CN101483662B - Packet forwarding device and method for virtual storage network switch - Google Patents

Abstract

A packet forwarding device and method for virtual storage network switch can be applied to switch environment established by iSCSI connection. The packet forwarding apparatus includes a header extractor, a distributor, and a forwarding unit. When the iSCSI conference connection finishes system authentication, the header extractor receives at least one packet on the iSCSI connection and extracts header data of the at least one packet; the distributor determines the flow direction of the received packet; the forwarding unit performs direct packet forwarding between the front end and the back end of the established connection, including an operation of converting a virtual address received by the distributor into a real address, an operation of establishing a corresponding relationship between the client interface and the storage interface, and an operation of transferring load data carried in the connection.

Description

Packet forwarding device and method for virtual storage network switch

technical field

The present invention relates to a packet forwarding device and method of a virtual storage network switch (virtualization switch), which can be applied to a switch established by an Internet Small Computer System Interface (iSCSI) connection Environment.

Background technique

The networking of storage devices is one of the trends in the development of the Internet, and the iSCSI transport protocol is an emerging network storage technology standard. Storage Area Network (SAN) refers specifically to the network connecting storage devices and servers, and uses the network to connect storage devices in series. Compared with redundant array of independent disks (Redundant Array of Independent Disks, RAID) and small computer system interface (Small Computer System Interface, SCSI) and other technologies that directly connect servers and storage devices, network storage technology not only has high-speed transmission capabilities, The expansion and backup of the storage device will not be restricted by the server itself, so the scalability and reliability are good.

Due to the extremely high bandwidth requirements, fiber channel (fiber channel1) is currently a commonly used storage network connection technology. Another emerging storage network technology is the iSCSI transmission protocol that has just completed standard formulation. The main function of the iSCSI transmission protocol is to transmit SCSI commands and data on the IP network through the iSCSI protocol data unit (Protocal Data Unit, PDU). The iSCSI protocol makes the IP network also a storage network. For the client, using the IP network as the storage network only needs to replace its servers and storage devices with iSCSI interfaces, and use the original IP network switching devices to build a storage network.

FIG. 1 is a schematic diagram of a general implementation example of an iSCSI virtual storage network. As shown in FIG. 1 , a client 102 installs a driver of an iSCSI initiator 110 and connects to a virtualization server 101 . The storage terminal 103 installs the driver program of the iSCSI target (target) 120 and connects with the virtual server 101 . In addition to installing iSCSI initiator and target drivers on the virtual server 101, a virtualization software layer 140 is added.

Based on the setting of the virtualization software layer 140, the change of the field of the iSCSI packet is observed from the perspective of the iSCSI packet data flow. At the beginning, the iSCSI target (target) 120 of the switch in the virtual server 103 receives the iSCSI packet and unpacks the packet, passes it to the SCSI layer 130 to read the command, calls the virtualization software layer 140 to convert the real access location, and generates SCSI Access command format, and finally encapsulate the SCSI command into an iSCSI packet and send it out. During this series of conversions, iSCSI removes the header from the initial input packet format, extracts the SCSI command (CDB) of the payload, accesses the SCSI layer, and goes through virtualization. The software layer maps (mapping) to find the actual access location of the data, regenerates the SCSI command, and finally adds the iSCSI header to transmit.

From the input (input) iSCSI packet to the final output (output) iSCSI packet, in appearance, both are iSCSI packets of the same protocol, but from the inside of the virtualization switch (virtualization switch), it needs to go through multiple layers With the help of professional software, this virtualized network storage based on iSCSI connection architecture can be completed. Finally, the iSCSI packet to be sent also undergoes redundant data movement operations repeatedly, including dismantling at the beginning, and returning to generating at the end.

Among the existing technologies, virtualized network storage technologies that use the iSCSI protocol as a connection method can be classified into the following categories.

The first type is shown in FIG. 2 . From the perspective of the relationship between the iSCSI switch 201 and the iSCSI user host (host) end 202, the agent mechanism (agent) for managing storage configuration is scattered outside the data path (datapath). The data can directly reach the network storage device 203 without going through the switch of the storage server. After the data leaves the user end, it is immediately transmitted to the storage device 203, and there is no redundant protocol processing and data duplication on the data path. However, an iSCSI agent mechanism (agent) needs to be installed on the user side and the storage device side, as indicated by reference numerals 210 and 220, to exchange access information with the storage server.

For example, in the document of US Pat. No. 6,845,403, the access packets of the iSCSI protocol are divided into control type and data type. Control packets are sent from the user end to an access management server, and the server then transmits the command to the real storage device, while the data type directly transmits data through the pre-established iSCSI connection between the user end and the real storage device end. This allows for the most direct data transfer path. However, a special iSCSI driver must be installed on the user end to have functional interaction with the management server, and when the number of storage devices increases, the iSCSI connection between the user end and the storage device also increases accordingly.

The second type is shown in FIG. 3 . The user host 302 and the storage 303 need an iSCSI proxy mechanism, as shown in labels 320 and 310, to assist, and the data is transferred to the storage through a switch router (switchrouter) 301. ). The router decides to send the data packet to the set storage terminal 303 by means of a simple and fast look-up table at the IP layer 311 . Therefore, the data path only needs to switch the IP location. However, the user end needs to install the iSCSI proxy mechanism 320, and the synchronization of access information is also difficult.

For example, in the literature of U.S. Patent No. 6,934,799, when the SCSI command is issued, the user end searches the unique virtual IP correspondence table at the user end to find out the IP and TCP location of the iSCSI packet transmission, and the packet is sent to the conversion network. Between the connector (virtualizationgateway), and then send the packet to the real network access location. The access load (1oading) is distributed between the user end and the real access location, and each packet transmission on the data path is reserved, and only the IP location is replaced. In this way, when changing the physical storage device, it is only necessary to change the correspondence between the virtual IP and the physical IP on the data network connector (virtualization gateway), and the transmitted packet only needs to change the fields of the IP and TCP positions on the data path, There is no redundant copying of the data packets.

However, the client needs to implement an access control module on the upper layer of the iSCSI driver to join the virtual access system of the gateway. The gateway manages the operation between virtual and real hard disks easily, but the process of modifying the corresponding relationship between the block range and the virtual TCP field on the access table of each client is complicated, and it is difficult to achieve unified management.

The third type is shown in FIG. 4 , the iSCSI switch 401 is interposed between the data path between the user end 402 and the real storage device 403, and completes the design of virtualized network storage in a stand alone mode, and the user end The data can be accessed through the switch 401 by installing the iSCSI driver (driver). The client 402 only needs to have a common iSCSI initiator (Initiator) 410 to access the network storage provided by the switch 401 . However, the data packets need to go through multi-layer protocol processing and data duplication during the transfer process of the switch 401, thus reducing the access performance.

For example, in the document of US Patent Publication No. US2005/0114464, a general iSCSI protocol plus a virtualization software layer is used to achieve a virtualized storage network. Between the received virtual location access and the real storage device, find out the corresponding relationship between virtual and actual access, optimize the data path, and connect two different transmission protocols between the user end and the storage network end, A protocol conversion is achieved. However, in the process of protocol conversion, no special handling method is disclosed for the buffering of transmitted data. The general data buffering method will increase redundant data copy operations and reduce access performance.

Existing technologies mainly implement virtualization in two ways: in-band and out-of-band. In the out-of-band mode, the virtual device is located outside the data path, which can provide a high-performance data transmission path between the client and the access device. However, when the configuration of the access device is changed, the communication between the update of the virtual device and the user will be time-consuming, and the client also needs to install a software interface to communicate with the virtual device, which is not suitable for direct access to the storage device. convenience. In the in-band method, the virtual device is placed in the data access path between the user and the storage device. Although the storage device group is changed, the management is more convenient, and it can provide the convenience of direct access to the storage device by the client. , however, the amount of data such as the client's data query to the virtual device will share the same bandwidth with the real data transmission path, which will reduce the data transmission efficiency.

In the above two methods, the in-band method is used to achieve virtualized access. The storage server on the data path usually completes the virtual and physical access address translation work alone, while the out-of-band method uses the virtual access work A proxy mechanism that distributes data between storage servers and clients.

Contents of the invention

In the implementation example of the present disclosure, a packet forwarding device and method in a virtual storage network switch can be provided, which can be applied to a switch environment established by an iSCSI connection. In the implementation example of this disclosure, the two ends of the iSCSI protocol connection, that is, the client and the storage end, use the iSCSI conference (session) connection lookup table corresponding method, and the packet is directly forwarded between the two ends of the conference (session) connection , to complete the data exchange between the client and the storage.

In an embodiment example, the present disclosure relates to a packet forwarding device in a virtual storage network switch. In the virtual storage network switch, a target module at the front end establishes a connection with at least one user end, and a startup at the back end The initiator module establishes a connection with at least one network storage terminal, and the packet forwarding device may include: a header extractor (HeaderExtractor), a distributor (Dispatcher), and a forwarding unit (Fordwarding Unit), the header extractor fetcher. After the iSCSIsession connection completes the system authentication, the header extractor takes over at least one packet on the iSCSI connection, and retrieves the header data of the at least one packet; the distributor determines the flow of these taken over packets; The action of directly forwarding packets between the front end and the back end of the established connection, including the action of converting the virtual address received by the distributor into a real address, establishing the client interface (client-sideinterface) and storage terminal (storage-sideinterface) ), and the action of transferring the load data carried on the connection.

In another embodiment, the present disclosure relates to a packet forwarding method of a virtual storage network switch, and the method may include: in at least one iSCSI packet received from a TCP connection used by an iSCSI session (session), only Extract the header of each iSCSI packet; for each iSCSI packet, check whether the format of the packet is the type of qualified forwarding; for the qualified forwarded packet, check whether it is the packet of the iSCSI command (Cmnd) type; if so, send this The corresponding virtual access address information of the packet of the iSCSI command (Cmnd) type is converted into a real address, and a new correspondence is added in the dynamic conference connection comparison table (DSMT); from the dynamic conference connection comparison table (DSMT) ), find out the corresponding destination session ID (session ID) and initiator task tag (ITT) fields; modify the iSCSI packet header according to the connection status of the destination session, And transmit the modified iSCSI header and the iSCSI payload data of the source TCP connection.

The packet forwarding mechanism disclosed in this disclosure can be implemented not only under the operating system with TCP protocol function, but also in the system with multi-processor architecture.

Description of drawings

In conjunction with the detailed description of the following drawings and embodiments, the above and other purposes and advantages of the present invention are described in detail below, wherein:

FIG. 1 is a schematic diagram of a typical implementation of an iSCSI virtual storage network.

FIG. 2 is a schematic diagram illustrating another technique for achieving virtualized network storage.

FIG. 3 is a schematic diagram illustrating another technique for achieving virtualized network storage.

FIG. 4 is a schematic diagram illustrating another technique for achieving virtualized network storage.

FIG. 5 is a schematic diagram of a process flow of direct transfer of iSCSI PDU packets from a client to a storage, and is consistent with certain embodiments of the present disclosure.

FIG. 6 is an exemplary schematic diagram illustrating the structure of a packet forwarding device of a virtual storage network switch, which is consistent with certain embodiments of the present disclosure.

FIG. 7 is an example schematic diagram illustrating the detailed actions of each module in the packet forwarding device, which is consistent with some implementation examples in the present disclosure.

FIG. 8 is a schematic diagram of an example of the content and establishment process of each element in the dynamic conference connection comparison table, which is consistent with some implementation examples in the present disclosure.

FIG. 9 illustrates an example schematic diagram of a virtual address access request corresponding to real storage locations for multiple different devices, and is consistent with certain embodiments of the present disclosure.

FIG. 10 illustrates an example schematic diagram in which an indicator bit is set behind each element in the conference connection table, and is consistent with some implementation examples in the present disclosure.

FIG. 11 illustrates that the PDU duplication takeover module plays the role of packet remanufacturing and integration in the flow direction of the packet, and is consistent with some implementation examples in this disclosure.

FIG. 12 is an exemplary flow chart illustrating how the PDU copy takeover module implements when the client sends a READ command, which is consistent with some implementation examples in this disclosure.

FIG. 13 is an exemplary flow chart illustrating how the PDU copy takeover module implements when the client sends a WRITE command, which is consistent with some implementation examples in this disclosure.

FIG. 14 is an example schematic diagram further illustrating the parameters of the input and output of the address converter interface, consistent with certain embodiments of the present disclosure.

FIG. 15 is a schematic flowchart illustrating a method of a packet forwarding device of a virtual storage network switch, consistent with certain embodiments of the present disclosure.

FIG. 16A is a schematic diagram of an example of the relationship among iSCSI, packet forwarding mechanisms, and core units on a single processor system.

FIG. 16B is a schematic diagram of an example of the relationship among iSCSI, packet forwarding mechanisms, and core units on a multiprocessor system.

FIG. 17 is a schematic diagram of an example in which several modules of the packet forwarding device are distributed and executed on independent microprocessors in a multi-processor system.

Detailed ways

In the implementation example of this disclosure, in a network protocol, such as iSCSI network protocol, in the switch environment established by the connection, a bridge for direct forwarding of packets is provided, and the processing flow of iSCSI packets from the user end to the storage end There is no need to go through complicated actions such as dismantling, reading, forwarding, and creating headers of redundant packets. FIG. 5 is a schematic diagram of a process flow of direct transfer of iSCSI PDU packets from a client to a storage, and is consistent with certain embodiments of the present disclosure.

Referring to FIG. 5 , as indicated by reference number 510 , in the processing flow from the client 502 to the storage 503 , the iSCSI PDU only receives the packet header for processing. Then, as shown by reference numeral 520, the corresponding iSCSI connection information is found through the direct forwarding mechanism. As shown by reference numeral 30 , the packet header is directly sent to the storage end 503 . The iSCSI protocol data unit sent back from the storage terminal 503 to the client 502 is the same as after receiving the header, as shown in the label 511, finds the connection of the client 502 that requires this data through the direct forwarding 520 mechanism, and then returns the packet header to The client 502 is shown by reference number 531 .

In other words, the present disclosure provides a connection bridge between the user end and the network storage device end. The connection object is the iSCSI session connection using the iSCSI network protocol as the connection method. The connection method is to generate a one-to-one connection relationship between the user and the two ends of the storage device in a table-building manner.

After the corresponding relationship is established, the headers of all iSCSI protocol packets on the connection are modified by fields, together with the entrained data, are directly forwarded to the network storage device, and the response data from the network storage device is the same Look up the table to find the iSCSI connection corresponding to the client, and directly send the packet to the client. Combined with the existing software layer approach of the virtualization switch, the implementation example disclosed in this disclosure can still provide the virtualized storage effect for the user during the packet forwarding process. Moreover, after the packet direct forwarding unit (Forwarding Unit) disclosed in the present disclosure is independent, better performance can be obtained on a platform with multiple processors.

FIG. 6 is an exemplary schematic diagram illustrating the structure of a packet forwarding device of a virtualization switch (virtualization switch), which is consistent with certain embodiments of the present disclosure. Referring to FIG. 6 , in the environment of this virtualized switch, a target module (target module) at the front end, such as an iSCSI target 620, can establish an iSCSI session (session) connection with at least one user end 602, and a back end An initiator module, such as an iSCSI initiator 610 , can establish an iSCSI session connection with at least one network storage terminal 603 . The exemplary structure 600 of the packet forwarding device of the present disclosure includes a header extractor 661 , a distributor 662 , and a forwarding unit 663 .

After the iSCSI session connection completes the system authentication, the header extractor 661 receives at least one packet on the iSCSI connection, and retrieves the header data of the at least one packet. Distributor 662 determines the flow direction of these taken over packets. The transfer unit 663 directly transfers packets between the front end and the back end of the two established iSCSI connections, including the action of converting the virtual access address of the packet received by the distributor 662 into a real address, establishing the client interface and storage end The corresponding relationship of the interface, such as establishing the corresponding relationship in a dynamic session connection comparison table (DynamicSession Mapping Table, DSMT), and the action of transmitting the iSCSI payload data (payload) carried on the iSCSI connection.

The header extractor 661 can perform data control on multiple iSCSI packets on the connection, and separate the iSCSI header from the data, so that only the data processing of the header field in the protocol data unit in the entire packet transfer process does not involve The data movement operation of the data field in the protocol data unit.

The transfer unit 663 may further include an address translator (Address Translator, AT), a dynamic conference connection comparison table (DSMT), a PDU duplication takeover module (PDU DuplicationHandler, PDH), and a header adapter (Header Adaptor) , and a data mover (Data Mover). The exemplary structure 600 of the packet forwarding device may also include a descriptor queue. FIG. 7 illustrates the detailed actions of these modules in the packet forwarding device, and is consistent with some embodiments of the present disclosure.

Referring to FIG. 7, after the header extractor 661 completes the authentication on the iSCSI connection system, it can find out the information of the connection ID (denoted as Conn) and the virtual access ID (denoted as VIun) from the connection information , combine these data into an access descriptor (descriptor), record it in the descriptor queue 711, and wait for the next processing procedure to proceed. The description sub-queue 711 is to separate the job of receiving headers from the job of analyzing headers, so that the two jobs can be performed at the same time, and can solve the competing problem that multiple connections need to be processed by the distributor 662 at the same time.

When the descriptor queue 711 has data to enter, the distributor 662 will take out an access descriptor from the descriptor queue 711, and the type of the received packet can be known from the descriptor information, such as iSCSI command (iSCSI Cmnd), iSCSI Data output (iSCSI DataOut), iSCSI data input (iSCSI DataIn), iSCSI reply (iSCSIRsp), iSCSI ready to transmit (iSCSI R2T) and other packet types, and determine whether the received packet is a qualified transfer type packet. If it is a command (Cmnd) type packet, the distributor 662 will send the data header of the packet to the next unit for further processing. If it is a packet other than the command (Cmnd) type, the distributor 662 will send the header and the entrained data of the packet back to the iSCSITarget on the system for processing; the distributor 662 can determine whether the packet is directly delivered to the transfer unit 663, let the transfer unit 663 directly transfer to the network storage end 603.

The function of the address translator 763a is to translate between virtual access and real location. After the address converter 763a receives the virtual access point (Vlun), the access address (block) and the access length (1ength), it can find out the real access location and the session to which the access location belongs. connection. That is to say, the address converter 763a can find out the corresponding relationship between the real access address and the session connection ID from the virtual access point and the access location.

In the iSCSI packet received from the distributor 662, after the access address is converted by the address converter 763a, the session connection ID of the real access location and the number of real storage devices accessed are obtained. The PDU Duplicate Handover Module (PDH) 763b determines the corresponding relationship between the source packet and the forwarded packet, and establishes the corresponding relationship in the Dynamic Conference Connection Map (DSMT) 763c.

The PDU duplication takeover module (PDH) 763b fills in the content of the corresponding entry in the dynamic conference connection table (DSMT) 763c. The content of each element in the dynamic conference connection table (DSMT) 763c will be illustrated in FIG. 7 . After the corresponding relationship is filled in the dynamic meeting connection table (DSMT) 763c, the PDU copy and takeover module (PDH) 763b transmits the header of the packet to the header adapter 763d. The header adapter 763d handles field modification and transmission of the iSCSI header. If the header has entrained payload data (payload), then use the data mover 763e to directly treat the TCP layer at the receiving end as transmission data, and send it out on the TCP connection at the storage end.

The PDU copy-and-handover module (PDH) 763b can solve the one-to-many correspondence between the virtual and the physical, allowing the user to expand the access of multiple real objects when triggering the access of a virtual object. The header adapter 763d can learn the replaced value of the iSCSI header field from the Dynamic Session Connection Map (DSMT) 763c, and then send the header over the TCP connection to which the iSCSI session (session) belongs. The data mover 763e can directly transmit the iSCSI payload data staying on the TCP connection as the data transmitted by iSCSI.

In this way, the user's access rights to the target and the settings of the virtual storage device are centralized on the switch for management, so that the user can install the iSCSI initiator link software that is generally easy to obtain, Directly treat the switch as an iSCSI target (target) for access. The real storage device connected by the iSCSI protocol on the switch can be configured in a virtualized manner and provided to the user for access through account and password confirmation. Use There is no need to install any additional access control software on the reader side. The reception of data is reserved at the lower network layer, and the location of data transmission is directly determined by the data mover, reducing redundant data duplication times.

FIG. 8 is a schematic diagram of an example of the content and establishment process of each element in the dynamic conference connection table (DSMT), which is consistent with some implementation examples in the present disclosure. Referring to Fig. 8, each element in the dynamic conference connection comparison table (DSMT) 763c mainly has three fields, that is, a client side interface (Client Side Interface) 831, a storage side interface (Storage Side Interface) 832, and a valid bit 833 , record the conference connection (sess) ID (recorded as src_sess) and the initiator tasktag (initiator tasktag) (denoted as src_ITT) of the user end, the conference connection (session) ID (denoted as dst_sess) and the start The envoy's work tag (denoted as dst_ITT) and the access sequence relationship corresponding to the same virtual access. The valid bit is convenient for the PDU Duplication Handling Module (PDH) 763b to manage the iSCSI packets sent back from the storage end.

The life cycle of each element in the dynamic meeting connection table (DSMT) 763c begins when the transfer unit 663 receives the iSCSI command protocol data unit (Cmd PDU) of the client, and ends when the transfer unit 663 receives the iSCSI response protocol data unit of the storage terminal (Rsp PDU). During the existence of the element, protocol data units (PDUs) received from the client can be sent to the target iSCSI storage device in a one-to-one or one-to-many session. The data packet sent back from the storage device will also look up the table through the forwarding unit 663 to find out the client requesting the data.

In the process of creating an element, the first iSCSI command protocol data unit (Cmd PDU) from the client to the conversion unit can know the session (sess) column dst_sess and the ITT column dst_ITT of the storage terminal through the assistance of the address translator 763a At the same time, record whether this correspondence is a one-to-many conference (session) correspondence in the effective bit field of the element. From then on, the protocol data unit (PDU) of the same iSCSI task (Task) enters the transfer unit 663, the meeting (session) and ITT field of the protocol data unit (PDU) will be used to search whether there is a corresponding element in the dynamic conference connection comparison table (DSMT) 763c, and the meeting (session) ID and ITT field of the destination end will be found. Bit, until the last iSCSI Response (Rsp) packet is sent back from the storage end, the corresponding relationship of this element will not end.

When the forwarding unit 663 receives an iSCSI command protocol data unit (Cmd PDU) from the client, after accessing it through the address translator 763a, it is known that the data volume of the virtual access is dispersed on multiple physical network storage devices, In addition to directing the packets, the forwarding unit 663 also needs to transmit the accessed iSCSI packets to a plurality of network storage devices covered by the access command.

For example, assume that a virtual address (Request Virtual Address, Request VA) 910 of a user’s request for access corresponds to a real storage location (Physical Address, PA) that is PA1 and PA2 of two different storage devices, as shown in FIG. 9 Show. When the forwarding unit 663 learns from the address translator 763a that the access target is two distributed network devices, the PDU duplication-handover module (PDH) 763b in the forwarding unit 663 will follow the target terminal number provided by the address translator 763a For the purpose information, a plurality of elements are established in the dynamic conference connection comparison table (DSMT) 763c, and each element represents the workload of the work (Task) that the transfer unit 663 needs to complete independently with the storage terminal.

As shown in Figure 10, when the PDU copy and takeover module (PDH) 763b creates elements, it will mark the valid bit (Vbit) behind each element, so that the iSCSI packet returned from the storage end can know whether there are many The last task access in a multi-task access. For example, V represents not the last working access and X represents the last working access.

FIG. 11 further illustrates that the PDU Duplicate Handover Module (PDH) not only plays the role of packet remanufacturing but also plays the role of packet integration in the flow direction of the packet. As shown in Figure 11, for example, when the user of the client sends a virtual access request, and the physical access data volume (block) is A+B packets, wherein the data volume A is on the target (Target) A Data, the data amount B is the data on the target B. At this time, the transfer unit 663 will generate individual access packets for the target A and target B at the storage end, and the iSCSI data protocol data unit (DataPDU) sent back from the storage end After modifying the lookup table, it will be sent back to the client user, but the iSCSI response protocol data unit (Rsp PDU) will not be sent back to the client user until the last target data access involved in this access is completed. Afterwards (known by the valid bit Vbit of the element of the dynamic conference connection comparison table (DSMT) 763c), the PDU replication takeover module (PDH) 763b just returns an iSCSI response protocol data unit (Rsp PDU) to the use of the client By. From the virtual access packet request sent by the client user, through the returned data packet, to the final received response protocol data unit, the client user does not feel the same as a storage target (StorageTarget) at all. ) for any difference in response. Therefore, the PDU Duplication Handling Module (PDH) 763b in the forwarding unit 663 can actually play the role of packet remaking and integration.

The exemplary flow charts in FIG. 12 and FIG. 13 respectively illustrate how the PDU duplication handler module (PDH) 763b implements when the client sends a READ/WRITE command request, and are consistent with some implementation examples in this disclosure.

When the client sends the request of the READ command, the iSCSI packet received by the forwarding unit 663 passes through the address converter 763a, and knows how many real read objects the iSCSI packet has, and the PDU copy takeover module (PDH) 763b is in the dynamic session The corresponding number of elements is established in the connection look-up table (DSMT) 763c. From the descriptor converted by the address translator 763a, it can be known that the data is scattered in the storage end. Then as shown in FIG. 12, in step 1201, the PDU duplication takeover module (PDH) 763b generates an iSCSI packet for data reading to the first storage terminal, and waits for the storage terminal to respond to iSCSI Data InPDU. After receiving the response iSCSI Data In PDU, modify the field and directly import it back to the client, as shown in step 1202. Repeat receiving the DataIn PDU and returning the user action until the storage end responds to the iSCSIRsp PDU, and end the data request to the storage end, as shown in step 1203. Then generate the data request of the next storage terminal, repeat the above receiving steps, and complete the receiving items corresponding to each element in the dynamic meeting connection table (DSMT) 763c, as shown in step 1204 . The iSCSI Rsp PDU is not sent back to the user until the valid bit field in the corresponding element shows the last correspondence of the storage terminal, as shown in step 1205.

When the client sends a WRITE command request, the action of the PDU copy takeover module (PDH) 763b is similar to the implementation of READ in Figure 12, the difference is that after the user sends the WRITE action, the user must wait for the storage terminal to respond and prepare to transmit ( Ready to Transfer (R2T) iSCSI packet, the PDU copy takeover module (PDH) 763b can start to send the iSCSI Data Out PDU to the storage device. The detailed implementation actions of the PDU copy takeover module (PDH) 763b are shown in FIG. 13 .

After the PDU copy and takeover module (PDH) 763b receives the command request across multiple storage devices WRITE from the user end, it fills in several numbers in the dynamic conference connection table (DSMT) 763c through the instruction of the address converter 763a. Items at the storage end, as shown in step 1301. Then send out the data request to the first storage terminal and wait to receive the iSCSI R2T PDU sent back, and guide it back to the user terminal, as shown in step 1302. The forwarding unit 663 starts to receive the iSCSI DataOut PDU, and the PDU copy and takeover module (PDH) 763b guides the received iSCSI packets to the real storage device one by one according to the storage items listed in the dynamic conference connection comparison table (DSMT) 763c, as in step 1303 shown. Until the storage devices involved in the WRITE command are all written, the PDU duplication and takeover module (PDH) 763b responds with an iSCSIRspPDU to the user end, as shown in step 1304 .

FIG. 14 is an example schematic diagram further illustrating the parameters of the input and output of the address converter interface, consistent with certain embodiments of the present disclosure. In this example, the address converter 763a can convert from the input virtual access information, such as information such as virtual access point (virtual disk ID), virtual access address (virtual block address) and access data length (access length). Output the number of physical storage devices (physical disknumber) corresponding to this virtual access, and the physical access point (physical disk ID), physical access address (physical block address), and data access length of each corresponding physical storage device (access length), the iSCSI session connection (iSCSI session) ID established by the physical storage device, etc. If the number of corresponding physical storage devices is more than one, there will be multiple groups of the same physical access parameters.

Based on the above, in the virtual storage network switch, the processing flow of transferring packets, such as iSCSI packets, from the user end to the network storage end can be illustrated by the example flow shown in FIG. 15 , and is consistent with some implementation examples in this disclosure. In the example shown in Figure 15, the network connection method and packet forwarding are illustrated by the TCP connection and iSCSI packet forwarding used by the iSCSI session respectively.

Referring to FIG. 15 , as shown in step 1510 , the header of each iSCSI packet is extracted from at least one iSCSI packet received in the TCP connection used by the iSCSI session. As shown in step 1520, for each iSCSI packet, check whether the format of the packet is a qualified forwarding type, such as packet types such as iSCSI Cmnd, iSCSI DataOut, iSCSI DataIn, iSCSI Rsp, iSCSI R2T, to determine the flow direction of the packet. As shown in step 1530, for the packets that are qualified to be forwarded, check whether they are packets of the iSCSI command (Cmnd) type. If so, as shown in step 1540, convert the virtual access address information corresponding to the packet of the iSCSI command (Cmnd) type into a real address, and add a new corresponding relationship in the dynamic conference connection table (DSMT) . For example, the corresponding virtual access address information (such as access point, address, length, etc.) of this type of packet can be known from the information in the TCP connection used by the iSCSI session, and the virtual access address information can be converted Into the real access location, and add the dynamic conference connection table (DSMT) as shown in the eighth figure. Then proceed to step 1550 .

If it is a packet of other qualified transfer types other than iSCSI Cmnd, as shown in step 1550, find out the corresponding destination conference (session) ID and ITT field from the dynamic conference connection table (DSMT). For example, from the dynamic conference connection table (DSMT), the session ID and ITT field of the source packet can be used to find out the corresponding destination session (session) ID and ITT field.

In step 1560, modify the iSCSI packet header according to the connection state of the destination session, and transmit the modified iSCSI header and the iSCSI payload of the source TCP connection. For example, the ITT field, CmdSN and other fields of the packet can be replaced according to the connection status of the destination session (session), and then the modified iSCSI packet header together with the iSCSI that stays on the source TCP connection can be replaced. Payload data, according to the TCP connection used by the destination session, send the data to the destination.

In this way, from the iSCSI packet received by the switch target (Switch Target) to the iSCSI packet to be transmitted by the switch initiator (Switch Initiator), the format of the iSCSI packet received by the target (Target) is The initiator (Initiator) will generate the same iSCSI packet format and send it to the destination, that is to say, both iSCSI headers have the same operation code (opcode), except that the destination of the received packet is different. The iSCSI session connection has a slight difference in the iSCSI header fields, and the other fields describing the data in the header will not change.

Because the data received by the switch will be completely transmitted to the destination after the virtualization software layer finds out the real network access location, so from the input (input) iSCSI packet to the output (output) iSCSI packet, In the above-mentioned implementation example of this disclosure, the iSCSI packet received from the target (Target) can be successfully replaced with the corresponding field, and an iSCSI packet suitable for the initiator (Initiator) can be generated, so many software layers can be dispensed with Between redundant packet disassembly and conversion actions.

The packet forwarding mechanism of the present disclosure can be implemented in a system with a single processor or a multi-processor architecture. FIG. 16A and FIG. 16B are schematic diagrams of examples of relationships among iSCSI, packet forwarding mechanism and kernel unit (Kerne1) on single-processor and multi-processor systems respectively.

Referring to FIG. 16A, on a single processor system, this example implements iSCSI and packet forwarding mechanisms in the host software layer. Referring to Figure 16B, on a multi-processor system, in this example, the main processor is used to operate the core unit and the iSCSI software layer, and the part of the packet forwarding mechanism is implemented by a microprocessor (MicroEngine, ME) other than the main processor. Execution, through a special communication interface (Core Component), the program executed on the microprocessor can exchange data with the program on the main processor.

In the example of a multi-processor system, one way of implementation is that several modules of the packet forwarding device in the implementation example shown in FIG. Show. Each independent microprocessor has its own local (local) variable access space, and also has a shared static random access memory (SRAM) access location, and also has its own program counter (programcounter), so that each microprocessor can operate simultaneously at the same time. A special communication channel is provided between different microprocessors, and data transmission or information notification between microprocessors can be achieved through the storage space specific to the hardware architecture. The advantage of the independence of each module is that it can be implemented on a multi-processor system architecture, allowing multiple modules to operate simultaneously, and reducing the load on the main processor, which can greatly increase the iSCSI protocol data unit (PDU) Transmission efficiency on a virtualization switch.

Therefore, the packet forwarding mechanism disclosed in the present disclosure can be implemented in various ways, such as software implementation, hardware implementation, implementation on a platform including a single processor or multiple processors, and so on.

However, what is described above is only an embodiment of the invention, and should not limit the implementation scope of the invention accordingly. That is, all equivalent changes and modifications made according to the claims of the present invention shall still fall within the scope covered by the patent of the present invention.

Claims (19)

1. the packet forwarding apparatus of a virtual storage network switch, in this virtual storage network switch, a target of front end and at least one user end are set up line, and one of rear end opens the beginning person and at least one network storage end is set up line, it is characterized in that, this packet forwarding apparatus comprises:

One header acquisition device receives at least one package on this line that builds on front end, and captures the gauge outfit data of this at least one package;

A distributor determines the flow direction of this received package; And

A transfer unit, between the front end of this built vertical line and rear end, do the action that package directly passes on, the virtual access address transition that comprises the package that this distributor reception is come becomes the action of true address, the corresponding relation of setting up client-side interface and storing end interface, and the action of transmitting the load data of carrying secretly on this line, wherein this transfer unit also comprises:

An address translator, the virtual access address transition that this distributor is received the package that comes becomes true address;

A dynamic conference line table of comparisons comprises at least one source package of a plurality of descriptions and passes on the element of corresponding relation of package that each element has client-side interface at least, stores end interface and three fields of significance bit;

A protocol Data Unit copies the adapter module, in this dynamic conference line table of comparisons, insert the content of corresponding element after, send out the gauge outfit that this distributor receives the package that comes;

A gauge outfit is adjusted device, processes the field of package gauge outfit and revises and transmit; And

A data Conveyer with resting on load data on the line and be used as the data of transmission, sends out.

2. the packet forwarding apparatus of virtual storage network switch as claimed in claim 1, it is characterized in that, the line of wherein setting up in this front end and the line of this rear end foundation are a kind of lines of world-wide web slim calculator system interface meeting of the host-host protocol that uses world-wide web slim calculator system interface, this target is a world-wide web slim calculator system interface target, and this Qi Shizhe is that a world-wide web slim calculator system interface opens the beginning.

3. the packet forwarding apparatus of virtual storage network switch as claimed in claim 1, it is characterized in that, wherein this client-side interface, store end interface and three fields of significance bit and record respectively the meeting line identification code of user's end and open beginning person work tap, store the meeting line of end and open envoy's work tap and corresponding to the access sequence relation of same virtual access.

4. the packet forwarding apparatus of virtual storage network switch as claimed in claim 1, it is characterized in that, wherein this header acquisition device produces the access descriptor of this at least one package gauge outfit, and each access descriptor comprises the information of corresponding line identification code and virtual access point.

5. the packet forwarding apparatus of virtual storage network switch as claimed in claim 1 is characterized in that, wherein this distributor is learnt the type of receiving package from the gauge outfit of this package, and determines that whether the package receive is the qualified package that passes on type.

6. the packet forwarding apparatus of virtual storage network switch as claimed in claim 1 is characterized in that, wherein this receives that the type of package is wherein a kind of packet type that order, data output, data input, answer and preparation transmit.

7. the packet forwarding apparatus of virtual storage network switch as claimed in claim 1, it is characterized in that, wherein this packet forwarding apparatus be with software realize, hardware is realized, contain the aforementioned wherein a kind of mode that realizes on single-processor or the platform of multiprocessor realizes.

8. the package forwarding method of a virtual storage network switch, in this virtual storage network switch, a target person of front end sets up line with at least one user's end, and one of rear end opens the beginning person and at least one network storage end is set up line, it is characterized in that, this package forwarding method comprises:

In at least one package that from this line that builds on front end, receives, capture the gauge outfit of each this package;

Whether the form of checking each this package is qualified type;

To the package of qualified type, check whether be the package of command type;

The words that are, convert the corresponding virtual access address information of the package of this command type to true address, and add a new corresponding relation in a dynamic conference line table of comparisons, wherein, the described dynamic conference line table of comparisons, comprise at least one source package of a plurality of descriptions and pass on the element of corresponding relation of package, each element has client-side interface at least, stores end interface and three fields of significance bit; Wherein this virtual access address information convert to true address comprise at least from the input virtual access point, virtual access address and access data length information, change out the physical storage device number that corresponding this virtual access comprises, the meeting line identification code that the entity access point of the physical storage device corresponding with each, entity access address, data access length, physical storage device are set up;

From this dynamic conference line table of comparisons, find out destination meeting identification code corresponding to each this package of qualified type and open beginning person work tap field; And

Revise the gauge outfit of each this package of qualified type, and send out amended gauge outfit and the load data that comes on the source line.

9. the package forwarding method of virtual storage network switch as claimed in claim 8, it is characterized in that, wherein this new corresponding relation comprises the meeting line identification code that the user holds and opens beginning person work tap, stores the meeting line of end and open envoy's work tap and corresponding to the information of the access sequence relation of same virtual access.

10. the package forwarding method of virtual storage network switch as claimed in claim 8 is characterized in that, wherein when the package of this command type be data read package the time, this package forwarding method also comprises:

The package that the generation data read stores end to first, and waits for that first stores the end response data and inputs protocol Data Unit;

After receiving this response data input protocol Data Unit, directly lead back corresponding user's end behind the modification field;

The repetition receive data inputs protocol Data Unit and this leads back the action of user's end until first stores the answer protocol Data Unit of end, and end is to the data demand of first storage end;

Produce the next data demand that stores end, repeat above-mentioned receiving step, finish reception project corresponding to each element in the dynamic conference line table of comparisons; And

Until the significance bit field demonstrates the correspondence that finishing touch stores end in the corresponding element, the answer protocol Data Unit that just this finishing touch is stored end returns to this corresponding user end.

11. the package forwarding method of virtual storage network switch as claimed in claim 8 is characterized in that, wherein when the package of this command type be data write package the time, this package forwarding method also comprises:

By address transition, in this dynamic conference line table of comparisons, fill out the project that at least one stores end;

Send the preparation transport protocol data unit that first is stored the data demand of end and waits for the reception passback, lead back corresponding user's end;

The package that receives according at least one the real storage device that leads one by one of listed save project in this dynamic conference line table of comparisons; And

Until after this writes storage device that order involves and all write, just respond one and reply protocol Data Unit to this corresponding user end.

12. the package forwarding method of virtual storage network switch as claimed in claim 8, it is characterized in that, wherein the line of the line set up of this front end and the foundation of this rear end is a kind of line of world-wide web slim calculator system interface meeting of the host-host protocol that uses world-wide web slim calculator system interface, this target is a world-wide web slim calculator system interface target, and this Qi Shizhe is that a world-wide web slim calculator system interface opens the beginning person.

13. the package forwarding method of virtual storage network switch as claimed in claim 12, it is characterized in that, wherein this dynamic conference line table of comparisons comprises an element at least, the lifetime of each this element starts from receiving the world-wide web slim calculator system interface bidding protocol data cell of client, finishes till receiving the world-wide web slim calculator system interface response protocol data cell that stores end.

14. the package forwarding method of virtual storage network switch as claimed in claim 13, it is characterized in that, wherein each this element exist during, be with one to one or the mode of a pair of a plurality of meetings is sent to the world-wide web slim calculator system interface storage device of destination from the protocol Data Unit of client.

15. the package forwarding method of virtual storage network switch as claimed in claim 8 is characterized in that, wherein the package of this qualified type is wherein one type the package that order, data output, data input, answer and preparation transmit.

16. the package forwarding method of virtual storage network switch as claimed in claim 8 is characterized in that, wherein is the connection state present situation according to this destination meeting, revises the gauge outfit of each this package of qualified type.

17. the package forwarding method of virtual storage network switch as claimed in claim 8, it is characterized in that, wherein the gauge outfit of each this package has a corresponding access descriptor, and each this access descriptor comprises the information of corresponding line identification code and virtual access point.

18. the package forwarding method of virtual storage network switch as claimed in claim 8 is characterized in that, wherein this package forwarding method is by carrying out in the host software layer on the single-processor.

19. the package forwarding method of virtual storage network switch as claimed in claim 8 is characterized in that, wherein the step of this package forwarding method is to carry out in primary processor at least one microprocessor in addition that is dispersed on the multiprocessor.

Images