TWI446194B - High-performance virtual disk management system - Google Patents

Description

High performance virtual disk management system

A high-performance virtual disk management system, especially for one-time computing of physical and virtual space address translation, to improve the performance of virtual disk access, and to achieve virtual disk-like storage on block-level storage media. A high-performance virtual disk management system for the utility of dynamic space configuration.

According to US Patent No. US 7,107,385 B2, the invention is based on the Storage Virtualization by Layering Virtual Disk Objects on a File System, which is a file system of an application operating system that supports a wide range of storage media, such as extended disks. Or a network storage medium, which is a volume of a file system that is used to store a virtual disk.

The Republic of China Patent No. I307026, for the purpose of proposing a storage management system and method, the invention uses a file system server, a file information (Metadata) server and an object storage device to achieve a virtual partition for file access. The file information server stores the management and index information of the file, and the object storage device accesses the file content data. Because the file management and index information and the real file content are separated and stored on the two servers, the virtual partition space can be dynamically grown by adding the storage unit of the object storage device.

The above existing virtual disk technology is intended to provide flexible physical disk space and virtual disk space correspondence, by using the operating system's file system management capabilities or separate file management and indexing information and real file content. The method is achieved, however, these techniques require a certain amount of additional computational cost to achieve, that is, a file access will undergo at least two layers of address translation, resulting in a performance bottleneck requiring high-speed virtual disk access applications.

The main purpose of the present invention is to use a virtual disk management system to plan a block level storage medium into a dynamic size set of physical segments, and then directly write virtual disk segments into a plurality of physical segments to avoid In the prior art, the address conversion required by the file system of the operating system becomes a one-time calculation of the physical and virtual space address conversion, thereby improving the performance of the virtual disk access and making the storage medium realize like a virtual disk. The utility of dynamic space configuration.

To achieve the above objective, the present invention is a virtual disk management system connected between a storage server and a storage medium, and the storage server is connected to the client computer through a network, and the virtual disk management system is provided with a disk segment. Planner, managed segment processor, disk access and scheduler, virtual disk drive, disk segment allocation and recycler, virtual disk storage segment table register, virtual disk write protection switch Site device, virtual disk accessor, and virtual disk data cache and buffer, wherein: the disk segment planner is configured to plan the storage medium into a plurality of physical segments of a given fixed size, an entity The segment is the smallest unit of use. When the disk segment planner plans to store the media, a physical segment index table is generated to store the state in which the physical segment of the storage medium is used, which is not configured to be represented by 0; Configured to be 1, in the initial state, the value in the index table is 0; in addition, if the virtual disk management system is connected to a plurality of storage media, the disk segment planner will be for each entity. Storage Media use a consistent programming mode, and a storage medium using the recording table group storage medium to store all the information can be used, each is assigned a unique storage medium identifier.

The control segment processor selects a plurality of unconfigured physical segments in each storage medium as a management segment for management purposes, and one storage medium stores a plurality of control segments, but only one is primary, and the other is For backup and inspection, to increase the security of important data, and the control segment processor will write the physical segment index table and the storage media group record table into the control section in the storage medium.

The disk access and scheduler is a drive component for accessing and scheduling the storage medium, so that the data command for access management purposes has a higher priority order than the data access command of the virtual disk.

The virtual disk drive is a controller for generating, deleting, and mounting a virtual disk on the storage medium. When the client computer issues a command to create, delete, or mount a virtual disk to the storage server, the storage server will The instruction is forwarded to the virtual disk drive, and the virtual disk drive obtains a new physical segment through the disk segment allocation and recycler, reclaims a physical segment occupied by a virtual disk, or queries the virtual disk for occupation. All physical segments.

The disk segment allocation and recovery device is configured to perform and complete the virtual disk drive to add, delete, and query a physical segment occupied by a virtual disk, when the disk segment allocation and the recycler execute When the instruction is executed, the latest entity segment index table is first obtained by the management segment processor to know which physical segments can be allocated or occupied; and, in addition, instructions or requirements for adding a virtual disk For the request to configure the physical segment, the disk segment allocation and recycler selects the physical segment allocation based on the principle of optimizing the access performance in the assignable entity segment, and the optimized access performance is optimized. The virtual disk entity segment occupancy table may be generated by preferentially selecting a neighboring physical segment, and the virtual disk physical segment occupancy table index is a unique identifier of the virtual disk in the virtual disk management system. The value corresponding to the identifier is a binary string, and the constituent elements of each group of data in the string are composed of an identifier of the storage media group record table and an entity segment index of the entity segment index table, and if the disk segment Distribution and The receiver changes the virtual disk physical segment occupation table or the physical segment index table due to the execution of the instruction to create and delete the virtual disk, and the virtual disk entity segment occupation table or entity segment index table to be updated. Transfer to the management of the managed zone processor update management zone.

The virtual disk storage segment table register is a cache register for accelerating the address virtual disk to the storage medium location, and when the space is sufficient, it will be the virtual magnetic disk of the virtual disk in use. The contents of the disc physical sector occupancy table are stored in a register that can be accessed at high speed (relative to disk access). The virtual disk write protection transcoder can be completed quickly when calculating the physical segment that the virtual disk can actually access. When any virtual disk physical segment occupies a table transaction event, the control segment processor will Update the virtual disk entity segment occupancy table in the virtual disk storage segment table register to keep the data consistent.

The virtual disk write protection relayer is two kinds of calculation work for address translation and check of the physical segment that the virtual disk can actually access; the client computer performs data access instruction on a virtual disk. And forwarded to the virtual disk write protection relayer to calculate the physical address of the virtual disk stored in the storage medium, and then the data access of the virtual disk is performed, but any access to a virtual disk is performed. The instruction should not exceed the physical segment occupancy table of the virtual disk to record all the physical segment ranges that the virtual disk can access to protect the data of multiple virtual disks, avoid reading and writing data to each other, but exceptions If a virtual disk is to be written to a new disk at a legal disk virtual location (not exceeding the capacity range of the virtual disk) of a physical storage space that is not yet configured but is about to be configured, the disk segment is allocated to the disk segment. And the recycler requests to configure a new entity segment. When the new entity segment is added, the original access instruction satisfies the protection condition and can be executed normally. Moreover, if an unconfigured real disk is read, When the legal position of the virtual storage space, that this read instruction reads the address data has not been written, it can simply reply to blank data without configuring a new physical section. In addition, the discontinuous physical segment configuration may cause the access instruction to operate on a plurality of physical segments. The operation of such cross-entity segments requires an instruction to be decomposed into a plurality of instructions for accessing a single entity segment data. After the instruction is completed, the original instruction is completed; further, the functions of the snapshot disk or the Snapshot or Differential Disks defined by the traditional virtual disk technology are also handled by the virtual disk write protection relayer. In the case of a differential disk, the differential disk is essentially regarded as a virtual disk. It is assumed that a virtual disk A already exists on the system and the user creates a different disk B for the virtual disk. Writing data to the virtual disk A will be dumped to the difference disk B; if the data is read from the virtual disk A, there are two possibilities: first, if the read is written on the differential disk The data of B will be read by the difference disk B. Secondly, if the data that has not been written to the difference disk B is read, it will be read directly by the virtual disk A. Therefore, the difference disk will Its corresponding virtual disk becomes the only one In order to achieve the above functions, the virtual disk write protection relayer establishes a differential disk relationship table, and the index field of the differential disk relation table is a unique number of a different disk, and the corresponding value is the difference magnetic The virtual disk number to which the disk is connected. If there are multiple records of the same virtual disk number, it means that a plurality of differential disks have been created on a virtual disk. To delete a disc, just delete the records in the disc relationship table, reset the corresponding position of the entity segment index table to release the physical segment occupied by it, and finally update the contents of the control segment. Just fine. If a virtual disk is deleted, all the differential disk records in the differential disk relation table that are created on the virtual disk must be deleted synchronously, and the contents of the control section can be updated synchronously; finally, if it is multi-level For differential disk associations, simply fill in a different disk number in the virtual disk field in the differential disk relationship table.

The virtual disk access device, which is a implemented component of the conventional virtual disk technology, is responsible for writing the (virtual) disk obtained by the client to the storage medium in a conventional virtual disk format such as VMDK, VHD, and VDI. When calculating the actual virtual storage format address of the traditional virtual disk format, only the offset of the physical segment obtained by the virtual disk write protection relayer lookup table is added, that is, the access of the data only needs to be performed once. The calculation of sex can be completed.

The virtual disk data cache and buffer are used to speed up the cache and buffer of the virtual disk data access, so as to reduce the number of times the storage medium is read, and the action of writing data can be asynchronously completed, reducing The response time of the write command; this virtual disk data cache and buffer differs from the traditional data cache and buffer in that the virtual disk write protector provides a differential disk function, resulting in a virtual disk and One to a plurality of related differential disks have different access frequencies, assuming that the same address is read to a plurality of differential disks associated with a virtual disk, if most of the differential disks have not been written The address, then the actual read address will be returned to the same address on the read virtual disk. Based on this feature, the cached virtual disk is cached faster than the cached data. It is more efficient in terms of cache utilization and hit rate; similarly, in the case of data access with locality and time locality, the data just written may be quoted immediately, so , write virtual disk data fast In addition to the conventional method, the access buffer can complete the write command in a non-synchronous manner, reducing the response time, and also allowing the newly written data to have a higher cache chance, especially for the data written on the differential disk; In other words, the virtual disk data cache and the buffer for the read command will enhance the associated virtual disk data cache, and for the write command, the data cache of the differential disk will be enhanced.

Referring to the first figure and the second figure, it can be clearly seen from the figure that the high-performance virtual disk management system of the present invention is used by the client computer 1 to connect one or more storage servers through the network 2. 3, the storage server 3 is connected to the storage medium 5 of the plurality of units through the virtual disk management system 4, and the virtual disk management system 4 is provided with a disk segment planner 41, a control segment processor 42, Disk access and scheduler 43, virtual disk drive 44, disk sector allocation and recycler 45, virtual disk storage section table register 46, virtual disk write protection translator 47, virtual The disk accessor 48 and the virtual disk data cache and buffer 49.

Referring to the first to fourth figures, when the storage medium 5 performs the initialization process, the disk segment planner 41, the control segment processor 42, and the disk access and scheduler 43 are combined. After completion, the disk access and scheduler 43 first detects the storage medium 5 connected to the high performance virtual disk management system, and transfers the number of the storage medium 5 to the disk segment planner 41; this embodiment The description is made with two sets of storage media 5, so that the disk access and scheduler 43 detects all the storage media 5 that are connected and usable, and the two groups are denoted by numbers 0 and 1, respectively. At this time, the disk segment planner 41 creates a storage media group recording table 53 (as shown in the third figure) according to the number of storage media 5 detected by the disk access and scheduler 43, and stores this. After the media group record table 53 is handed over to the management segment processor 42, and the disk segment planner 41 knows the number of storage media 5, the disk segment planner 41 will have a predetermined fixed segment capacity value. Each storage medium 5 is planned into a plurality of physical segments 51 in a consistent planning manner, and a physical segment index table 52 is generated for each storage medium 5; in this embodiment, when the storage medium 5 has a capacity size In the case of 1 Terabyte, each entity segment 51 is the smallest unit of use for the physical segment 51 of 1024 fixed Gigabyte (GB) fixed capacity; the segment index value can be used to quickly locate a region. The segment is at the actual address of the storage medium 5, for example, the i-th segment starts at the address of ix 1 GB, and terminates at the address of (i+1) x 1 GB-1, which determines the physical segment 51. After the capacity, the disk section planner 41 then generates a physical section index table 52 to store the physical section 51 to be The status used is not configured with 0; otherwise it is configured with 1. In the initial state, the values in the index table are all 0. The disk section planner 41 is again caused to pass the physical section index table 52 to the managed section processor 42, and the governing section processor 42 selects a plurality of each storage medium 5 (this example is 0th and last). A) the physical segment 51 becomes the primary and backup management segment 511, and the 0th and the last physical segment index in the physical segment index table 52 is filled with 1 to represent the allocated disk partition. The planner 41 writes the physical section index table 52 and the storage media group recording table 53 to the management section 511 of the storage medium 5.

Referring to the first to third and fifth figures, after the initialization of the storage medium 5 is completed, the virtual disk management system 4 can process the creation of the virtual disk from the client computer 1 through the storage server 3. Operation instructions for deleting, deleting, and mounting, as shown in Figure 5, this type of operation is the same as the traditional network hard disk being mounted on the file system as a storage space that can be operated, for example, a storage server 3 can be the iSCSI Target side, and the client computer 1 can be the iSCSI Initiator side. The operation instructions for performing the creation, deletion, and mounting of the virtual disk are operated by the virtual disk drive 44 of the virtual disk management system 4, the disk sector allocation and recycler 45, the management segment processor 42, and the disk access. And the scheduler 43 cooperates to complete. When the client computer 1 issues an operation command through the storage server 3, the storage server 3 transmits an operation command to the virtual disk management system 4. At this time, the virtual disk drive 44 first determines the operation instruction type, and if it is a new virtual The instructions of the disk will request the configuration of the new physical segment 51 to the disk segment allocation and recycler 45. This embodiment is illustrated by the physical segment 51 of the 0th storage medium 5 and the physical segment 51, and will The newly created virtual disk is given a unique pipeline code, which is assumed to be number 0, and the index value of the entity segment 51 it has just been assigned is recorded in the virtual disk entity segment occupancy table 54 and filled in the instruction to create a new magnetic The size of the disc, that is, the maximum address space available for the disc (in this case, 2 GB, as shown in the third figure), and the capacity of the virtual disk can exceed the capacity of the actually allocated storage medium 5, as long as it is used again. It is sufficient to expand the capacity of the actual storage medium 5. Next, the governing section processor 42 updates the relative position of the physical section index table 52 to the occupied state, that is, to 1. Furthermore, if the instruction to create a new differential disk is assumed to be a newly created virtual disk 1 as a differential disk of the virtual disk 0, the difference between the new disk and the virtual disk 0 specified by the parameter needs to be increased. The disk relationship table 55 (as shown in the third figure) establishes a differential disk whose capacity value is equal to its associated virtual disk, and the difference disk relationship table 55 is updated through the control segment processor. 42 and the disk access and scheduler 43 write to the control section 511, the process of which is the same as the above-mentioned initial storage medium 5 write control section 511, and therefore will not be described later. If the virtual disk drive 44 receives the instruction to delete the virtual disk, the instruction must specify the deleted target disk number as the parameter, first release the occupied physical segment 51 and update the physical segment index table 52 to the unconfigured state. If the physical segment 51 from the plurality of storage media 5 is occupied, the corresponding number of physical segment index tables 52 are updated, and then, the record only needs to be found and deleted in the virtual disk physical segment occupancy table 54. The virtual disk can be destroyed; if the deleted object is a differential disk, the difference disk record 55 is deleted from the difference disk table; if the virtual disk of the differential disk is created, the associated disk is associated with it. All discs need to be deleted as well. Also, for the mount virtual disk command, the virtual disk drive 44 checks whether the target virtual disk already exists in the virtual disk physical segment occupying table 54 through the management segment processor 42 and checks the differential disk relation table 55. Whether it has an associated differential disk, and then loads the occupied physical segment 51 into the system. If the instructions received by the virtual disk drive 44 are not of the above types, such as may be the status query instructions of the storage medium 5, these instructions will be consistent with the traditional standard component workflow, and are not emphasized in the present invention. Some functions required for management purposes, such as enumerating all registered virtual disks or differential disks, adding or deleting new storage media 5, etc., although not described in detail, can be easily analogized from the above, nor in the present invention. Emphasized.

Referring to the first to third and sixth figures, when the virtual disk is mounted by the client computer 1, the virtual disk access command can be served. This workflow consists of a managed segment processor 42, a disk access and scheduler 43, a disk segment allocation and recycler 45, a virtual disk storage segment table register 46, and a virtual disk write protection switch. The addresser 47 is completed. When the virtual disk drive 44 receives the instruction to fetch the data from the virtual disk, the transfer is processed by the virtual disk write protection relayer 47 (as shown in FIG. 5), and the virtual disk write protection protector 47 The virtual magnetic disk can be known from the virtual disk storage segment table register 46 (faster) or the disk segment allocation and recycler 45 (slower speed because it needs to be read by the control area 511 on the storage medium 5) The disc physical sector occupation table 54 obtains all the physical segments 51 of the storage medium 5 that the virtual disk can access, and then checks the legality first, and any access instructions for the virtual disk should not exceed these. Scope, otherwise the instruction should not be executed. When the virtual disk will write new data on the legal virtual location of an unconfigured physical storage space (not exceeding the capacity range of the virtual disk), it will first be magnetic. The disc section allocation and recycler 45 requests configuration of the new entity section 51 and updates the material in the management section 511 so that the originally accessed instructions still fulfill the protection conditions and are normally executed. The above operation can calculate that the virtual disk is stored in the physical address of the storage medium 5, and then the virtual disk accessor 48 performs real data access to the storage medium 5 according to the virtual disk format; the virtual disk The format is consistent with the traditional approach and is therefore not represented in the flow chart. Furthermore, the access command is processed differently depending on whether the target disk is a differential disk. When a virtual disk is created with a different disk, the virtual disk is read-only and is protected from writing data. Protection, the written data will be transferred to its differential disk; however, there are two possibilities for reading the data: if the data that was written on the differential disk is read, the difference disk will be Read; if it is to read data that has not been written to the differential disk, it will be read directly from the virtual disk.

Finally, the virtual disk data cache and buffer 49 in the virtual disk access operation principle and the traditional cache method are not detailed, and the virtual disk data cache and buffer 49 is characterized by its operation For the read instruction, the virtual disk data of the differential disk that has been created is preferentially cached; for the write command, the data written to the differential disk is preferentially cached, and if the cache space is insufficient, The data obtained by the read command takes precedence over the data just written.

In summary, the high-performance virtual disk management system of the present invention can realize the effect of dynamic space allocation on a plurality of block-level storage media 5 without the support of the file system of the operating system for storage and management. Multiple virtual disks. Moreover, the present invention can avoid the computational performance loss of the operating system and its file system, and reduce the computational cost of an expensive file system address conversion, thereby providing high-speed access performance. Therefore, by utilizing the technology of the present invention, the storage space utilization rate of the conventional storage medium 5 technology can be improved without causing a large loss of access performance.

1. . . Client computer

2. . . network

3. . . Storage server

4. . . Virtual disk management system

41. . . Disk sector planner

42. . . Control segment processor

43. . . Disk access and scheduler

44. . . Virtual disk drive

45. . . Disk segment allocation and recycler

46. . . Virtual disk storage section table register

47. . . Virtual disk write protection translator

48. . . Virtual disk accessor

49. . . Virtual disk data cache and buffer

5. . . Storage medium

51. . . Physical section

511. . . Control section

52. . . Entity segment index table

53. . . Storage media group record table

54. . . Virtual disk entity segment occupancy table

55. . . Differential disk relationship table

The first figure is a schematic diagram of the connection storage server of the present invention.

The second figure is a block diagram of the present invention.

The third figure is the content of the control section of the present invention.

The fourth figure is a flow chart of the operation of the invention in the initialization of the storage medium.

The fifth figure is a flow chart of the operation of the invention for creating, deleting and mounting a virtual disk.

The sixth figure is a flow chart of the operation of the present invention in processing access commands.

3. . . Storage server

4. . . Virtual disk management system

41. . . Disk sector planner

42. . . Control segment processor

43. . . Disk access and scheduler

44. . . Virtual disk drive

45. . . Disk segment allocation and recycler

46. . . Virtual disk storage section table register

47. . . Virtual disk write protection translator

48. . . Virtual disk accessor

49. . . Virtual disk data cache and buffer

5. . . Storage medium

51. . . Physical section

Claims (4)

A high-performance virtual disk management system is provided with a client computer, the client computer is connected to the storage server through the network, and the storage server is connected to the storage medium of the block level through the virtual disk management system, and the virtual The disk management system is used for creating and accessing a plurality of virtual disks with dynamic space configuration effects on the storage medium, and the information package used by the virtual disk management system has a physical segment index table and a storage media group record table. a virtual disk physical segment occupation table and a differential disk relation table, wherein: the physical segment index table is a virtual disk management system with a predetermined fixed segment capacity value, and each storage medium is consistent Planning mode, planning into a plurality of entity segments, entering the generated entity segment index table; the storage media group record table is a virtual disk management system detecting the number of all connected and usable storage media in the virtual disk management system After that, each storage medium is assigned a unique identification code, and the created storage media group record table; the virtual disk physical segment occupation In order to include the virtual disk number, the physical segment index, and the virtual disk number for the virtual disk management system to process the operation command from the client computer and execute the virtual disk through the storage server, the virtual disk management system will The created virtual disk is generated by a unique pipeline code, and the physical sector index is configured to record a new physical segment sequence for the virtual disk management system; the differential disk relational table is processed by the virtual disk management system. From the client computer, when the storage server performs the operation instruction of the differential disk, the virtual disk management system sets the created differential disk with a unique serial number and associates the water code and the difference of the virtual disk. The water flow code of the disk is recorded in the differential disk relationship table. The high-performance virtual disk management system of claim 1, wherein the virtual disk management system selects a primary and a plurality of secondary physical segments to be stored on the storage medium as a control segment. Entity segment index table, storage media group record table, virtual disk entity segment occupancy table, and differential disk relationship table. The high-performance virtual disk management system according to claim 2, wherein the virtual disk management system obtains a physical segment index table by the management segment to know that those physical segments can be allocated or have been Occupied, when the virtual disk management system executes the instruction to create, delete, and mount a virtual disk, it can allocate a physical segment to the virtual disk, recycle the physical segment occupied by the virtual disk, or query the virtual disk. The occupied physical segment. The high-performance virtual disk management system according to claim 2, wherein the virtual disk management system stores the recorded content of the virtual disk physical sector occupancy table of the virtual disk in use in high-speed access. In the scratchpad, and any virtual disk physical segment occupancy table event occurs, the virtual disk physical segment occupancy table in the control segment and the scratchpad is updated.