KR102015188B1 - Method for probabilistic file optial recovering in helper nodes with limited cache memory and distributed storage method and apparatus - Google Patents

Abstract

A stochastic file optimal recovery method and a file distributed storage method and apparatus in a secondary node having limited memory are disclosed. In a method for distributing and storing files in a plurality of help nodes belonging to a cluster, a limited memory size and success transfer probability for each specific node are targeted to a specific file to be divided into a plurality of file pieces. determining a number to be divided based on a transmission probability, dividing the specific file into a plurality of file pieces based on maximum distance separation (MDS) coding based on the determined number, and dividing the divided file pieces The method may include distributedly storing the plurality of secondary nodes.

Description

TECHNICAL FOR PROBABILISTIC FILE OPTIAL RECOVERING IN HELPER NODES WITH LIMITED CACHE MEMORY AND DISTRIBUTED STORAGE METHOD AND APPARATUS}

The present invention relates to a technique for providing a file requested by a user terminal belonging to a cluster based on cooperative transmission between demodulation nodes belonging to a cluster. More specifically, it is composed of a plurality of helper nodes having a limited size of cache memory, a user equipment (UE) and a base station (UE) receiving files from the auxiliary nodes. In a wireless communication environment, the present invention relates to a technique for deciding how many pieces of a particular file should be divided and distributed to secondary nodes in order to maximize transmission gain in cooperative transmission.

Recently, as the Internet of Things (IOT) and big data era enter, the consumption of mobile video is exploding, and the demand for mobile video is not expected to stop. According to Cisco, global mobile data volume reached 7.2 exabytes (EB) at the end of 2016, and from these trends, more than 700% of traffic is expected between 2016-2021, resulting in more than 700% of zettabytes per year. zetabyte) shows that the arrival is not long. Due to the rapidly growing mobile traffic, a technique called caching is being used to prefetch frequently used contents and load them into a device having a memory.

Caching can alleviate the load on the backhaul link from the upper base station, and preload data that may be required from the user from the lower layer base station in nano or pico units. It is also expected to improve.

 However, it is not easy to improve the performance or quality of the memory device due to soaring data volume due to social or technical factors such as price. Accordingly, research on how to efficiently load data within a limited memory size is required. In addition, a method for more reliably transmitting desired data or files through cooperation between multiple auxiliary nodes is being studied, rather than a single node closest to a user being solely responsible for the service for the user.

In the wireless caching method, same file caching, which is a technique of storing a file in the same manner for each base station, was first noticed with uncoded caching, which stores the file itself as it is, but the same file caching method may serve an end user. There is a disadvantage in that various files may not be transmitted from various base stations, and unnecessary files may be repeatedly transmitted several times. To overcome this drawback, different file caching methods have been developed to load different files for each base station. [1] JY Song, HJ Song and W. Choi , "Optimal content placement for wireless femto- caching network," in IEEE Transactions on Wireless Communications, vol. 16, no. 7, pp. 4433-4444, Jul . According to the different file caching method presented in 2017 , the bit error rate (BER) can be minimized in consideration of the tradeoff relationship between the channel diversity as well as the file diversity. In addition, non-patent literature [2] SH Chae , W. Choi , "Caching placement in stochastic wireless caching helper networks: channel selection diversity via caching," in IEEE Transactions on Wireless Communications, vol. 15, no. 10, pp. 6626-6637, Oct. The probabilistic caching method proposed in 2016. is a technique that stores stochastic different files in limited memory, and shows better performance than the existing method.

However, rather than uncoded caching, which saves the whole file, coded caching has been developed in which a secondary node partially stores a part of the file so that the file is stably provided to the user while avoiding duplication of the file. For example, MDS (Maximum Distance Separable) code, which is easy to implement among coded caching techniques, divides the file itself into several file fragments and encodes based on the divided fragments so that a certain number of file fragments are transmitted to the user terminal. When received from, it is a coding technique to recover the file. [3] X. Xu and M. Tao, "Modeling, Analsis , and optimization of coded caching in small-cell networks," in IEEE Transactios on Communications, vol. 65, no. 8, pp. 3415-3428, May 2017. In this paper, we propose the same file coded caching technique where several auxiliary nodes store different packets of the same file and cooperatively transmit it.The same file coded caching technique can serve the end user more efficiently than the uncoded caching technique. Suggests that there is.

In the existing uncoded caching technique, coded different file caching technique is similar to coded caching technique with more file diversity gain than other file caching technique. It can be expected to be superior to the file caching technique.

Therefore, there is a need for a technology capable of optimally serving end users in coded different file caching techniques.

Korean Patent Laid-Open Publication No. 10-2010-00048130 relates to a cluster-based distributed storage system and a method of operation thereof, and describes a technique of multicasting a cluster after distributing and storing data in a cache.

[1] J. Y. Song, H. J. Song and W. Choi, "Optimal content placement for wireless femto-caching network," in IEEE Transactions on Wireless Communications, vol. 16, no. 7, pp. 4433-4444, Jul. 2017. [2] S. H. Chae, W. Choi, "Caching placement in stochastic wireless caching helper networks: channel selection diversity via caching," in IEEE Transactions on Wireless Communications, vol. 15, no. 10, pp. 6626-6637, Oct. 2016. [3] X. Xu and M. Tao, "Modeling, Analsis, and optimization of coded caching in small-cell networks," in IEEE Transactios on Communications, vol. 65, no. 8, pp. 3415-3428, May 2017.

The present invention relates to a technique for determining whether an end user can be optimally serviced when a file is divided into several files and cooperatively transmitted in a coded different file caching technique. In other words, files are stored only on secondary nodes that store MDS coded packets (ie, fragmented file fragments based on Maximum Distance Separable (MDS) coding) of different files through probabilistic caching. The present invention relates to a technique for determining whether a transmission gain is maximized when cooperating transmission by dividing into several pieces.

In a method for distributing and storing files in a plurality of help nodes belonging to a cluster, a limited memory size and success transfer probability for each specific node are targeted to a specific file to be divided into a plurality of file pieces. determining a number to be divided based on a transmission probability, dividing the specific file into a plurality of file pieces based on maximum distance separation (MDS) coding based on the determined number, and dividing the divided file pieces The method may include distributedly storing the plurality of secondary nodes.

According to one aspect, the limited memory size for each secondary node may be smaller than the size of the specific file.

According to another aspect, the determining of the number to be split may determine the number to be split based on the number of auxiliary nodes capable of storing the file fragment of the specific file to be requested in the user terminal belonging to the cluster. .

According to another aspect, the determining of the number to be divided, if the number of auxiliary nodes that can store the file fragments of the specific file is less than the number of pieces of files to be divided, the specific file without partitioning the specific file And determining to distribute the entirety to the secondary node.

According to another aspect, the determining of the number to be divided comprises: calculating successive transmission probability for each number of file fragments to be divided when the specific file is divided into a plurality of file fragments, the calculated Updating a maximum success transfer probability to a maximum value among success transfer probabilities, and targeting a file hitting probability in the cluster that changes as the specific file is divided, and has a file corresponding to the maximum success transfer probability Determining the probability, and determining the number of file fragments to be divided corresponding to the determined file existence probability.

According to another aspect, the determining of the number to be divided includes: specifying a search range in a range from when the number of file fragments is 1 to the number of file fragments to be divided corresponding to the determined file existence probability; And finally determining the number of file fragments having a maximum transmission probability within the specified search range as the number to be divided.

According to another aspect, the file hitting probability indicating the probability that the specific file or a part of the specific file is stored in a cache of a plurality of secondary nodes belonging to the cluster is a target of the specific file. As the number of divided file fragments increases, the size of the file fragments can be relatively reduced.

According to another aspect, the determining of the number to be divided may include: when the number of auxiliary nodes capable of storing file fragments of the specific file is equal to or greater than the number of file fragments to be divided, the specific file within a predetermined maximum probability range. May be determined as the number to be divided.

According to another aspect, as the request for providing the specific file is requested from a user terminal belonging to the cluster, the specific file is based on the cooperative transmission between the plurality of auxiliary nodes storing the file fragment in which the specific file is divided. It may be provided to the user terminal.

In a file distributed storage device for distributing and storing files in a plurality of help nodes belonging to a cluster, a limited memory size and success transfer for each specific node are directed to a specific file to be divided into a plurality of file fragments. A determination unit for determining a number to be divided based on a successive transmission probability, a file division unit for dividing the specific file into a plurality of file pieces based on MDS (Maximum Distance Separable) coding based on the determined number; And a storage controller configured to distribute and store the divided file fragments in the plurality of auxiliary nodes.

According to one aspect, the limited memory size for each secondary node may be smaller than the size of the specific file.

According to another aspect, the determination unit may determine the number to be divided based on the number of auxiliary nodes capable of storing the file fragment of the specific file to be requested in the user terminal belonging to the cluster.

According to another aspect, when the number of auxiliary nodes capable of storing file fragments of the specific file is smaller than the number of file fragments to be divided, the determining unit may store and store the entire specific file in the auxiliary node without dividing the specific file. Can be.

According to another aspect, the determining unit, when the specific file is divided into a plurality of file fragments, calculates a successful transmission probability for each number of file fragments to be divided, and as a maximum value of the calculated successful transmission probability Update a maximum success transfer probability, determine a file existence probability corresponding to the maximum success transfer probability, and target a file hitting probability in the cluster that changes as the specific file is divided, and determine the determined file existence probability The number of file fragments to be divided corresponding to may be determined.

According to another aspect, the determining unit specifies a search range in a range from when the number of file fragments is 1 to the number of file fragments to be divided corresponding to the determined file existence probability, and succeeds within the specified search range. The number of file fragments having the maximum transmission probability may be finally determined as the number to be divided.

According to another aspect, the file hitting probability indicating the probability that the specific file or a part of the specific file is stored in a cache of a plurality of secondary nodes belonging to the cluster is a target of the specific file. As the number of divided file fragments increases, the size of the file fragments can be relatively reduced.

According to another aspect, the determining unit, when the number of auxiliary nodes capable of storing the file fragments of the specific file is equal to or greater than the number of file fragments to be divided, the determination unit may determine the maximum number of possible division of the specific file within a predetermined maximum probability range. It can be determined by the number to be divided.

According to another aspect, as the request for providing the specific file is requested from a user terminal belonging to the cluster, the specific file is based on the cooperative transmission between the plurality of auxiliary nodes storing the file fragment in which the specific file is divided. It may be provided to the user terminal.

According to embodiments of the present invention, the number of file fragments for dividing a specific file is determined in consideration of the successive transmission probability, and the file fragments (eg, MDS codes) of the specific file are divided based on the determined number. ) Is distributed to secondary nodes, and by providing the corresponding file to the user terminal through cooperative transmission between the auxiliary nodes to the user terminal requesting the file, the transmission gain can be maximized.

1 is a diagram for a wireless communication environment for performing cooperative transmission between secondary nodes with limited memory according to an embodiment of the present invention.
2 is a flowchart illustrating a relationship for file distribution storage between a base station, an auxiliary node, and a user terminal according to an embodiment of the present invention.
3 is a flowchart illustrating a method of determining a file fragment of a specific file based on a successful transmission probability according to an embodiment of the present invention.
4 is a block diagram illustrating an internal configuration of a file distributed storage device according to an embodiment of the present invention.
5 is a diagram illustrating a graph of a file existence probability and a successful transmission probability according to one embodiment of the present invention.
6 is a graph illustrating a MAC capacity region according to an embodiment of the present invention.

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

According to the present invention, in a wireless communication environment in which two or more auxiliary nodes having a limited memory size and a single user terminal form a communication cluster, a file successfully requested by the user terminal by one transmission through cooperative transmission between the auxiliary nodes is provided. It relates to a technique for distributing, storing and recovering files so that they can be provided. In particular, the present invention relates to a technique for determining the number of file fragments (ie, MDS code length) in which the transmission gain is maximized through cooperative transmission between secondary nodes when the file is divided into several fragments and distributed to the secondary nodes. It is about.

For example, the present invention provides MDS (Maximum Distance Separable) coding to distribute and store at least one file among a plurality of files stored in a base station to a plurality of secondary nodes corresponding to a specific cluster among clusters belonging to the base station. Can be used. In addition, the present invention is the number of file fragments, that is, the length of the MDS code when distributed storage of the file fragments (that is, MDS code) divided by the MDS coding in each of a plurality of secondary nodes having a limited cache memory It is about the technique of obtaining size.

In the present embodiments, a transmitting node represents a base station, and a helper node is a relay device such as a gateway existing between the base station and user equipment, and the burden of providing data to the receiving node by the base station. It can be used to perform an auxiliary function to relieve the problem. For example, the transmitting node may be represented as a file distributed storage device. The receiving node indicates a user terminal (UE) that receives a packet corresponding to a request file (ie, a file fragment) from at least one of a transmitting node (ie, a base station) belonging to the cluster and auxiliary nodes belonging to the cluster. Can be. For example, the receiving node may include a smartphone, a tablet PC, a notebook, and the like.

In addition, the secondary node has a cache memory and stores in advance files frequently requested by a user station (UE) from a base station corresponding to a cluster to which the secondary node belongs, that is, a transmitting node. You may be doing For example, the files may include data files stored in a general file system, multimedia data such as video and music, control data for connection of a communication system, and the like.

In the present embodiments, the file hitting probability may indicate a probability that the entire file or a part of the file requested by the user terminal (UE) is present (ie, stored) in the secondary node's cache.

In the present embodiments, the secondary node has limited cache memory, wherein the limited cache memory has a larger amount of memory space allocated for storing a particular file or portion of a particular file on the secondary node than the total size of the file. It can represent a small memory.

이내에 해당하는 영역을 나타낼 수 있다. In the present embodiments, the cluster is a predetermined maximum radius around the user terminal (UE)

The corresponding area can be indicated within the range.

1 is a diagram for a wireless communication environment for performing cooperative transmission between secondary nodes with limited memory according to an embodiment of the present invention.

라고 할 때, 해당 반경 내에 속하는 보조 노드, 즉, 클러스터에 속하는 보조 노드 간 협력 전송을 수행하는 무선 통신 환경을 도시한 도면이다.That is, FIG. 1 shows a maximum radius that can serve a UE (ie, provide a file requested by the UE).

In this case, it is a diagram showing a wireless communication environment for performing cooperative transmission between secondary nodes belonging to the radius, that is, secondary nodes belonging to the cluster.

According to FIG. 1, a state capable of recovering a file requested by a terminal during cooperative transmission between auxiliary nodes may be divided according to the number of auxiliary nodes belonging to a cluster.

In FIG. 1, it is assumed that a specific file is divided into four equal sizes. For example, MDS coding can generate a large number of file fragments (i.e., packets) without duplication using logical operations such as XOR. Four or more types of packets must be received by the terminal to recover a file.

)에 속하는 보조 노드의 개수가 4개 미만인 경우, 단말(114)이 요청한 파일을 복구 가능한 상태가 not hit에 해당할 수 있다. 즉, not hit는 상기 보조 노드들(111, 112, 113) 간의 협력 전송을 통해서 하나의 특정 파일을 복구할 수 없는 경우에 해당할 수 있다.Referring to 110, if a particular file is divided into four equal sizes, the cluster (i.e. radius)

If the number of auxiliary nodes belonging to the subordinate) is less than four, the state in which the terminal 114 may recover the requested file may correspond to not hit. That is, not hit may correspond to a case in which one specific file cannot be recovered through cooperative transmission between the auxiliary nodes 111, 112, and 113.

) 내에 속하는 보조 노드(121,122,123,124)의 개수가 분할된 파일 조각의 개수인 4개 이상인 경우, 해당 파일이 복구될 수 있다. 이에 따라, 보조 노드의 개수가 분할된 파일 조각의 개수 이상인 경우는 해당 파일을 복구 할 수 있는 전제 조건으로서 히팅 이벤트(hitting event)로 정의될 수 있다.At this time, as in 120, the cluster (i.e. radius

If the number of subnodes 121, 122, 123, and 124 belonging to the subframe is 4 or more, that is, the number of divided file fragments, the corresponding file may be recovered. Accordingly, when the number of auxiliary nodes is equal to or greater than the number of split file fragments, the auxiliary node may be defined as a heating event as a precondition for recovering the corresponding file.

) 내에 4개 이상의 보조 노드들(131, 132, 133, 134)이 존재하더라도, 경로 손실, 페이딩, 간섭 등의 채널 조건으로 인해 4개가 아닌 3개 이하의 일부 보조 노드만이 전송 가능한 상황이 발생할 수 있다. 이처럼, 클러스터에 속하는 보조 노드의 개수는 분할된 파일 조각의 개수(예컨대, 4개) 이상이지만, 채널 조건으로 인해 일부 보조 노드만(즉, 분할된 파일 조각의 개수 미만의 보조 노드만)이 실제 전송이 가능한 상황인 경우를 비성공 전송 이벤트(not successful transmission event)로 나타낼 수 있다. In the wireless communication environment, power from the secondary nodes to the user terminal meets a certain threshold or more necessary for transmission due to path loss and fading effects and interference due to simultaneous transmission between the auxiliary nodes. Failure to do so may occur. Then, as 130, the cluster (i.e. radius

Even if there are four or more secondary nodes 131, 132, 133, and 134 within the channel), a channel condition such as path loss, fading, or interference may cause only some of the secondary nodes, not four, to transmit. Can be. As such, the number of secondary nodes that belong to the cluster is greater than the number of split file fragments (for example, four), but due to channel conditions, only some secondary nodes (i.e., less than the smaller number of split file fragments) are actually due to channel conditions. A case in which transmission is possible may be represented as a not successful transmission event.

) 내에 4개 이상의 보조 노드들(131, 132, 133, 134)이 존재하고, 채널 상태를 고려해도 4개 이상의 보조 노드가 실제 전송이 가능인 상황(예컨대, 130에서 4개의 보조 노드 모두가 전송 가능한 상황)을 성공 전송 이벤트(successful transmission event)로 정의할 수 있다. And, like 130, the cluster (i.e. radius

) 4 or more secondary nodes 131, 132, 133, and 134 exist within the network, and even if the channel state is considered, 4 or more secondary nodes can actually transmit (for example, all of the 130 to 4 secondary nodes transmit). Possible situation) can be defined as a successful transmission event.

2 is a flowchart illustrating a relationship for file distribution storage between a base station, an auxiliary node, and a user terminal according to an embodiment of the present invention.

In step 210, the base station 201, which is a transmitting node, splits a specific file among a plurality of files to be stored in the cache of the secondary nodes based on the limited memory size and successive transmission probability for each secondary node. You can decide how many file fragments you want. That is, when a particular file is to be divided into a plurality of file fragments (MDS codes) based on MDS coding, the size / length of the MDS codes may be determined.

In step 220, the base station 201 may divide the specific file into a plurality of file pieces by performing MDS coding based on the determined number (ie, the size of the MDS code).

In operation 230, the base station 201 may control to divide and store the plurality of divided file pieces in a cache of the plurality of auxiliary nodes. For example, the base station 201 may broadcast the divided plurality of file fragments (ie, the plurality of MDS codes) to the plurality of secondary nodes.

In operation 240, each of the secondary nodes 202 may store a file fragment (ie, MDS code) provided from the base station 201 in its cache.

In operation 250, when the user terminal 203 wants to receive a specific file (that is, when a request is made to provide a specific service such as music or a movie), the auxiliary node 202 requests a transmission of the specific file from the user terminal ( 203).

In step 260, as the file fragment (that is, the MDS code) which is part of the file requested from the user terminal 203 is previously stored in the cache of the secondary node 202, the secondary node 202 corresponds to the requested file. The file fragment (that is, MDS code) may be provided directly to the user terminal 203 without separately requesting the base station 201.

Hereinafter, a method of maximizing successful transmission probability for secondary nodes having a limited memory size will be described. That is, as shown in 130, in order to maximize the probability of successful transmission, a method of determining whether it is optimal to divide into several file fragments when MDS coding a specific file will be described.

개 존재함을 가정할 수 있다. 이때, 파일의 크기를 1이라고 하면, 복수의 보조 노드들 각각은 1보다 작은 동등한 메모리 크기를 가질 수 있다. 즉, 보조 노드의 캐시 메모리의 크기는 분할하고자 하는 특정 파일의 전체 크기(예컨대, 1)보다 작으며, 각 보조 노드의 캐시 메모리의 크기는 동일할 수 있다.First, secondary nodes with limited memory are on average per unit area according to the Poisson point process in a 2D environment.

It can be assumed that there are dogs. At this time, if the size of the file is 1, each of the plurality of auxiliary nodes may have an equivalent memory size of less than one. That is, the size of the cache memory of the secondary node is smaller than the total size (eg, 1) of the specific file to be split, and the size of the cache memory of each secondary node may be the same.

(

: i번째 파일을 저장할 확률, T: 보조 노드의 메모리 크기, F: 파일의 총 개수)을 제한 조건으로 할 수도 있고, 파일의 분할 조건까지 고려하여 제한 조건을 규정할 수도 있다. 예를 들어, 파일의 분할 조건까지 고려한 경우,

는 파일의 크기가

인 파일

를

로 저장할 확률로 정의할 수 있다. 그러면, 보조 노드의 메모리 제약 조건은

로 규정/정의될 수 있다. 여기서,

는 i번째 파일을 나눈 개수(즉, 분할된 파일 조각의 개수)를 나타낼 수 있다.When using probabilistic methods, probability caching

(

: the probability of storing the i-th file, T: the memory size of the auxiliary node, F: the total number of files, or the like may be defined in consideration of the partitioning condition of the file. For example, if you consider split conditions for files,

Is the size of the file

File

To

It can be defined as the probability to store as. Then, the memory constraints of the secondary node are

It can be defined / defined as here,

May represent the number of i-th files divided (that is, the number of divided file fragments).

If pathloss is α in a Rayleigh channel environment, it may be assumed that α is greater than 2, which is a free space state. Then, in the Rayleigh channel environment, the probability that the target user receives the i-th file successfully may be expressed as in Equation 1 below.

[Equation 1]

는 파일 i의 인기도를 나타내고, 집 분포(zipf distribution)를 따름을 가정할 수 있다. 여기서,

이고, 사용자 단말이 클러스터(즉, 반경

)의 중심(origin)에 존재하는 경우,

는 해당 클러스터 내에 j개의 보조 노드가 존재할 확률을 나타내고,

는 j개의 SINR 값 중에서

번째까지 가장 큰 SINR 값을 가지는 값들의 집합을 나타낼 수 있다. 그러면,

는 대칭적 맥 레이트(symmetric mac rate)

로 전송하는 경우, 사용자 단말에서 i번째 파일을 성공적으로 전송받을 확률, 즉, 성공 전송 확률을 나타낼 수 있다. 이때, 도 1의 120과 같이, 히팅 이벤트(hitting event)가 가능하기 위해서는, 해당 클러스터(즉, 반경

) 내에

개 이상의 보조 노드들이 존재해야 할 수 있다. 이에 따라, 수학식 1에서 sum 형태가

부터 무한대까지 되어야 할 수 있다. 그리고, 모든 파일 인덱스(index)에 따라 인기도를 고려하는 경우, 평균 합(average sum)으로서, 평균 전송 확률을 고려할 수 있다.In Equation 1,

Denotes the popularity of file i and can be assumed to follow a zipf distribution. here,

Where the user terminal is a cluster (i.e. radius

In the center of the

Denotes the probability that there are j secondary nodes in the cluster,

Is the j SINR value

It can represent the set of values having the largest SINR value to the first. then,

Is the symmetric mac rate

In this case, the i-th file may be successfully transmitted from the user terminal, that is, the success transmission probability may be indicated. In this case, as shown in 120 of FIG. 1, in order to enable a heating event, a corresponding cluster (ie, a radius) may be used.

Within)

There may be more than one secondary node. Accordingly, the sum form in Equation 1

May have to be infinity. In addition, when considering popularity according to all file indexes, as an average sum, an average transmission probability may be considered.

가 클 때 평균 전송 확률이 큰 값을 가질 수 있다. 이에 따라,

이 클 때 성공 전송 확률(successful transmission probability) 값이 커질 수 있다. 본 발명에서는 파일 간 인기도에 따른

의 변화가 아닌, 파일 하나가 존재하는 경우에 해당 파일을 몇 개로 분할 시 성공 전송 확률(successful transmission probability)이 최대가 되는지를 결정하고자 한다. 이처럼, 특정 파일을 대상으로 몇 조각으로 분할 시 성공 전송 확률이 최대인지를 결정하기 위해, 상기 수학식 1에 표현된 문제는 아래의 수학식 2와 같이 단순화될 수 있다. 수학식 2는 하나의 파일에 대해서

이 고정된 경우에 성공 전송 확률(successful transmission probability)을 최대화하는

(즉, 분할하고자 하는 파일 조각의 개수)를 구하기 위해 이용될 수 있다.In this case, the constraint and the objective function are expressed as individual sums between files, and the more popular files are

When is large, the average transmission probability may have a large value. Accordingly,

When this value is large, a successful transmission probability value may increase. In the present invention according to the popularity between files

In the case of a single file, the successive transmission probability is determined when the file is divided into several files. As such, the problem represented by Equation 1 above may be simplified as shown in Equation 2 below to determine whether the maximum probability of successful transmission when splitting into a number of pieces of a specific file is maximum. Equation 2 is for one file

To maximize the successive transmission probability

(Ie, the number of file fragments to be split).

[Equation 2]

에 대해서 보조 노드의 캐시 메모리가

로 할당된 경우, 파일

을 몇 개로 분할 시 성공 전송 확률이 최대화할 수 있으며, 성공 전송 확률이 최대일 때의 저장 확률

이 무엇인지를 풀고자 하는 문제를 나타낼 수 있다.The problem represented by Equation 2 above (ie, the problem of simplifying the problem of Equation 1) can be solved using a stochastic caching method.

The secondary node's cache memory

File, if assigned

Can be maximized to maximize the probability of successful transmission, and the storage probability when the probability of successful transmission is maximum.

This may indicate a problem to solve.

3 is a flowchart illustrating a method of determining a file fragment of a specific file based on a success transfer probability according to an embodiment of the present invention. FIG. 4 is a flowchart illustrating a file distribution storage device according to an embodiment of the present invention. It is a block diagram which shows the internal structure.

In FIG. 4, the file distributed storage device 400 may include a determiner 410, a file divider 420, and a storage controller 430. 3 may be performed by the determination unit 410, the file division unit 420, and the storage control unit 430.

를 결정할 수 있다.In operation 310, the determining unit 410 may determine the number of auxiliary nodes capable of storing the file fragment of the specific file to be requested by the user terminal belonging to the cluster, in consideration of the limited memory size and the success transmission probability of the plurality of auxiliary nodes belonging to the cluster. The number to split based on

Can be determined.

와 확률론적 지오메트리 툴(stochastic geometry tool)을 이용하여 성공 전송 확률의 상계(upper bound) 및 하계(lower bound)를 결정할 수 있다. 그리고, 결정부(410)는 결정된 성공 전송 확률의 상계 내지 하계 점위에서 성공 전송 확률을 최대로 하는 상기 분할하고자 하는 개수

를 결정할 수 있다.At this time, the determination unit 410 is asymmetric mac capacity (symmetric mac capacity)

And the stochastic geometry tool can be used to determine the upper bound and lower bound of the success transfer probability. And, the determination unit 410 is the number to be divided to maximize the success transmission probability in the upper or lower points of the determined success transmission probability

Can be determined.

개 이상의 랜덤 변수(random variable)로 이루어진 다변수(multi variable) 문제에 해당할 수 있다. 이에 따라, 도 6을 참고하면, 맥 캐패시티 영역(mac capacity region, 610) 중 비대칭 맥 캐패시티(symmetric mac capacity)

(620)의 타이트한 상계(upper bound)에 해당하는 최소 개별 레이트(minimum individual rate)인

(630)과 하계(lower bound)인

(640)을 사용하여 상기 다변수 문제를 동일한 개별 레이트(individual rate)로 생각할 수 있으며, 아래의 수학식 3과 같이 표현될 수 있다. 여기서,

는

중 최소 SINR 값을 나타낼 수 있다.For example, in Equation 2 above, the asymmetric mac capacity is

This may correspond to a multivariable problem consisting of more than one random variable. Accordingly, referring to FIG. 6, asymmetric mac capacity of the mac capacity region 610 is shown.

The minimum individual rate corresponding to the tight upper bound of (620)

(630) and lower bound

Using 640, the multivariate problem may be considered at the same individual rate, and may be expressed as Equation 3 below. here,

Is

It may represent the minimum SINR value.

[Equation 3]

From Equation 3, an upper bound of success transmission probability may be derived using a stochastic geometry tool, and the derived upper bound may be as shown in Table 1 below.

대신에

을 대체함으로써, 성공 전송 확률의 하계(lower bound)를 구할 수 있다. 이때, 표 1의 부등식의 Pr[] 안의 우변 값이

에서

로 대체될 수 있다. 그러면, SINR 값의 dB 파워(power)가 -15dB로부터 5dB를 고려한 상황에서

값은 작은 값을 갖게 되어 상계(upper bound)와 하계(lower bound)의 차이(gap)가 작아질 수 있다.The lower bound of the success transmission probability can be obtained in a manner similar to theorem 1 of Table 1. For example, in Table 1

Instead of

By substituting, the lower bound of the success transmission probability can be obtained. At this time, the right side value in Pr [] of the inequality of Table 1

in

Can be replaced with Then, in a situation where the dB power of the SINR value is taken into account from -15 dB to 5 dB

The value may be small so that the gap between upper bound and lower bound may be small.

을 대상으로(즉, 하나의 파일을 대상으로 최대 확률 범위 내에서 분할 가능한 최대 개수를 대상으로) 1차원 검색(search)을 하여 성공 전송 확률이 최대인 파일 조각 개수

를 결정할 수도 있으나, 계산 복잡도(complexity)를 줄이기 위해 분할 가능한 모든 개수(예컨대, 10개까지 분할 가능한 경우 1내지 10)가 아닌 일부 개수를 포함하도록 검색 범위를 특정할 수 있다. 클러스터 내에 사용자 단말에서 요청한 파일

을 저장 가능한 보조 노드의 수와 분할될 파일 조각의 개수에 기초하여, 상기 검색 범위는 특정될 수 있다.At this time, all of the fall within the determined upper and lower ranges

The number of file fragments with the highest probability of successful transmission by performing a one-dimensional search on the target file (ie, the maximum number of files that can be split within the maximum probability range for one file).

However, in order to reduce the computational complexity, the search range may be specified to include some numbers other than all the dividable numbers (eg, 1 to 10 when dividing up to 10). File requested by user terminal in cluster

The search range may be specified based on the number of secondary nodes capable of storing the number and the number of file fragments to be divided.

을 저장 가능한 보조 노드의 수와 분할될 파일 조각의 개수에 기초하여, 상기 검색 범위를 특정할 수 있다. 예컨대, 결정부(410)는

이 1보다 클 때와 1보다 작을 때, 히팅 이벤트(hitting event)를 고려하여 1차원 상에서 검색해야 되는 분할될 파일 조각의 개수를 줄이는 방향으로 검색 범위를 특정할 수 있다. 파일 분할부(420)는 특정 검색 범위 내에서 결정된 분할하고자 하는 개수에 기초하여 파일을 복수개의 파일 조각(즉, MDS 코드들)으로 분할할 수 있으며, 저장 제어부(430)는 분할된 파일 조각을 클러스터 내에 속하는 복수의 보조 노드들의 캐시 메모리에 저장할 수 있다. 예컨대, 동일 파일에 해당하는 서로 다른 파일 조각이 보조 노드들로 브로드캐스됨에 따라, 보조 노드들 각각의 캐시 메모리에 파일 조각이 저장될 수 있다.In operation 320, the determination unit 410 may request a file requested by a user terminal in a cluster.

The search range may be specified based on the number of secondary nodes that can be stored and the number of file fragments to be divided. For example, the determination unit 410 is

When larger than 1 and smaller than 1, the search range may be specified in a direction of reducing the number of file fragments to be divided in one dimension in consideration of a heating event. The file dividing unit 420 may divide a file into a plurality of file fragments (that is, MDS codes) based on the number of segments to be determined within a specific search range, and the storage control unit 430 divides the divided file fragments. It may be stored in the cache memory of a plurality of secondary nodes belonging to the cluster. For example, as different file fragments corresponding to the same file are broadcast to the secondary nodes, the file fragments may be stored in the cache memory of each of the secondary nodes.

인 클러스터에 속하는 보조 노드의 개수가 특정 파일을 대상으로 분할하고자 하는 개수와 동일한 경우),

이 0으로 간주되어 사용자 단말이 모든 보조 노드들로부터 해당 파일의 파일 조각을 제공받을 수 있는 상황이 되므로 아래의 표 2와 같이 푸아송 분포의 정규 분포 근사를 기반으로 M값이 고정되어 있을 때,

에 따른 경향성으로부터 아래의 표 3과 같은 정리(theorem) 2가 유도될 수 있다. 여기서,

은

로 분할된 패킷을 전송 영역

안의 클러스터 내에서 평균적으로 저장하고 있는 양을 나타낼 수 있다. 이에 따라, M이 1보다 큰 경우(M>1)는

로 바꾸어 표현될 수 있으며, 전송 영역 안에 파일 l을 평균적으로 저장한 양이 파일을 분할한 개수

개 보다 클 경우를 나타낼 수 있다. M이 1보다 작거나 같은 경우(M

1인 경우)는

로 바꾸어 표현될 수 있으며, 전송 영역

안의 파일 l을 평균적으로 저장한 양이 파일을 분할한 개수

개 보다 적은 경우를 나타낼 수 있다.At this time, when the heating event (hitting event) (that is, the radius is

If the number of secondary nodes belonging to the cluster is equal to the number you want to split into specific files),

Since this is considered to be 0 and the UE can receive the file fragment of the corresponding file from all secondary nodes, when the M value is fixed based on the normal distribution approximation of the Poisson distribution as shown in Table 2 below,

Theorem 2 as shown in Table 3 can be derived from the tendency according to. here,

silver

Divided packet into transmission area

It can show how much you are storing on average within a cluster. Accordingly, when M is greater than 1 (M> 1)

The number of files divided by the average amount of files stored in the transmission area.

Can be larger than If M is less than or equal to 1 (M

1)

Can be expressed as

The average number of files l stored inside divided the number of files

Fewer cases than may appear.

은

을 변수로 하면서 최대화하고자 하는 목적함수를 의미할 수 있다. 그리고,

은

와

일 때 다른 경향성을 보이는 것으로 위의 정리 2가 증명될 수 있다. In theorem 2 of Table 3,

silver

Can be used to mean the objective function to maximize. And,

silver

Wow

Theorem 2 can be proved to show a different tendency when

의 파일 조각을 저장 가능한 보조 노드의 개수가 분할될 파일조각의 개수보다 적은(

) 경우, 결정부(410) 특정 파일

을 분할하지 않고 특정 파일 전체를 통째로 저장하는 것으로 결정할 수 있다. 이처럼, 통째로 파일을 저장하는 것이 최적이지만, 특정 파일의 파일 조각을 저장 가능한 보조 노드의 개수가 분할될 파일조각의 개수 이상(

)인 경우, 결정부(410)는 미리 지정된 최대 확률 범위 내에서 특정 파일을 분할 가능한 최대 개수를 분할하고자 하는 개수로 결정할 수 있다. 즉, 위의 정리 2에 따르면, 파일을 통째로 보조 노드에 저장하는 것뿐만 아니라, 최대 확률 범위 내에서 분할할 수 있는 만큼 해당 파일을 분할하는 것이 최적일 수 있다(즉, 사용자 단말에서 요청한 파일을 보조 노드 간 협력 전송을 통해 정상적으로 복구하는 것에 해당할 수 있다). According to theorem 2 of Table 3 above, certain files

The number of secondary nodes that can store a file fragment of is less than the number of file fragments to be split (

), The decision unit 410 specific file

You can decide to save the entire specific file without splitting it. As such, although it is optimal to save a file as a whole, the number of secondary nodes capable of storing file fragments of a specific file is equal to or greater than the number of pieces of files to be divided (

), The determination unit 410 may determine that the maximum number of possible division of a specific file within a predetermined maximum probability range is desired. That is, according to theorem 2, it may be optimal not only to store the file entirely on the secondary node, but also to divide the file as much as possible so that the file can be split within the maximum probability range (ie, a file requested by the user terminal). It may correspond to normal recovery through cooperative transmission between secondary nodes).

인 경우, 파일이 분할될 수록(즉, 많은 조각으로 분할될 수록), 해당 파일에 대한 파일 존재 확률(hitting probability)은 계속해서 감소할 수 있다. 또한, 성공 전송 확률(STP)은 상기 파일 존재 확률에 무선 채널 환경까지 반영한 것이므로, 항상 파일 존재 확률의 하계(lower bound)와 관련될 수 있다.Then, by theorem 2 above

If the file is split (i.e., split into many pieces), the file hitting probability for that file may continue to decrease. Also, since the success transmission probability (STP) reflects the radio channel environment in the file existence probability, it may always be related to the lower bound of the file existence probability.

5 is a diagram illustrating a graph of a file existence probability and a successful transmission probability according to one embodiment of the present invention.

의 파일 조각을 저장 가능한 보조 노드의 개수가 분할될 파일조각의 개수보다 적은 경우(

), 결정부(410)는 특정 파일을 분할하는 경우에 성공 전송 확률(STP)을 최대화하는 분할될 파일 조각 개수를 결정할 수 있다.Referring to Figure 5, a specific file

If the number of secondary nodes that can store a file fragment of is less than the number of file fragments to be split (

The determination unit 410 may determine the number of file fragments to be divided to maximize the success transmission probability (STP) when the specific file is divided.

을 대상으로

을 시작으로 2, 3, 4 등으로 계속 파일을 분할하면서, 분할될 각 파일 조각 개수 별 성공 전송 확률(STP)을 계산할 수 있다. 즉, 성공 전송 확률(STP)이 최대가 되는 지점(즉, 성공 전송 확률의 최대값)을 파일을 분할함에 따라, 분할될 파일 조각 개수에 해당하는 성공 전송 확률의 값으로 계속하여 갱신할 수 있다. 즉, 최대 성공 전송 확률을 갱신할 수 있다. 도 4에서는

일 때 성공 전송 확률(STP)이 최대값을 가질 수 있다. 이때, 결정부(410)는 파일을 분할할수록 변화하는 해당 파일

의 파일 존재 확률(hitting probability)을 대상으로, 해당 파일의 성공 전송 확률의 최대값(

, 501)과 동일해 지는 지점(502)에 해당하는 파일 존재 확률을 결정할 수 있다. 그리고, 결정부(410)는 결정된 파일 존재 확률(502)에 해당하는 파일 조각의 개수(503, 즉, 63개)를 결정할 수 있다. 결정부(410)는 파일을 분할하지 않을 때(

)부터 결정된 상기 파일 조각의 개수(503,

)까지를 검색 범위로 특정할 수 있다. 즉, 결정부(410)는 최대 확률 범위 내에서 분할 가능한 파일 조각의 개수인 120개 모두를 대상으로 성공 전송 확률을 구하지 않고,

내지

의 범위 내에서 성공 전송 확률을 계산하고, 계산된 성공 전송 확률 중 최대값(예컨대, 0.325)에 해당하는 파일 조각의 개수(

, 501)를 특정 파일

을 대상으로 분할하고자 하는 최종 파일 조각의 개수로 결정할 수 있다.In one example, the determination unit 410 is a specific file

To target

By starting to divide the file into 2, 3, 4, etc., the success transfer probability (STP) for each number of file fragments to be split can be calculated. That is, by dividing the file at a point where the success transfer probability STP becomes the maximum (that is, the maximum value of the success transfer probability), the success transfer probability corresponding to the number of file fragments to be split may be continuously updated. . That is, the maximum successful transmission probability can be updated. In Figure 4

When the success transmission probability (STP) may have a maximum value. In this case, the determination unit 410 changes the corresponding file as the file is divided.

For the file hitting probability of, the maximum value of the success transfer probability of the file (

, 501 may determine the file existence probability corresponding to the point (502). The determination unit 410 may determine the number of file fragments 503 (ie, 63) corresponding to the determined file existence probability 502. When the determination unit 410 does not divide the file (

The number of the file fragments determined from

) Can be specified as a search range. That is, the determination unit 410 does not calculate the success transfer probability for all 120 pieces of file fragments that can be split within the maximum probability range.

To

Calculate the success transfer probability within the range of, and determine the number of file fragments corresponding to the maximum value of the calculated success transfer probability (eg, 0.325)

501 specific files

This can be determined by the number of final file fragments to be split into.

)까지 분할될 파일 조각의 개수를 검색하는 경우, 성공 전송 확률(STP)은 파일 존재 확률에 lower bound되어 있으므로 더 이상 성공 전송 확률을 크게 하는

값이 존재할 수 없다. 이에 따라, 결정부(410)는 검색 범위를

내지

로 특정하고, 해당 범위 내에서 해당 파일을 대상으로 분할하고자 하는 파일 조각의 개수를 최종적으로 결정할 수 있다. 즉,

내지

중 어느 하나를 분할하고자 하는 파일 조각의 개수로 최종 결정할 수 있다. 예컨대, 성공 전송 확률을 최대화하는

을 파일 조각의 개수로 최종 결정할 수 있다. 이외에, 파일의 인기도 등의 추가 파라미터를 고려하여 상기

내지

중

이 아닌 다른 값을 최종 파일 조각의 개수로 결정할 수도 있다. 이처럼,

내지

까지만 검색함에 따라, 120까지 검색할 때 대비 상대적으로 복잡도가 절반으로 감소할 수 있다. 이때, 도 5의 510을 참고하면, 성공 전송 확률의 하계(lower bound)와 상계(upper bound)가 많이 타이트(tight)해서 구별이 안될 수 있으며, 이 경우, 성공 전송 확률의 하계(lower bound) 및 상계(upper bound) 중 어느 것을 선택해도 문제가 되지 않을 수 있다.As such, the point at which the file hitting probability becomes equal to the maximum successful transfer probability of the file (ie, FIG. 5).

When searching for the number of file fragments to be divided up to), the success transfer probability (STP) is lower bound to the file existence probability, so

The value cannot exist. Accordingly, the determination unit 410 determines the search range.

To

In this example, the number of file fragments to be split into the corresponding file within the corresponding range may be finally determined. In other words,

To

One can finally determine the number of file fragments to be split. For example, to maximize the probability of successful transmission

Can be determined by the number of file fragments. In addition, considering the additional parameters such as the popularity of the file

To

medium

You can also determine a value other than the number of final file fragments. Likewise,

To

As far as searching only, the complexity may be reduced by half compared to when searching up to 120. In this case, referring to 510 of FIG. 5, the lower bound and the upper bound of the success transmission probability may be indistinguishable from each other and in this case, the lower bound of the success transmission probability may be indistinguishable. And either upper bound may not be a problem.

인 경우, 즉, 클러스터 내에서 특정 파일

또는 특정 파일

의 파일 조각을 저장하고 있는 평균 보조 노드의 개수가 분할된 파일 조각의 개수 이하인 경우, 계산 복잡도를 줄이면서 성공 전송 확률(STP)을 최대화하는 알고리즘은 아래의 표 4와 같을 수 있다.5 described

That is, specific files within the cluster

Or specific files

If the average number of secondary nodes that store the file fragments is equal to or less than the number of fragmented file fragments, an algorithm for maximizing the success transmission probability (STP) while reducing the computational complexity may be as shown in Table 4 below.

인 경우(즉, 클러스터 내에서 특정 파일

또는 특정 파일

의 파일 조각을 저장하고 있는 평균 보조 노드의 개수가 분할된 파일 조각의 개수보다 많은 경우), 표 1의 정리 1의 상계(upper bound)를 전개하면 아래의 수학식 4와 같을 수 있다.Referring back to Table 1 above,

(That is, specific files within the cluster

Or specific files

If the average number of secondary nodes that store the file fragments of is larger than the number of fragmented file fragments), the upper bound of theorem 1 of Table 1 may be expanded as shown in Equation 4 below.

[Equation 4]

이 0일 경우에

=1이 되고, 결과적으로 히팅 목적 함수(hitting objective function)가 파일 존재 확률(hitting probability)로 정의될 수 있다. 그러면, 파일 존재 확률(hitting probability)은

와 같이 표현될 수 있으며, 채널 환경을 반영한 경우인 위의 수학식 4를 전개해서 풀어쓰면

와 같이 표현될 수 있다. 즉, 위의 표 2 에서 보완 히팅 확률(complementary hitting probability)(=1-hitting probability)인

와 유사한 형태로 반영되어 있음을 알 수 있다. 다시 말해, 위의 수학식 4는

의 형태와 유사하며,

을

로 대체하고 정규 근사(normal approximation)를 이용하면 위의 수학식 4로부터 아래의 수학식 5가 도출될 수 있다.The objective function to optimize in Equation 2 above

If is 0

= 1, and as a result, a heating objective function may be defined as a file hitting probability. Then the file hitting probability is

It can be expressed as, and if you expand the equation (4) above to reflect the channel environment

It can be expressed as That is, in Table 2, the complementary hitting probability (= 1-hitting probability)

It can be seen that it is reflected in the form similar to. In other words, Equation 4 above

Similar to the form of,

of

By substituting for and using normal approximation, Equation 5 below can be derived from Equation 4 above.

[Equation 5]

에 근사하게 되고, 결국,

에 대해서

를 만족하는

이 존재할 수 있다.

를 잘 결정하면, 위의 수학식 5는 아래의 수학식 6 및 수학식 7과 같이 표현될 수 있다. 즉,

에 대해

이 1에 수렴하도록

를 미리 지정된 기준값 이하의 상당히 작은 값으로 만들면 위의 수학식 5는 아래의 수학식 6 및 7과 같이 표현될 수 있다. 예컨대,

는

에 반비례하는 함수에 해당할 수 있다.In Equation 5, as N increases

To be cool, eventually,

about

To satisfy

This may exist.

When well determined, Equation 5 above may be expressed as Equation 6 and Equation 7 below. In other words,

About

To converge on this one

When E is made to be a significantly smaller value below a predetermined reference value, Equation 5 above may be expressed as Equations 6 and 7 below. for example,

Is

This may correspond to a function inversely proportional to.

에서

이 되므로,

에서 증가할 수 있다. 이에 따라,

로부터

인 경우에 성공 전송 확률(STP)가 증가할 수 있다. 이때,

에 해당할 수 있다. 이처럼,

인 경우에도

일 때와 마찬가지로 모든 가능한 범위를 대상으로

에 대해서 검색을 하는 것이 아니라,

의 집합 내에서 검색이 수행될 수 있다. 즉, 결정부(410)는

인 경우, 검색 범위를

로 특정할 수 있다. 여기서, N_F는 가능한 확률범위 내에서 분할 가능한 최대 개수(즉, 최대 분할의 수)를 나타낼 수 있다.

인 경우에도 특정 파일을 여러 조각으로 분할 시 전체가 아닌 검색 범위를

로 특정하고, 특정된 검색 범위 내에서 최종적으로 분할에 적용할 파일 조각의 개수를 결정함으로써, 계산 복잡도는 낮추면서 STP를 최대화하는

을 찾을 수 있다. Also, the file hitting probability is

in

This becomes

Can increase from Accordingly,

from

In this case, the success transmission probability (STP) may increase. At this time,

It may correspond to. Likewise,

Even if

To all possible ranges

Rather than searching for,

The search may be performed within the set of. That is, the determination unit 410 is

If, the search scope

Can be specified as Here, N _F may represent the maximum number of possible divisions (ie, the maximum number of divisions) within a possible probability range.

Even if you split a specific file into pieces,

By determining the number of file fragments to be finally applied to the partition within the specified search range, we maximize STP while reducing computational complexity.

Can be found.

은

에 대해

값이 0에 가까운 값으로 고정된 경우,

을 만족하는 값에 해당할 수 있다. 예컨대,

이지만,

인 상황을 나타낼 수 있다. 그리고,

는 파일 존재 확률(hitting probability)의 극값을 0으로 만드는 값으로서,

일 때

인 경우에 파일 존재 확률(hitting probability)은 감소하고,

인 경우에 파일 존재 확률(hitting probability)은 증가할 수 있다. 그리고, M이 커질 경우

는 작아질 수 있다. 이에 따라 파일 l을 분할한 개수

(단,

)에 대해서 STP가 계속적으로 증가하게 되고, 결과적으로

범위만 검색(search)을 하여 복잡도를 낮추면서 STP를 최대화하는 파일 l의 분할

을 찾을 수 있다.Specifically,

silver

About

If the value is fixed to near zero,

It may correspond to a value satisfying. for example,

as,

May indicate a situation. And,

Is the value that makes the extreme value of the file hitting probability zero.

when

If the file hitting probability decreases,

If the file hitting probability (Hitting probability) may increase. And if M becomes large

Can be made smaller. The number of files l divided accordingly

(only,

), The STP will continue to increase

Partitioning of files l to maximize STP while reducing complexity by searching only ranges

Can be found.

일 때(즉, 사용자 단말에서 요청한 파일을 서비스할 수 있는 반경 내에서 해당 파일(예컨대, 파일 l)을 저장하고 있는 평균 보조 노드의 수가 분할된 파일의 개수보다 적을 경우)와

경우(사용자 단말에서 요청한 파일을 서비스할 수 있는 반경 내에서 해당 파일(예컨대, 파일 1)을 저장하고 있는 평균 보조 노드의 수가 분할된 파일의 개수보다 많을 경우)로 구분하여 검색 범위가 특정될 수 있다. 그리고, 특정된 검색 범위 내에서 성공 전송 확률(STP)을 최대화하는 해당 파일을 분할할 파일 조각의 개수 최종 결정함으로써, 검색 범위를 특정하기 이전보다 상대적으로 계산 복잡도는 감소시키면서 성공 전송 확률을 최대화하는 파일 조각의 개수를 찾을 수 있다. 즉, 계산 복잡도는 줄이면서 단말에서는 요청한 파일을 보조 노드들 간의 협력 전송을 통해 성공적으로 복구할 수 있다.As described above, in dividing a file into pieces based on MDS coding, in order to determine the number of partitions to maximize the success transmission probability (STP) in consideration of file diversity.

(I.e., if the average number of secondary nodes storing the file (eg, file l) is smaller than the number of partitioned files within a radius that can service the file requested by the user terminal).

The search range can be specified by dividing into cases (when the average number of secondary nodes storing the file (eg, file 1) is larger than the number of divided files within a radius in which the file requested by the user terminal can be serviced). have. By finally determining the number of file fragments for dividing the corresponding file maximizing the success transfer probability (STP) within the specified search range, maximizing the success transfer probability while reducing the computational complexity relative to before specifying the search range. Find the number of file fragments. That is, while reducing the computational complexity, the terminal can successfully recover the requested file through cooperative transmission between secondary nodes.

The method according to the embodiment may be embodied in the form of program instructions that can be executed by various computer means and recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the media may be those specially designed and constructed for the purposes of the embodiments, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

Although the embodiments have been described by the limited embodiments and the drawings as described above, various modifications and variations are possible to those skilled in the art from the above description. For example, the described techniques may be performed in a different order than the described method, and / or components of the described systems, structures, devices, circuits, etc. may be combined or combined in a different form than the described method, or other components. Or even if replaced or substituted by equivalents, an appropriate result can be achieved.

Therefore, other implementations, other embodiments, and equivalents to the claims are within the scope of the claims that follow.

Claims (18)

In the method in which the file distributed storage device distributed the file to a plurality of help nodes belonging to a cluster (cluster),
Determining at least one of the plurality of secondary nodes based on the limited memory size and successive transmission probability for each secondary node for the file;
Determining the number of files to be divided based on the determined number of auxiliary nodes and the limited memory size;
Dividing the file into file fragments corresponding to the determined number based on Maximum Distance Separable (MDS) coding; And
Distributing and storing the file fragments in the determined secondary node such that each of the determined secondary nodes stores at least one of the file fragments in accordance with the limited memory size.
Distributed storage method comprising a. The method of claim 1,
And the limited memory size is less than the size of the file. delete The method of claim 1,
If the determined number of secondary nodes is 1, storing the file in the determined secondary node without dividing the file
Distributed storage method comprising more. The method of claim 1,
Determining the number to be divided,
Calculating a successive transmission probability for each number of file fragments to be divided;
Updating the maximum success transmission probability to the maximum of the calculated success transmission probabilities;
Determining a file existence probability corresponding to the maximum successful transmission probability based on a file hitting probability within the cluster that changes as a file is divided; And
Determining the number to be divided according to the determined file existence probability
Distributed storage method comprising a. The method of claim 5,
Determining the number to be divided,
Specifying a search range in a range from when the number of file fragments is 1 to the number of file fragments to be divided corresponding to the determined file existence probability; And
Finally determining the number of file fragments having a maximum transmission probability within the search range as the number to be divided
Distributed storage method comprising more. The method of claim 5,
The file hitting probability, which indicates the probability that the file or a part of the file is stored in the plurality of auxiliary nodes, is a file that decreases as the number of file fragments divided for the file increases. Distributed storage method. The method of claim 1,
Determining the number to be divided,
If the determined number of auxiliary nodes is equal to or greater than the number of file fragments to be divided, determining the maximum number of possible file divisions within a predetermined maximum probability range as the number to be divided. The method of claim 1,
When the file is requested from a user terminal belonging to the cluster, the file is distributed to the user terminal based on cooperative transmission between the determined secondary nodes. A file distributed storage device for distributing and storing files in a plurality of help nodes belonging to a cluster,
Based on the limited memory size and successive transmission probability of each secondary node for the file, at least one of the plurality of secondary nodes is determined, and the determined number of secondary nodes and the limited memory size are determined. A determining unit determining a number of files to be divided;
A file division unit for dividing the file into file fragments corresponding to the determined number based on MDS (Maximum Distance Separable) coding; And
A storage controller configured to distribute and store the file fragments in the determined secondary node such that each of the determined secondary nodes stores at least one of the file fragments according to the limited memory size
Distributed storage device including a. The method of claim 10,
And the limited memory size is less than the size of the file. delete The method of claim 10,
The determination unit,
If the determined number of secondary nodes is 1, the file distribution storage device for storing the determined secondary node without dividing the file. The method of claim 10,
The determination unit,
Calculate the successive transmission probability for each number of file fragments to be split, update the maximum successful transmission probability to the maximum of the calculated successive transmission probabilities, and change the file existence probability in the cluster as the file is divided. and determining a file existence probability corresponding to the maximum successful transmission probability, and determining the number to be divided according to the determined file existence probability. The method of claim 14,
The determination unit,
To specify the search range from the time when the number of file fragments is 1 to the number of file fragments to be divided corresponding to the determined file existence probability, and to divide the number of file fragments having the maximum success transfer probability within the search range. File distributed storage that ultimately determines by count. The method of claim 14,
The file hitting probability, which indicates the probability that the file or a part of the file is stored in the plurality of auxiliary nodes, is a file that decreases as the number of file fragments divided for the file increases. Distributed storage. The method of claim 10,
The determination unit,
And when the determined number of auxiliary nodes is equal to or greater than the number of file fragments to be divided, determining the maximum number of possible file divisions within a predetermined maximum probability range as the number to be divided. The method of claim 10,
When the file is requested from a user terminal belonging to the cluster, the file is distributed to the user terminal based on the determined cooperative transmission between the auxiliary nodes.

Images