KR20070061259A - Object-based storage system using pmem useful for high speed transmission with dma and method thereof - Google Patents

Abstract

An object-based storage system using a PMEM capable of high speed transfer applied to a DMA mode and a transfer method thereof are provided to perform the high speed transfer among a block storage device, a network device, and the PMEM through the DMA mode by installing the PEMM and a dedicated bus between the block storage device and the network device. Each user process part(210) performs a file transfer service. A data storing part(230) stores object data transferred to the network device and permits DMA access to the user process part. A driver part(220) manages the data storing part to enable the DMA access by reporting a storage state. The data storing part includes a block storing part storing the object data in a block unit, a network interface part connecting to the network to transfer the object data from the user process part, and a PEMM part. The PEMM part temporarily stores and outputs the object data in the block unit by respectively connecting to the block storing part and the network interface part through the DMA mode.

Description

Object-based storage system using PMEM useful for high speed transmission with DMA and method approximately}

1 is a diagram showing the overall configuration of an object-based storage system.

FIG. 2 is a block diagram illustrating a configuration of an object-based storage system using PMEM (PCI Memory) capable of high-speed transmission using the direct access method according to the present invention.

3 is a detailed functional block diagram of the data storage unit of FIG. 2.

4 is a block diagram showing another configuration of an object-based storage system using PMEM capable of high-speed transmission using the direct access method according to the present invention.

5 is a block diagram illustrating a configuration of a hardware platform of FIG. 4.

6 is a view for explaining a PMEM block arrangement according to the present invention.

7 is a layout view of available lists for PMEM block management according to the present invention.

8 is a layout view of allocation list for PMEM block management according to the present invention.

9 is a diagram for explaining zero-copy of a user process according to the present invention.

10 is a view for explaining a file write zero-copy using PMEM according to the present invention.

11 is a flowchart illustrating a process of a transmission method according to the present invention.

12 is a flowchart showing the process of the transmission method according to the present invention in more detail.

The present invention relates to an object-based storage system using a PMEM (PCI Memory) capable of high-speed transmission using a direct access method, and a data transmission method in the storage system, and more particularly to a user file manager in an object-based storage system. The present invention relates to a system and a method capable of minimizing the load on the CPU and the system bus by placing a PMEM device between these devices in order to directly transfer the data from the block-based storage device to the network device.

Conventionally, systems consisting of storage area network (SAN) storage and dozens of high-performance CPUs have been implemented through clustered middleware software. Another method is to provide efficient object service through configuration and arrangement for a specific file type (eg, large multimedia object) service. However, since there is a storage structure optimized for a specific transmission, there is a problem that the generality is poor because it cannot support several types. In addition, when the data transmission process from a general computer system to the network, there is a problem that spends a lot of time to process the data copy of the user address space and the operating system address space.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and includes a PMEM memory device between a block storage device and a network device, and a dedicated bus independent of the CPU between the block storage device and the network device. The present invention provides a system capable of high-speed transfer through a direct transfer method (DMA) between PMEM memory devices and a transfer method using the system.

In order to achieve the above technical problem, a storage system using PMEM (PCI Memory) capable of high-speed transmission using the direct access method according to the present invention includes at least one user processor for performing a file transfer service; A data storage unit for storing object data for transmission to a network device and accessible by the user process unit in a direct connection method; And a driver unit for managing the data storage unit to inform the storage state so that the direct access method can be accessed.

In order to achieve the above technical problem, a transmission method in a storage system capable of high-speed transmission using the direct access method according to the present invention includes a block-based storage device and a PMEM connected to a bus independent of a system bus and an input / output bus inside the system. In the method for performing file transfer in a storage system capable of high-speed transfer using a direct access method including a network interface device, an available list listing available blocks among at least one block of the PMEM and data Providing block information including a list of the full blocks as an allocation list; Searching for an available block by referring to the available list when a file write request is received from a user process, mapping the user block to a user space address, and updating the block information if present; Transmitting, by the user process, the object data stored in the storage device based on the user space address to the mapped available block and then through the network interface; And if there is no data requiring further transmission, releasing the address space of the mapped available blocks and deleting them from the allocation list to return to the available block state.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. FIG. 1 is a diagram illustrating an overall configuration of an object-based storage system. As shown in FIG. 1, an object-based storage system includes n user file managers 100 and 101, a metadata management server 110, and n ′ object-based storage devices. 120,121). The user file managers 100 and 101 receive a user's file-related request and request metadata from a metadata management server and actual data of a file from object-based storage devices. The metadata management server that manages the configuration information of the file composed of objects checks the validity of the user file manager request and sends the requested metadata (that is, the object configuration information of the file) through the network. Similarly, object-based storage that manages objects passes object data to the user's file manager.

FIG. 2 is a block diagram showing the configuration of an object-based storage system using PMEM (PCI Memory) capable of high-speed transmission according to the present invention. FIG. 3 is a detailed functional block diagram of the data storage unit of FIG. . First, referring to FIG. 2, the user processor 210 performs a file transfer service, and the data storage 230 stores object data for transmission to a network device connected to an object-based storage system. 220 manages the data storage unit 230 to inform the storage state of the data to enable direct access (DMA) access. In addition, the user processor 210 accesses the data storage 230 in a direct connection manner through an input / output bus independent of the system bus.

Look at the detailed configuration of the data storage unit 230. First, the block storage unit 310 stores the object data in units of blocks. The network interface 320 performs a function of accessing a network in order to transmit object data to be transmitted by the user processor 210. The PMEM unit 330 is connected to the block storage unit 310 and the network interface unit 320 in a direct connection method, respectively, and temporarily stores and outputs the object data in block units. The PMEM unit 330 may include a block information unit 331 containing information including a start address of the block, a free block unit 333 indicating whether a block of an address referred to by the block information unit is empty, and the data is different from each other. It is composed of an allocation block unit 335 that informs the list of blocks, and ultimately divides the memory into one or more block units, and provides the user processor 210 with the availability of each block unit memory. The PMEM unit 330 determines the block unit (eg, 128 KByte, 256 Kbyte, ..., etc.) according to the characteristics of the object data to be stored.

Now, look in more detail with reference to Figures 4 to 10. FIG. 4 is a block diagram showing another configuration of an object-based storage system using PMEM capable of high-speed transmission using the direct access method according to the present invention. FIG. 5 is a block diagram showing the configuration of the hardware platform of FIG. 6 to 8 relate to a PMEM block, FIG. 6 is a diagram illustrating a PMEM block arrangement according to the present invention, FIG. 7 is an available list arrangement diagram for PMEM block management according to the present invention, and FIG. An allocation list arrangement diagram for managing PMEM blocks according to the present invention. 9 is a view for explaining the zero-copy of the user process according to the present invention, Figure 10 is a view for explaining a file write zero-copy using PMEM according to the present invention.

Referring to Figure 4 may also be divided into the software side (400 to 460) and the hardware side (470). In terms of software, there are a PMEM driver 440 that manages PMEM and a block device driver 430 that controls block data. The TCP / IP Offload Engine (TOE) driver 460 uses the PMEM block to directly transmit the PMEM data through the network in a zero-copy manner. The iSCSI initiator driver 450 transmits object data to the object-based storage device through zero-copy through the TOE driver 460. The MEM character driver 410 delivers a PMEM driver service request in an OS address space to user processes 400 and 401 providing a file transfer service. In addition, there is a / proc 420 to inform the user process about multiple PMEM memory devices.

A more detailed configuration of the hardware aspect is shown in FIG. A PMEM memory device 511 is provided in the block-based storage device 510 and the network device 509, and a dedicated bus 506 independent of the host CPUs 500 to 501 is disposed between the PMEM memory devices. It is possible to transmit at high speed through the direct transfer method (DMA) without interfering with the CPU 500 to 501 and the system buses 502 and 504 used by the CPU.

The PMEM (600 to 630) memory is divided into blocks (601 to 606, 611 to 616, 621 to 626, 631 to 636) of various sizes (2M, 1M, 512K, 256K, and 128K) according to the characteristics of the object. Manage as follows. The PMEM block size is a unit of direct transfer (DMA) transfer from the PMEM memory to the block storage device 510 or a network device.

More than one user process must be able to access the PMEM memory device simultaneously to provide file transfer services. If a PMEM block reads / writes several user processes at the same time, the common resource PMEM block cannot contain accurate information. For this purpose, information about a PMEM block such as an owner identifier and a physical start address is expressed through pmem_list_entry (701 to 704). Each pmem_list_entry 701 to 704 designates blocks 711 to 715 of PMEM, respectively. For unused PMEM blocks, pmem_list_entry for this is put in free_list_head (700). This free_list_head 700 manages only the PMEM block for each PMEM device. This allows the user processes to know which PMEM blocks are available on which PMEM devices. The PMEM blocks 821, 822, and 831 allocated among the blocks 821 to 835 of the PMEM take out pmem_list_entry 810 to 812 from the free_list_head, and put them in the alloc_list_head 800 to manage them as shown in FIG. Alloc_list_head 800 has only one list different from free_list_head 700 shown in FIG. 7.

PMEM blocks allocated through the PMEM driver cannot be directly accessed without copying into the user address space 910 without going through the operating system address space 920. For this purpose, the PMEM 931 blocks allocated from the PMEM blocks 930, 931 and 932 in the PMEM physical address space are provided to the file transfer service processes by using a system call called mmap. This approach can eliminate copying from the user address space 910 to the operating system and copying from the operating system to the user.

10 shows a file transfer service process sending object data via PMEM block 1041. This approach allows user processes that are allocated and directly accessible to the PMEM block to transmit via direct transfer (DMA) without interfering with the CPU, the system bus used by the CPU, and other I / O buses.

Referring to FIG. 7, the object data 1040, 1041, 1042 in the block storage device 1033 through the PMEM block 1034 is transferred to the direct transfer method through the network interface card 1035 by reading and writing commands. You can see how. Each block storage device 1033 is connected by a block storage controller 1031 and a PMEM 1034 by a PMEM controller 1032 to the network interface card 1035 by a direct connection method. Since the mapping between the PMEM blocks 1010, 1011, and 1012 and the hardware PMEM block 1034 in the user address space 1001 and the PMEM physical address space is the same as in FIG. 9, description thereof is omitted.

Now, a preferred embodiment of a transmission method according to the present invention will be described with reference to FIGS. 11 to 12. 11 is a flowchart illustrating a process of the transmission method according to the present invention, and FIG. 12 is a flowchart showing the process of the transmission method according to the present invention in more detail.

As described above, a PMEM capable of high-speed transfer using a direct access method including a block-based storage device, a PMEM (PCI Memory), and a network interface device connected to a bus independent of the system bus and the I / O bus inside the system is provided. In the storage system used, first, the PMEM is divided into at least one block according to the characteristics of object data, and then an available list listing the available blocks among at least one block of the PMEM and a list of blocks filled with data are listed. The block information included in the allocated allocation list is prepared (1111).

Next, if there is a file write request from a user process, it is determined whether the available block exists, and if so, the information of the available block is transferred to the allocation list and deleted from the available list, and then the physical of the available block is deleted. After the start address is mapped to the user process space and the user space address mapped to the user process is provided, the block information is updated (1113).

Now, the user process transmits the object data stored in the storage device to the mapped available block based on the user space address and then transmits it through the network interface, but directly from the storage device to the mapped available block. The transmission method 1115. Finally, if there is no data requiring further transmission, the address space of the mapped available block is released, deleted from the allocation list, returned to the available block state, and returned to the initial state (1117).

Now, look at in more detail with reference to FIG. When a write request for a file comes from a file transfer user process (1200), the metadata management server requests information about the requested object (1201). These information files contain information about which objects are organized in which object-based storage devices. It is checked whether the object exists in the object-based storage device (1202). If not, a message is generated (1204) that the request does not exist for the file write request. If the requested object exists, a PMEM block allocation request is sent to the PMEM driver through the PMEM character driver (1203). Here, the PMEM driver retrieves the first available PMEM block from the free_list_head 700 (1205 to 1206), and, if there is, free_pmem_entry for it, puts it into alloc_list_head 800 (1207 to 1209). The PMEM character driver gives the start address of the mapped user process space to the user process requesting the writing of the PMEM request file through mmap (1211 to 1213). This address allows the user process to access the allocated PMEM block. However, there is a case where all the limited resources, PMEM blocks, are used up and cannot be allocated. In this case, wait for the given time and request the PMEM block again (1208).

The file transfer user process that can access the allocated PMEM block sends a read request as many as PMEM blocks to the block storage device driver in which data to be transferred is stored (1215). The device driver instructs the storage device to transmit a PMEM block to the PMEM block through a direct transfer method (DMA) (1217). The PMEM block receiving the data transmitted by the direct transmission method (DMA) sends a command to the TOE driver to transmit the data directly from the PMEM block to the network device by the direct transmission method (DMA) (1218). If the size of the requested file is larger than the size of the allocated PMEM block, it is continuously read to disk and exported to the network device by DMA (1219). After the service for the write request of the object is finished, the user address area of the mapped PMEM block is released through unmap (1210), and then all_oclist_head (800) finds list_pmem_entry of the allocated PMEM block and free_list_head (700) to which the PMEM block belongs. To release the PMEM block (1212). After all of these procedures have been successfully completed, a message is generated (1214) that the user has successfully read, and the procedure is terminated by sending them to the user (1216).

The method according to the invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD_ROM, magnetic tape, floppy disks, and optical data storage, and may also include those implemented in the form of carrier waves (eg, transmission over the Internet). . The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

As described above, optimal embodiments have been disclosed in the drawings and the specification. Although specific terms have been used herein, they are used only for the purpose of describing the present invention and are not intended to limit the scope of the present invention as defined in the claims or the claims. Therefore, those skilled in the art will understand that various modifications and equivalent other embodiments are possible from this. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

As described above, a storage system capable of high-speed transmission using the direct access method according to the present invention and a transmission method in the system, when the user file manager of the object-based storage system is implemented, the interference between the CPU and the system bus In order to write object data from the storage device to the network device with the minimum amount of data, a PMEM device is placed between the block-based storage device and the network device, and the user process can access the managed PMEM memory and transfer it directly. Through the DMA, the CPU load can be reduced, and a high performance user file manager can be realized at a relatively low cost.

Claims (12)

At least one user processor for performing a file transfer service; A data storage unit for storing object data for transmission to a network device and accessible by the user process unit in a direct connection method; And And a driver unit for managing the data storage unit to inform the storage state so that the direct access method can be accessed. The storage system capable of high-speed transmission using PMEM (PCI Memory) to apply the direct access method. The method of claim 1, wherein the data storage unit Block storage unit for storing the object data in block units; A network interface unit connected to a network to access the network for transmitting the object data to be transmitted by the user processor; And A PMEM unit configured to connect to the block storage unit and the network interface unit in a direct connection method, respectively, and temporarily store and output the object data in block units; Storage system using PCI Memory. The method of claim 1, wherein the user processor A storage system using a PMEM (PCI Memory) capable of high-speed transfer applying a direct connection method characterized in that the data storage unit is accessed by a direct connection method with an input / output bus independent of a system bus. The method of claim 2, wherein the PMEM unit A storage system using a PMEM (PCI Memory) capable of high-speed transfer applying a direct access method characterized by dividing a memory into at least one block unit and providing the availability of each block unit memory to the user processor. The method of claim 4, wherein the PMEM unit A storage system using PMEM (PCI Memory) capable of high-speed transmission applying the direct access method characterized in that the block unit is determined according to the characteristics of object data. The method of claim 4, wherein the PMEM unit A block information unit containing information including a start address of the block; A free block unit for indicating whether a block of an address referred to by the block information unit is empty; And Allocating block unit for indicating a list of blocks filled with data; Storage system using a PMEM (PCI Memory) capable of high-speed transmission applying a direct access method comprising a. In a storage system using PMEM that enables high-speed transfer using a direct access method that includes a block-based storage device connected to the system bus and an I / O bus inside the system, and a PMEM (PCI Memory) and a network interface device. In the method of performing the transmission, (a) having an available list listing available blocks among at least one block of the PMEM, and block information including an allocation list listing the blocks filled with data; (b) searching for an available block by referring to the available list when a file write request is received from a user process, mapping the user block to a user space address if present, and updating the block information; (c) the user process transmitting object data stored in the storage device to the mapped available block based on the user space address and then transmitting the data through the network interface; And (d) if there is no data requiring further transmission, releasing the address space of the mapped available block and deleting it from the allocation list to return to the available block state; Transmission method in storage system using PMEM (PCI Memory) which enables high speed transmission. The method of claim 7, wherein step (a) Dividing the PMEM into at least one block according to the characteristics of the object data; A transmission method in a storage system using a PMEM (PCI Memory) capable of high-speed transmission applying a direct access method characterized in that it comprises a. The method of claim 7, wherein The step of transmitting the direct access method further comprises the step; the step of proceeding the transmission process only if the request is obtained by requesting the object information about the file to the server managing the metadata before the step (a) Transfer method in a storage system using PMEM (PCI Memory). The method of claim 7, wherein step (b) (b1) determining whether the available block exists; (b2) if present, transplanting information of the available blocks to the allocation list and deleting them from the available list; And (b3) mapping a physical start address of the available block to a user process space and providing a user space address mapped to the user process; PMEM capable of high speed transmission using a direct access method comprising: Transmission method in storage system using (PCI Memory). The method of claim 7, wherein step (c) The transmission method in a storage system using a PMEM (PCI Memory) capable of high-speed transmission, characterized in that the transmission from the storage device to the mapped available block by a direct transmission method. In a storage system using PMEM (PCI Memory) that enables high-speed transfer using a direct connection method that includes a block-based storage device connected to the system bus and an I / O bus inside the system, and a PMEM and a network interface device. In the transmission method, (a) having an available list listing available blocks among at least one block of the PMEM, and block information including an allocation list listing the blocks filled with data; (b) searching for an available block by referring to the available list when a file write request is received from a user process, mapping the user block to a user space address if present, and updating the block information; (c) the user process transmitting object data stored in the storage device to the mapped available block based on the user space address and then transmitting the data through the network interface; And (d) if there is no data requiring further transmission, releasing the address space of the mapped available block and deleting it from the allocation list to return to the available block state; A computer-readable storage medium that records a program that enables a computer to execute a transfer method in a storage system using PMEM (PCI Memory) capable of high speed transfer.

Images