDMA management module 414 can exist according to the mark of the second port It is scanned, is inquired on the port of corresponding virtual ethernet network interface card and the port of the virtual ethernet network interface card in advance in bus The virtual ip address of configuration, and it is based on the virtual ip address and pre-stored GID generating algorithm, it is corresponding to obtain the first port The first GID.
Step 604, the address of first port is set the first GID by the first virtual machine 410.
It in an implementation, can be by the first end after the first GID of the determining first port of virtual RDMA management module 414 The address of mouth is set as the first GID.
In the embodiment of the present invention, when administrator modifies the virtual ip address of second port, virtual RDMA management module 414 Can also the GID to the first port configured accordingly, specific steps can be such that
Step 1, when the first virtual machine 410 detects the ethernet address modification information for the mark for carrying second port When, according to the mark of second port and pre-stored third mapping table, determine corresponding virtual RDMA network interface card 412 mark and The mark of first port.
Wherein, third mapping table includes the mark of the mark of the port of virtual ethernet network interface card 411, virtual RDMA network interface card 412 Know three's corresponding relationship with the mark of the port of virtual RDMA network interface card 412.
In an implementation, third mapping table has been can store in virtual RDMA management module 414, the third mapping table is by virtual RDMA management module 414 is configured.Virtual RDMA management module 414 receives looking into for the virtual transmission of RDMA control module 421 Response message is ask, which is parsed, the mark of second port is obtained, the virtual RDMA management module 414 It can be by three's corresponding relationship of the mark of the mark of the virtual RDMA network interface card 412, the mark of the first port and second port It is stored in third mapping table.
In order to which the IP address change to second port is monitored in real time, virtual RDMA management module 414 can be to void The operating system of quasi- machine registers ethernet address altering event, so that administrator is modifying the virtual ip address in the second port When, virtual RDMA management module 414 can the GID to the first port configured accordingly.For example, virtual RDMA manages mould Block 414 calls register_inetaddr_notifier () interface to register in an operating system and changes thing to ethernet address The call back function of part.
When the operating system of virtual machine detects the management instruction of certain corresponding virtual ip address, the ether entoilage can be triggered Location altering event, wherein management instruction can be deletion instruction or configuration-direct etc..
When operating system detects that administrator deletes the deletion instruction of the virtual ip address of second port, operating system can To determine that the type of the ethernet address altering event is closed as network interface card.When operating system detects the virtual ether of administrator configurations When the configuration-direct of the virtual ip address of the port of net network interface card, operating system can determine the class of the ethernet address altering event Type is network interface card starting.Operating system can send ethernet address altering event information, ether to virtual RDMA management module 414 The mark of the type for the ethernet address altering event determined is carried in net address altering event information.
Operating system can send ethernet address modification information to virtual RDMA management module 414, which becomes The mark of the port of the virtual ethernet network interface card and the type of ethernet address altering event are carried in more information.
Virtual RDMA management module 414 receives ethernet address modification information, to the ethernet address modification information into Row parsing, obtains the mark of the port of the virtual ethernet network interface card, and then inquiry is virtual with the presence or absence of this in third mapping table The mark of the port of the mark and virtual RDMA network interface card of the corresponding virtual RDMA network interface card 412 of mark of the port of ethernet nic. If it is, otherwise executing step 2 does not execute other processing.
Step 2, the first virtual machine 410 is according to the event type carried in ethernet address modification information, to first port GID configured.
In an implementation, virtual RDMA management module 414 obtains the event type carried in the ethernet address modification information Afterwards, it can be updated accordingly according to GID of the event type to first port.
Optionally, virtual RDMA management module 414 is according to the type of ethernet address altering event, to first port GID carries out configuration update and is specifically divided into the following two kinds situation treatment process are as follows:
The first situation: when event type is that network interface card is closed, the first virtual machine 410 deletes the address of first port.
Second situation: when event type is that network interface card starts, it is corresponding that the first virtual machine 410 obtains second port mark Virtual ip address, the first virtual machine 410 determine the second of first port according to virtual ip address and preset GID generating algorithm GID, and the 2nd GID is set by first port address.
In an implementation, after virtual RDMA management module 414 receives ethernet address altering event information, event class is obtained The mark of type deletes the corresponding first end in port of the virtual ethernet network interface card if the mark is used to indicate network interface card closing The address of mouth.
If the mark is used to indicate network interface card starting, virtual RDMA management module 414 can be according to the mark of the second port Know in the bus, inquire the virtual ip address configured in the second port, then according to the virtual ip address and pre-stored GID generating algorithm determines the 2nd GID of the corresponding first port in the port of the virtual ethernet network interface card.Virtual RDMA management After module 414 determines the 2nd GID of the first port, the 2nd GID can be set by the address of the first port.
The present invention also provides the correlating methods of a kind of port of virtual ethernet network interface card and the port of virtual RDMA network interface card Example, as shown in Figure 5, Figure 7, concrete processing procedure is as follows:
Step 701, after virtual RDMA network interface card 412 starts, virtual RDMA management module 414 is to RDMA trawl performance module Logon message is sent, which is registered, the virtual RDMA network interface card 412 is carried in the logon message Mark.
Step 702, RDMA trawl performance module registers the virtual RDMA network interface card 412, and stops virtual to this The management of the GID of the port of each virtual RDMA network interface card on RDMA network interface card 412.
Step 703, virtual RDMA management module 414 obtains the mark of the virtual RDMA network interface card 412 and the mark of first port Know.
Step 704, virtual RDMA management module 414 sends inquiry request information to virtual RDMA control module 421, this is looked into It askes in solicited message and carries the mark of virtual RDMA network interface card 412 and the mark of first port.
Step 705, virtual RDMA control module 421 returns to inquiry response information to virtual RDMA management module 414, this is looked into Ask the mark that the second port is carried in response message.
Step 706, virtual RDMA management module 414 inquires whether second port is configured with virtual ip address, if it is, Step 707-708 is executed, otherwise, executes step 709.
Step 707, virtual RDMA management module 414 is corresponding according to the mark of preset GID generating algorithm and second port Virtual ip address, determine the first GID of first port.
Step 708, the address of first port is set the first GID by virtual RDMA management module 414.
Step 709, virtual RDMA management module 414 registers ethernet address altering event to operating system.
Step 710, when virtual RDMA management module 414 detects ethernet address modification information, according to second port Mark and pre-stored third mapping table, determine the mark of corresponding virtual RDMA network interface card 412 and the mark of first port.
Step 711, virtual RDMA management module 414 obtains the event type carried in the ethernet address modification information, If the event type is network interface card starting, 712-713 is thened follow the steps, if the event type is network interface card closing, executes step Rapid 714.
Step 712, virtual RDMA management module 414 obtains second port and identifies corresponding virtual ip address, according to virtual IP address and preset GID generating algorithm, determine the 2nd GID of first port.
Step 713, first port address is set the 2nd GID by virtual RDMA management module 414.
Step 714, virtual RDMA management module 414 deletes the address of first port.
The concrete processing procedure of step 701 to step 714 can refer to step 601 to step 604, and the present invention repeats no more.
In the embodiment of the present invention, when virtual RDMA management module 414 determines the first port of virtual RDMA network interface card 412 After GID, the QP attribute setting information that virtual RDMA control module 421 can send virtual machine is managed, such as Fig. 5, Fig. 8 institute Show, step can be such that
Step 801, the first virtual machine 410 determines that the first virtual machine 410 is right according to the virtual address of the first virtual machine 410 The global identity GID for the RDMA network answered.
In an implementation, when certain application program in the first virtual machine 410 needs and certain application program in the second virtual machine When carrying out data transmission, the application program in the first virtual machine 410 according to local virtual ip address and can be stored in advance GID generating algorithm, determine the GID of corresponding first port.
Step 802, the first virtual machine 410 sends QP attribute setting letter to host 420 by the corresponding first port of GID Breath, the first port are the port of the virtual RDMA network interface card of first virtual machine, and QP attribute setting information includes the One instruction and the second instruction, the first instruction are used to indicate host 420 and set the first virtual machine 410 for the source address in QP queue Virtual address, the second instruction is used to indicate host 420 and sets the virtual of the second virtual machine for the destination address in QP queue Address, the QP queue are in the corresponding QP queue of HCA card, and the first virtual machine communicates used QP queue with the second virtual machine.
In an implementation, the application program in the first virtual machine 410 can be by the corresponding first port of the GID, to void Quasi- RDMA control module 421 sends QP attribute setting information.Wherein, QP attribute setting information includes that the first instruction and second refer to It enables, wherein the first instruction can serve to indicate that the virtual ip address for setting source address to the first virtual machine 410, the second instruction It can serve to indicate that the virtual ip address for setting destination address to the second virtual machine.The virtual ip address of second virtual machine can To be arranged in the application program of the first virtual machine 410 in advance by administrator.For example, the first virtual machine 410 passes through ib_ Modify_qp (qp_attr) interface sends QP attribute setting information to host 420, and it is first that wherein qp_attr, which specifies source address, The virtual ip address of virtual machine 410, destination address are the virtual ip address of the second virtual machine.
Step 803, host 420 deletes the first instruction, is third instruction by the second instruction modification, and third instruction is used to indicate Host 420 sets the destination address in QP queue to the physical address of the second virtual machine.
In an implementation, if in the attribute of QP queue, source address is the virtual ip address of the first virtual machine 410, destination Location is the virtual ip address of the second virtual machine, then the application program of subsequent 430 pair of first virtual machine 410 of HCA card is virtual to second When the data that the application program of machine is sent carry out protocol encapsulation, using the virtual ip address of the first virtual machine 410 as source address, with The virtual ip address of second virtual machine is purpose address, then passes through the physics between the first virtual machine 410 and the second virtual machine Network is transmitted.Due to there is no the first virtual machine 410 and the corresponding virtual ip address of the second virtual machine in the physical network Routing entry, the only routing entry of the first virtual machine 410 and the corresponding physical IP address of the second virtual machine, therefore, virtually RDMA control module 421 needs to delete the first instruction after receiving QP attribute setting information, meanwhile, it will be in the second instruction The corresponding virtual ip address of second virtual machine is revised as the corresponding physical IP address of the second virtual machine, obtains third instruction.This Three instructions are used to indicate the physical IP address for setting destination address to the second virtual machine.
It optionally, is third instruction by the second instruction modification, step can be such that
Step 1, first mapping table according to the pre-stored data of host 420 and second instruction in the second virtual machine it is virtual IP address determines the physical IP address of the second virtual machine.
Wherein, the first mapping table includes the corresponding relationship of virtual ip address and physical IP address.
In an implementation, it can be previously stored with the first mapping table in virtual RDMA control module 421, which can To be preset by administrator.
After virtual RDMA control module 421 receives QP attribute setting information, which is parsed, The virtual ip address of the second virtual machine in second instruction is obtained, second virtual machine is then inquired in the first mapping table The corresponding physical IP address of virtual ip address.
It should be noted that in order to improve search efficiency, administrator can only be cached in virtual RDMA control module 421 Complete first mapping table is stored in the controller 423 of host 420 by the list item for the first mapping table inquired.It is empty Quasi- RDMA control module 421 receives QP attribute setting information, with obtaining the virtual IP address of the second virtual machine in second instruction Then location first whether there is the corresponding physical IP address of virtual ip address of second virtual machine in local search.If it is, Execute step 2;Otherwise, virtual RDMA control module 421 sends address lookup solicited message, the address lookup to controller 423 The virtual ip address of second virtual machine is carried in solicited message.Controller 423 receives virtual RDMA control module 421 and sends out The address lookup solicited message sent, obtains the virtual ip address of second virtual machine, and then controller 423 is inquired pre-stored The corresponding physical IP address of virtual ip address that whether there is second virtual machine in first mapping table, if it is, to virtual RDMA control module 421 sends address lookup response message, carries the physical IP address in the address lookup response message;It is no Then, to virtual 421 return address query failure message of RDMA control module.When virtual RDMA control module 421 receives the ground When the inquiry response information of location, the physical IP address is obtained, then executes step 2, while by the virtual IP address of second virtual machine The corresponding relationship of location and the physical IP address is stored in local, so as to subsequent offer quick search.When virtual RDMA control module 421 when receiving the address lookup failure information, administrator can be notified by modes such as display or short messages, to manage Member in time modifies to mistake.
The virtual ip address of the second virtual machine in second instruction is revised as the second virtual machine by step 2, host 420 Physical IP address obtains third instruction.
In an implementation, virtual RDMA control module 421 determines the corresponding physical IP of virtual ip address of second virtual machine Behind address, the virtual ip address of the second virtual machine in the second instruction can be deleted, then by the purpose in second instruction Address is set as the physical IP address of the second virtual machine, obtains third instruction.
Step 804, the attribute of QP queue is arranged according to modified QP attribute setting information for host 420, so that HCA card 430 send data to the second virtual machine based on the QP queue after setting.
In an implementation, virtual RDMA control module 421 is according to modified QP attribute setting information, by the mesh in QP queue Address be set as the physical IP address of the second virtual machine.Meanwhile virtual RDMA control module 421 is scanned in the bus, The physical IP address configured on corresponding Ethernet interface is inquired, then sets source address to the physical IP of the Ethernet interface Address.It is subsequent, it, should when application program application program transmission data into the second virtual machine in the first virtual machine 410 RDMA interface can read data from the corresponding memory of application program of first virtual machine 410, with the object of Ethernet interface Managing IP address, address is packaged the data as a purpose as source address, the physical IP address of the second virtual machine, is counted According to message, the data message is then sent to second by the physical network between the first virtual machine 410 and the second virtual machine On the port of the corresponding HCA card 430 of virtual machine.
The embodiment of the present invention, when host 420 receives the QP attribute setting information of the first virtual machine 410 transmission, which belongs to Property setting information include the first instruction and the second instruction, the first instruction is used to indicate host 420 and sets the source address in QP queue It is set to the virtual address of the first virtual machine 410, the second instruction is used to indicate host 420 and sets the destination address in QP queue to The virtual address of second virtual machine, host 420 delete the first instruction, are third instruction, third instruction by the second instruction modification It is used to indicate the physical address that host 420 sets the destination address in QP queue to the second virtual machine;Then according to modification after QP attribute setting information setting QP queue attribute so that HCA card 430 is sent out based on the QP queue after setting to the second virtual machine Send data.Based on the present invention, it can it realizes that the virtual network of virtual machine is isolated with physical network, guarantees the safety of virtual machine, It can be compatible with RoCE agreement again.
Based on the same technical idea, the embodiment of the invention also provides a kind of equipment, as shown in figure 5, equipment includes the One virtual machine 410, host 420 and host channel adapter HCA card 430, the first virtual machine 410 include virtual ethernet network interface card 411 and virtual RDMA network interface card 412, virtual ethernet network interface card 411 includes the port of virtual ethernet network interface card, virtual RDMA network interface card 412 include the port of virtual RDMA network interface card;Wherein: the first virtual machine 410 connects for establishing data with the second virtual machine when needs When connecing, according to the virtual internet interconnection protocol address of the first virtual machine 410, the corresponding RDMA net of the first virtual machine 410 is determined The global identity GID of network;First virtual machine 410, is also used to through the corresponding first port of GID, sends QP to host 420 Attribute setting information, QP attribute setting information include the first instruction and the second instruction, and the first instruction is used to indicate host 420 for QP Source address in queue is set as the virtual address of the first virtual machine 410, and the second instruction is used to indicate host 420 will be in QP queue Destination address be set as the virtual address of the second virtual machine;Host 420, for deleting the first instruction, by the second instruction modification For third instruction, third instruction is used to indicate the physics that host 420 sets the destination address in QP queue to the second virtual machine Address;Host 420, for the attribute of QP queue to be arranged according to modified QP attribute setting information, so that HCA card 430 is based on QP queue after setting sends data to the second virtual machine.
In one possible implementation, host 420 are specifically used for: the first mapping table according to the pre-stored data and The virtual address of the second virtual machine in two instructions determines that the physical address of the second virtual machine, the first mapping table include virtually The corresponding relationship of location and physical address;The virtual address of the second virtual machine in second instruction is revised as to the object of the second virtual machine Address is managed, to obtain third instruction.
In one possible implementation, the first virtual machine 410, is also used to: obtaining preset virtual RDMA network interface card 412 Mark and first port mark;According to pre- in the mark, the mark of first port and host 420 of virtual RDMA network interface card 412 The second mapping table first stored, determines the mark of corresponding second port, and the second mapping table includes the mark of virtual RDMA network interface card 412 Know, three's corresponding relationship of the mark of the port of the mark of the port of virtual RDMA network interface card and virtual ethernet network interface card;According to pre- If GID generating algorithm and second port the corresponding virtual address of mark, determine the first GID of first port;
The first GID is set by the address of first port.
In one possible implementation, the first virtual machine 410, is specifically used for: sending inquiry request letter to host 420 It ceases, carries the mark of virtual RDMA network interface card 412 and the mark of first port in inquiry request;Host 420, is specifically used for: root It is investigated that asking solicited message and pre-stored second mapping table, the mark of corresponding second port is determined;To the first virtual machine 410 Inquiry response information is sent, the mark of second port is carried in inquiry response information.
In one possible implementation, the first virtual machine 410, is also used to: carrying second port when detecting When the ethernet address modification information of mark, according to the mark of second port and pre-stored third mapping table, determines and correspond to Virtual RDMA network interface card 412 mark and first port mark, third mapping table includes the port of virtual ethernet network interface card 411 Mark, the mark of virtual RDMA network interface card 412 and virtual RDMA network interface card 412 port mark three's corresponding relationship;According to The event type carried in ethernet address modification information is updated the address of first port.
In one possible implementation, the first virtual machine 410, is specifically used for: when event type is that network interface card starts, It obtains second port and identifies corresponding virtual address;According to virtual address and preset GID generating algorithm, first port is determined 2nd GID, and the 2nd GID is set by the address of first port.
In one possible implementation, the first virtual machine 410, is specifically used for: when event type is that network interface card is closed, Delete the address of first port.
The embodiment of the present invention, when host 420 receives the QP attribute setting information of the first virtual machine 410 transmission, which belongs to Property setting information include the first instruction and the second instruction, the first instruction is used to indicate host 420 and sets the source address in QP queue It is set to the virtual address of the first virtual machine 410, the second instruction is used to indicate host 420 and sets the destination address in QP queue to The virtual address of second virtual machine, host 420 delete the first instruction, are third instruction, third instruction by the second instruction modification It is used to indicate the physical address that host 420 sets the destination address in QP queue to the second virtual machine;Then according to modification after QP attribute setting information setting QP queue attribute so that HCA card 430 is sent out based on the QP queue after setting to the second virtual machine Send data.Based on the present invention, it can it realizes that the virtual network of virtual machine is isolated with physical network, guarantees the safety of virtual machine, It can be compatible with RoCE agreement again.
Based on the same technical idea, the embodiment of the invention also provides a kind of equipment, as shown in figure 9, the equipment 900 It include: memory 910, processing circuit 920 and host channel adapter HCA card 930.The equipment 900 can execute the application note Any of the above-described kind of method carried, and, system architecture involved in above-mentioned any method, explanation of nouns, beneficial effect etc., Relevant paragraph is referred to, details are not described herein again.
Wherein, the memory 910 is used for store code.
The processing circuit 920, for running the first virtual machine and host according to store code in the memory 910.
The HCA card 930, for sending data to second virtual machine.
The processing circuit 920 is also used to realize following methods by first virtual machine and host:
First virtual machine includes the port of virtual long immediate data access RDMA network interface card, which comprises
First virtual machine determines that first virtual machine is corresponding according to the virtual address of first virtual machine The global identity GID of RDMA network;
First virtual machine sets QP attribute by the corresponding first port of the GID, Xiang Suoshu host transmit queue Confidence breath, the first port are the port of the virtual RDMA network interface card of first virtual machine, the QP attribute setting letter Breath includes the first instruction and the second instruction, and first instruction is used to indicate the host and sets the source address in QP queue to The virtual address of first virtual machine, second instruction are used to indicate the host for the destination address in the QP queue It is set as the virtual address of second virtual machine, the QP queue is in the corresponding QP queue of the HCA card, and described first is empty Quasi- machine communicates used QP queue with second virtual machine;
The host deletes first instruction, is third instruction by second instruction modification, and the third instruction is used The destination address in QP queue is set in the instruction host physical address of second virtual machine;
The attribute of the QP queue is arranged according to modified QP attribute setting information for the host, so that the HCA card Data are sent to second virtual machine based on the QP queue after setting.
In one possible implementation, it is described by second instruction modification be third instruct, comprising:
The void of host first mapping table and second virtual machine in second instruction according to the pre-stored data Quasi- address, determines that the physical address of second virtual machine, first mapping table include pair of virtual address and physical address It should be related to;
The virtual address of second virtual machine in second instruction is revised as described second virtually by the host The physical address of machine, to obtain third instruction.
In one possible implementation, first virtual machine further includes the port of virtual ethernet network interface card, described Method further include:
First virtual machine obtains the mark of the preset virtual RDMA network interface card and the mark of the first port;
First virtual machine is according to the mark, the mark of the first port and the host of the virtual RDMA network interface card In pre-stored second mapping table, determine the mark of corresponding second port, second mapping table includes virtual RDMA net The mark of card, virtual RDMA network interface card port mark and virtual ethernet network interface card port mark three's corresponding relationship, The second port is the port of the virtual ethernet network interface card of first virtual machine;
First virtual machine is corresponding virtually according to the mark of preset GID generating algorithm and the second port Location determines the first GID of the first port;
The address of the first port is set the first GID by first virtual machine.
In one possible implementation, first virtual machine is according to the mark of the virtual RDMA network interface card, described Pre-stored second mapping table, determines the mark of corresponding second port in the mark of first port and the host, comprising:
First virtual machine sends inquiry request information to the host, carries in the inquiry request described virtual The mark of RDMA network interface card and the mark of the first port;
The host determines corresponding second port according to the inquiry request information and pre-stored second mapping table Mark;
The host sends inquiry response information to first virtual machine, carries in the inquiry response information described The mark of second port.
In one possible implementation, the method also includes:
When first virtual machine detects the ethernet address modification information for the mark for carrying the second port, According to the mark of the second port and pre-stored third mapping table, mark and the institute of corresponding virtual RDMA network interface card are determined The mark of first port is stated, the third mapping table includes the mark of the port of virtual ethernet network interface card, virtual RDMA network interface card Three's corresponding relationship of the mark of the port of mark and virtual RDMA network interface card;
First virtual machine is according to the event type carried in the ethernet address modification information, to the first end The address of mouth is updated.
In one possible implementation, first virtual machine is carried according in the ethernet address modification information Event type, the address of the first port is updated, comprising:
When the event type is that network interface card starts, first virtual machine obtains the second port and identifies corresponding void Quasi- address;
First virtual machine determines the first port according to the virtual address and preset GID generating algorithm 2nd GID, and the 2nd GID is set by the address of the first port.
In one possible implementation, first virtual machine is carried according in the ethernet address modification information Event type, the address of the first port is updated, comprising:
When the event type is that network interface card is closed, first virtual machine deletes the address of the first port.
In the above-described embodiments, can come wholly or partly by software, hardware, firmware or any combination thereof real It is existing.When implemented in software, it can entirely or partly realize in the form of a computer program product.The computer program Product includes one or more computer instructions.When loading on computers and executing the computer program instructions, all or It partly generates according to process or function described in the embodiment of the present invention.The computer can be general purpose computer, dedicated meter Calculation machine, computer network or other programmable devices.The computer instruction can store in computer readable storage medium In, or from a computer readable storage medium to the transmission of another computer readable storage medium, for example, the computer Instruction can pass through wired (such as coaxial cable, optical fiber, number from a web-site, computer, server or data center User's line (DSL)) or wireless (such as infrared, wireless, microwave etc.) mode to another web-site, computer, server or Data center is transmitted.The computer readable storage medium can be any usable medium that computer can access or It is comprising data storage devices such as one or more usable mediums integrated server, data centers.The usable medium can be with It is magnetic medium, (for example, floppy disk, hard disk, tape), optical medium (for example, DVD) or semiconductor medium (such as solid state hard disk Solid State Disk (SSD)) etc..
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.

Claims (17)

1. a kind of method that data are sent, which is characterized in that the method is applied to equipment, and the equipment includes first virtual Machine, host and host channel adapter HCA card, wherein first virtual machine includes virtual long immediate data access RDMA The port of network interface card, which comprises
First virtual machine determines the corresponding RDMA net of first virtual machine according to the virtual address of first virtual machine The global identity GID of network;
By the corresponding first port of the GID, QP attribute is arranged in Xiang Suoshu host transmit queue believes first virtual machine Breath, the first port are the port of the virtual RDMA network interface card of first virtual machine, the QP attribute setting information packet The first instruction and the second instruction are included, first instruction is used to indicate the host and sets described for the source address in QP queue The virtual address of first virtual machine, second instruction are used to indicate the host and the destination address in the QP queue are arranged For the virtual address of second virtual machine, the QP queue is first virtual machine in the corresponding QP queue of the HCA card QP queue used in being communicated with second virtual machine;
The host deletes first instruction, is third instruction by second instruction modification, and the third instruction is for referring to Show that the host sets the destination address in QP queue to the physical address of second virtual machine;
The attribute of the QP queue is arranged according to modified QP attribute setting information for the host, so that the HCA card is based on The QP queue after setting sends data to second virtual machine.
2. the method according to claim 1, wherein it is described by second instruction modification be third instruct, packet It includes:
Second virtual machine in host first mapping table according to the pre-stored data and second instruction is virtually Location determines that the physical address of second virtual machine, first mapping table include that virtual address and the corresponding of physical address are closed System;
The virtual address of second virtual machine in second instruction is revised as second virtual machine by the host Physical address, to obtain third instruction.
3. the method according to claim 1, wherein first virtual machine further includes virtual ethernet network interface card Port, the method also includes:
First virtual machine obtains the mark of the preset virtual RDMA network interface card and the mark of the first port;
First virtual machine is according to pre- in the mark, the mark of the first port and the host of the virtual RDMA network interface card The second mapping table first stored, determines the mark of corresponding second port, and second mapping table includes virtual RDMA network interface card Mark, virtual RDMA network interface card port mark and virtual ethernet network interface card port mark three's corresponding relationship, it is described Second port is the port of the virtual ethernet network interface card of first virtual machine;
First virtual machine is according to the corresponding virtual address of mark of preset GID generating algorithm and the second port, really First GID of the fixed first port;
The address of the first port is set the first GID by first virtual machine.
4. according to the method described in claim 3, it is characterized in that, first virtual machine is according to the virtual RDMA network interface card Pre-stored second mapping table in mark, the mark of the first port and the host, determines corresponding second port Mark, comprising:
First virtual machine sends inquiry request information to the host, carries the virtual RDMA in the inquiry request The mark of network interface card and the mark of the first port;
The host determines the mark of corresponding second port according to the inquiry request information and pre-stored second mapping table Know;
The host sends inquiry response information to first virtual machine, carries described second in the inquiry response information The mark of port.
5. according to the method described in claim 3, it is characterized in that, the method also includes:
When first virtual machine detects the ethernet address modification information for the mark for carrying the second port, according to The mark of the second port and pre-stored third mapping table determine the mark and described of corresponding virtual RDMA network interface card The mark of Single port, the third mapping table include the mark of the port of virtual ethernet network interface card, the mark of virtual RDMA network interface card With three's corresponding relationship of the mark of the port of virtual RDMA network interface card;
First virtual machine is according to the event type carried in the ethernet address modification information, to the first port Address is updated.
6. according to the method described in claim 5, it is characterized in that, first virtual machine is changed according to the ethernet address The event type carried in information is updated the address of the first port, comprising:
When the event type is that network interface card starts, it is corresponding virtually that first virtual machine obtains the second port mark Location;
First virtual machine determines the second of the first port according to the virtual address and preset GID generating algorithm GID, and the 2nd GID is set by the address of the first port.
7. according to the method described in claim 5, it is characterized in that, first virtual machine is changed according to the ethernet address The event type carried in information is updated the address of the first port, comprising:
When the event type is that network interface card is closed, first virtual machine deletes the address of the first port.
8. a kind of equipment, which is characterized in that the equipment includes the first virtual machine, host and host channel adapter HCA card, institute State the port that the first virtual machine includes virtual long immediate data access RDMA network interface card;Wherein:
First virtual machine determines that first virtual machine is corresponding for the virtual address according to first virtual machine The global identity GID of RDMA network;
First virtual machine is also used to through the corresponding first port of the GID, and Xiang Suoshu host sends QP attribute setting letter Breath, the first port are the port of the virtual RDMA network interface card of first virtual machine, the QP attribute setting information packet The first instruction and the second instruction are included, first instruction is used to indicate the host and sets described for the source address in QP queue The virtual address of first virtual machine, second instruction are used to indicate the host and the destination address in the QP queue are arranged For the virtual address of second virtual machine, the QP queue is first virtual machine in the corresponding QP queue of the HCA card QP queue used in being communicated with second virtual machine;
Second instruction modification is third instruction for deleting first instruction by the host, and the third instruction is used The destination address in QP queue is set in the instruction host physical address of second virtual machine;
The host, for the attribute of the QP queue to be arranged according to modified QP attribute setting information, so that the HCA card Data are sent to second virtual machine based on the QP queue after setting.
9. equipment according to claim 8, which is characterized in that the host is specifically used for:
The virtual address of first mapping table according to the pre-stored data and second virtual machine in second instruction, determines institute The physical address of the second virtual machine is stated, first mapping table includes the corresponding relationship of virtual address and physical address;
The virtual address of second virtual machine in second instruction is revised as to the physical address of second virtual machine, To obtain third instruction.
10. equipment according to claim 8, which is characterized in that first virtual machine further includes virtual ethernet network interface card Port, first virtual machine is also used to:
Obtain the mark of the preset virtual RDMA network interface card and the mark of the first port;
It is reflected according in the mark, the mark of the first port and the host of the virtual RDMA network interface card pre-stored second Firing table, determines the mark of corresponding second port, and second mapping table includes the mark of virtual RDMA network interface card, virtual RDMA net Three's corresponding relationship of the mark of the port of the mark and virtual ethernet network interface card of the port of card, the second port are described the The port of the virtual ethernet network interface card of one virtual machine;
According to the corresponding virtual address of mark of preset GID generating algorithm and the second port, the first port is determined The first GID;
The first GID is set by the address of the first port.
11. equipment according to claim 10, which is characterized in that first virtual machine is specifically used for:
Inquiry request information is sent to the host, mark and the institute of the virtual RDMA network interface card are carried in the inquiry request State the mark of first port;
The host, is specifically used for: according to the inquiry request information and pre-stored second mapping table, determining corresponding The mark of Two-port netwerk;Inquiry response information is sent to first virtual machine, described the is carried in the inquiry response information The mark of Two-port netwerk.
12. equipment according to claim 10, which is characterized in that first virtual machine is also used to:
When detecting the ethernet address modification information for the mark for carrying the second port, according to the second port Mark and pre-stored third mapping table determine the mark of corresponding virtual RDMA network interface card and the mark of the first port, The third mapping table includes the mark of the port of virtual ethernet network interface card, the mark of virtual RDMA network interface card and virtual RDMA network interface card Port mark three's corresponding relationship;
According to the event type carried in the ethernet address modification information, the address of the first port is updated.
13. equipment according to claim 12, which is characterized in that first virtual machine is specifically used for:
When the event type is that network interface card starts, obtains the second port and identify corresponding virtual address;
According to the virtual address and preset GID generating algorithm, the 2nd GID of the first port is determined, and by described The address of Single port is set as the 2nd GID.
14. equipment according to claim 12, which is characterized in that first virtual machine is specifically used for:
When the event type is that network interface card is closed, the address of the first port is deleted.
15. a kind of equipment, which is characterized in that the equipment includes: memory, processing circuit and host channel adapter HCA card;
Wherein, the memory is used for store code;
The processing circuit, for running the first virtual machine and host according to store code in the memory;
The processing circuit is also used to realize claim 1 to 7 any one institute by first virtual machine and the host The method stated;
The HCA card, for sending data to second virtual machine.
16. a kind of computer readable storage medium, including instruction, when described instruction is run on computers, make the calculating Machine executes method according to any one of claims 1 to 7.
17. a kind of computer program product comprising instruction, when run on a computer, so that computer executes such as right It is required that 1 to 7 described in any item methods.
CN201810064378.6A 2018-01-23 2018-01-23 Data sending method and device Active CN110071900B (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201810064378.6A CN110071900B (en) 2018-01-23 2018-01-23 Data sending method and device
CN202011203280.8A CN112398817B (en) 2018-01-23 2018-01-23 Data sending method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810064378.6A CN110071900B (en) 2018-01-23 2018-01-23 Data sending method and device

Related Child Applications (1)

Application Number Title Priority Date Filing Date
CN202011203280.8A Division CN112398817B (en) 2018-01-23 2018-01-23 Data sending method and device

Publications (2)

Publication Number Publication Date
CN110071900A true CN110071900A (en) 2019-07-30
CN110071900B CN110071900B (en) 2020-11-17

Family

ID=67365132

Family Applications (2)

Application Number Title Priority Date Filing Date
CN201810064378.6A Active CN110071900B (en) 2018-01-23 2018-01-23 Data sending method and device
CN202011203280.8A Active CN112398817B (en) 2018-01-23 2018-01-23 Data sending method and device

Family Applications After (1)

Application Number Title Priority Date Filing Date
CN202011203280.8A Active CN112398817B (en) 2018-01-23 2018-01-23 Data sending method and device

Country Status (1)

Country Link
CN (2) CN110071900B (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112583655A (en) * 2020-11-02 2021-03-30 北京迈格威科技有限公司 Data transmission method and device, electronic equipment and readable storage medium
WO2021135471A1 (en) * 2019-12-31 2021-07-08 腾讯科技(深圳)有限公司 Data transmission method and apparatus, network card and storage medium
CN113194045A (en) * 2020-01-14 2021-07-30 阿里巴巴集团控股有限公司 Data flow analysis method and device, storage medium and processor
CN113395359A (en) * 2021-08-17 2021-09-14 苏州浪潮智能科技有限公司 File currency cluster data transmission method and system based on remote direct memory access
CN113535370A (en) * 2020-04-09 2021-10-22 深圳致星科技有限公司 Method and equipment for realizing multiple RDMA network card virtualization of load balancing
CN114050998A (en) * 2022-01-07 2022-02-15 阿里巴巴(中国)有限公司 Method, device, electronic equipment and medium for realizing remote direct memory access
CN115277640A (en) * 2022-07-29 2022-11-01 迈普通信技术股份有限公司 Data processing method and device, intelligent network card and storage medium
CN115314437A (en) * 2022-06-16 2022-11-08 阿里巴巴(中国)有限公司 Container virtual network communication method and container virtual network

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113448634B (en) * 2021-05-31 2022-07-19 山东英信计算机技术有限公司 ROCE network card resource management method, device, equipment and readable medium
CN113766005B (en) * 2021-07-29 2023-04-28 苏州浪潮智能科技有限公司 RDMA-based method and system for batch creation of cloud hosts
CN113312155B (en) * 2021-07-29 2022-02-01 阿里云计算有限公司 Virtual machine creation method, device, equipment, system and computer program product
CN115834296B (en) * 2023-02-16 2023-04-18 北京融为科技有限公司 Data transmission method and device, electronic equipment and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7886139B2 (en) * 2007-02-23 2011-02-08 International Business Machines Corporation Method to enable firmware to boot a system from an ISCSI device
CN104753816A (en) * 2015-03-27 2015-07-01 华为技术有限公司 RDMA (remote direct memory access) connection message processing method and related device
CN107193699A (en) * 2017-05-22 2017-09-22 郑州云海信息技术有限公司 One kind tests the wide time-delay method of HCA cassette tapes automatically by RDMA modes

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8832216B2 (en) * 2011-08-31 2014-09-09 Oracle International Corporation Method and system for conditional remote direct memory access write
WO2014106321A1 (en) * 2013-01-04 2014-07-10 华为技术有限公司 Pcie network-based virtual machine communication method, server and system
US10230794B2 (en) * 2013-03-15 2019-03-12 Oracle International Corporation System and method for efficient virtualization in lossless interconnection networks
US9577928B2 (en) * 2013-08-27 2017-02-21 Oracle International Corporation System and method for supporting data service addressing in an engineered system for middleware and application execution
US9723009B2 (en) * 2014-09-09 2017-08-01 Oracle International Corporation System and method for providing for secure network communication in a multi-tenant environment
US9747249B2 (en) * 2014-12-29 2017-08-29 Nicira, Inc. Methods and systems to achieve multi-tenancy in RDMA over converged Ethernet
CN104636185B (en) * 2015-01-27 2018-03-02 华为技术有限公司 Business contexts management method, physical host, PCIE device and migration management equipment
US9928093B2 (en) * 2015-02-24 2018-03-27 Red Hat Israel, Ltd. Methods and systems for establishing connections associated with virtual machine migrations
US9813338B2 (en) * 2015-12-10 2017-11-07 Cisco Technology, Inc. Co-existence of routable and non-routable RDMA solutions on the same network interface

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7886139B2 (en) * 2007-02-23 2011-02-08 International Business Machines Corporation Method to enable firmware to boot a system from an ISCSI device
CN104753816A (en) * 2015-03-27 2015-07-01 华为技术有限公司 RDMA (remote direct memory access) connection message processing method and related device
CN107193699A (en) * 2017-05-22 2017-09-22 郑州云海信息技术有限公司 One kind tests the wide time-delay method of HCA cassette tapes automatically by RDMA modes

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021135471A1 (en) * 2019-12-31 2021-07-08 腾讯科技(深圳)有限公司 Data transmission method and apparatus, network card and storage medium
CN113194045A (en) * 2020-01-14 2021-07-30 阿里巴巴集团控股有限公司 Data flow analysis method and device, storage medium and processor
CN113194045B (en) * 2020-01-14 2023-11-17 阿里巴巴集团控股有限公司 Data traffic analysis method, device, storage medium and processor
CN113535370A (en) * 2020-04-09 2021-10-22 深圳致星科技有限公司 Method and equipment for realizing multiple RDMA network card virtualization of load balancing
CN112583655A (en) * 2020-11-02 2021-03-30 北京迈格威科技有限公司 Data transmission method and device, electronic equipment and readable storage medium
CN113395359A (en) * 2021-08-17 2021-09-14 苏州浪潮智能科技有限公司 File currency cluster data transmission method and system based on remote direct memory access
CN113395359B (en) * 2021-08-17 2021-10-29 苏州浪潮智能科技有限公司 File currency cluster data transmission method and system based on remote direct memory access
CN114050998A (en) * 2022-01-07 2022-02-15 阿里巴巴(中国)有限公司 Method, device, electronic equipment and medium for realizing remote direct memory access
CN115314437A (en) * 2022-06-16 2022-11-08 阿里巴巴(中国)有限公司 Container virtual network communication method and container virtual network
CN115314437B (en) * 2022-06-16 2024-03-01 阿里巴巴(中国)有限公司 Container virtual network communication method and container virtual network
CN115277640A (en) * 2022-07-29 2022-11-01 迈普通信技术股份有限公司 Data processing method and device, intelligent network card and storage medium
CN115277640B (en) * 2022-07-29 2023-11-24 迈普通信技术股份有限公司 Data processing method, device, intelligent network card and storage medium

Also Published As

Publication number Publication date
CN112398817B (en) 2022-02-25
CN112398817A (en) 2021-02-23
CN110071900B (en) 2020-11-17

Similar Documents

Publication Publication Date Title
CN110071900A (en) The method and apparatus that data are sent
US11372802B2 (en) Virtual RDMA switching for containerized applications
US10700979B2 (en) Load balancing for a virtual networking system
US9311230B2 (en) Local direct storage class memory access
US10574477B2 (en) Priority tagging based solutions in fc sans independent of target priority tagging capability
US11689631B2 (en) Transparent network service migration across service devices
US11379405B2 (en) Internet small computer interface systems extension for remote direct memory access (RDMA) for distributed hyper-converged storage systems
US10599356B2 (en) Aggregating memory to create a network addressable storage volume for storing virtual machine files
US9015333B2 (en) Apparatus and methods for handling network file operations over a fibre channel network
JP7023845B2 (en) Systems and methods to provide dynamic tenant relocation in a multi-tenant database environment
US20220222016A1 (en) Method for accessing solid state disk and storage device
WO2022143714A1 (en) Server system, and virtual machine creation method and apparatus
US20220200952A1 (en) Network address translation between networks
US20130013755A1 (en) Diskless pc network communication agent system
WO2022028092A1 (en) Vnf instantiation method and apparatus
US20180139176A1 (en) PaaS CONNECTION METHOD AND PaaS CONNECTION DEVICE
CN114726657A (en) Method and device for interrupt management and data receiving and sending management and intelligent network card
US8271258B2 (en) Emulated Z-series queued direct I/O
JP7212158B2 (en) Provider network service extension
WO2020233205A1 (en) Container service management method and device
WO2018098767A1 (en) Device and method for performing network interface bonding
CN114911411A (en) Data storage method and device and network equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant