KR20120020811A - Storage apparatus of network video recorder - Google Patents

Abstract

PURPOSE: A storage device of a network based video recorder is provided to supply a real time photographing environment even though an unexpected error occurs in system data. CONSTITUTION: A file system(310) generates a code word by channel coding to restore media data. The file system restores lost data by using the code word. The file system writes a copy of system data in a separate second storage medium. The file system restores lost data of the system using the copy.

Description

Storage Apparatus of Network-Based Video Recorder

The present invention relates to a storage device of a network-based video recorder, and more particularly, to a storage device of a network-based video recorder capable of recovering damage to media data and responding to errors in system data in real time.

Recently, due to the development of photographing technology, the development of network technology, and the decrease of costs, the video surveillance system is more widely used than before.

In other words, in addition to the general-purpose security applications, it is being used in various fields such as elementary school students' way to school, school production line monitoring, shopping mall visitor status analysis, etc. and is continuously developing in recent years. In the past, the video cassette recorder (VCR) method, which records images on a video tape by analog recording, has evolved into a DVR (Digital Video Recorder) system that changes recording media from video tape to digital media. In addition, as network technologies have been developed and infrastructures have been established, network-based video recorders (NVRs) capable of remotely accessing and controlling IP cameras and video servers have emerged as shown in FIG. 1.

On the other hand, the existing network-based video recorder has a hard disk as a storage device, although the quantity varies depending on the size of the entire system. We also used the same file system that we used in the personal PC environment.

However, these hard disks cause errors for various reasons. Types of errors can be classified into data loss and loss. Loss of data occurs due to loss of Servo area, head track deviation, SMART time limit, controller failure, etc., mainly in hardware or firmware area. Data loss errors are easy to detect because the area where the error occurs is large and clear. Therefore, much research has been done and sufficient measures have been taken.

Data loss occurs locally due to device aging, bit transmission errors, data loss due to overheating or corrosion, and loss due to scratches on the platters, which are potential errors that are not detected unless actual input / output of the data is attempted. Representative data loss errors are bad sector errors in the hard disk, which can result in complete loss of data in a particular hard disk sector.

In order to solve such a data loss error of a hard disk, a RAID (Redundant Arrays of Inexpensive Disks) method using a plurality of hard disks in general has been proposed. The RAID method is a striping method for distributing data to a plurality of hard disks to improve I / O performance, a mirroring method for making a copy, an error correction code (ECC) suitable for detecting and recovering bit errors, and a block error. There is a parity method for detecting and recovering the data. By using this, the performance and reliability of the hard disk can be improved.

However, when configuring a small system, it may be difficult to use multiple storage devices. In addition, since the RAID method is generally implemented in the device driver or firmware area, there is a problem in that a different reliability method cannot be applied depending on the characteristics of the data.

In addition, file I / O requests in a personal PC environment and file I / O requests in a network-based video recorder environment have different characteristics. Most file I / O requests in personal PC environment are small in size and randomly occur without a common pattern. However, file I / O requests of network-based video recorders are large in size and repeat a certain pattern. . Therefore, in consideration of these characteristics, a file system capable of maximizing the input / output performance of a hard disk in accordance with the characteristics of a network-based video recorder is required.

However, the network-based video recorder repeatedly performs a large amount of write operations on the storage device, which causes the storage device to easily age. As a result, the number of bad sectors increases, the possibility of data corruption increases, and image data and system data captured by an unstable network situation, a wrong operation of a user, an external shock, etc. may be lost.

However, according to the file system used in existing network-based video recorders, the loss of reliability is not guaranteed because it is not sufficiently prepared for such losses.

In addition, since network-based video recorders store and preserve video data recorded and stored in storage devices, they are used as important clues for cause analysis in the event of an accident or crime. Therefore, high reliability is required for the stored data.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems of the prior art, and an object of the present invention is to differentiate the recovery of lost data according to the data type of the storage device of the network-based video recorder as follows. That is, a codeword is generated through channel coding for media data, the damaged data is recovered using the generated codeword, and a copy of the system data is stored in a separate storage medium using a mapping table for system data. The present invention provides a storage device of a network-based video recorder capable of restoring damaged data by using a copy of stored system data, thereby enabling reliable recovery of data and maintaining system real-time.

An application program created by a user, a virtual file system generating an input / output request corresponding to an invocation of the application program, and a file system for inputting and outputting a file corresponding to the input / output request of the virtual file system. In the storage device, the file system, the media data including the video and audio to generate a codeword for the media data recovery by channel coding, recover the lost data using the generated codeword, operating system Alternatively, the system data related to a dedicated application program may perform mirroring to create a copy of the system data on a second storage medium separately from the first storage medium in which the system data is stored, and use the copy of the system data. RS to recover lost data It is a PM (Reliable Storage using Parity and Mirroring) file system.

In the storage device of a network-based video recorder according to the present invention, the generation of the codeword may include generating a first codeword by performing an XOR (Exclusive OR) operation on a plurality of striped media data, and generating the first codeword. The second codeword is generated by performing XOR operation on each corresponding coding distribution matrix and multiplying the XOR-operated data, wherein the code distribution matrix is calculated as Equation 2 below. Where I → i is the value of the matrix shifted 1 bit to the right as i increases, and Oy, y + i-1 is the value of the row and column corresponding to (y, y + i-1) is 1 and the rest It is characterized by being a matrix.

In the storage device of a network-based video recorder according to the present invention, the codeword generation of the media data may include storing the media data in a buffer in a predetermined unit and then releasing the media data stored in the predetermined unit, Another feature is that codewords are generated for a predetermined unit of media data.

In the storage device of a network-based video recorder according to the present invention, the first storage medium is a hard disk, and the second storage medium is another feature of the flash memory.

In the storage device of a network-based video recorder according to the present invention, the mirroring is another feature that a copy of system data is dynamically allocated to and stored in a second storage medium through a mapping table.

In the storage device of a network-based video recorder according to the present invention, when the system data is changed, a copy of the system data is also changed through the mapping table.

In a storage device of a network-based video recorder according to the present invention, the mapping table is another feature of a binary tree.

As described in detail above, according to the storage device of the network-based video recorder, a large amount of space is required and media data having a secondary restoration means generates code words based on a parity code and a Liberation code. By recovering the data, an average of 11.289% of data corruption can be prevented, and the data of the system data can be replaced in real time by providing a dynamic mirroring technique that can replace the data immediately. It has an effect.

1 is a diagram illustrating a conventional network-based video recorder (NVR).
2 illustrates a storage device of a network-based video recorder according to the present invention.
3 is a diagram illustrating the structure of the EXT3 file system used in Linux.
4 is a flow chart for the classification of data and the write operation of system data.
5 is a diagram illustrating delayed writing of media data in an RSPM file system according to the present invention.
6 is a flowchart showing a write operation sequence of delayed writing.
7 illustrates an example of an encoding operation of a bit matrix coding system performed in an RSPM file system according to the present invention.
8 is a diagram illustrating an example of calculating a CDM unique to a Liberation code performed in an RSPM file system according to the present invention.
9 is a diagram illustrating an example of a decoding operation for the recovery of media data performed in the RSPM file system according to the present invention.
10 illustrates an example of a binary tree used in a dynamic mirroring method in an RSPM file system according to the present invention.
FIG. 11 is an example of searching for damaged media data based on the data of FIG. 9 and Table 3. FIG.
FIG. 12 is a diagram illustrating a ratio of encoding times when assuming a 60-minute shooting time is 100%.
FIG. 13 is a diagram illustrating a ratio of an error generated in an entire hard disk by a stress test to a RSPM file system according to the present invention, and a recovery rate of a parity code and a Liberation code according to the present invention.
14 is a diagram illustrating a data corruption rate according to a Weibull distribution for the RSPM file system according to the present invention.

Hereinafter, with reference to the accompanying drawings, preferred embodiments of the present invention will be described in detail.

The video server of a network-based video recorder basically has one hard disk. Often bad sectors in the hard disk occur and some of the data is corrupted. There is no countermeasure against existing network-based video recorders. In order to overcome this problem, the present invention provides a Reliable Storage Using Parity and Mirroring (RSPM) file system in which one flash memory is additionally installed. The added flash memory stores a copy of the system data and a part of a codeword of a channel coding method. If the hard disk is damaged, the data stored in the flash memory can be repaired by using the data stored in the flash memory. The specific configuration of the present invention is as follows.

2 shows a storage device 10 of a network-based video recorder according to the invention. According to FIG. 2, the storage device 10 of the network-based video recorder according to the present invention includes an application program layer 100, a virtual file system 200, and an RSPM layer 300.

The RSPM layer 300 includes an RSPM file system 310, an RSPM block device driver 320, and an RSPM block device 330. The RSPM block device 330 includes a flash memory 331 and a hard disk ( 332).

Looking at a specific configuration, the application program layer 100 represents an application written by the user and can access the virtual file system 200 through a system call. The virtual file system 200 interacts with the RSPM file system 310 to generate an input / output request suitable for the RSPM block device driver 320. The RSPM file system 310 applies a channel coding and dynamic mirroring method such as a parity code or a Liberation code to an input / output request based on the EXT3 file system. Channel coding methods and dynamic mirroring will be described later.

The RSPM block device driver 320 controls the RSPM block device 330 that regards one flash memory 331 and one hard disk 332 as one unit device.

Here, the EXT3 file system is a Linux general-purpose file system that is adopted by many network-based video recorders operating in an embedded Linux environment. EXT3 is an advanced version of EXT2, and as shown in FIG. 3, a journaling function is added to improve reliability.

Journaling is intended for cases where errors occur during the writing of file system metadata (U), inodes (N _i ), and data blocks (N _d ), adversely affecting the entire file system.

As a detailed operation, as shown in FIG. 3, before writing data to the physical storage medium, the data to be written in the journal is copied first, and then the data writing operation is performed to the physical storage medium when the writing operation is normally completed in the journal. . If a problem occurs while writing to the journal, the write request itself is discarded to ensure the integrity of the file system. In addition, if a problem occurs during the write process to the physical storage medium, the integrity of the file system can be guaranteed because the problem is attempted to write from the journal to the physical storage medium after the problem is solved.

On the other hand, RSPM file system 310 according to the present invention can be processed according to a separate method according to the data type, unlike a typical PC environment. In other words, files of general PC environment storage devices have different roles and characteristics, and they cannot be clearly distinguished and thus cannot be processed appropriately for the characteristics of each file.

However, network-based video recorders can be broadly classified into media data including video and audio and system data necessary to maintain the system according to data input / output characteristics and capacities.

Table 1 shows the reliability method for each data type. Media data refers to the video, photo, and audio taken by the camera. It is essential to network-based video recorders.

In addition, because media data is large and there are secondary restoration measures such as frame restoration (when a particular frame is damaged, restoring the damaged frame compared to adjacent frames), smaller media sizes can be used for space efficiency and overload protection than high restoration rates. A channel coding method such as a parity code or a Liberation code that generates codewords is applied. The codeword encoded by the channel coding method is stored in the flash memory and used to recover data blocks of the hard disk when they are lost.

System data are files used by the operating system or dedicated application of the network-based video recorder and are essential files for the normal operation of the network-based video recorder. Therefore, we apply a mirroring method that writes a copy to flash memory so that it can be recovered immediately in the event of a crash. The mirroring method will also be described later.

Data type

size
Maintain real time
Channel coding
Mirroring
media

Big.
-
○
-
system

small.
○
-
○

4 is a flow chart for the classification of data and the write operation of system data. The RSPM file system classifies the type according to the directory in which the file is stored, which can be modified by the file system designer (S100).

Therefore, an application program that stores camera data must be able to arrange the data accordingly. For example, data written to the / bin, / sbin, / lib, / etc, / boot, / initrd, / sys, and / usr directories may be classified as system data.

After the data classified as the system data is checked whether the corresponding data exists in the flash memory 331 (S110), if the corresponding data exists, the write processing of the system data is simultaneously updated in the flash memory 331 according to the mirroring method (S120). ). If the corresponding data does not exist, a new area for the corresponding data is allocated to the flash memory 331 to create a mirror in the flash memory 331, and a write request is performed on the hard disk 332 to correspond thereto (S130).

Meanwhile, the media data may be a file written in / multimedia, and the characteristics of the media data and the characteristics of the hard disk used in the network-based video recorder should be considered.

In other words, the hard disk reads data by using a arm on a rotating platter. Due to this mechanical property, the performance of discontinuous read / write is poor and relatively good for continuous read / write. Therefore, the network-based video recorder should induce continuous read / write in consideration of the characteristics of the hard disk.

In the RSPM file system 310 according to the present invention, media data may be continuously written for each channel as shown in FIG. 5, and delayed writing may be used to efficiently encode codewords.

Delayed writing is a method of generating a codeword by collecting a certain amount of media data by using the main memory as a buffer and then bundling it in a certain unit when releasing it. The camera data collected in the buffer is emitted by sorting by channel number, thereby maintaining spatial locality. Codewords for delayed writing are generated using data written continuously regardless of channel. Here, the code words P and Q are stored in the flash memory.

The write operation of lazy writing is shown in FIG. Delayed writing is performed by collecting camera data A, B, C, and D from each channel and loading them until the buffer is full. The buffer is dual and can be configured to load into another buffer even when the data in the buffer is released.

When the buffer is full, the data is released, and codewords P and Q are generated and stored in the flash memory. After the write operation of the code words P and Q is completed, the actual media data is stored.

The following describes a Liberation code used for data recovery of media data in the RSPM file system 310 according to the present invention.

Liberation code is a RAID-6 based channel coding method that creates a codeword using XOR (Exclusive OR) operation on striped data, and uses this codeword to recover from data corruption.

Here, striping refers to writing or reading at the same time alternately of different data for each HDD when writing continuous data using at least two HDDs.

Liberation codes show high performance because of the small number of XOR operations performed in encoding and decoding operations among the proposed channel coding methods. Therefore, the RSPM method employs a Liberation code.

7 is a diagram illustrating an example of an encoding operation of a bit matrix coding system. According to FIG. 7, the Libration code is a binary distribution matrix (BDM) based code and generates codewords P and Q for recovering data of media data. Each vector consists of w bits, and each matrix consists of w × w bits. The I matrix of BDM is the matrix that produces the same result even if the XOR multiplication operation is performed on the identity matrix.

X _p and X _q of the BDM are code distribution matrices (CDMs) for generating codewords and used to generate codewords P and Q. P may be calculated by performing a mutual XOR operation on each striped data stored in each k hard disks. The specific calculation is based on the following equation (1). The generated codeword P may be stored in a flash memory.

Before calculating the codeword Q, it is necessary to generate a CDM unique to the Liberation code as shown in FIG. CDM may be generated by Equation 2 below.

8 is a diagram illustrating an example of calculating a CDM unique to a Liberation code. In the example of FIG. 8, it is assumed that media data is 7 bits and that media data is striped (w = 7, k = 4) on four hard disks.

In Equation 2, the first matrix X _{q , 0} is an identity matrix, and L _{→ i} indicates that the matrix value moves by one bit to the right as i increases. In addition, add the matrix O _{y , y + i- 1} where the value of the row and column of (y, y + i-1) is 1 and the value of the remaining row and column is 0.

For example, assuming each stripe as D0 = 1001101, D1 = 0010110, D2 = 1011001, and D3 = 0110111, P = 0110101 through an XOR operation between the stripes. In addition, when X (OR) Di (0≤i <k) and X _{q , i} (0≤i <k) are cross-multiplied by the XOR operation, Q = 0101101.

On the other hand, if data is corrupted, it should be possible to recover it using codewords. This operation is called a decoding operation and restores the original data by performing the operation of FIG. 7 in reverse. An example of the decoding operation is shown in FIG. As shown in FIG. 9 (intact data (survivor) + codeword) * a matrix corresponding to a damaged region of the inverse matrix of the matrix from which the portion corresponding to the damaged region of the BDM is deleted by calculating Restore the damaged area.

Next, dynamic mirroring used for data recovery of system data in the RSPM file system 310 according to the present invention will be described.

The mirroring method of the RSPM file system 310 operates in units of blocks of the RSPM file system 310 and dynamically allocates and stores a copy differently from the mirroring method generally used in the storage area.

Recently, with the development of flash memory, the storage space has been continuously expanded. However, the required storage space of the embedded-based operating system that runs the video server is not large, and the increase is not expected to be large in the future. The existing mirroring method is a 1: 1 mapping method between a device and a hardware device. When employing this method, a storage device having the same capacity as a storage device including system data should be used.

However, dynamically allocating the storage area of a copy allows you to use an area other than the area where the copy is stored, and to increase utilization by utilizing a different type of storage device than the operating system.

In the present invention, a mirror block mapping table (MBMT) is introduced for a dynamic mirroring method. MBMT is created when the flash memory where the copy is to be stored is mounted, changes are made to the copy when the original system data block is updated, and MBMT is also updated.

MBMT includes the following information. That is, the original block address (start block address of the original system data), the copy block address (start block address of the copy system data), and the length of the target block.

If the system data area is damaged and there is a problem with the read request, the replacement block should be returned as the result of the read request. For this purpose, the address of the replacement block must be known, which is recorded in the MBMT. However, since MBMT does not record all block addresses, it only records the start block address of a contiguous block. Therefore, it should be possible to search the start block of a contiguous block to which a target block belongs.

To solve this problem, MBMT is composed of a binary tree as shown in FIG. 10 illustrates a binary tree based on an original block address of MBMT as an example of a binary tree used in a dynamic mirroring method. The retrieval of the replacement block is contiguous with the damaged block and the first block described in the MBMT must be searched first.

FIG. 11 is a search example created based on the data shown in FIG. 9 and Table 3 below. The replacement block search may find a replacement block by traversing the binary tree and searching the start block of the damaged block as shown in FIG. 11.

Specifically, assuming that the damaged area is x = 0x00548378, the data value corresponding to the damaged area is compared with the data values of the first tree. Since the x value is smaller than the data value stored in 0x01435584 of the upper tree, the x value is compared with the data values stored in the lower tree connected to 0x01435584. Since the value of x is larger than the data stored in 0x00545070 and 0x00548370 of the subtree, the value of x is skipped and the value of x is smaller than the data stored in 0x01422754. Therefore, 0x00548370, which is the address immediately before 0x01422754, is obtained as the value y near the x. Next, after obtaining yx, the address difference between y and x, increase and decrease the address as much as yx in the Mirror Block of y and return it to the address of x.

The following describes the results of measuring the reliability of the RSPM block device 330 according to the present invention and the case of using a single hard disk.

Before describing the measurement results, a method of measuring reliability of a storage device will be described. The reliability unit of storage uses Mean Time To Failure (MTTF). This is the average time between a fault and the next fault and is usually included in the product's data sheet. Mean Time To Data Loss (MTTDL) is a unit that emerged from a method of securing reliability using a plurality of storage devices, and is an average time between loss of actual data and loss of next data.

All hard disks and flash memory include MTTF in hardware specifications, and in the RSPM method according to the present invention, since the flash memory and the hard disk are used at the same time, MTTDL can be calculated as in Equation 3 below.

_PrRSPM occurs when data is lost during data recovery and can occur in two cases. The hard disk fails after the flash memory fails and the flash memory fails after the hard disk fails. Therefore, the probability of data loss can be calculated as shown in Equation 4 below.

로 가정한다.

이므로 데이터 복구 중 하드 디스크가 오류를 일으킬 가능성은

이다. 이와 비슷하게

이다. 이를 수학식 4에 적용하면 하기의 수학식 5와 같다.
_Pref FLASH is the possibility that flash memory will fail first of the two devices. P _h is the probability that the hard disk will fail during data recovery and P _f is the probability that the flash memory will fail. If t _rFLASH is defined as the time taken to recover the error from the flash memory, the error rate according to the exponential distribution

Assume that

The possibility of hard disk causing errors during data recovery

to be. Similarly

to be. If this is applied to the equation (4) is equal to the following equation (5).

Applying Equation 5 to Equation 3 leads to the same conclusion as Equation 6 below.

Table 3 below is a result of comparing and calculating the MTTDL when applying the conventional single disk and the RSPM method by applying the equation (6). It is assumed here that the MTTF _HDD is 300000 hours and MTTF _FLASH = 50000 hours.

As shown in Table 3, when the RSPM block device according to the present invention is used rather than using a single disk, the MTTDL is significantly higher, indicating that the error occurrence rate is significantly reduced.

MTTDL (year)

Single disk
5.7

RSPM
5265097

Next, the performance of the RSPM file system 310 and the existing file system according to the present invention will be described.

For the experiment, the RSPM method partially modified the source code of the Linux kernel 2.6.24, and modified the EXT3 file system to build the RSPM file system. In addition, the experiment was performed by creating a video server simulator that receives camera data transmitted from an IP camera.

According to the RSPM method according to the present invention, one flash memory is utilized as a parity disk for storing codewords. However, the Liberation code applied to the RSPM method was originally written with two parity disks, which can cause unnecessary overload of the entire system.

Table 4 below compares the number of XOR operations of a parity code that generates only codeword P and a Liberation code that generates both codewords P and Q. When the number of XOR operations of a single parity code is 1, the Liberation code produces both codewords P and Q, which requires 2.2 times more operations than a single parity code. In addition, about 2.54 times more computation is needed for decoding.

Encoding
decoding

Parity  code
1.0

1.0
Liberation  code
2.225

2.54

In order to measure the effects of the Liberation code and the parity code on the performance of the network-based video recorder, the following experiment was performed. Six IP cameras were shot at 24 frames per second for 60 minutes, and they were transferred to the network and stored using a video server simulator. During 60 minutes of shooting, the Liberation code performed an encoding operation of about 187.013 seconds, and the parity code performed an encoding operation of about 89.054 seconds.

FIG. 12 is a diagram illustrating a ratio of encoding times when assuming a 60-minute shooting time is 100%. As shown in FIG. 12, an additional overload of 5.195% was applied when applying the Liberation code and about 2.474% when the parity code was applied. Liberation code generates parity code P with inherent code Q, resulting in about twice as much overhead as parity code.

However, while the load on the system during encoding operation occupies a low percentage of the load on the operation of the entire system, it can secure higher data recovery reliability than when applying Liberation code.

In addition, no additional overload was generated due to codeword generation, and the processing power of the video server simulator was enough to perform distributed processing. Therefore, even though the Liberation code and the parity code are applied to the network-based video recorder, the load on the network-based video recorder is extremely small.

The following is an evaluation of how the data coding reliability is applied to the RSPM file system 310 according to the present invention.

In the RSPM file system according to the present invention, data in the system is classified into two types, system data and media data, and a parity code and a Liberation code are applied to the media data, and a dynamic mirroring method is applied to the system data, respectively. Given k data blocks, the RSPM method with Liberation code can recover two lost blocks, and the RSPM method with parity code can recover only one lost block. Therefore, parity code is more vulnerable to continuous error than Liberation code.

FIG. 13 is a diagram illustrating a percentage of errors generated among all hard disks and a recovery rate of a parity code and a liberation code according to a kind of stress test. Liberation code showed an average recovery rate of 15.177% higher than parity code.

14 is a diagram illustrating a damage rate of data according to a Weibull distribution. The Weibull distribution is a continuous probability distribution mainly used for life data analysis, and is mainly used for life data analysis. The RSPM file system according to the present invention attempts to cope with data loss errors such as bad sectors, and the scale factor of the Weibull distribution is set to η = 9259, and the Weibull coefficient is set to β = 1 and k = 16.

As a result, the RSPM method with Liberation code showed an average damage rate of 11.29% lower than that of general EXT3 file system, and 5.21% lower than the RSPM method with parity code.

In the above description of the preferred embodiment of the present invention by way of example, the scope of the present invention is not limited to this specific embodiment, and those skilled in the art within the scope of the claims of the present invention Changes may be made as appropriate.

10: storage device of a network-based video recorder
100: Application Layer 200: Virtual File System
300: RSPM layer 310: RSPM file system
320: RSPM block device driver 330: RSPM block device
331: flash memory 332: hard disk

Claims (7)

An application program created by a user, a virtual file system generating an input / output request corresponding to an invocation of the application program, and a file system for inputting and outputting a file corresponding to the input / output request of the virtual file system. In the storage device,
The file system,
The media data including video and audio generates a codeword for the media data recovery by channel coding, and recovers the lost data using the generated codeword.
System data related to an operating system or a dedicated application may be mirrored to create a copy of the system data on a second storage medium separately from the first storage medium on which the system data is stored, and by using the copy of the system data. And a reliable storage using parity and mirroring (RSPM) file system for recovering lost data of the system.
The method of claim 1,
Generation of the codeword,
A first codeword is generated by performing XOR (Exclusive OR) operation on a plurality of striped media data.
After performing XOR operation on the striped media data and a coding distribution matrix corresponding thereto, a second codeword is generated by mutually multiplying the XOR-operated data.
The sign variance matrix is calculated as in Equation 2 below, wherein I _{→ i} is shifted by one bit to the right as i increases, and O _{y , y + i- 1} is (y, y + i- Storage device of a network-based video recorder, characterized in that the matrix corresponding to the row and column of 1) is 1 and the rest of the matrix is 0.

The method of claim 1,
Codeword generation of the media data,
And storing the media data in a buffer for a predetermined unit, releasing the stored media data, and generating a codeword for the released predetermined media data.
The method of claim 1,
And wherein the first storage medium is a hard disk and the second storage medium is a flash memory.
The method according to claim 1 or 4,
The mirroring is a storage device of a network-based video recorder, characterized in that a copy of the system data is dynamically allocated to the second storage medium through a mapping table (Mapping Table).
The method of claim 5,
And when the system data is changed, a copy of the system data is changed through the mapping table.
The method of claim 5,
The mapping table is a storage device of a network-based video recorder, characterized in that consisting of a binary tree.

Images