WO2016187813A1 - 一种光电混合网络的数据传输方法及装置 - Google Patents

一种光电混合网络的数据传输方法及装置 Download PDF

Info

Publication number
WO2016187813A1
WO2016187813A1 PCT/CN2015/079829 CN2015079829W WO2016187813A1 WO 2016187813 A1 WO2016187813 A1 WO 2016187813A1 CN 2015079829 W CN2015079829 W CN 2015079829W WO 2016187813 A1 WO2016187813 A1 WO 2016187813A1
Authority
WO
WIPO (PCT)
Prior art keywords
transmission
link
data
rdma
switching link
Prior art date
Application number
PCT/CN2015/079829
Other languages
English (en)
French (fr)
Inventor
沈伟锋
赵秀楚
张丰伟
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to PCT/CN2015/079829 priority Critical patent/WO2016187813A1/zh
Priority to CN201580001273.0A priority patent/CN106664290B/zh
Publication of WO2016187813A1 publication Critical patent/WO2016187813A1/zh
Priority to US15/821,620 priority patent/US10868767B2/en

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00—Traffic control in data switching networks
    • H04L47/10—Flow control; Congestion control
    • H04L47/36—Flow control; Congestion control by determining packet size, e.g. maximum transfer unit [MTU]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00—Data switching networks
    • H04L12/28—Data switching networks characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks]
    • H04L12/46—Interconnection of networks
    • H04L12/4633—Interconnection of networks using encapsulation techniques, e.g. tunneling
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00—Data switching networks
    • H04L12/28—Data switching networks characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks]
    • H04L12/46—Interconnection of networks
    • H04L12/4641—Virtual LANs, VLANs, e.g. virtual private networks [VPN]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/10—Protocols in which an application is distributed across nodes in the network
    • H04L67/1097—Protocols in which an application is distributed across nodes in the network for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
    • H04L69/14—Multichannel or multilink protocols
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04Q—SELECTING
    • H04Q11/00—Selecting arrangements for multiplex systems
    • H04Q11/0001—Selecting arrangements for multiplex systems using optical switching
    • H04Q11/0003—Details
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04Q—SELECTING
    • H04Q11/00—Selecting arrangements for multiplex systems
    • H04Q11/0001—Selecting arrangements for multiplex systems using optical switching
    • H04Q11/0062—Network aspects
    • H04Q11/0066—Provisions for optical burst or packet networks
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L2212/00—Encapsulation of packets
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00—Arrangements for monitoring or testing data switching networks
    • H04L43/16—Threshold monitoring
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00—Packet switching elements
    • H04L49/60—Software-defined switches
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
    • H04L69/30—Definitions, standards or architectural aspects of layered protocol stacks
    • H04L69/32—Architecture of open systems interconnection [OSI] 7-layer type protocol stacks, e.g. the interfaces between the data link level and the physical level
    • H04L69/321—Interlayer communication protocols or service data unit [SDU] definitions; Interfaces between layers
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04Q—SELECTING
    • H04Q11/00—Selecting arrangements for multiplex systems
    • H04Q11/0001—Selecting arrangements for multiplex systems using optical switching
    • H04Q11/0062—Network aspects
    • H04Q2011/0073—Provisions for forwarding or routing, e.g. lookup tables

Definitions

  • the present invention relates to the field of computers, and in particular, to a data transmission method and apparatus for an opto-electric hybrid network.
  • RDMA Remote Direct Memory Access
  • the RDMA communication is based on three queue implementations: a send queue (SQ), a receive queue (RQ), and a completion queue (Completion Queue, CQ).
  • SQ send queue
  • RQ receive queue
  • CQ completion queue
  • the Work Queue Element (WQE) of the SQ indicates the storage location of the data to be sent
  • WQE of the RQ indicates the storage location of the received data
  • CQE Complete Queue Element (CQE) indicates the transmission. Or the completion of the reception.
  • SQ and RQ are collectively referred to as a Queue Pair (QP), as shown in FIG. 2.
  • QP Queue Pair
  • RDMA eliminates external memory copying and text exchange operations, it can free up memory bandwidth and Central Processing Unit (CPU) cycles and improve application system performance.
  • CPU Central Processing Unit
  • Optical switching means that the optical signal at the input end is directly switched to any optical output without any optical/electrical conversion.
  • Optical switching is one of the key technologies of all-optical networks.
  • the all-optical network In the modern communication network, the all-optical network is the development direction of the future broadband communication network.
  • the all-optical network can overcome the bottleneck limitation of the electronic exchange capacity, save a lot of network construction costs, and improve the flexibility and reliability of the network.
  • RDMA RDMA based on optical switching
  • advantages of optical switching are high throughput, but the switching time is longer, the switching time is greater than 10ms, and the burst traffic support is not good.
  • the advantage of electrical switching is that it supports good burst traffic and has no switching problem.
  • the disadvantage is that the throughput is relatively small and the capacity encounters bottleneck restrictions.
  • the switch end performs prediction analysis based on the traffic statistics result of the network data packet, and selects an optical switching path or an electrical switching path for the subsequent data packet according to the analysis result.
  • analysis is performed on the switch side. Since the switch side needs to perform a large number of network data packet traffic statistics analysis, and allocates a transmission path for subsequent network data packets, the system may be inefficient in transmitting.
  • the selection of the transmission path is based on the statistical analysis of the traffic of the previous network data packet, and the prediction of the subsequent data packet traffic is determined, so that a prediction error may occur, causing an error in the path selection, resulting in waste of resources.
  • the embodiment of the invention provides a data transmission method and device for an opto-electric hybrid network, so as to avoid the problem of path selection error and resource waste caused by statistical analysis of traffic of a large number of network data packets.
  • the first aspect provides a data transmission method of an opto-electric hybrid network, where the data transmission method of the opto-electric hybrid network includes:
  • Receiving a remote direct data access RDMA transmission request determining an RDMA transmission type and a transmission data amount for performing data transmission according to the RDMA transmission request;
  • the method before the receiving the RDMA transmission request, the method further includes:
  • the electrical switching link is selected to send the signaling data packet to the opposite end.
  • the determining, according to the RDMA transmission request, determining an RDMA transmission type for performing data transmission including:
  • the transmitting the data packet according to the Size, the exchange link selected for data transmission in the electrical switching link and the optical switching link including:
  • an optical switch link is available, an optical switch link is selected in the electrical switch link and the optical switch link as the switch link for performing data transmission.
  • the switching link for data transmission is selected in the electrical switching link and the optical switching link according to the transmission cost of the transport data packet.
  • the selecting according to a transmission cost of the transport data packet, selecting the data transmission in an electrical switching link and an optical switching link Switching links, including:
  • an electrical switching link is selected as the switching link for data transmission in the electrical switching link and the optical switching link.
  • the transmitting by the exchange link of the selected transmission data, the transmitting the data packet to the peer end, including:
  • the intelligent network interface card Transmitting, by the intelligent network interface card, the transport data packet to the switch, and sending, by the port of the switch, the transport data packet to the opposite end, based on the selected exchange link of the transport data, where the switch
  • the ports correspond to optical switching links and electrical switching links, respectively.
  • Adding a link flag to the transport data packet wherein the link flag is used to indicate that the transmission is performed by using an electrical switching link or by using an optical switching link.
  • a second aspect provides a data transmission device of an opto-electric hybrid network, where the data transmission device of the opto-electric hybrid network includes:
  • a receiving unit configured to receive a remote direct data access RDMA transmission request, and determine, according to the RDMA transmission request, an RDMA transmission type and a transmission data volume size for performing data transmission;
  • a determining unit configured to determine a preset encapsulation format corresponding to the RDMA transmission type for performing data transmission and the transmission data amount for performing data transmission;
  • An encapsulating unit configured to receive data to be transmitted, and encapsulate the to-be-transmitted data into the determining a transport packet of a preset package format
  • a selecting unit configured to select a switching link for data transmission in the electrical switching link and the optical switching link according to the size of the transport data packet
  • a sending unit configured to send the transport data packet to the opposite end based on the exchange link of the selected transmission data.
  • the method before the receiving the RDMA transmission request, the method further includes:
  • the receiving unit is configured to receive an RDMA transmission setup request message sent to the opposite end, and determine an RDMA transmission type corresponding to the RDMA transmission setup request message;
  • the determining unit and the encapsulating unit are configured to encapsulate the RDMA transmission setup request message into a signaling data packet according to an RDMA transmission type corresponding to the RDMA transmission setup request message, and
  • the selecting unit and the sending unit are configured to select an electrical switching link to send the signaling data packet to the opposite end.
  • the determining, according to the RDMA transmission request, determining an RDMA transmission type for performing data transmission When transmitting the amount of data the receiving unit is used to:
  • the transmitting the data packet according to the Size when selecting a switching link for data transmission in an electrical switching link and an optical switching link, the selecting unit is used to:
  • an optical switch link is available, an optical switch link is selected in the electrical switch link and the optical switch link as the switch link for performing data transmission.
  • the switching link for data transmission is selected in the electrical switching link and the optical switching link according to the transmission cost of the transport data packet.
  • the selection unit is used to:
  • an electrical switching link is selected as the switching link for data transmission in the electrical switching link and the optical switching link.
  • the sending unit when the exchange link that is based on the selected transmission data sends the transmission data packet to the opposite end, the sending unit is configured to:
  • the intelligent network interface card Transmitting, by the intelligent network interface card, the transport data packet to the switch, and sending, by the port of the switch, the transport data packet to the opposite end, based on the selected exchange link of the transport data, where the switch
  • the ports correspond to optical switching links and electrical switching links, respectively.
  • Adding a link flag to the transport data packet wherein the link flag is used to indicate that the transmission is performed by using an electrical switching link or by using an optical switching link.
  • a data transmission method of an opto-electric hybrid network is proposed: receiving an RDMA transmission request, determining an RDMA transmission type and a transmission data amount for performing data transmission according to an RDMA transmission request; determining an RDMA transmission type corresponding to performing data transmission and a preset encapsulation format for transmitting data amount of data transmission; receiving data to be transmitted, and encapsulating the data to be transmitted into a transmission packet of a predetermined preset encapsulation format; according to the size of the transmission data packet, in the electrical exchange link and the light A switching link for performing data transmission is selected in the switching link, and the transmission data packet is sent to the opposite end based on the switched link of the selected transmission data.
  • the intelligent network interface card on the host analyzes the RDMA transmission request for data transmission, determines the RDMA transmission type and the amount of transmission data for data transmission, and determines the reasonable size according to the corresponding package size.
  • the switching link transmits, avoids sending the data packet to the switch, and then selects the switching link according to the traffic statistics result, which causes the system to transmit low efficiency and predict the error of subsequent data packet traffic, and realizes data every time.
  • the transmitted RDMA transmission can select a reasonable exchange link, avoiding the problem of resource waste, and making the resources of the opto-electric hybrid network get the most efficient use.
  • FIG. 1 is a schematic diagram of RDMA data transmission in the prior art
  • FIG. 2 is a schematic diagram of three working queues of RDMA in the prior art
  • FIG. 3 is a flow chart showing an overview of data transmission of an opto-electric hybrid network according to an embodiment of the present invention
  • FIG. 4 is a schematic structural diagram of an RDMA transmission request in an embodiment of the present invention.
  • FIG. 5 is a schematic diagram of selection of a photoelectric hybrid network according to an embodiment of the present invention.
  • FIG. 6 is a schematic diagram of RDMA transmission between a host A and a host b according to an embodiment of the present invention
  • FIG. 7 is a schematic diagram of an intelligent network interface card selecting a transmission path according to different RDMA transmission requests according to an embodiment of the present invention
  • FIG. 8 is a schematic diagram of RDMA transmission between a host 1 and a host 2 according to an embodiment of the present invention.
  • FIG. 9 is a schematic structural diagram of a data transmission apparatus of an opto-electric hybrid network according to an embodiment of the present invention.
  • an embodiment of the present invention provides a data transmission method and apparatus for an opto-electric hybrid network, which is: receiving an RDMA transmission request, and determining an RDMA transmission type and transmission for performing data transmission according to the RDMA transmission request.
  • the amount of data is determined; a preset encapsulation format corresponding to an RDMA transmission type for performing data transmission and a transmission data amount for performing data transmission; a transmission packet for receiving data to be transmitted and encapsulating the data to be transmitted into a determined preset encapsulation format And selecting, according to the size of the transport data packet, a switching link for performing data transmission in the electrical switching link and the optical switching link, and transmitting the transport data packet to the opposite end based on the switched link of the selected transmission data.
  • the intelligent network interface card on the host analyzes the RDMA transmission request for data transmission, determines the RDMA transmission type and the amount of transmission data for data transmission, and determines the reasonable size according to the corresponding package size.
  • the switching link transmits, avoids sending the data packet to the switch, and then selects the switching link according to the traffic statistics result, which causes the system to transmit low efficiency and predict the error of subsequent data packet traffic, and realizes data every time.
  • the transmitted RDMA transmission can select a reasonable exchange link, avoiding the problem of resource waste, and making the resources of the opto-electric hybrid network get the most efficient use.
  • the specific process of data transmission of the opto-electric hybrid network in the present invention is:
  • Step 300 Receive an RDMA transmission request, and determine to perform data transmission according to the RDMA transmission request.
  • the RDMA transfer type and the amount of transferred data are included in Step 300.
  • an application of the application layer calls the intelligent network interface card to send or receive data through the RDMA interface
  • the application sends an RDMA transmission request to the intelligent network interface card according to the protocol, and uses the opcode to clarify the data.
  • a transfer is a read or write process of bulk data, or a send of a single data message.
  • RDMA is the ability to transfer data directly from the network to the computer's storage area, moving data from a system to a remote system memory without any impact on the operating system, thus eliminating the need for much computer processing. It eliminates external memory copying and protocol processing operations, freeing up bus space and CPU cycles to improve application performance.
  • the RDMA request is sent from the application running in the user space to the local intelligent network interface card without any kernel memory involvement ( Network Interface Card, NIC).
  • the local smart NIC reads the buffered content and transmits it to the target NIC over the network.
  • the RDMA information transmitted on the network includes the target virtual address, the memory key, and the data itself. Therefore, RDMA operations enable applications to read data from or write data to the memory of a remote application.
  • the target NIC receives the RDMA message sent by the local NIC, it confirms the memory key and completes the RDMA read or write request for the data.
  • the intelligent network interface card Before receiving the RDMA transmission request for the transmission of the large data block, that is, when receiving the RDMA transmission setup request message sent by the local application layer to the opposite end for the first time, the intelligent network interface card determines the RDMA transmission type corresponding to the RDMA transmission setup request message, according to
  • the RDMA transmission type corresponding to the RDMA transmission setup request message is generally a transmission type of send, and the RDMA transmission setup request message is encapsulated into a corresponding signaling data packet, and the electrical switching link is preferably selected to send a signaling data packet to the opposite end.
  • the signaling data packet includes an interface negotiation data packet (single data packet), and the data transmission request data packet (single data packet).
  • the signaling data packet at this time is used for negotiating the RDMA transmission for the subsequent data transmission with the opposite end, and determining the authentication information such as the address and the memory key of the RDMA transmission for subsequent data transmission, the signaling data packet is basically For a single data message, and the amount of data is small, preferred
  • the power exchange link sends data, which can make good use of the electrical switching link to support burst traffic, and has no switching problem.
  • the send message directly sends the data of the sender application to a buffer that has not been explicitly declared by the data receiver application, and determines the address of the subsequent RDMA transmission and the authentication information of the memory key by performing necessary negotiation of the transmission.
  • the application of host A needs to perform RDMA transmission with the application of host B.
  • the application of host A sends an RDMA transmission establishment request message to the intelligent network interface card of host A. Determining the RDMA transmission type corresponding to the RDMA transmission setup request message, and the intelligent network interface card of the host A encapsulates the RDMA transmission setup request message into a corresponding signaling data packet according to the RDMA transmission type corresponding to the RDMA transmission setup request message, and preferentially selects the power
  • the switching link sends the signaling packet to the opposite end.
  • the authentication information such as the address and the memory key of the RDMA transmission for subsequent data transmission is determined.
  • the RDMA transmission request for data transmission sent by the application in the local application layer is received, and the RDMA transmission type for data transmission is determined according to the operation code in the RDMA transmission request. And determining the amount of RDMA transmission data for data transmission according to the number of data link tables to be transmitted and the number of data blocks in the RDMA transmission request for data transmission, as shown in FIG.
  • the RDMA request for data transmission includes a request ID, a data link table to be transmitted, a number of data blocks, an operation code, and other necessary information, wherein the data link to be transmitted includes at least a data offset and a data segment length. And access rights.
  • Step 310 Determine a preset encapsulation format corresponding to an RDMA transmission type for performing data transmission and a transmission data amount for performing data transmission.
  • a preset encapsulation format corresponding to the RDMA transmission type and the transmission data amount of the data transmission is determined.
  • RDMA has two infinite broadband (infiniband, IB) and Internet wide-area RDMA protocol (iWARP) protocol suites, but this Most of them are similar.
  • the RDMA protocol stack of the iWARP protocol family consists of three protocols: RDMA, Direct Data Placement (DDP) protocol, and Label-based (Protocol-data-unit-Aligned, MPA) protocol. Together, it is used to ensure high-speed network interoperability.
  • the RDMA layer is used to convert the RDMA read, write, and send request messages into RDMA messages, that is, first convert to RDMA messages according to the RDMA transmission type, and transfer the RDMA messages to the DDP layer, and the DDP encapsulates the RDMA messages into DDP segments.
  • the data packet is forwarded to the MPA layer of the lower layer, that is, the RDMA message generated by the conversion is encapsulated into several data packets according to the amount of data transmission of the RDMA, and the MPA layer inserts the DDP data packet into the identifier, the length and the cyclic redundancy check code (Cyclic). Redundancy Check (CRC) checksum, which constitutes the MPA data segment.
  • CRC Redundancy Check
  • the Transmission Control Protocol (TCP) is responsible for scheduling TCP segments to ensure that packets are delivered to the target location.
  • the protocol Internet Protocol, IP
  • IP Internet Protocol
  • Step 320 Receive data to be transmitted, and encapsulate the data to be transmitted into a transmission data packet of a predetermined preset encapsulation format.
  • the data to be transmitted is encapsulated into a transmission packet of the determined preset encapsulation format by using the determined preset encapsulation format described above.
  • Step 330 Select, according to the size of the transport data packet, an exchange link for data transmission in the electrical switch link and the optical switch link.
  • the electrical switch link is selected as an exchange link for data transmission in the electrical switch link and the optical switch link.
  • the optical switch link is selected as an exchange link for data transmission in the electrical switch link and the optical switch link.
  • the optical switching link is unavailable, proceeding according to the transmission cost of the transport packet A switching link for data transmission is selected among the electrical switching link and the optical switching link.
  • the switching link for data transmission is selected in the electrical switching link and the optical switching link according to the transmission cost of the transmission data packet, wherein the calculation method of the transmission cost is: calculating the power consumption The first length of time required for the switching link to transmit the data packet and the second time required to establish a new optical switching link.
  • the optical switching link is selected as the switching link for data transmission in the electrical switching link and the optical switching link.
  • the electrical switching link is selected as the switching link for data transmission in the electrical switching link and the optical switching link.
  • the link-switching time of a fixed path is a fixed value, which is related to the physical properties of the device.
  • the second time is the time when the current optical switch link is removed and the optical switch link is established.
  • the intelligent network interface card of the host A determines the RDMA transmission type and the amount of the transmitted data amount for performing data transmission according to the received RDMA transmission request, and determines corresponding correspondence.
  • the size of the transport packet of the host A is compared with a preset threshold.
  • the threshold is greater than the preset threshold, it is further determined whether the current optical switch link is available, that is, whether the current optical switch link is The optical switching link between the host A and the host B, if it is determined that the current optical switching link is unavailable, continues to calculate the transmission cost of transmitting the transport data packet, that is, calculating the first duration and the second duration, and determining the shorter one.
  • the link corresponding to the duration is used as the switching link for the final data transmission.
  • Step 340 Send a transmission data packet to the opposite end based on the exchange link of the selected transmission data.
  • the switched link based on the selected transmitted data can send the data transmission packet to the opposite end in two ways:
  • the transmission data packet is sent to the opposite end through the port of the intelligent network interface card, where the ports of the intelligent network interface card respectively correspond to the optical switching link and the electrical switching link.
  • the transmission network packet is sent to the switch through the intelligent network interface card, and the transmission link is sent to the opposite end through the port of the switch based on the exchange link of the selected transmission data, wherein the ports of the switch respectively correspond to the optical switching link and the electrical Exchange link.
  • the intelligent network interface card transmits data according to the exchange link of the selected transmission data before the transmission data packet is sent to the opposite end through the port of the switch.
  • the packet adds a link tag, wherein the link tag is used to indicate whether to transmit using an electrical switching link or an optical switching link.
  • the tag here can be a vlan tag or the like.
  • the intelligent network interface card adds vlan 100 to the data packet sent by the optical switching link, adds vlan200 to the data packet sent by the electrical link, and the intelligent network interface card transmits the data packet to the switch.
  • the switch After the switch receives the data packet of the intelligent network interface card, the switch sends the data packet of the vlan 100 to the optical switching link interface according to the marking of the data packet, and the data packet of the vlan 200 is sent to the electrical switching link interface.
  • the local intelligent network interface card receives the RDMA transmission request sent by the application X for data transmission, the request type is write, and the transmission data size is a certain value E, and the request is An RDMA transfer from host A to host b, host A and host b have previously performed RDMA transfers.
  • the arbiter module in the intelligent network interface card determines the RDMA transmission request type and the data volume size of the current data transmission, determines a preset encapsulation format corresponding to the RDMA transmission type and the transmission data amount of the data transmission, and receives the data to be transmitted, and The data to be transmitted is encapsulated into a transport packet of a determined encapsulation format.
  • the arbiter module in the intelligent network interface card determines whether the optical switching link is available, if not available. , calculating the first duration required to transmit the transport data packet by using the electrical switching link, that is, the data length Dividing the bandwidth by the bandwidth and the second duration, for example, 200 ms, comparing the two durations. If the first duration is greater than or equal to the second duration, the optical switching link is selected as the data transmission link in the electrical switching link and the optical switching link. The switching link, if the first duration is less than the second duration, selects the electrical switching link as the switching link for data transmission in the electrical switching link and the optical switching link.
  • host b feeds back an ACK to host A. It is assumed here that the second duration is less than the first duration, the transmission data packet is transmitted by using the optical switching link, and the host b feeds back the ACK to the host A through the electrical switching link, where the ACK is a signaling data packet.
  • the intelligent network interface card is for different RDMA request types, the read and write type requests are generally large data block transmissions, and the send type requests are small data blocks signaling transmissions, however, It should be noted here that the request of the send type only preferentially selects the electrical switching link to send, instead of only transmitting through the electrical switching link. If the optical switching link between the two hosts is available at this time, the optical switching link can also be used. . Therefore, the intelligent network interface card transmits the appropriate switching link for each RDMA transmission request, so that the resources of the opto-electric hybrid network are utilized most efficiently.
  • the read and write type requests are generally large data block transmissions, and the send type requests are small data block signaling transmissions, it is also possible to pre-set the switching links used for each type of transmission, as shown in FIG.
  • the request for the read and write types is preferably transmitted by using an optical switching link
  • the request for the send type is preferably transmitted by using an electrical switching link, wherein the connection request is a specific application type of the send, and is established.
  • do not negotiate the address first use the default configuration.
  • the RDMA request of the send type initiated by the host 1 and the host 2 transmits signaling data packets using an electrical switching link, and all signaling data packets transmitted by using the electrical switching link are indicated by solid lines.
  • RDMA requests of the read and write types use the optical switching link to transmit transport packets, and all transport packets sent over the optical switch link are indicated by dashed lines.
  • a data transmission device of an opto-electric hybrid network includes:
  • the receiving unit 90 is configured to receive a remote direct data access RDMA transmission request, and determine, according to the RDMA transmission request, an RDMA transmission type and a transmission data volume size for performing data transmission;
  • a determining unit 91 configured to determine a preset encapsulation format corresponding to an RDMA transmission type for performing data transmission and a transmission data amount for performing data transmission;
  • the encapsulating unit 92 is configured to receive data to be transmitted, and encapsulate the data to be transmitted into a transmission data packet of a predetermined preset encapsulation format
  • the selecting unit 93 is configured to select, according to the size of the transport data packet, an exchange link for performing data transmission in the electrical switch link and the optical switch link, and
  • the sending unit 94 is configured to send a transport data packet to the opposite end based on the exchange link of the selected transmission data.
  • the method before receiving the RDMA transmission request, the method further includes:
  • the receiving unit 90 is configured to receive an RDMA transmission setup request message sent to the opposite end, and determine an RDMA transmission type corresponding to the RDMA transmission setup request message.
  • a determining unit 91 and an encapsulating unit 92 configured to encapsulate the RDMA transmission setup request message into a signaling data packet according to an RDMA transmission type corresponding to the RDMA transmission setup request message
  • the selecting unit 93 and the sending unit 94 are configured to select an electrical switching link to send a signaling data packet to the opposite end.
  • the receiving unit 90 when determining an RDMA transmission type and a transmission data volume size for performing data transmission according to the RDMA transmission request, the receiving unit 90 is configured to:
  • the amount of RDMA transmission data for data transmission is determined according to the number of data link tables to be transmitted and the number of data blocks in the RDMA transmission request.
  • the selecting unit 93 when selecting a switching link for data transmission in the electrical switching link and the optical switching link according to the size of the transport data packet, is configured to:
  • the electrical switch link is selected as an exchange link for data transmission in the electrical switch link and the optical switch link;
  • the optical switching link is selected as an exchange link for data transmission in the electrical switching link and the optical switching link, and the optical switching is performed.
  • the link is not available, according to the transmission cost of the transmitted data packet in the electrical switching link and light A switching link for data transmission is selected in the switching link.
  • the selecting unit 93 is configured to:
  • the optical switching link is selected as the switching link for data transmission in the electrical switching link and the optical switching link;
  • the electrical switching link is selected as the switching link for data transmission in the electrical switching link and the optical switching link.
  • the sending unit 94 is configured to:
  • the sending unit 94 before sending the transport data packet to the switch through the intelligent network interface card, and based on the exchange link of the selected transport data, before sending the transport data packet to the opposite end through the port of the switch, the sending unit 94 is further configured to:
  • a link flag is added for transmitting the data packet, wherein the link flag is used to indicate that the transmission is performed by using an electrical switching link or by using an optical switching link.
  • the embodiment of the present invention provides a data transmission method and apparatus for an opto-electric hybrid network, which is: receiving an RDMA transmission request, and determining data transmission according to the RDMA transmission request.
  • RDMA transmission type and transmission data amount determining a preset encapsulation format corresponding to an RDMA transmission type for performing data transmission and a transmission data amount for performing data transmission; receiving data to be transmitted, and encapsulating the data to be transmitted into a determined preset a transport packet of an encapsulated format; selecting, according to the size of the transport packet, a switching link for data transmission in the electrical switching link and the optical switching link, and transmitting the transmission to the opposite end based on the switched link of the selected transmission data data pack.
  • the intelligent network interface card on the host analyzes the RDMA transmission request for data transmission, determines the RDMA transmission type and the amount of transmission data for data transmission, and determines the reasonable size according to the corresponding package size.
  • the switching link transmits, avoids sending the data packet to the switch, and then selects the switching link according to the traffic statistics result, which causes the system to transmit low efficiency and predict the error of subsequent data packet traffic, and realizes data every time.
  • the transmitted RDMA transmission can select a reasonable exchange link, avoiding the problem of resource waste, and making the resources of the opto-electric hybrid network get the most efficient use.
  • embodiments of the present invention can be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or a combination of software and hardware. Moreover, the invention can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) including computer usable program code.
  • computer-usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc.
  • the computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture comprising the instruction device.
  • the apparatus implements the functions specified in one or more blocks of a flow or a flow and/or block diagram of the flowchart.
  • These computer program instructions can also be loaded onto a computer or other programmable data processing device such that a series of operational steps are performed on a computer or other programmable device to produce computer-implemented processing for execution on a computer or other programmable device.
  • the instructions provide steps for implementing the functions specified in one or more of the flow or in a block or blocks of a flow diagram.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Computer Security & Cryptography (AREA)
  • Optical Communication System (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

本发明涉及计算机领域，特别涉及一种光电混合网络的数据传输方法及装置，该方法为：接收RDMA传输请求，根据RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小；确定对应进行数据传输的RDMA传输类型和进行数据传输的传输数据量大小的预设封装格式；接收待传输数据，并将待传输数据封装为确定的预设封装格式的传输数据包；根据传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路，并基于选择的传输数据的交换链路向对端发送所述传输数据包。因此，实现了每次进行数据传输的RDMA传输都能选择合理的交换链路，避免了资源浪费的问题，使得光电混合网络的资源得到了最大效率的利用。

Description

一种光电混合网络的数据传输方法及装置 技术领域
本发明涉及计算机领域，特别涉及一种光电混合网络的数据传输方法及装置。
背景技术
远程直接数据存取(Remote Direct Memory Access，RDMA)技术，是为了解决网络传输中服务器端数据处理的延迟而产生的。RDMA通过网络把资料直接传入计算机的存储区，将数据从一个系统快速移动到远程系统存储器中，而不对操作系统造成任何影响，这样就不需要用到多少计算机的处理功能，如图1所示。
具体的，RDMA通信基于三种队列实现包括：发送队列(Send Queue，SQ)、接收队列(Receive Queue，RQ)和完成队列(Completion Queue，CQ)。其中，SQ存储的工作队列元素(Work Queue Element，WQE)指示需要发送的数据的存储位置，RQ存储的WQE指示接收数据的存放位置，CQ存储的完成队列元素(Completion Queue Element，CQE)指示发送或接收的完成情况。此外，SQ和RQ合称为队列对(Queue Pair，QP)，如图2所示。
因此，由于RDMA消除了外部存储器复制和文本交换操作，因而能解放内存带宽和中央处理器(Central Processing Unit，CPU)周期，改进应用系统性能。
光交换是指不经过任何光/电转换，将输入端光信号直接交换到任意的光输出端。光交换是全光网络的关键技术之一。在现代通信网中，全光网是未来宽带通信网的发展方向，全光网可以克服电子交换在容量上的瓶颈限制，大量节省建网成本，提高网络的灵活性和可靠性。
虽然传统的RDMA主要应用电交换，但是由于光交换的显著优越性，基于光交换的RDMA已经越来越受到重视，其中，基于电交换的RDMA与基 于光交换的RDMA相比各有优劣，光交换的优点在于高吞吐量，但是切换时间较长，切换时间大于10ms，且对突发流量支持不好。电交换的优点在于对突发流量支持好，无切换问题，缺点为吞吐量相对小，容量遇到瓶颈限制。
因此，现在业界越来越倾向于光电混合网络，以充分利用光交换和电交换的各自优势。
现有技术的光电混合网络中，交换机端基于对网络数据包的流量统计结果进行预测分析，根据分析结果为后续的数据包选择光交换路径或电交换路径发送。但是这样的分析是在交换机端进行的，由于交换机端需要进行大量的网络数据包的流量统计分析，为后续的网络数据包分配发送路径，因此可能造成系统的发送效率较低。此外，发送路径的选择是根据之前的网络数据包的流量统计分析，并对后续数据包流量做出预测确定的，因此可能出现预测错误的情况，使路径选择出现错误，造成资源浪费。
发明内容
本发明实施例提供一种光电混合网络的数据传输方法及装置，用以避免对大量网络数据包的流量统计分析造成路径选择出现错误、资源浪费的问题。
第一方面，提供一种光电混合网络的数据传输方法，所述光电混合网络的数据传输方法，包括：
接收远程直接数据存取RDMA传输请求，根据所述RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小；
确定对应所述进行数据传输的RDMA传输类型和所述进行数据传输的传输数据量大小的预设封装格式；
接收待传输数据，并将所述待传输数据封装为所述确定的预设封装格式的传输数据包；
根据所述传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路，并
基于选择的所述传输数据的交换链路向对端发送所述传输数据包。
结合第一方面，在第一种可能的实现方式中，在所述接收RDMA传输请求之前，进一步包括：
接收向所述对端发送的RDMA传输建立请求消息，确定所述RDMA传输建立请求消息对应的RDMA传输类型；
根据所述RDMA传输建立请求消息对应的RDMA传输类型，将所述RDMA传输建立请求消息封装为信令数据包，并
选择电交换链路向所述对端发送所述信令数据包。
结合第一方面，或第一方面的第一种可能的实现方式中的任一种，在第二种可能的实现方式中，所述根据所述RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小，包括：
根据所述RDMA传输请求中的操作码确定所述进行数据传输的RDMA传输类型，以及
根据所述RDMA传输请求中的待发送数据链表和数据块个数确定所述进行数据传输的RDMA传输数据量大小。
结合第一方面，或第一方面的第一种可能的实现方式至第二种可能的实现方式中的任一种，在第三种可能的实现方式中，所述根据所述传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路，包括：
若确定所述传输数据包的大小小于预设的门限，则在电交换链路和光交换链路中选择电交换链路作为所述进行数据传输的交换链路；
若确定所述传输数据包的大小大于等于预设的门限，则在光交换链路可用时，在电交换链路和光交换链路中选择光交换链路作为所述进行数据传输的交换链路，在光交换链路不可用时，根据所述传输数据包的传输代价在电交换链路和光交换链路中选择所述进行数据传输的交换链路。
结合第一方面的第三种可能的实现方式，在第四种可能的实现方式中，所述根据所述传输数据包的传输代价在电交换链路和光交换链路中选择所述进行数据传输的交换链路，包括：
计算采用电交换链路发送所述传输数据包需要的第一时长，以及计算建立新的光交换链路需要的第二时长；
若所述第一时长大于等于所述第二时长，则在电交换链路和光交换链路中选择光交换链路作为所述进行数据传输的交换链路；
若所述第一时长小于所述第二时长，则在电交换链路和光交换链路中选择电交换链路作为所述进行数据传输的交换链路。
结合第一方面，在第五种可能的实现方式中，所述基于选择的所述传输数据的交换链路向对端发送所述传输数据包，包括：
基于选择的所述传输数据的交换链路，通过智能网络接口卡的端口向所述对端发送所述传输数据包，其中，所述智能网络接口卡的端口分别对应光交换链路和电交换链路；或者
通过智能网络接口卡发送所述传输数据包至交换机，基于选择的所述传输数据的交换链路，通过所述交换机的端口向所述对端发送所述传输数据包，其中，所述交换机的端口分别对应光交换链路和电交换链路。
结合第一方面的第五种可能的实现方式，在第六种可能的实现方式中，在通过智能网络接口卡发送所述传输数据包至交换机，基于选择的所述传输数据的交换链路，通过所述交换机的端口向所述对端发送所述传输数据包之前，进一步包括：
为所述传输数据包添加链路标记，其中，所述链路标记用于指示采用电交换链路发送或是采用光交换链路发送。
第二方面，提供一种光电混合网络的数据传输装置，所述光电混合网络的数据传输装置，包括：
接收单元，用于接收远程直接数据存取RDMA传输请求，根据所述RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小；
确定单元，用于确定对应所述进行数据传输的RDMA传输类型和所述进行数据传输的传输数据量大小的预设封装格式；
封装单元，用于接收待传输数据，并将所述待传输数据封装为所述确定 的预设封装格式的传输数据包；
选择单元，用于根据所述传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路，并
发送单元，用于基于选择的所述传输数据的交换链路向对端发送所述传输数据包。
结合第二方面，在第一种可能的实现方式中，在所述接收RDMA传输请求之前，进一步包括：
所述接收单元，用于接收向所述对端发送的RDMA传输建立请求消息，确定所述RDMA传输建立请求消息对应的RDMA传输类型；
所述确定单元和所述封装单元，用于根据所述RDMA传输建立请求消息对应的RDMA传输类型，将所述RDMA传输建立请求消息封装为信令数据包，并
所述选择单元和所述发送单元，用于选择电交换链路向所述对端发送所述信令数据包。
结合第二方面，或第二方面的第一种可能的实现方式中的任一种，在第二种可能的实现方式中，所述根据所述RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小时，所述接收单元用于：
根据所述RDMA传输请求中的操作码确定所述进行数据传输的RDMA传输类型，以及
根据所述RDMA传输请求中的待发送数据链表和数据块个数确定所述进行数据传输的RDMA传输数据量大小。
结合第二方面，或第二方面的第一种可能的实现方式至第二种可能的实现方式中的任一种，在第三种可能的实现方式中，所述根据所述传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路时，所述选择单元用于：
若确定所述传输数据包的大小小于预设的门限，则在电交换链路和光交换链路中选择电交换链路作为所述进行数据传输的交换链路；
若确定所述传输数据包的大小大于等于预设的门限，则在光交换链路可用时，在电交换链路和光交换链路中选择光交换链路作为所述进行数据传输的交换链路，在光交换链路不可用时，根据所述传输数据包的传输代价在电交换链路和光交换链路中选择所述进行数据传输的交换链路。
结合第二方面的第三种可能的实现方式，在第四种可能的实现方式中，所述根据所述传输数据包的传输代价在电交换链路和光交换链路中选择所述进行数据传输的交换链路时，所述选择单元用于：
计算采用电交换链路发送所述传输数据包需要的第一时长，以及计算建立新的光交换链路需要的第二时长；
若所述第一时长大于等于所述第二时长，则在电交换链路和光交换链路中选择光交换链路作为所述进行数据传输的交换链路；
若所述第一时长小于所述第二时长，则在电交换链路和光交换链路中选择电交换链路作为所述进行数据传输的交换链路。
结合第二方面，在第五种可能的实现方式中，所述基于选择的所述传输数据的交换链路向对端发送所述传输数据包时，所述发送单元用于：
基于选择的所述传输数据的交换链路，通过智能网络接口卡的端口向所述对端发送所述传输数据包，其中，所述智能网络接口卡的端口分别对应光交换链路和电交换链路；或者
通过智能网络接口卡发送所述传输数据包至交换机，基于选择的所述传输数据的交换链路，通过所述交换机的端口向所述对端发送所述传输数据包，其中，所述交换机的端口分别对应光交换链路和电交换链路。
结合第二方面的第五种可能的实现方式，在第六种可能的实现方式中，在通过智能网络接口卡发送所述传输数据包至交换机，基于选择的所述传输数据的交换链路，通过所述交换机的端口向所述对端发送所述传输数据包之前，所述发送单元进一步用于：
为所述传输数据包添加链路标记，其中，所述链路标记用于指示采用电交换链路发送或是采用光交换链路发送。
本发明有益效果如下：
本发明实施例中，提出一种光电混合网络的数据传输方法：接收RDMA传输请求，根据RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小；确定对应进行数据传输的RDMA传输类型和进行数据传输的传输数据量大小的预设封装格式；接收待传输数据，并将待传输数据封装为确定的预设封装格式的传输数据包；根据传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路，并基于选择的传输数据的交换链路向对端发送所述传输数据包。因此，主机上的智能网络接口卡通过对接收到进行数据传输的RDMA传输请求进行分析，确定进行数据传输的RDMA传输类型和传输数据量大小，并根据最后封装成的对应的数据包大小确定合理的交换链路进行传输，避免了将数据包发送至交换机处再根据流量统计结果选择交换链路，造成系统的发送效率较低以及对后续数据包流量预测错误的问题，实现了每次进行数据传输的RDMA传输都能选择合理的交换链路，避免了资源浪费的问题，使得光电混合网络的资源得到了最大效率的利用。
附图说明
图1为现有技术中RDMA数据传输示意图；
图2为现有技术中RDMA三种工作队列示意图；
图3为本发明实施例中光电混合网络的数据传输概述流程图；
图4为本发明实施例中RDMA传输请求结构示意图；
图5为本发明实施例中光电混合网络选择示意图；
图6为本发明实施例中主机A与主机b之间的RDMA传输示意图；
图7为本发明实施例中智能网络接口卡根据不同的RDMA传输请求选择发送路径的示意图；
图8为本发明实施例中主机1与主机2之间的RDMA传输示意图。
图9为本发明实施例中光电混合网络的数据传输装置结构示意图。
具体实施方式
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
为了充分利用光电混合网络发送数据，本发明实施例提供了一种光电混合网络的数据传输方法及装置，该方法为：接收RDMA传输请求，根据RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小；确定对应进行数据传输的RDMA传输类型和进行数据传输的传输数据量大小的预设封装格式；接收待传输数据，并将待传输数据封装为确定的预设封装格式的传输数据包；根据传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路，并基于选择的传输数据的交换链路向对端发送所述传输数据包。
因此，主机上的智能网络接口卡通过对接收到进行数据传输的RDMA传输请求进行分析，确定进行数据传输的RDMA传输类型和传输数据量大小，并根据最后封装成的对应的数据包大小确定合理的交换链路进行传输，避免了将数据包发送至交换机处再根据流量统计结果选择交换链路，造成系统的发送效率较低以及对后续数据包流量预测错误的问题，实现了每次进行数据传输的RDMA传输都能选择合理的交换链路，避免了资源浪费的问题，使得光电混合网络的资源得到了最大效率的利用。
下面结合附图对本发明优选的实施方式进行详细说明。
参阅图3所示，本发明中光电混合网络的数据传输的具体流程为：
步骤300：接收RDMA传输请求，根据RDMA传输请求确定进行数据传输 的RDMA传输类型和传输数据量大小。
具体的，在应用层的某个应用程序通过RDMA接口调用智能网络接口卡发送或接收数据时，按照协议规定，应用程序向智能网络接口卡发送RDMA传输请求，使用操作码(opcode)，明确数据传输是大块数据的读取(read)或者写出(write)过程，或者是单个数据报文的发送(send)。
RDMA是通过网络把资料直接传入计算机的存储区，将数据从一个系统快速移动到远程系统存储器中，而不对操作系统造成任何影响，这样就不需要用到多少计算机的处理功能。它消除了外部存储器复制和协议处理操作，因而能腾出总线空间和CPU周期用于改进应用系统性能。
具体的，当一个应用执行RDMA读(read)或写(write)请求时，在不需要任何内核内存参与的条件下，RDMA请求从运行在用户空间中的应用中发送到本地智能网络接口卡(Network Interface Card，NIC)。本地智能NIC读取缓冲的内容，并通过网络传送到目标NIC。其中，在网络上传输的RDMA信息包含目标虚拟地址、内存钥匙和数据本身。因此，RDMA操作使应用可以实现从一个远程应用的内存中读数据或向这个内存写数据。当目标NIC接收到本地NIC发送的RDMA消息时，确认内存钥匙，完成对数据的RDMA读或写请求。
而在接收RDMA传输请求进行大数据块的传输之前，即首次接收到本地应用层向对端发送的RDMA传输建立请求消息时，智能网络接口卡确定RDMA传输建立请求消息对应的RDMA传输类型，根据RDMA传输建立请求消息对应的RDMA传输类型，一般为传输类型为send，将RDMA传输建立请求消息封装为对应的信令数据包，并优选选择电交换链路向对端发送信令数据包。其中，信令数据包包括接口协商数据包(单个的数据报文)，数据发送请求数据包(单个的数据报文)。
由于此时的信令数据包用于与对端建立后续进行数据传输的RDMA传输必要的协商，并确定后续进行数据传输的RDMA传输的地址和内存钥匙等认证信息，因此，信令数据包基本为单个的数据报文，且数据量较小，优先选 择电交换链路发送数据，能够很好地利用电交换链路对突发流量支持好，无切换问题的特点。
一般地，send消息将发送方应用的数据直接发送到数据接收方应用尚未明确声明的缓冲区中，通过进行传输必要的协商，确定后续RDMA传输的地址和内存钥匙等认证信息。
例如，主机A的应用程序需要与主机B的应用程序进行RDMA传输，在主机A首次发起与主机B进行RDMA传输时，主机A的应用程序向主机A的智能网络接口卡发送RDMA传输建立请求消息，确定RDMA传输建立请求消息对应的RDMA传输类型，主机A的智能网络接口卡根据RDMA传输建立请求消息对应的RDMA传输类型将RDMA传输建立请求消息封装为对应的信令数据包，并优先选择电交换链路向对端发送该信令数据包。当主机A与主机B的RDMA传输协商建立成功后，确定后续进行数据传输的RDMA传输的地址和内存钥匙等认证信息。
在本地发起RDMA的应用程序与对端协商成功之后，此时接收本地应用层中应用程序发送的进行数据传输的RDMA传输请求，根据RDMA传输请求中的操作码确定进行数据传输的RDMA传输类型，以及根据进行数据传输的RDMA传输请求中的待发送数据链表和数据块个数确定进行数据传输的RDMA传输数据量大小，参阅图4所示。
如图4中所示，进行数据传输的RDMA请求包括请求ID，待发送数据链表，数据块个数，操作码等其他必要信息，其中，待发送数据链至少表包括数据偏移，数据段长度和访问权限。
步骤310：确定对应进行数据传输的RDMA传输类型和进行数据传输的传输数据量大小的预设封装格式。
根据RDMA协议栈的协议内容，确定对应进行数据传输的RDMA传输类型和传输数据量大小的预设封装格式。
目前，RDMA有无限宽带(infiniband，IB)和互联网广域远程直接内存访问协议(Internet wide-area RDMA protocol，iWARP)两种协议族，不过这 两者大部分类似。具体的，iWARP协议族的RDMA协议栈由RDMA、直接数据存放(Direct Data Placement，DDP)协议、基于标识的协议数据单元集合(Marker-based，Protocol-data-unit-Aligned，MPA)三种协议共同实现，用来保证高速网络的互操作性。RDMA层用于将RDMA的read、write及send的请求消息转化成RDMA消息，即首先根据RDMA的传输类型转化成RDMA消息，并将RDMA消息传送至DDP层，DDP将RDMA消息分段封装成DDP数据包转发到下层的MPA层，即根据RDMA的传输数据量大小将转化生成的RDMA消息封装成若干个数据包，MPA层将DDP数据包插入标识符，长度及循环冗余校验码(Cyclic Redundancy Check，CRC)校验，构成MPA数据段。传输控制协议层(Transmission Control Protocol，TCP)负责对TCP数据段进行调度，确保发包能够顺利到达目标位置。网络之间互连的协议(Internet Protocol，IP)层则在数据包中增加必要的网络路由数据信息，即添加对端的网络路由数据信息。
步骤320：接收待传输数据，并将待传输数据封装为确定的预设封装格式的传输数据包。
采用上述的确定的预设封装格式将待传输数据封装为确定的预设封装格式的传输数据包。
步骤330：根据传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路。
参阅图5所示，在分析进行数据传输的RDMA请求类型和数据量大小并封装成对应的传输数据包之后，具体的，首先将该传输数据包的大小与预设的门限值进行比较：
若确定传输数据包的大小小于预设的门限，则在电交换链路和光交换链路中选择电交换链路作为进行数据传输的交换链路。
若确定所述传输数据包的大小大于等于预设的门限，则在光交换链路可用时，在电交换链路和光交换链路中选择光交换链路作为进行数据传输的交换链路，在光交换链路不可用时，继续进而根据所述传输数据包的传输代价 在电交换链路和光交换链路中选择进行数据传输的交换链路。
进一步地，当确定光交换链路不可用时，根据传输数据包的传输代价在电交换链路和光交换链路中选择进行数据传输的交换链路，其中，传输代价的计算方法为：计算采用电交换链路发送传输数据包需要的第一时长，以及计算建立新的光交换链路需要的第二时长。
根据传输数据包的传输代价选择电交换链路或光交换链路向对端发送传输数据包的具体过程为：
若第一时长大于等于第二时长，则在电交换链路和光交换链路中选择光交换链路作为进行数据传输的交换链路。
若第一时长小于第二时长，则在电交换链路和光交换链路中选择电交换链路作为进行数据传输的交换链路。
这里选择第一时长和第二时长中的较小值作为该传输数据包的传输代价。一般的固定路径的链路切换时间是一个固定值，和设备的物理性质有关，其中，第二时长为拆除当前光交换链路和建立本条光交换链路的时间和。
例如，当主机A的应用程序非首次与主机B进行RDMA传输时，主机A的智能网络接口卡根据接收到的RDMA传输请求，判断进行数据传输的RDMA传输类型和传输数据量大小，确定对应进行数据传输的RDMA传输类型和传输数据量大小的预设的封装格式，接收待传输数据，并将待传输数据封装为确定的封装格式的传输数据包。
首先，将主机A的传输数据包的大小与预设的门限值进行比较，当确定大于预设的门限时，进一步判断当前的光交换链路是否可用，即当前的光交换链路是否为主机A到主机B的光交换链路，若确定当前的光交换链路不可用时，则继续计算发送该传输数据包的传输代价，即计算上述的第一时长和第二时长，确定其中较短时长所对应的链路作为最终进行数据传输的交换链路。
步骤340：基于选择的传输数据的交换链路向对端发送传输数据包。
在根据传输数据包的大小，从电交换链路和光交换链路中选择进行数据 传输的交换链路之后，基于选择的传输数据的交换链路可以通过两种方式向对端发送数据传输数据包：
第一，通过智能网络接口卡的端口向对端发送传输数据包，其中，智能网络接口卡的端口分别对应光交换链路和电交换链路。
第二，通过智能网络接口卡发送传输数据包至交换机，基于选择的传输数据的交换链路，通过交换机的端口向对端发送传输数据包，其中，交换机的端口分别对应光交换链路和电交换链路。
在第二种方式中，在通过智能网络接口卡发送传输数据包至交换机，基于选择的传输数据的交换链路，通过交换机的端口向对端发送传输数据包之前，智能网络接口卡为传输数据包添加链路标记，其中，链路标记用于指示采用电交换链路发送还是采用光交换链路发送。例如，这里的标记可以为vlan tag等。
又例如，智能网络接口卡为采用光交换链路发送的数据包添加vlan 100，为采用电链路发送的数据包添加vlan200，智能网络接口卡将数据包传输至交换机。交换机收到智能网络接口卡的数据包之后，交换机根据数据包的标记，将vlan100的数据包发往光交换链路接口，vlan200的数据包发往电交换链路接口。
实施例1中，参阅图6所示，本地的智能网络接口卡收到应用X发送的进行数据传输的RDMA传输请求，该请求类型为write，传输数据量大小为某一个值E，该请求是一个从主机A到主机b的RDMA传输过程，主机A和主机b之前已经进行过RDMA传输。智能网络接口卡中的arbiter模块确定当前进行数据传输的RDMA传输请求类型和数据量大小，确定对应进行数据传输的RDMA传输类型和传输数据量大小的预设的封装格式，接收待传输数据，并将待传输数据封装为确定的封装格式的传输数据包。进一步地，判断该数据包的大小大于预设门限64K时，其中，预设的门限与计算机和网络的实际性能相关，智能网络接口卡中的arbiter模块判断光交换链路是否可用，若不可用，则计算采用电交换链路发送该传输数据包需要的第一时长，即数据长 度除以带宽，和第二时长，例如200ms，比较两个时长大小，若第一时长大于等于第二时长，则在电交换链路和光交换链路中选择光交换链路作为进行数据传输的交换链路，若第一时长小于第二时长，则在电交换链路和光交换链路中选择电交换链路作为进行数据传输的交换链路。
在传输完毕后，主机b反馈一个ACK给主机A。这里假设第二时长小于第一时长，采用光交换链路发送该传输数据包，主机b通过电交换链路反馈ACK给主机A，其中，ACK为信令数据包。
参阅图7所示，一般地，智能网络接口卡针对不同的RDMA请求类型，read和write类型的请求一般是大数据块的传输，而send类型的请求是小数据块的信令传输，但是，这里需要指出，send类型的请求只是优先选择电交换链路发送，而不是只能通过电交换链路发送，如果此时两个主机之间光交换链路可用，也可采用光交换链路发送。因此，智能网络接口卡通过对每一次RDMA传输请求分配合适的交换链路进行传输，使得光电混合网络的资源得到了最大效率的利用。
此外，由于read和write类型的请求一般是大数据块的传输，而send类型的请求是小数据块的信令传输，因此也可以预先设置每种传输类型使用的交换链路，参阅图8所述，对于read和write类型的请求优选采用光交换链路进行传输，对于send类型的请求优选采用电交换链路进行传输，其中，连接请求(Connection request)是send的一个具体应用类型，在建立连接的时候，不用协商地址，先用默认配置。例如，图8中，主机1发起的与主机2之间的send类型的RDMA请求，采用电交换链路发送信令数据包，所有采用电交换链路发送的信令数据包用实线表示。read和write类型的RDMA请求采用光交换链路发送传输数据包，所有采用光交换链路发送的传输数据包用虚线表示。
参阅图9所示，一种光电混合网络的数据传输装置，包括：
接收单元90，用于接收远程直接数据存取RDMA传输请求，根据RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小；
确定单元91，用于确定对应进行数据传输的RDMA传输类型和进行数据传输的传输数据量大小的预设封装格式；
封装单元92，用于接收待传输数据，并将待传输数据封装为确定的预设封装格式的传输数据包；
选择单元93，用于根据传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路，并
发送单元94，用于基于选择的传输数据的交换链路向对端发送传输数据包。
可选的，在接收RDMA传输请求之前，进一步包括：
接收单元90，用于接收向对端发送的RDMA传输建立请求消息，确定RDMA传输建立请求消息对应的RDMA传输类型；
确定单元91和封装单元92，用于根据RDMA传输建立请求消息对应的RDMA传输类型，将RDMA传输建立请求消息封装为信令数据包，并
选择单元93和发送单元94，用于选择电交换链路向对端发送信令数据包。
可选的，根据RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小时，接收单元90用于：
根据RDMA传输请求中的操作码确定进行数据传输的RDMA传输类型，以及
根据RDMA传输请求中的待发送数据链表和数据块个数确定进行数据传输的RDMA传输数据量大小。
可选的，根据传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路时，选择单元93用于：
若确定传输数据包的大小小于预设的门限，则在电交换链路和光交换链路中选择电交换链路作为进行数据传输的交换链路；
若确定传输数据包的大小大于等于预设的门限，则在光交换链路可用时，在电交换链路和光交换链路中选择光交换链路作为进行数据传输的交换链路，在光交换链路不可用时，根据传输数据包的传输代价在电交换链路和光 交换链路中选择进行数据传输的交换链路。
可选的，根据传输数据包的传输代价在电交换链路和光交换链路中选择进行数据传输的交换链路时，选择单元93用于：
计算采用电交换链路发送传输数据包需要的第一时长，以及计算建立新的光交换链路需要的第二时长；
若第一时长大于等于第二时长，则在电交换链路和光交换链路中选择光交换链路作为进行数据传输的交换链路；
若第一时长小于第二时长，则在电交换链路和光交换链路中选择电交换链路作为进行数据传输的交换链路。
可选的，基于选择的传输数据的交换链路向对端发送传输数据包时，发送单元94用于：
基于选择的传输数据的交换链路，通过智能网络接口卡的端口向对端发送传输数据包，其中，智能网络接口卡的端口分别对应光交换链路和电交换链路；或者
通过智能网络接口卡发送传输数据包至交换机，基于选择的传输数据的交换链路，通过交换机的端口向对端发送传输数据包，其中，交换机的端口分别对应光交换链路和电交换链路。
可选的，在通过智能网络接口卡发送传输数据包至交换机，基于选择的传输数据的交换链路，通过交换机的端口向对端发送传输数据包之前，发送单元94进一步用于：
为传输数据包添加链路标记，其中，链路标记用于指示采用电交换链路发送或是采用光交换链路发送。
综上所述，为了充分利用光电混合网络发送数据，本发明实施例提供了一种光电混合网络的数据传输方法及装置，该方法为：接收RDMA传输请求，根据RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小；确定对应进行数据传输的RDMA传输类型和进行数据传输的传输数据量大小的预设封装格式；接收待传输数据，并将待传输数据封装为确定的预设 封装格式的传输数据包；根据传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路，并基于选择的传输数据的交换链路向对端发送所述传输数据包。
因此，主机上的智能网络接口卡通过对接收到进行数据传输的RDMA传输请求进行分析，确定进行数据传输的RDMA传输类型和传输数据量大小，并根据最后封装成的对应的数据包大小确定合理的交换链路进行传输，避免了将数据包发送至交换机处再根据流量统计结果选择交换链路，造成系统的发送效率较低以及对后续数据包流量预测错误的问题，实现了每次进行数据传输的RDMA传输都能选择合理的交换链路，避免了资源浪费的问题，使得光电混合网络的资源得到了最大效率的利用。
本领域内的技术人员应明白，本发明的实施例可提供为方法、系统、或计算机程序产品。因此，本发明可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本发明可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本发明是参照根据本发明实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
尽管已描述了本发明的优选实施例，但本领域内的技术人员一旦得知了基本创造性概念，则可对这些实施例作出另外的变更和修改。所以，所附权利要求意欲解释为包括优选实施例以及落入本发明范围的所有变更和修改。
显然，本领域的技术人员可以对本发明实施例进行各种改动和变型而不脱离本发明实施例的精神和范围。这样，倘若本发明实施例的这些修改和变型属于本发明权利要求及其等同技术的范围之内，则本发明也意图包含这些改动和变型在内。

Claims (14)

  1. 一种光电混合网络的数据传输方法，其特征在于，包括：
    接收远程直接数据存取RDMA传输请求，根据所述RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小；
    确定对应所述进行数据传输的RDMA传输类型和所述进行数据传输的传输数据量大小的预设封装格式；
    接收待传输数据，并将所述待传输数据封装为所述确定的预设封装格式的传输数据包；
    根据所述传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路，并
    基于选择的所述传输数据的交换链路向对端发送所述传输数据包。
  2. 如权利要求1所述的方法，其特征在于，在所述接收RDMA传输请求之前，进一步包括：
    接收向所述对端发送的RDMA传输建立请求消息，确定所述RDMA传输建立请求消息对应的RDMA传输类型；
    根据所述RDMA传输建立请求消息对应的RDMA传输类型，将所述RDMA传输建立请求消息封装为信令数据包，并
    选择电交换链路向所述对端发送所述信令数据包。
  3. 如权利要求1或2所述的方法，其特征在于，所述根据所述RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小，包括：
    根据所述RDMA传输请求中的操作码确定所述进行数据传输的RDMA传输类型，以及
    根据所述RDMA传输请求中的待发送数据链表和数据块个数确定所述进行数据传输的RDMA传输数据量大小。
  4. 如权利要求1-3任一项所述的方法，其特征在于，所述根据所述传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路， 包括：
    若确定所述传输数据包的大小小于预设的门限，则在电交换链路和光交换链路中选择电交换链路作为所述进行数据传输的交换链路；
    若确定所述传输数据包的大小大于等于预设的门限，则在光交换链路可用时，在电交换链路和光交换链路中选择光交换链路作为所述进行数据传输的交换链路，在光交换链路不可用时，根据所述传输数据包的传输代价在电交换链路和光交换链路中选择所述进行数据传输的交换链路。
  5. 如权利要求4所述的方法，其特征在于，所述根据所述传输数据包的传输代价在电交换链路和光交换链路中选择所述进行数据传输的交换链路，包括：
    计算采用电交换链路发送所述传输数据包需要的第一时长，以及计算建立新的光交换链路需要的第二时长；
    若所述第一时长大于等于所述第二时长，则在电交换链路和光交换链路中选择光交换链路作为所述进行数据传输的交换链路；
    若所述第一时长小于所述第二时长，则在电交换链路和光交换链路中选择电交换链路作为所述进行数据传输的交换链路。
  6. 如权利要求1所述的方法，其特征在于，所述基于选择的所述传输数据的交换链路向对端发送所述传输数据包，包括：
    基于选择的所述传输数据的交换链路，通过智能网络接口卡的端口向所述对端发送所述传输数据包，其中，所述智能网络接口卡的端口分别对应光交换链路和电交换链路；或者
    通过智能网络接口卡发送所述传输数据包至交换机，基于选择的所述传输数据的交换链路，通过所述交换机的端口向所述对端发送所述传输数据包，其中，所述交换机的端口分别对应光交换链路和电交换链路。
  7. 如权利要求6所述的方法，其特征在于，在通过智能网络接口卡发送所述传输数据包至交换机，基于选择的所述传输数据的交换链路，通过所述交换机的端口向所述对端发送所述传输数据包之前，进一步包括：
    为所述传输数据包添加链路标记，其中，所述链路标记用于指示采用电交换链路发送或是采用光交换链路发送。
  8. 一种光电混合网络的数据传输装置，其特征在于，包括：
    接收单元，用于接收远程直接数据存取RDMA传输请求，根据所述RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小；
    确定单元，用于确定对应所述进行数据传输的RDMA传输类型和所述进行数据传输的传输数据量大小的预设封装格式；
    封装单元，用于接收待传输数据，并将所述待传输数据封装为所述确定的预设封装格式的传输数据包；
    选择单元，用于根据所述传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路，并
    发送单元，用于基于选择的所述传输数据的交换链路向对端发送所述传输数据包。
  9. 如权利要求8所述的装置，其特征在于，在所述接收RDMA传输请求之前，进一步包括：
    所述接收单元，用于接收向所述对端发送的RDMA传输建立请求消息，确定所述RDMA传输建立请求消息对应的RDMA传输类型；
    所述确定单元和所述封装单元，用于根据所述RDMA传输建立请求消息对应的RDMA传输类型，将所述RDMA传输建立请求消息封装为信令数据包，并
    所述选择单元和所述发送单元，用于选择电交换链路向所述对端发送所述信令数据包。
  10. 如权利要求8或9所述的装置，其特征在于，所述根据所述RDMA传输请求确定进行数据传输的RDMA传输类型和传输数据量大小时，所述接收单元用于：
    根据所述RDMA传输请求中的操作码确定所述进行数据传输的RDMA传输类型，以及
    根据所述RDMA传输请求中的待发送数据链表和数据块个数确定所述进行数据传输的RDMA传输数据量大小。
  11. 如权利要求8-10任一项所述的装置，其特征在于，所述根据所述传输数据包的大小，在电交换链路和光交换链路中选择进行数据传输的交换链路时，所述选择单元用于：
    若确定所述传输数据包的大小小于预设的门限，则在电交换链路和光交换链路中选择电交换链路作为所述进行数据传输的交换链路；
    若确定所述传输数据包的大小大于等于预设的门限，则在光交换链路可用时，在电交换链路和光交换链路中选择光交换链路作为所述进行数据传输的交换链路，在光交换链路不可用时，根据所述传输数据包的传输代价在电交换链路和光交换链路中选择所述进行数据传输的交换链路。
  12. 如权利要求11所述的装置，其特征在于，所述根据所述传输数据包的传输代价在电交换链路和光交换链路中选择所述进行数据传输的交换链路时，所述选择单元用于：
    计算采用电交换链路发送所述传输数据包需要的第一时长，以及计算建立新的光交换链路需要的第二时长；
    若所述第一时长大于等于所述第二时长，则在电交换链路和光交换链路中选择光交换链路作为所述进行数据传输的交换链路；
    若所述第一时长小于所述第二时长，则在电交换链路和光交换链路中选择电交换链路作为所述进行数据传输的交换链路。
  13. 如权利要求9所述的装置，其特征在于，所述基于选择的所述传输数据的交换链路向对端发送所述传输数据包时，所述发送单元用于：
    基于选择的所述传输数据的交换链路，通过智能网络接口卡的端口向所述对端发送所述传输数据包，其中，所述智能网络接口卡的端口分别对应光交换链路和电交换链路；或者
    通过智能网络接口卡发送所述传输数据包至交换机，基于选择的所述传输数据的交换链路，通过所述交换机的端口向所述对端发送所述传输数据包， 其中，所述交换机的端口分别对应光交换链路和电交换链路。
  14. 如权利要求13所述的装置，其特征在于，在通过智能网络接口卡发送所述传输数据包至交换机，基于选择的所述传输数据的交换链路，通过所述交换机的端口向所述对端发送所述传输数据包之前，所述发送单元进一步用于：
    为所述传输数据包添加链路标记，其中，所述链路标记用于指示采用电交换链路发送或是采用光交换链路发送。
PCT/CN2015/079829 2015-05-26 2015-05-26 一种光电混合网络的数据传输方法及装置 WO2016187813A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
PCT/CN2015/079829 WO2016187813A1 (zh) 2015-05-26 2015-05-26 一种光电混合网络的数据传输方法及装置
CN201580001273.0A CN106664290B (zh) 2015-05-26 2015-05-26 一种光电混合网络的数据传输方法及装置
US15/821,620 US10868767B2 (en) 2015-05-26 2017-11-22 Data transmission method and apparatus in optoelectronic hybrid network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2015/079829 WO2016187813A1 (zh) 2015-05-26 2015-05-26 一种光电混合网络的数据传输方法及装置

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US15/821,620 Continuation US10868767B2 (en) 2015-05-26 2017-11-22 Data transmission method and apparatus in optoelectronic hybrid network

Publications (1)

Publication Number Publication Date
WO2016187813A1 true WO2016187813A1 (zh) 2016-12-01

Family

ID=57392306

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/079829 WO2016187813A1 (zh) 2015-05-26 2015-05-26 一种光电混合网络的数据传输方法及装置

Country Status (3)

Country Link
US (1) US10868767B2 (zh)
CN (1) CN106664290B (zh)
WO (1) WO2016187813A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111885038A (zh) * 2020-07-16 2020-11-03 深圳思凯微电子有限公司 数据的封装方法、装置、系统及计算机可读存储介质

Families Citing this family (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105141603B (zh) * 2015-08-18 2018-10-19 北京百度网讯科技有限公司 通信数据传输方法及系统
US10733137B2 (en) * 2017-04-25 2020-08-04 Samsung Electronics Co., Ltd. Low latency direct access block storage in NVME-of ethernet SSD
US10659376B2 (en) * 2017-05-18 2020-05-19 International Business Machines Corporation Throttling backbone computing regarding completion operations
US11095961B2 (en) * 2019-07-15 2021-08-17 Viscore Technology Inc. Remote data multicasting and remote direct memory access over optical fabrics
CN110750486A (zh) * 2019-09-24 2020-02-04 支付宝(杭州)信息技术有限公司 Rdma数据流控方法、系统、电子设备及可读存储介质
CN113076280B (zh) * 2019-12-18 2024-03-01 华为技术有限公司 一种数据传输方法及相关设备
US12047114B2 (en) 2020-02-17 2024-07-23 National Institute Of Advanced Industrial Science And Technology Communication control method and communication control apparatus
CN111314731A (zh) * 2020-02-20 2020-06-19 上海交通大学 面向视频文件大数据的rdma混合传输方法、系统及介质
CN113452608A (zh) * 2020-03-25 2021-09-28 阿里巴巴集团控股有限公司 数据传输方法、装置、电子设备及计算机可读介质
CN112187760B (zh) * 2020-09-22 2022-11-08 宏图智能物流股份有限公司 一种基于数据拆分的网络请求防篡改方法
US11467992B1 (en) 2020-09-24 2022-10-11 Amazon Technologies, Inc. Memory access operation in distributed computing system
US11409685B1 (en) 2020-09-24 2022-08-09 Amazon Technologies, Inc. Data synchronization operation at distributed computing system
US11354258B1 (en) * 2020-09-30 2022-06-07 Amazon Technologies, Inc. Control plane operation at distributed computing system
CN112653634A (zh) * 2020-12-10 2021-04-13 苏州浪潮智能科技有限公司 一种流量控制方法、装置、设备及可读存储介质
CN112600826B (zh) * 2020-12-10 2022-02-22 中国科学院深圳先进技术研究院 一种虚拟化安全网关系统
US20230066835A1 (en) * 2021-08-27 2023-03-02 Keysight Technologies, Inc. Methods, systems and computer readable media for improving remote direct memory access performance
WO2023032200A1 (ja) * 2021-09-06 2023-03-09 日本電信電話株式会社 通信システム、計算機、通信方法、および、プログラム
CN115174484A (zh) * 2022-06-16 2022-10-11 阿里巴巴（中国）有限公司 基于rdma的数据传输方法、装置、设备及存储介质
US12001352B1 (en) 2022-09-30 2024-06-04 Amazon Technologies, Inc. Transaction ordering based on target address
CN117318642B (zh) * 2023-11-29 2024-02-23 合肥博律微波技术有限公司 一种lc滤波器金属外壳封装设备
CN118075135B (zh) * 2024-04-18 2024-06-25 清华大学 光电混合数据中心网络的构建方法、装置及介质

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140019574A1 (en) * 2012-07-12 2014-01-16 International Business Machines Corp. Remote Direct Memory Access Socket Aggregation
CN103636181A (zh) * 2011-06-29 2014-03-12 微软公司 通过远程直接存储器访问的任意大小的传输操作
CN103889054A (zh) * 2014-04-11 2014-06-25 中国人民解放军信息工程大学 一种ap、移动终端以及移动终端定位系统和方法

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0330546A (ja) * 1989-06-27 1991-02-08 Nec Corp パケットスイッチ
CN103685033B (zh) 2013-12-19 2017-01-04 武汉邮电科学研究院 Sdn架构中支持分组交换和电路交换的通用流表及方法
US9313147B1 (en) * 2014-01-27 2016-04-12 Google Inc. Cutting and adding links within a switch
CN103997458B (zh) 2014-03-18 2017-04-12 华信咨询设计研究院有限公司 基于光电混合交叉的节点成本最优路由选择方法
US9892075B2 (en) * 2015-12-10 2018-02-13 Cisco Technology, Inc. Policy driven storage in a microserver computing environment
US10187292B2 (en) * 2016-04-15 2019-01-22 Microsoft Technology Licensing, Llc Data center topology having multiple classes of reliability

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103636181A (zh) * 2011-06-29 2014-03-12 微软公司 通过远程直接存储器访问的任意大小的传输操作
US20140019574A1 (en) * 2012-07-12 2014-01-16 International Business Machines Corp. Remote Direct Memory Access Socket Aggregation
CN103889054A (zh) * 2014-04-11 2014-06-25 中国人民解放军信息工程大学 一种ap、移动终端以及移动终端定位系统和方法

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111885038A (zh) * 2020-07-16 2020-11-03 深圳思凯微电子有限公司 数据的封装方法、装置、系统及计算机可读存储介质
WO2022012210A1 (zh) * 2020-07-16 2022-01-20 深圳思凯微电子有限公司 数据的封装方法、装置、系统及计算机可读存储介质
CN111885038B (zh) * 2020-07-16 2022-06-10 深圳思凯微电子有限公司 数据的封装方法、装置、系统及计算机可读存储介质

Also Published As

Publication number Publication date
US20180102978A1 (en) 2018-04-12
CN106664290A (zh) 2017-05-10
US10868767B2 (en) 2020-12-15
CN106664290B (zh) 2019-12-06

Similar Documents

Publication Publication Date Title
WO2016187813A1 (zh) 一种光电混合网络的数据传输方法及装置
US20220311544A1 (en) System and method for facilitating efficient packet forwarding in a network interface controller (nic)
US10880204B1 (en) Low latency access for storage using multiple paths
US11949576B2 (en) Technologies for out-of-order network packet management and selective data flow splitting
US11909642B2 (en) Offload of acknowledgements to a network device
US8953631B2 (en) Interruption, at least in part, of frame transmission
CN106034084B (zh) 一种数据传输方法及装置
EP3298739B1 (en) Lightweight transport protocol
CN113490927B (zh) 具有硬件集成和乱序放置的rdma输送
WO2016210210A1 (en) In-line network accelerator
WO2017028399A1 (zh) 通信数据传输方法及系统
CN113347017B (zh) 一种网络通信的方法、装置、网络节点设备及混合网络
US9203895B1 (en) System and method for lateral connection between interface devices with bypass of external network
WO2015055008A1 (zh) 一种存储控制芯片及磁盘报文传输方法
CN115396372B (zh) 数据流的速率控制方法、智能网卡、云端设备及存储介质
CN105512075A (zh) 高速输出、输入接口电路及数据传输方法
EP4393131A1 (en) System for storage of received messages
CN103701717A (zh) 云基站用户数据处理方法、装置和系统
CN116266800A (zh) 组播传输方法、装置和系统
US12093571B1 (en) Accelerating request/response protocols
KR102716942B1 (ko) 터널 Ctrl 쓰레드를 이용한 데이터 통신에서의 터널 데이터 업데이트 처리방법
Wang et al. An Optimized RDMA QP Communication Mechanism for Hyperscale AI Infrastructure
CN118714183A (zh) 报文传输的方法和装置
CN118869775A (zh) 报文传输的方法和装置
CN118200253A (zh) 面向rdma ud传输的可靠通信方法、电子设备及可读介质

Legal Events

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

Ref document number: 15892894

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 15892894

Country of ref document: EP

Kind code of ref document: A1