JP5514041B2 - Identifier assignment method and program - Google Patents

Description

  The present invention relates to a distributed system in which a plurality of nodes operate in cooperation with each other, and more particularly to a method and program for assigning identifiers to nodes that enable load distribution among nodes in the distributed system.

  Generally, in a distributed system in which a plurality of nodes operate in cooperation, an identifier for recognizing and identifying each node in the distributed system is assigned. In some distributed systems, identifiers only have a role that needs to be able to identify individual nodes. In that case, the identifier may be assigned as long as it assigns a unique value so that the same value is not used between nodes. On the other hand, as in the case of consistent hashing (Non-patent Document 1), the role of the identifier is more than the identification of the individual node, such as the amount of processing load assigned to the individual node is determined by the value of the identifier. Some distributed systems have

  In such a distributed system, a data management method based on consistent hashing or the like is used. According to the consistent hashing, an identifier for identifying each node and an identifier for identifying each data managed by these nodes are assigned. The range of identifiers of data managed by a node is typically determined between the identifier of the node and the identifier of the node having the largest identifier among the nodes having identifiers smaller than the node. It is done. By assigning these nodes and data identifiers with random values returned using a hash function, etc., the amount of data handled by each node is stochastically equalized, and such distributed data management It is possible to equalize the load between nodes by the program.

  FIG. 1 shows an example of a single identifier assignment according to a random identifier assignment method according to the prior art. In FIG. 1, the nodes are arranged in a circle from the one with the smallest identifier value. The reason why they are arranged in a circle is that in the consistent hashing or the like, the minimum value comes after the maximum value of the identifier. This circle is called an ID space following the consistent hashing terminology. Symbols such as circles and squares arranged along with numbers in the ID space represent node identifiers, and the numbers represent the number of nodes added to the distributed system. In the ID space, it is assumed that the identifier value increases as it goes clockwise. Here, the area in charge of a node corresponds to an arc from the identifier of the node to the identifier of another node counterclockwise. In the example shown in FIG. 1, the area in charge of node 1 is an arc between node 1 and node 2. In proportion to the size of this area, the data handled by the node and the amount of processing load are determined. FIG. 1 is a specific example of a method of assigning the most basic node identifier in consistent hashing. Here, up to five nodes are added, and each is given a random value identifier and exists in the ID space. Since the identifier value is randomly determined, the size of the area between the nodes also varies.

  Thus, in this consistent hashing, the node and data identifiers are assigned by a hash function. A hash function is a function that returns a random output in response to an input. Since the same output is always returned for a certain input, it is not a function that returns a completely random value, but a random output in the sense that it returns an output that ignores the meaning of the original input. It is a function. For this hash function, a unique value determined for each node or data, for example, a value obtained by inputting a node name or address, a data name, or the like is used as an identifier.

  When such an identifier allocation method is used, the range of identifiers of data handled by individual nodes is higher than the average value with high probability. If this is described by Landau's asymptotic notation, it can be described that the amount of load in charge of each node is Θ (logN) times the average load (N is the total number of nodes). This means that as the number of nodes increases, there are nodes that have to handle more loads than the average. In order to alleviate this problem, consistent hashing probabilistically equalizes the load in charge by assigning a plurality of identifiers to one node.

  FIG. 2 shows an example of assignment of a plurality of identifiers by a random identifier assignment method according to the prior art. In FIG. 2, a plurality of identifiers are assigned to each node. In consistent hashing, it is mentioned to use this method for load distribution. FIG. 2 shows a state in which three nodes are added when three identifiers are assigned to one node. Each node is in charge of an area obtained by bundling three areas in charge of three identifiers. As the number of identifiers assigned to one node is increased according to the law of large numbers, the area in units of nodes is averaged.

D. Karger, E. Lehman, T. Leighton, M. Levine, D. Lewin, and R. Panigrahy, "Consistent Hashing and Random Trees: Distributed Caching Protocols for Relieving Hot Spots on the World Wide Web," in Proceedings of the 29th ACM Symposium on Theory of Computing (STOC'97), pp.654-663, May 1997.

  Although the above-described method of assigning a plurality of identifiers to one node according to the prior art has a certain effect, further improvement is required in a distributed system in which load distribution with higher uniformity is desired. .

  When considering a more effective method for assigning identifiers, it is necessary to consider the following items. For one thing, after assigning an identifier for a node, it should not be changed. This is because when the identifier of a node is changed, the amount handled by some nodes changes, and processing with a high communication cost between the nodes, such as movement of data between the nodes, occurs. This cost may hinder the data management process itself in some cases, and must be avoided before the load is equalized.

  The other is that in a distributed system, the number of nodes constituting the distributed system can change dynamically. When the amount of data to be processed in the entire distributed system is increasing, nodes are added to increase the number of nodes. When the amount of data to be processed decreases or when a node cannot continue processing as a node due to some failure, the node is deleted to reduce the total number of nodes. Here, when the number of nodes decreases, the range handled by the deleted node is determined by the method of determining the range (the identifier range of the data managed by a certain node is the node's own identifier and a value smaller than that node) Between the nodes having the largest identifier among those having the largest identifier, and the node having the largest value is inherited by the nearest node having an identifier larger than the deleted node. Therefore, even if the amount handled by each node is completely uniform, if one of the nodes is deleted, the amount handled by that node is taken over by the next node, and the node that took over is The amount of charge will be doubled.

  In view of the above problems, an object of the present invention is to provide an identifier assignment method and program that make it possible to equalize the load of individual nodes even if the number of nodes in a distributed system varies.

  In order to solve the above-described problem, an aspect of the present invention provides a method for assigning an identifier to a node device in a distributed system in which a plurality of node devices operate in cooperation, and an identifier assignment event for the node device in the distributed system Detecting the identifier assignment event, accessing the identifier list storing the assigned identifiers in the distributed system, referring to the identifier list, and storing the identifier list in the identifier list A method comprising: determining an adjacent identifier having a maximum area between adjacent identifiers; and assigning any identifier in the area between the determined adjacent identifiers to the node device. About.

  According to the present invention, it is possible to equalize the processing amount and data amount handled by each node by equalizing the intervals of the identifiers of the distributed system, and to equalize the burden amount between the nodes. In addition, regardless of the situation of addition / deletion of nodes in the distributed system, the burden between the nodes can be made fair, and the utilization efficiency of the entire system can be improved.

FIG. 1 shows an example of a single identifier assignment according to a prior art identifier assignment method. FIG. 2 shows an example of assignment of a plurality of identifiers by the identifier assignment method according to the prior art. FIG. 3 shows a configuration of an identifier assigning device according to an embodiment of the present invention. FIG. 4 shows a configuration of a distributed system including an identifier assigning device according to an embodiment of the present invention. FIG. 5 shows another configuration of a distributed system including an identifier assigning device according to an embodiment of the present invention. FIG. 6 illustrates an example of a single identifier assignment according to an identifier assignment method according to an embodiment of the present invention. FIG. 7 shows an example of assignment of a plurality of identifiers by an identifier assignment method according to an embodiment of the present invention. FIG. 8 shows an example of the change of the assigned area in the assignment of a single identifier when a node is deleted by the identifier assignment method according to an embodiment of the present invention. FIG. 9 shows an example of a change of a responsible area in assignment of a plurality of identifiers when a node is deleted by the identifier assignment method according to one embodiment of the present invention. FIG. 10 is a flowchart showing the process of the identifier assignment method according to one embodiment of the present invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  First, the principle of the present invention will be described. In the present invention, when assigning identifiers to nodes, it is possible to assign a plurality of identifiers to each node instead of one identifier. Each identifier has a range in charge of each identifier, and one node collectively manages them. The range assigned to each identifier has a high probability of Θ (logN) times as described in the related art, but by combining the ranges handled by a plurality of identifiers, uniformity can be achieved by the law of a large number. improves.

  Further, when a node is deleted, the assigned range of each identifier is inherited by the latest identifier having a value larger than each identifier. There is a high possibility that a node having these takeover destination identifiers is not one node but several nodes. As a result, it is possible to reduce the problem caused by the prior art that the load at the time of node deletion is taken over by the adjacent node and the uniformity of the load is lost.

  In addition, when assigning individual identifiers, a random value is not assigned using a hash function as in the prior art, but a value that averages a portion where deviation occurs is assigned intentionally. That is, according to the method of determining the assigned range of a node by selecting an area to be assigned to a node by the number of identifiers in descending order of the interval between identifiers at the time of assigning the identifier, and assigning a certain value in the area as an identifier. The area having a large interval until then is divided into two.

  This means that the intervals between identifiers are not deterministically averaged as in the prior art, but the intervals between identifiers are deterministically uniform. As a result, even if there is a region with a larger interval than the average in the prior art, it is this decisive method whether the next added node has an identifier that divides this region is random. Then, it will have an identifier that will definitely divide a large area.

  Here, if only the uniformity of identifier intervals when adding nodes is considered, there is no need to assign a plurality of identifiers to one node, and only one identifier can be assigned to a node by the above decisive method. That's fine. On the other hand, if only one identifier is assigned, the non-uniformity that occurs when a node is deleted is not equalized until the next node is added. In the present invention, the load of individual nodes can be obtained without causing the lack of uniformity in the fluctuation of the number of nodes, which is a problem with the prior art, by using the assignment of a plurality of identifiers together with the assignment of a definitive identifier. Can be made uniform.

  Next, an identifier assigning apparatus according to an embodiment of the present invention will be described. FIG. 3 shows a configuration of an identifier assigning device according to an embodiment of the present invention. The identifier assignment device 100 is typically composed of one or more of various hardware resources such as an auxiliary storage device, a memory device, a CPU, and a communication device. The auxiliary storage device is composed of a hard disk, a flash memory, and the like, and stores programs and data for realizing various processes and functions described later. The memory device is composed of a RAM (Random Access Memory) or the like, and reads and stores the program from the auxiliary storage device when an instruction to start the program is given. The CPU functions as a processor for processing information, and realizes various functions described later according to a program stored in the memory device. The communication device includes various communication circuits for connecting to other devices via a network. The communication device typically includes a communication interface for connecting to various networks. The identifier assigning apparatus 100 is not limited to the hardware configuration described above, and may have any other appropriate hardware configuration having a function for bidirectional communication with each node via a network. Good.

  The identifier assignment device 100 may be arranged as a server in a server / client type distributed system as shown in FIG. That is, the identifier assignment device 100 functions as a server that receives an identifier assignment request from a node and assigns an identifier in response to the request by connecting to each node of the distributed system.

  Alternatively, the identifier assignment device 100 may be arranged in each node of the distributed system as shown in FIG. That is, the identifier assignment device 100 may be incorporated as an independent device or a program or a part of a program of the distributed system on the node of the distributed system.

  The identifier assignment device 100 of the present invention does not need to be an independent physical device, and is realized as a computer program for causing a general-purpose computer having a processor and a memory to execute an identifier assignment function and method described later. May be. In this case, the operating position of the program is not limited, and any connection configuration can be used as long as an identifier assignment request is received from the distributed system and a list of identifiers of nodes of the distributed system is held or can be referred to. Even if it exists, it is operable.

  The node described in this specification is a component of processing of the distributed system. Specifically, each computer constituting the distributed system is realized as a node. On the other hand, by using a plurality of virtual machines or by starting a plurality of distributed system programs, it is possible to realize a form in which a plurality of nodes exist on one computer.

  As shown in FIG. 3, the identifier assignment device 100 includes an identifier list storage unit 101 and an identifier determination unit 110.

  The identifier list storage unit 101 stores an identifier list describing the identifiers already assigned to the nodes in the distributed system. As described above, the identifier according to the present invention not only identifies an individual node but also means more than the identification of the individual node, such as the amount of processing load handled by the individual node is determined by the value of the identifier. have. Typically, the identifier list stores identifiers assigned to existing nodes in the distributed system in association with node names and the order in which the nodes are connected to the distributed system. Thereby, when deleting a node, an identifier assigned to the node to be deleted can be quickly identified.

  In response to an identifier allocation event such as addition of a new node, the identifier determination unit 110 refers to the identifier list and determines an identifier to be allocated by processing to be described later. Further, after determining the identifier, the identifier determination unit 110 notifies the identifier list storage unit 101 of the identifier, and updates the identifier list.

  The identifier determining unit 110 includes a deterministic identifier determining unit 111 and a multiple identifier assigning unit 112. The deterministic identifier determination unit 111 first acquires identifiers of all nodes of the distributed system by referring to the identifier list, arranges them in the order of the size of the identifiers, and calculates the distance between the identifiers. The distance between the identifiers is, for example, the distance on the ID space shown in FIGS. 1 and 2, and the distance between adjacent identifiers depends on the length of the arc between the identifiers, the identifier and the center of the circle. It may be defined as the area of the fan-shaped region to be defined. Note that expressions such as “adjacent”, “value interval is large”, and “region is large” in this specification are expressions based on the definition of the distance between identifiers. In general, Euclidean norm, exclusive OR, or the like is used to define the distance. In this specification, the Euclidean norm is used as a distance definition without particular notice. However, the present invention is not limited to this, and an exclusive OR or other definition of distance may be used.

  Based on the calculated distance information, the deterministic identifier determination unit 111 creates regions (arcs between identifiers in the ID space described above) by the number of identifiers requested from the distributed system in descending order. select. Each value included in each selected area is set as an identifier, and the plurality of identifiers are notified to the distributed system. According to such an identifier determination method, by taking any value in the area as an identifier, the nodes to which this identifier has been assigned exist in the descending order of the areas that have existed so far and caused non-uniformity. Will be divided.

  In addition, when assigning identifiers, the regions are selected in descending order of identifier intervals. When there are a plurality of regions having the same interval size, a selection method for determining which one to select is selected. Is not particularly limited in the present specification. Here, as an example, it is assumed that an area having a small identifier value, that is, an ID space as shown in FIGS. 1 and 2, is selected from an area in a counterclockwise direction.

  Here, several methods are conceivable as to how to divide the selected area, that is, which identifier in the selected area is determined. For example, the simplest example is a method of equally dividing into two. In this case, every time a new node is added to the distributed system, the size of the maximum area is divided in half, thereby correcting the deviation in the interval between identifiers. In another embodiment, the allocated area may be adjusted according to the calculation capability and storage capacity of each node. That is, the area may be divided at an arbitrary ratio. However, the present invention is not limited to the above-described division ratio, and any other appropriate division method may be used.

  According to the identifier determination method described above, the area in charge of each node is made uniform, and the amount of processing and the amount of data in charge of each node are also made uniform. In some cases, the consistent hashing mechanism is used to determine the location for processing the distributed system rather than determining the data storage location. In the present specification, the description will be made on the premise that the identifier assigning apparatus according to the present invention is applied in the data distribution mechanism, but the present invention is also applied to the determination of the position where the processing of the distributed system is performed. It is possible, and the application target of the present invention is not limited only to the determination of the data storage position.

  The multiple identifier assigning unit 112 assigns a plurality of identifiers to one node by using the deterministic identifier determining unit 111 as many as requested from the distributed system. Specifically, the multiple identifier assigning unit 112 executes the process of the deterministic identifier determining unit 111 described above a plurality of times in response to the number of identifiers requested from the distributed system, and determines the determined multiple identifiers. Assign to a node.

  Next, a specific example of identifier assignment by the above-described identifier assignment device of the present invention will be described with reference to FIGS. In FIGS. 6 to 9, similarly to FIGS. 1 and 2, nodes are arranged in a circle from the smallest identifier value. Symbols such as circles and squares arranged with numbers in the ID space indicated by the circles represent node identifiers, and the numbers represent the number of the node added to the distributed system. In the ID space, it is assumed that the identifier value increases as it goes clockwise. Here, the area in charge of a node corresponds to an arc from the identifier of the node to the identifier of another node counterclockwise. In proportion to the size of this area, the data handled by the node and the amount of processing load are determined.

  FIG. 6 shows a state when a single identifier is assigned to five nodes using the definitive identifier assignment method of the present invention. Here, as an example, the identifier is assigned in such a manner that the area is equally divided into two. In an initial state where no area exists, an arbitrary value is assigned to the assignment of an identifier to the node 1, and a new identifier is assigned to the middle of the largest area among the existing areas after the node 2. In the example of FIG. 6, when there are a plurality of regions of the same size, the region with the smallest identifier is selected starting from the uppermost point of the circle (where node 1 is present in FIG. 6). . For this reason, in the ID space that has been accurately divided into four at the time when the node 4 is added, the next node 5 is in the middle of the area managed by the node 3, that is, the area between the node 1 and the node 3. The area assigned with the identifier and handled by the node 3 is divided into two.

  FIG. 7 shows a state when three identifiers are assigned to each of three nodes using the definitive identifier assignment method shown in FIG. As in the case of FIG. 2, it can be seen that the areas handled by the individual nodes are more uniform than the example of FIG. 6 by the law of large numbers. Further, unlike the example of FIG. 2, since the identifier value is determined deterministically, the identifier is always arranged in the ID space in the same form without variation depending on the probability. Therefore, compared with the conventional identifier assignment method of FIGS. 1 and 2 and the single identifier assignment method of FIG. 6, the multiple identifier assignment method of the present invention shown in FIG. 7 is stable regardless of the probability. It can be seen that this is the method in which the region is most uniform.

  8 and 9 show changes in the assigned area of the node when the node is deleted. Such node deletion is triggered by, for example, a node deletion instruction from the administrator of the distributed system, a node failure, or a node disconnection from the distributed system. FIG. 8 shows a change in the assigned area of the node when node 1 is deleted in the example of FIG. 1 in which a single identifier is assigned to each node. As shown on the left side of FIG. 8, before node 1 was deleted, node 1 was responsible for the area of arc 201 and node 4 was responsible for the area of arc 200. As shown on the right side of FIG. 8, when node 1 is deleted, the entire area of arc 201 that node 1 was in charge of is directly transferred to adjacent node 4, and the area in charge of node 4 is arc 202. Increase to

  On the other hand, FIG. 9 shows a change in the assigned area of the node when node 1 is deleted in the example of FIG. 2 in which a plurality of identifiers are assigned to each node. As shown on the left side of FIG. 9, before node 1 is deleted, node 1 is in charge of three arcs 301, 303, and 305. As shown in the right side of FIG. 9, when node 1 is deleted, the three areas in charge of the arc 301 are inherited by the arc 300 in charge of the node 2 and become an arc 306. . The region of the arc 303 is taken over by the arc 302 in charge of the node 2 and becomes an arc 307. The area of the arc 305 is taken over by the arc 304 in charge of the node 3 and becomes an arc 308. In this way, by assigning a plurality of identifiers to each node, compared to the example of FIG. 8, in the example of FIG. Although the uniformity is increased, in the example of FIG. 9, since the three responsible areas of the node 1 are taken over by the two nodes 2 and 3, the uneven increase is suppressed. Further, by increasing the number of identifiers assigned to one node, it is possible to further suppress non-uniformity when deleting nodes.

Here, the advantage of the identifier assignment method for assigning a plurality of identifiers to one node when the node is deleted has been described using the examples of FIGS. 1 and 2, but the example of FIGS. 6 and 7 is also used. It will be apparent that a similar advantage can be obtained by assigning multiple identifiers.
Even in the examples so far, the example of FIG. 7 using the multi-identifier allocation method of the present invention distributes the load assigned to each node most evenly in both the node addition state and the node deletion state. It will be understood that

  Next, an identifier assignment method according to an embodiment of the present invention will be described with reference to FIG. FIG. 10 is a flow diagram illustrating processing of a definitive multiple identifier assignment method according to one embodiment of the present invention. Note that the processing flow shown in FIG. 10 is merely exemplary, and the present invention is not limited to the illustrated processing flow.

  The identifier assignment method according to the present embodiment is started by an identifier assignment request for a newly added node (step S101). Next, in step S102, the number v of identifiers assigned to each node, the total number a of all identifiers existing at this time, and an identifier list describing these identifiers are input. Thereafter, in step S103, 0 is set to the counter variable i as an initialization process.

  Next, in step 104, it is determined whether the total number of identifiers allocated at present is 0, that is, if a = 0. If the total number of currently assigned identifiers is 0 (step S104: yes), that is, if the identifier to be currently assigned is the first identifier of the first node in this distributed system, it is assigned in step S105. Set the identifier value to 0. For example, this corresponds to the identifier of the uppermost node 1 shown in FIG.

  On the other hand, if the total number of currently assigned identifiers is not 0 (step S104: no), it is further determined in step S106 whether the total number of currently assigned identifiers is 1, that is, a = 1. The If the total number of identifiers assigned at present is 1 (S106: yes), that is, if the identifier to be currently assigned is the second identifier in this distributed system, the value of the assigned identifier is set in step S107. Set to (upper limit of identifiers divided by 2). For example, this corresponds to the identifier of node 1 at the lower end shown in FIG.

On the other hand, when the total number of identifiers allocated at the present time is not 1 (S106: no), the multiple identifier allocation method of the present embodiment determines the identifier value to be allocated with reference to the identifier list in step S108 and subsequent steps. In step S108, the identifier list is referred to determine a region where the distance between adjacent identifiers is maximum. Next, in step S109, when identifiers corresponding to the start point and end point of the determined area are ID _k and ID _{k + 1} , respectively, the assigned identifier is a value that bisects this area, that is, (ID It is set to the identifier calculated by _k + ID _{k + 1} ) / 2. Strictly speaking, the maximum area | ID _k −ID _{k + 1} | may exist across the upper limit of the identifier. In this case, although the above equation cannot be divided into two equal parts, an equation that appropriately returns the median value of the region is applied.

  After a new identifier is acquired in steps S105, S107, and S109 in this way, in step S110, the counter variable i and the total number of identifiers a are each incremented by 1, and the identifier list is updated.

  In step S111, if the counter variable i is less than the number of identifiers v assigned to each node (S111: no), the process returns to step S106 to perform further procedures by assigning further identifiers and executing the same procedure. Is assigned. On the other hand, when the counter variable i is equal to the number of identifiers v assigned to each node (S111: yes), the necessary processing of v identifiers has been completed, so the processing is terminated (step S112). .

  In the identifier assigning apparatus and method of the present invention, an identifier is assigned to each of the data managed by the node, the range of the identifier of the data managed by a certain node, the identifier of the node itself, and an identifier having a value smaller than that of the node. It is determined between the identifier of the node having the largest identifier among the possessed nodes. In addition, a plurality of identifiers are assigned to each node in order from the region having the largest identifier interval so that the load such as the amount of data handled by each node is evenly distributed.

  In the above-described embodiment, the case where the identifier assigning apparatus and method of the present invention are used mainly in the consistent hashing framework is described. However, the present invention is not limited to this, and the identifier is not limited to this. Can be applied to any distributed program that has more meaning than identification.

  According to the identifier assigning apparatus and method of the present invention, it is possible to make the intervals of identifiers in the distributed system uniform, thereby making the amount of processing and data handled by the nodes uniform, and the burden amount between the nodes becomes fair. Also, the problem of uniformity being lost when a node is deleted is alleviated by assigning a plurality of identifiers to each node. With these characteristics, in a distributed system, the burden between nodes can be equalized regardless of the situation of addition / deletion of nodes, and the utilization efficiency of the entire system can be increased.

  As mentioned above, although the Example of this invention was explained in full detail, this invention is not limited to the specific embodiment mentioned above, In the range of the summary of this invention described in the claim, various deformation | transformation・ Change is possible.

DESCRIPTION OF SYMBOLS 100 Identifier allocation apparatus 101 Identifier list storage part 110 Identifier determination part 111 Deterministic identifier determination part 112 Multiple identifier allocation part

Claims (5)

A plurality of node devices operate in a coordinated manner , an identifier is assigned to each node device and each data managed by the node device, and a node device in charge of managing each data is determined by the value of the identifier. In a distributed system in which a plurality of identifiers are assigned to a device and each data identifier is assigned by a hash function having a unique value as an input, the identifier is assigned to the node device,
A first step of detecting an identifier assignment event for a node device of the distributed system;
Upon detecting the identifier assignment event, a second step of accessing an identifier list storing assigned identifiers in the distributed system;
Referring to the identifier list, a third step of determining adjacent identifiers having a maximum area between adjacent identifiers stored in the identifier list;
A fourth step of assigning any identifier in the area between the determined adjacent identifiers to the node device;
I have a,
How to assign multiple identifiers to the third step and the fourth step and the node device number of repetitions of the predetermined number. Detecting a node device deletion event of deleting a node device to which an identifier has been assigned in the distributed system;
When the node device deletion event is detected, the identifier assigned to the node device is deallocated, and the processing load of the deleted node device is a node having an identifier adjacent to the deallocated identifier in the identifier list. Reassigning to the device;
Further comprising The method of claim 1. The method according to claim 1 or 2 , wherein the fourth step assigns the identifier to the node device so as to equally divide an area between the determined adjacent identifiers. The fourth step allocates the identifier to the node device so as to divide an area between the determined adjacent identifiers according to the computing capacity and / or storage capacity of the node device. 3. The method according to any one of 3 . Computer program for executing the claims 1 to 4 The method of any one claim in the computer.

Images