CN1997033B - A method and system for network storage - Google Patents

Abstract

A protocol and system for network storage, belonging to the field of computer information storage technology, aims to overcome the shortcomings of existing network storage protocols, construct a network storage protocol that can meet the requirements of price, bandwidth, deployment complexity, etc., and Reduce server load and improve I/O service performance. The protocol of the present invention is based on VI, including initiator process and target device process; the system of the present invention includes an initiator and N target devices, and the initiator and target devices are connected to the network through VI network interface and VI connection channel; Data is transferred between the initiator and target through one or more VI connection channels. The present invention uses VI to replace the traditional TCP/IP protocol as the communication basis for constructing the storage protocol, shortens the critical path of network storage data, improves the utilization rate of physical network bandwidth and the response speed of network I/O, and effectively solves the problems of network bandwidth, Network storage issues such as storage access speed and interoperability.

Description

A method and system for network storage

technical field

The invention belongs to the technical field of computer information storage, and in particular relates to a protocol for network storage and a system thereof.

Background technique

With the popularization of the network and the surge in the amount of modern information data, many important computer application fields, especially storage-intensive applications, put forward requirements for performance technical indicators such as capacity, bandwidth, I/O response time, and scalability of information storage systems. higher requirements. For this reason, various network storage architectures such as Network Attached Storage System (NAS) and Storage Area Network (SAN) have been proposed and realized, which improve the performance of storage systems from the perspective of storage systems.

The basis of building a network storage system - the storage protocol, directly affects the performance of the storage system. Today there are many storage protocols to choose from, including SCSI Serial protocol, Fiber Channel protocol, and iSCSI protocol. In today's industry, Firbre Channel is the first choice to build a high-performance storage area network (SAN), but its high price and deployment complexity are not suitable for low-end users. On the other hand, for popular network Ethernet, with the recent development of technologies such as Gigabit or even 10 Gigabit, it becomes more and more attractive and feasible to construct a large-scale storage system based on Ethernet. The traditional method of building an Ethernet network storage system is to use a TCP/IP-based storage protocol, such as the iSCSI protocol, to encapsulate SCSI commands and data into TCP/IP packets for transmission. Although the TCP/IP technology is quite mature, and its functions such as flow control and congestion control have brought great convenience to data transmission, but in the LAN environment, when you want to obtain higher transmission performance, the TCP/IP protocol Complexity becomes a stumbling block.

Therefore, it is necessary to design a new storage protocol based on Ethernet, make full use of the cheap and popular advantages of Ethernet, and bypass the complicated TCP/IP protocol at the same time, so that the storage system built on Ethernet can maintain the price advantage , with higher bandwidth and faster I/O response speed. Virtual interface (VI) is a high-bandwidth and low-latency communication mechanism for cluster systems. The main idea is to provide each user process with a protected and directly accessible virtual interface to save system processing overhead in traditional network models. . Each virtual interface represents a communication endpoint, and two virtual interfaces can be logically connected to realize point-to-point bidirectional data transmission. Using VI instead of TCP/IP as a communication protocol can achieve higher throughput and keep low overhead. On this basis, the new storage protocol vSCSI designed and implemented can make the storage system achieve better performance.

Contents of the invention

The present invention provides a protocol for network storage and its system, aiming at overcoming the price and performance deficiencies of existing network storage protocols when deploying network storage systems, and constructing a network storage protocol that can meet the requirements of price, bandwidth, deployment complexity, etc. A network storage protocol based on a virtual interface, which reduces server load and improves I/O service performance.

A protocol for network storage of the present invention includes an initiator process and a target process,

(1) The launcher process sequence includes:

(1.1) Initiator initialization step, the initiator selects and loads the SCSI adapter driver module or vSCSI API module according to the mode specified by the user, and completes the relevant initialization;

(1.2) The initiator initiates a login step, and the initiator invokes the connection management module to send authentication information to one or more targets designated by the user to request login;

(1.3) The initiator waits for the authentication step, the initiator waits for the authentication response from the target, if the authentication response is received within the time threshold, proceed sequentially, otherwise turn to the authentication timeout processing step;

(1.4) VI connection channel establishment step, establish one or more VI connection channels between the initiator and one or more targets, register corresponding VI send/receive buffers, and associate send/receive descriptors;

(1.5) The initiator waits for the command step. After the connection is established, the initiator waits to receive the SCSI command from the system or the upper-layer application. After receiving the command, proceed in sequence;

(1.6) The initiator parses the command step, the initiator parses the current SCSI command, generates one or more vSCSI requests according to the data distribution rule or strategy specified by the user, and the type of the current SCSI command; if it is a SCSI read command, proceed to step (1.7) , otherwise go to step (1.8);

(1.7) Initiator read command preprocessing step, the initiator uses one or more targets used in the data distribution rule or strategy specified by the user, and according to the size of data in each vSCSI read request generated in step (1.6), Execute the PostRecv operation on the VI receiving buffer corresponding to each target, prepare each VI receiving queue, and proceed sequentially;

(1.8) The initiator sends a request step, and the initiator concurrently executes the PostSend operation for one or more vSCSI requests generated by the step (1.6) for the target device, and sends a vSCSI request to the target device; if what is sent is a write request, proceed to the step ( 1.9), otherwise go to step (1.10);

(1.9) Initiator write command preprocessing step, the initiator calculates and locates the data to be written in the SCSI command request_buffer according to the rules or strategies used, determines the corresponding relationship between the data to be written and each target device, and then encrypts the data according to the user-specified data The level performs encryption operation processing on the corresponding data of each target device determined above, and finally copies the encrypted data to the respective corresponding VI sending buffer queues in parallel, waiting for sending, and proceeding sequentially;

(1.10) The initiator waits for a response step. The initiator waits for a response from the target. After receiving the response, it enters different processing routines according to different response types: the response to the read command enters step (1.11), and the response to the write command enters Step (1.12), the response of the status query class command enters step (1.13); if no response is received within the time threshold, then proceed to the response overtime processing step;

(1.11) Initiator read command processing step, the initiator receives data from one or more targets, and performs decryption operation on the data in the receiving buffer of each VI according to the data encryption level specified by the user, and then according to the used rules or The strategy calculates and locates the space to be backfilled in the SCSI command request_buffer, determines the data from each target device and its corresponding relationship, and finally copies the decrypted data to the corresponding position of the SCSI command calculated above in request_buffer; after backfilling, go to step (1.14);

(1.12) Initiator write command processing step, the initiator receives responses from one or more target devices, executes the PostSend operation concurrently on each data queue to be written, and sends the data to be written to the corresponding target device; after the data is sent, Go to step (1.14);

(1.13) Initiator status query command processing steps, the initiator receives one or more vSCSI status query responses corresponding to one or more targets, extracts and returns SCSI response information to the system or upper-layer application; after the feedback is completed, go to the step (1.14);

(1.14) In the step of completing the processing of the initiator request, the initiator finishes executing the current SCSI command, and turns to step (1.5), waiting for the execution of the next SCSI command.

(2) The target process sequence includes:

(2.1) The target device initialization step, the target device starts the device service thread initialization, and connects the management module after initialization;

(2.2) The target device waits for the login step, and the target device waits to receive a login authentication request from the initiator;

(2.3) The target device receives the login step, the target device receives the login request, performs identity verification to it, and the illegal login request is transferred to the illegal processing step, and the legal request passed through verification accepts its login, and sends an authentication response to the initiator;

(2.4) The target device waits for a command step, and the target device waits to receive a vSCSI command from the initiator;

(2.5) The target device parses the command step, the target device receives the vSCSI command, parses it, and performs different processing according to the difference of the current vSCSI command type: vSCSI read command transfer step (2.6), write command transfer step (2.7), Status query command to step (2.10);

(2.6) Target device read command processing steps. Under the control of the read-write speed-up module, the target device copies data from the physical storage device or Cache to the VI sending buffer queue, and executes the PostSend operation on it, and sends the data to the initiator ;All the data has been sent, go to step (2.11);

(2.7) The target device write command preprocessing step, the target device executes the PostRecv operation according to the request size, prepares the VI receiving queue, generates and calls the PostSend operation to send a ready (ready for writing) response to the initiator, and proceeds sequentially;

(2.8) The target device waits for the data step, the target device waits for the data from the initiator, if the data is received within the time threshold, the sequence is carried out; otherwise, the receiving timeout processing step is transferred to,

(2.9) The target device write command processing step, the target device receives the data from the initiator, and under the control of the read-write speed-up module, writes the data directly from the VI receiving buffer to the physical storage device or Cache; step (2.11);

(2.10) The processing steps of the target device state query class command, the target device extracts the SCSI query command encapsulated in the vSCSI command, executes the query operation according to the SCSI protocol specification, and encapsulates the feedback information obtained from the query into the vSCSI query response, and calls the PostSend operation to send To the initiator; after the query response is sent, go to step (2.11);

(2.11) The target device requests to complete the processing step, and the target device completes the successful execution of the current vSCSI command, and then proceeds to step (2.4), waiting to execute the next vSCSI command;

The described protocol for network storage is characterized in that: (1) the authentication timeout processing step is responsible for returning an error status to the system or upper-layer application, and terminating the initiator process; (2) the response timeout processing Step, is responsible for reporting current SCSI command to system or upper stratum application and carries out mistake, and terminates the execution of current SCSI command, changes over to starter to wait for command step, waits to execute next SCSI command; (3) described receiving overtime processing step, is responsible for sending to The initiator reports that the execution of the current vSCSI command is wrong, and terminates the execution of the current vSCSI command, and turns to the target device to wait for the command step and wait for the execution of the next vSCSI command; (4) the illegal processing step is responsible for the rejection of the illegal login, and Go to the target waiting for login step, and continue to wait for the next initiator to login.

A system for the network storage protocol of the present invention includes an initiator and N targets, the initiator is connected to the network through a VI network interface and a VI connection channel; the target device is connected to the network through a VI network interface and a VI connection channel. into the network; data is transmitted between the initiator and the target through one or more VI connection channels, and N is a natural number;

(1) The initiator includes: SCSI adapter driver module, vSCSI API module, connection management module, command processing module and data encryption and decryption module;

(1.1) The SCSI adapter driver module registers one or more virtual SCSI disk devices with the operating system, receives standard SCSI commands from the SCSI middle layer of the operating system and returns execution status, and provides traditional disk devices to user space and accepts access;

(1.2) The vSCSIAPI module provides a direct call interface to upper-layer developers, bypassing the operating system SCSI layered system to directly receive upper-layer commands and return to the execution status;

(1.3) The connection management module in the initiator initiates one or more VI connection authentication requests to one or more targets, completes initialization work such as connection establishment, parameter negotiation, memory registration, and friendly logout work. Re-establish the connection and transfer the service to the new VI connection;

(1.4) The command processing module is responsible for converting the command received from the SCSI adapter driver module or vSCSI API module into a vSCSI command, calling PostSend to execute the delivery of data in the VI sending buffer queue, calling PostRecv to execute the preparation of the VI receiving buffer, Send and receive vSCSI commands, generate corresponding commands or status, and return to the upper SCSI adapter driver module or vSCSI API module;

(1.5) The data encryption and decryption module is responsible for the encryption processing of the data to be transmitted and the decryption processing of the received data; when writing a request, the data is encrypted according to the encryption and decryption level and then copied from the system buffer to the VI registration buffer; the read request , copy the data from the VI registration buffer to the system buffer after decrypting the data;

(2) The target device includes: a device management module, a connection management module, a request processing module and a read-write speed-up module;

(2.1) The device management module is responsible for establishing device service threads, collecting physical storage device information, completing the initialization of local device service status and control data, and interacting with devices in real time;

(2.2) The connection management module in the target monitors and verifies the connection authentication request from the initiator node, establishes one or more VI connections with the legal initiator, and completes initialization work such as parameter negotiation and memory registration;

(2.3) The request processing module is responsible for receiving and processing vSCSI requests sent by the initiator, including read and write requests, device status query requests, etc., and starts the corresponding processing process according to the type of arrival request, wherein the data interaction between the read and write requests and the physical storage device will be determined by Read and write speed-up module for control;

(2.4) The read-write speed-up module controls whether the data in the VI registration buffer is read from or written to the Cache or the physical storage device during the processing of the read-write request, and handles the consistency between the data in the Cache and the data in the physical storage device.

The system is characterized in that, the initiator and the target running vSCSI are equipped with a network adapter supporting VI transmission; the VI network adapter provides at least one of two reliability levels of reliable sending or reliable receiving ; Both the initiator and the target device can run in the same or distributed among different computer host nodes; the virtual SCSI device and the operating system in the node where the initiator is located, and the vSCSI API all adopt the SCSI interface; the target device The physical storage device in the node is a disk drive or disk array with SCSI, FC, IDE or SATA interface; the network is a local area network of Ethernet medium.

Above, the network storage protocol vSCSI of the present invention, full name is VI-attached SCSI, wherein VI refers to virtual interface, and SCSI refers to small computer system interface; The callable function PostRecv provided by the layer application realizes the function of preparing the VI receiving buffer; the callable function PostSend provided by the virtual interface (VI) to the upper layer application realizes the function of delivering the data in the VI send buffer; Cache is a memory buffer , used to cache data on the physical storage device; request_buffer is an area in the SCSI command structure used to store data to be read and written.

The present invention uses VI to replace the traditional TCP/IP protocol as the communication basis for constructing the storage protocol, shortens the critical path of network storage data, improves the utilization rate of physical network bandwidth and the response speed of network I/O, and effectively solves the problem of network bandwidth , storage access speed, interoperability and other major network storage issues.

Specifically, the present invention has the following advantages:

1. Utilize VI's high bandwidth and low latency features to obtain higher transmission performance than traditional TCP/IP-based storage protocols; measured data shows that vSCSI has 115% higher sequential write performance than iSCSI (intel version) under the same conditions, Up to 35% improvement in sequential read performance.

2. Using the multi-level reliability transmission of VI, the important function of guaranteeing transmission reliability is moved down to the internal implementation of VI, thus greatly avoiding the complexity of upper-layer storage protocols in realizing storage reliability, reducing storage protocol overhead, and improving storage protocol efficiency.

3. Use the performance characteristics of VI in data transmission behavior to realize data encryption without loss of transmission performance during data storage; conversely, data decryption without loss of transmission performance can also be realized. That is, the time required to deliver data packets (PostSend operation) to VI is a step function with the amount of data continuously transmitted by VI. When the continuous amount exceeds the second critical value, the delivery time will stabilize at a relatively high value. The delivery time of a data packet can encrypt the next data packet; vice versa can also be decrypted in the same way.

4. The protocol supports the establishment of multiple VI connections between the initiator and multiple targets. These VI connections are highly parallel. In addition to parallel access operations, parallel network transmission of data can also be achieved, ensuring a high level of storage protocol. efficiency and high performance.

5. The protocol internally supports multiple data distribution rules (rules) or policies (policy) in a multi-target environment to realize functions such as load balancing between targets and data backup.

6. The present invention can flexibly provide multiple storage services for different types of application environments and requirements: standard SCSI device services, API call services.

Description of drawings

Fig. 1 is a schematic diagram of the composition and structure of the system of the present invention;

Fig. 2 is a schematic diagram of modules and their relationships in the initiator and target device of the present invention;

Figure 3 is a schematic diagram of the protocol of the present invention.

Detailed ways

Fig. 1 is a schematic diagram of the composition and structure of the system of the present invention. The present invention includes an initiator 100 and n target devices 200.1, 200.2, ..., 200.n, which are interconnected through a VI communication network 300, and n is a natural number. The initiator 100 is connected to the network 300 through VI network interfaces 121.1-121.m (m is a natural number), and the targets 200.1, 200.2, ..., 200.n are respectively connected to Enter the network 300. The initiator 100 establishes k (k is a natural number) VI connection channels 120.1 to 120.k through m VI network interfaces 121.1 to 121.m, and the targets 200.1, 200.2, ..., 200.n respectively through their respective VI network interfaces 221.1, 221.2, ..., 221.n establish a total of k VI connection channels 220.1-220.k; among them, the VI connection channels 120.1-120.k ₁ of the initiator 100 and the VI connection channels 220.1-220.k of the target device 200.1 ₁ Establish k ₁ VI connections, responsible for the request response and data transmission between the initiator 100 and the target 200.1; the VI connection channel 120.k of the initiator 100 ₁ +1~120.k ₂ and the VI of the target 200.2 The connection channel 220.k ₁ +1～220.k ₂ establishes k ₂ -k ₁ VI connections, responsible for request response and data transmission between the initiator 100 and the target 200.2; and so on, the VI of the initiator 100 The connection channel 120.k _n-1 +1 ~ 120.k and the VI connection channel 220.k _n-1 +1 ~ 220.k of the target 200.2 establish kk _n-1 VI connections, responsible for the initiator 100 and the target The request response and data transmission between the initiators 200.2; the size of k in the initiator 100 is not limited to the size of m, but k≥m, using m VI network interfaces can increase the physical bandwidth of the initiator to access the network, but not Affecting the logic implementation, the same is true for the target device.

Fig. 2 shows the modules in the initiator and the target device of the present invention and the relation diagram among the modules. SCSI adapter driver module 150, vSCSI API module 160, connection management module 110, command processing module 130 and data encryption and decryption module 140 run in the initiator 100. The SCSI adapter driver module 150 interacts internally with the command processing module 130, and externally interacts with the operating system (SCSI middle layer) of the node host where the initiator is located; the vSCSIAPI module 160 interacts internally with the command processing module 130, and externally interacts with the node host where the initiator is located The running third-party program interaction; SCSI adapter driver module 150 and vSCSI API module 160 are two parallel functional modules, which provide a program interface for the use of the present invention by the outside world. Generally, only one of the modules is used to communicate with the outside world at the same time for the same initiator. interact.

The device management module 250 , the connection management module 210 , the request processing module 230 and the read/write acceleration module 240 run in the target device 200 . The device management module 250 interacts internally with the request processing module 230, and externally interacts with the physical storage device mounted on the node host where the target device is located; the read-write speed-up module 240 interacts internally with the request processing module 230 and the connection management Interact with the physical storage device mounted on the node host where the target device is located; the device management module 250 mainly performs the interaction of control information such as status query with the physical storage device, and the read-write speed-up module 240 mainly performs data access operations with the physical storage device .

Fig. 3 is a schematic diagram of the protocol of the present invention, which mainly shows the process of safe login and three types of commands (query, read, write).

Claims (4)

1. a method that is used for the network storage comprises starter process and object machine process,

(1) the starter procedural sequences comprises:

(1.1) starter initialization step, starter select to load small computer system interface scsi adapter driver module or based on the small computer system interface protocol application DLL (dynamic link library) vSCSI API module of virtual interface, finishes relevant initialization according to user's designated mode;

(1.2) starter initiates to login step, and starter calls the connection management module and sends authentication information to one or more object machines of user's appointment, and request is logined;

(1.3) starter is waited for authenticating step, and starter is waited for the authentication response from object machine, if wait until authentication response in time threshold, order is carried out, otherwise changes authentication timeout treatment step over to;

(1.4) virtual interface VI interface channel establishment step is set up one or more VI interface channels between starter and one or more object machine, register corresponding VI transmission/reception buffering area and related transmission/reception descriptor;

(1.5) after starter wait command step, connection were set up and finished, starter was waited for the scsi command that receives from system or upper layer application, receives order, and order is carried out;

(1.6) starter resolve command step, starter is resolved current scsi command, and according to user's data designated distribution rule or strategy, and the type of current scsi command generates one or more vSCSI requests; If step (1.7) is carried out in the SCSI read command, otherwise change step (1.8);

(1.7) starter read command pre-treatment step, starter is according to the one or more object machines that use in user's data designated distribution rule or the strategy, according to data volume size in each the vSCSI read request that generates in the step (1.6), VI to each object machine correspondence receives buffering area execution beginning to receive PostRecv operation, be ready to each VI and receive formation, order is carried out;

(1.8) starter sends request step, and the one or more vSCSI at object machine that starter generates step (1.6) ask concurrent execution to begin to send the PostSend operation, send the vSCSI request to object machine; If what send is write request, carries out step (1.9), otherwise change step (1.10);

(1.9) starter write order pre-treatment step, starter carries out the computing location according to rule of using or strategy to treating write data among the scsi command request buffer request_buffer, determine to treat the corresponding relation of write data and each object machine, according to user's data designated encryption level above each definite object machine corresponding data being carried out cryptographic calculation respectively then handles, at last the data parallel after the encryption is copied to each self-corresponding VI and send the buffering area formation, etc. to be sent, order is carried out;

(1.10) starter wait-for-response step, starter is waited for the response from object machine, receive response, then according to different respond styles, enter the different disposal routine: the response of read command enters step (1.11), the response of write order enters step (1.12), and the response of status poll class order enters step (1.13); If in time threshold, do not receive response, then change the response timeout treatment step over to;

(1.11) starter read command treatment step, starter receives the data from one or more object machines, and according to user's data designated encryption level each VI is received that data are decrypted calculation process in buffering area, carry out the computing location according to rule of using or strategy to treating the backfill space among the scsi command request_buffer then, determine data and its corresponding relation, at last the data after the decryption processing are copied in the request_buffer correspondence position of the above scsi command that calculates from each object machine; Backfill finishes, and changes step (1.14);

(1.12) starter write order treatment step, starter receives the response from one or more object machines, to the concurrent execution PostSend of each data queue to be written operation, will treat that write data sends to corresponding object machine; Data send and finish, and change step (1.14);

(1.13) starter status poll class command process step, starter receive the one or more vSCSI status polls response of corresponding one or more object machines, extract and return the SCSI response message to system or upper layer application; Feedback finishes, and changes step (1.14);

(1.14) the initiator request step that disposes, the starter current scsi command that is finished changes step (1.5), waits pending next scsi command.

(2) the object machine procedural sequences comprises:

(2.1) connection management module after object machine initialization step, the initialization of object machine starting device service thread, initialization finish;

(2.2) object machine is waited for and is logined step, and object machine is waited for the login authentication request that receives self-starter;

(2.3) object machine receives and logins step, and object machine receives the request of logining, and it is carried out authentication, and the request of illegally logining changes illegal treatment step over to, accepts it by the legitimate request of checking and logins, and sends authentication response to starter;

(2.4) object machine wait command step, object machine are waited for the vSCSI order that receives self-starter;

(2.5) object machine resolve command step, object machine receive the vSCSI order, and it is resolved, difference according to current vSCSI command type, carry out different disposal: step (2.6) is changeed in the vSCSI read command, and write order changes step (2.7), and step (2.10) is changeed in the order of status poll class;

(2.6) object machine read command treatment step, object machine copy data the formation of to VI transmission buffering area from physical storage device or core buffer Cache, and it are carried out the PostSend operation under read-write speedup module controls, and data are sent to starter; Data all send and finish, and change step (2.11);

(2.7) object machine write order pre-treatment step, object machine is carried out the PostRecv operation according to request size, is ready to VI and receives formation, generates and calls the PostSend operation and send to write to starter and be ready to the ready response, and order is carried out;

(2.8) object machine is waited for data step, and object machine waits for the data of self-starter, if receive data in time threshold, order is carried out; Otherwise change the receive time-out treatment step over to,

(2.9) object machine write order treatment step, object machine receives the data of self-starter, under read-write speedup module controls, data is received buffering area from VI directly write physical storage device or Cache; Data all receive, and change step (2.11);

(2.10) object machine status poll class command process step, object machine extracts the SCSI querying command of encapsulation in the vSCSI order, carry out query manipulation according to the SCSI protocol specification, and the feedback information that will inquire about gained encapsulates in the vSCSI inquiry response, call PostSend and operate and be sent to starter; Inquiry response sends and finishes, and changes step (2.11);

(2.11) the object machine request step that disposes, object machine finishes current vSCSI order successful execution, changes step (2.4) over to, waits pending next vSCSI order;

2. a kind of method that is used for the network storage as claimed in claim 1 is characterized in that: (1) described authentication timeout treatment step, and be responsible for returning error condition, and stop the starter process to system or upper layer application; (2) described response timeout treatment step is responsible for reporting that to system or upper layer application current scsi command execution makes mistakes, and is stopped the execution of current scsi command, changes starter wait command step over to, waits pending next scsi command; (3) described receive time-out treatment step is responsible for reporting that to starter current vSCSI command execution makes mistakes, and is stopped the execution of current vSCSI order, changes object machine wait command step over to, waits pending next vSCSI order; (4) described illegal treatment step is responsible for this refusal of illegally logining, and changes object machine over to and wait for and login step, continues to wait for that next starter logins.

3. a system that is used for claim 1 or 2 described network storage methods comprises a starter and N object machine, and starter is connected into network by virtual interface VI network interface and VI interface channel; Object machine is connected into network by VI network interface and VI interface channel; By one or more VI interface channel transmission data, N is a natural number between starter and object machine;

(1) comprises in the starter: small computer system interface scsi adapter driver module, based on small computer system interface protocol application DLL (dynamic link library) vSCSI API module, connection management module, command processing module and the data encrypting and deciphering module of virtual interface;

(1.1) the scsi adapter driver module is registered one or more virtual scsi disk equipment to operating system, receives from the standard scsi command in operating system SCSI intermediate layer and returns executing state, traditional magnetic disk equipment is provided and accepts the interview to user's space;

(1.2) vSCSI API module provides direct calling interface to the upper strata developer, and the SCSI of workaround system layered system directly receives upper layer commands and returns executing state;

(1.3) the connection management module in the starter is initiated one or more VI connection authentication request to one or more object machines, finish connect, parameter negotiation and the work of internal memory register initial and close friend publish work, under the unexpected situation about disconnecting of VI connection, rebulid connection, and new VI connection is transferred in service;

(1.4) command processing module is responsible for the command conversion from scsi adapter driver module or the reception of vSCSI API module is become the vSCSI order, call and begin to send the delivery work that PostSend carries out data in the formation of VI transmission buffering area, call and begin to receive the preparation that PostRecv carries out VI reception buffering area, send and receive the vSCSI order, and generate the corresponding command or state, return upper strata scsi adapter driver module or vSCSI API module;

(1.5) the data encrypting and deciphering module is responsible for to the encryption of data to be transmitted with to the decryption processing of data accepted; During write request, data are carried out behind the cryptographic calculation copying to from the system buffer VI registration buffer district according to the encryption and decryption rank; During read request, copy the system buffer to from VI registration buffer district after data are decrypted computing;

(2) comprise in the object machine: device management module, connection management module, request processing module and read-write speedup module;

(2.1) device management module is responsible for the apparatus for establishing service thread, collects the physics storing device information, finishes the initialization of local device service state and control data, carries out real-time information interaction with equipment;

(2.2) the connection management module in the object machine is monitored the connection authentication request that verification comes the self-starter node, sets up one or more VI with legal starter and is connected, and finishes parameter negotiation and the work of internal memory register initial;

(2.3) request processing module is responsible for receiving and handling the vSCSI request that starter sends, comprise read-write requests and device status inquiries request, start the respective handling process according to the kind that arrives request, wherein the data interaction of read-write requests and physical storage device will be controlled by read-write speedup module;

(2.4) in the read-write speedup module controls read-write requests processing procedure in the VI registration buffer district data read from or write core buffer Cache or physical storage device, and handle among the Cache consistency problem of data in the data and physical storage device.

4. system as claimed in claim 3 is characterized in that, described starter and object machine all are equipped with the network adapter of supporting the VI transmission; The network adapter of described support VI transmission provides a kind of in reliable transmission or two kinds of reliability class of reliable reception at least; Described starter and object machine all can run on same or be distributed among the various computing machine host node; Interface in the node of described starter place between virtual scsi device and operating system, and vSCSI API all adopts scsi interface; Physical storage device in the node of described object machine place is for adopting the disc driver or the disk array of SCSI, optical-fibre channel FC, Integrated Device Electronics IDE or Serial Advanced Technology Attachment SATA interface; Described network is the local area network (LAN) of ether medium.

Images