TWI414162B - An optimal method for placing storage nodes in a network - Google Patents

Abstract

An optimal method for placing storage nodes in a network, more particularly in a tree-network with heterogeneous channel costs, is provided to determine the minimum overall communication cost in the tree-network. A node is fully covered if data from all source nodes are pushed to this node. If at least one such node exists, the method determines the minimum set of fully covered nodes and makes them the storage nodes. If there are no nodes that can be fully covered, then a node with the minimum communication cost is set as storage node. The present invention decreases the access cost by retrieving data from storage nodes. In addition, the present invention determines the optimal locations of storage nodes which render a minimum overall communication cost for a tree-network with heterogeneous channel costs.

Description

Method for optimizing storage nodes in a network

The present invention is a method for finding a data storage node group with the lowest overall communication cost in a tree network, and in particular, a method for finding the lowest overall communication cost in a tree network of heterogeneous channel costs.

According to the rapid development of wireless communication and microelectronics technology, and widely used in many wireless networks, common applications such as MEMS (Micro-Electro-Mechanical Systems) technology, wireless sensor networks and Zigbee Wait. The above related technologies are generally applied to the collection, storage, and processing of environmental materials.

The tree network structure is a widely used network topology. The aforementioned Zigbee is a tree network structure like this. In Zigbee's application, access is an important function. Its main technical features are collected and stored on a storage device by using the data collected and sensed by each node in the network. A related application can be found in the disclosure of US Patent Application No. US 2008/0007403, which is based on Zigbee wireless transmission technology and conforms to the IEEE 802.15.4 specification and can be connected to an external communication network. Family security system. The home security system uses a receiving module to receive the information collected by each sensing unit, and stores the collected information in a storage unit, and transmits the collected information to the external terminal device when needed. . However, such a master-slave architecture can only send data through a primary storage device or actively request the primary storage device to obtain data by each terminal device.

Such an architecture usually operates in traditional resource sharing, that is, the resource provider places all available resources under a fixed server or server group, and the user needs to search through the server to obtain Resources. However, the master-slave network architecture has many shortcomings, such as: it is easy to interrupt the resource sharing service due to server failure and damage; or the server may be attacked by a person with a heart, such as a service. Blocking or being invaded by hackers to destroy and stealing jacquards; or, because the server must bear a large amount of network bandwidth, it is limited by the hardware of the server, so when the number of resources is too large It will affect the efficiency of resource sharing and search.

In addition, the method used in the prior art to reduce the overall communication cost of the communication network is directed to a homogenous network, wherein the homogenous network refers to the communication cost of the channel is symmetrical. This type of network is also known as a single-link cost network. However, some existing wireless network systems are mostly Heterogeneous networks, that is, the communication cost of their channels is heterogeneous. Therefore, the aforementioned conventional methods cannot be applied to the heterogeneous wireless networks, and thus the overall communication cost of the wireless networks cannot be effectively reduced. That is, the conventional technology cannot address the communication cost of the asymmetric channels. The network to find out the best overall communication costs.

Therefore, there is a need for a method for finding the location of at least one optimal storage node in the network structure of the communication cost of the heterogeneous channel for storing the data collected by the data source nodes in the network, and letting each By querying the closest storage node data in the storage node, the node can develop a method for finding the storage node with the lowest overall communication cost in the minimum linear time.

In view of the above-mentioned conventional techniques that cannot solve the problem of the overall minimum communication cost of the heterogeneous network, the present invention discloses a method for determining an optimized storage node in a network to effectively reduce communication costs.

One object of the present invention is to determine at least one storage node for storing data collected by all nodes in the network and providing the data for reading and use by all nodes.

Another object of the present invention is to develop a method for finding a storage node with the lowest overall communication cost in a minimum linear time.

A further object of the present invention is to improve the speed of calculation in a network storage location application.

To achieve the foregoing objects, the present invention discloses a method of optimizing a storage node, which is a method for determining a minimum communication cost in a tree network of a single source node. The method comprises the following steps: first, obtaining structural information of the tree network, such as the structure of the tree, the data generation rate of each data source node, and the query rate of each node, wherein the tree network includes a data source node, the data source node is used to generate data; then, the node having a single path is found as an edge node; and then, starting from any node in the tree network, the node is associated with the node The path of the neighbor node is removed to form two subtrees. If the sum of the data source node generation rates of one subtree is less than the sum of the data query rates of all the other subtrees, the node is a storage node; then, the recovery is performed. Removing the path, repeating the above steps until all the paths in the tree network perform the above steps; finally, if the result has at least one storage node, the at least one storage node and the data source node form a storage node set; Any node in the tree network can query the data on any node in the storage node set. Using the aforementioned method, the lowest overall communication cost can be determined.

The present invention further discloses another method for determining an optimized storage node, which is a method for determining the overall minimum communication cost in a tree network of a plurality of data source nodes. The method comprises the following steps: first, obtaining structural information of the tree network, such as the structure of the tree, the data generation rate of each data source node, and the query rate of each node, wherein the tree network includes a plurality of data source nodes, the plurality of data source nodes generating data; and then finding a node having a single path as an edge node; and thereafter, a path from any node in the tree network to a neighbor node of the node Removed to form two subtrees. If the sum of the data source node generation rates of one subtree is less than the sum of all node data query rates of another subtree, the node is the covered node; then, the removal is resumed. Path, and repeating the above steps until all the paths in the tree network perform the above steps, wherein if one node is completely covered by the neighbor node, the node is a storage node; finally, if at least one storage node is included The storage node forms a set of storage nodes; and any node in the tree network can query any node in the storage node set Material. Using the aforementioned method, the lowest overall communication cost can be determined.

In some embodiments of the present invention, if at least one fully covered node cannot be found in the tree network of the plurality of data source nodes, the node that can achieve the lowest overall communication cost is found as the at least one storage node.

In some embodiments of the present invention, if the storage node set includes more than one node, and the storage node set has a node that is not queried by any node, the node having no query by any node is removed from the node. Store the node collection and make the node not a storage node.

The above communication cost includes the traffic when the node in the foregoing storage node set returns the data.

The aforementioned source node is a node for generating data, mainly collecting surrounding information of the node, such as temperature, humidity, power and other environmental parameters; and the storage node is used to collect all data source nodes in the foregoing tree network. The information produced. When at least one storage node is determined, the data generated by all data source nodes will be actively pushed to all storage nodes by the aforementioned data source node.

In addition, any node in the tree network is eligible to be a storage node; and any node in the tree network can query the data of at least one of the foregoing storage nodes or data source nodes.

The above query node is a node that wants to query the storage node, and each node may be a query node. When a query node wants to query the network data, the query node first sends an inquiry message to the nearest storage node, and then the node The most recent storage node passes the stored data back to the query node.

Finally, the communication cost of all adjacent nodes in the tree network can be different. If the communication cost of any channel is different, the tree network is a heterogeneous network. A homogenous network is the same as the cost of going back and forth between all nodes.

One advantage of the present invention is that the present invention solves the problem of storage nodes that are currently unable to measure the minimum overall communication cost in a minimum of linear time in a network of heterogeneous channel costs.

Another advantage of the present invention is that the present invention can be applied to a network storage location to increase the associated computing speed.

The advantages of the present invention will be apparent from the following description of the preferred embodiments of the invention.

The present invention will be described in detail with reference to the preferred embodiments and the accompanying claims Therefore, the invention may be embodied in other embodiments in addition to the preferred embodiments described herein.

In one embodiment of the invention, the present invention discloses a method of determining a storage node whose overall minimum communication cost is determined in a tree network having a single source node. The steps of the method are described in detail as follows: First, a tree network T with a single data source node is provided with a data source node s and a plurality of nodes, and all nodes thereof form a tree structure. The data generation rate of the data source node s is G _s (times/second), and the rate at which each node i queries the aggregated data is r _i (times/second). The nodes in the tree network T having only a single path are defined as edge nodes. If there are two neighboring nodes i and j in the tree network T , the path between the two nodes is defined as link(i,j) , and in addition, the node j is referred to as a neighbor node of the node i . When the path link(i,j) of the node i and its neighbor node j is removed, the tree network T forms a subtree T(i,j) rooted at the node i and the other node j The root tree T(j,i) . In addition, if there is a node x in the tree network, and there is only a single neighbor node in the path of the node x to the data source node, the neighbor node is referred to as the data source of the node x to connect to the neighbor node (Source- Bound-Neighbor). Here, the single neighbor node is defined as the node sn(x) . In this embodiment, the sum of query rates of all nodes in a subtree T(x, sn(x)) is first defined as R _x and can be expressed as the following formula. . Therefore, if the final result is R _x > G _{s ',} it means that the query rate required for the T(x,sn(x)) query data is greater than the rate at which the data is generated. In other words, it takes a lot of communication costs to ask for information. Therefore, the node x is defined as a storage node for collecting data of the data source node and providing all node queries.

Then, after repeating the foregoing steps and calculating the communication cost after each path of the entire tree network T is removed, at least one storage node may be obtained. The at least one storage node and the data source node are a set of storage nodes, and any node in the tree network T can query a storage node or a data source node that is closest to the node in the storage node set. The information is used to reduce the overall cost of inquiry and reduce the overall communication cost.

Referring to the first figure, the tree network T has 9 nodes, wherein the node s is a data source node, and the data generation rate is G _s = 5, and the data inquiry rate of each node is represented by each node. It can be seen from the first figure that there are five edge nodes which are nodes a , g , h , i and f , respectively, and the node b is defined as sn(a) . Next, referring to the second figure, removing the path link(a,b) of nodes a and b, the sum of the interrogation rates required to calculate the subtree T(a,sn(a)) is node a . The inquiry rate R _x = R _a = 5, and the data source node generation rate is G _s = 5, and it can be known that R _x = G _s does not meet the requirement of R _x > G _s , and it can be known that the node a is not one. Storage node. And so on, as shown in the third figure, if the path of node d and node s is removed, then R _x is the sum of the interrogation rates of nodes d , e , f, and i , which can be obtained R _x = R( _{d , s )} = 10, whose data source node is generated at a rate of G _s = 5, and it can be known that R _x > G _s , so d is a storage node. After the link(a, b) is restored , the foregoing steps are repeated and the communication cost after each path of the total tree network is removed, and finally the node b and the node d are the storage nodes of the tree network. For the final result, please refer to the fourth figure. In this embodiment, the nodes f , i , e can query the data on the storage node d , the node a can query the data on the storage node b , and the nodes g , h can query the data source. The data on the node s .

In another embodiment of the present invention, the present invention discloses a method for determining the overall minimum communication cost in a tree network structure T having a plurality of data source nodes. In this embodiment , the paths of the two adjacent nodes i, j are also defined as link(i, j) . In addition, the paths of the two adjacent nodes in this embodiment have a directional relationship, that is, two direction channels ( Directional Channels). The direction of nodes i to j is defined as ch(i,j) and the direction of nodes j to i is defined as ch(ji) . The sum of the interrogation rates of all nodes in a tree network T(i,j) is defined as R(i,j) and the sum of the generation rates of all data source nodes is defined as S(i,j) . The steps of the method disclosed in this embodiment are as follows: First, a tree network T' having a plurality of data source nodes includes a plurality of data source nodes and a plurality of nodes, and all the nodes form a tree. structure. The nodes in the tree network with only a single path are defined as edge nodes, and all the paths of the tree network are sequentially found according to the state of the edge node and its neighbor nodes. If there is a node j and its neighbor node i in the tree network, when the path ch(i, j) is removed, two subtrees are formed. If S(i,j) < R(j,i ) holds, the node i is said to cover the node j ; and if the node j is covered by all its neighbor nodes, the node j is said to be completely covered ( The node of Fully-Covered), and let the node j be a storage node. Therefore, S(i,j) and R(i,j) after the path of compute node i and node j are removed can be expressed as:

and

After restoring the removed path ch(i,j) , the foregoing steps are repeated and the communication cost after each path of the entire tree network is removed is calculated. If a plurality of storage nodes are determined in the tree network T , all of the storage nodes are clustered together to form a subtree. Moreover, if the storage nodes are a set of storage nodes, any node in the tree network can query the storage node or the data source node closest to the node in the storage node set. In this way, the overall inquiry cost can be reduced and the overall communication cost can be reduced.

Referring to the fifth figure, the tree network T' has 9 nodes, wherein S ₁ and S ₂ are data source nodes, and the data generation rates are G ₁ =8, G ₂ =10, and each node. The data query rate is indicated next to each node. Also known by the fifth figure of the tree network T 'edge a total of five nodes, respectively a, g, h, i and f, where S ₁ is at the node of the neighboring node c. Next, referring to the sixth figure, when the path ch ( S ₁ , c ) between the nodes c and S ₁ is removed, the sum of the storage node generation rates S ( S ) can be obtained from the equations (1) and (2). ₁ , c )=8, and the sum of the node interrogation rates is the sum of the interrogation rates of the nodes c , g , h , d , S ₂ , f and i , so R ( c, S ₁ )=20, ie S ( S ₁ , c ) < R ( c, S ₁ ). Therefore, node S ₁ covers node c .

In addition, as shown in the seventh figure, if the path ch ( c,d ) is removed , S ( d,c )=10, R ( c,d )=18, that is, S ( d,c )< R ( c,d ). Therefore, node d also covers node c . In addition, the edge node g , h also covers the node c . Therefore, it can be determined from the foregoing that the node c is a completely covered node, that is, the node c is a storage node. Then, after recovering the path ch ( S ₁ , c ) and repeating the foregoing steps to calculate the communication cost after each path of the entire tree network is removed, the available node d is also a completely covered node. The final result, as shown in the eighth figure, can be determined that the node c and the node d are the storage nodes of the tree network T' .

In another embodiment of the present invention, a method for determining an overall minimum communication cost in a tree network T' having a plurality of data source nodes is disclosed, wherein the tree network T' is not completely The covered node exists. Please refer to the fifth figure, where S ₁ , S ₂ are data source nodes, and the data generation rate G ₁ =21, G ₂ =20, and the data inquiry rate of each node is represented by each node. Repeat the previous steps and calculate the communication cost after each path of the entire tree network is removed. However, in the present embodiment, it is not possible to find at least one node as a completely covered node. Therefore, in the present embodiment, it is sufficient to find a single node that can achieve the lowest overall communication cost.

According to the foregoing method, nodes that are not completely covered in the tree network T' are found, and the nodes form a Residual Tree T _U . Assuming that one edge node of node i T _U residue of the tree, and comprising more than one node within the tree residual _U T, the assumed T _U is a set of reference storage node. At this time, if the node i is removed, and the communication cost of the channel of the node i 's unique neighbor node j to the node i is defined as c _j,i , the saved data generation cost is c _j,i × S(j , i) , and the increased data query cost is c _j,i × R(i,j) , and the communication cost that can be reduced by the last channel is defined as b(i,j) , which can be expressed as:

b(i,j) = c _j,i ‧( S ( j,i )- R ( i,j )) (3)

Repeat the foregoing steps to calculate the communication cost of the residual tree network T _U from its edge nodes and remove the paths one by one, to find a single node that can achieve the overall minimum communication cost. Referring to the ninth figure, the nodes that are not completely covered in the tree network T' are found according to the foregoing method, and the nodes are nodes S ₁ , c and d respectively , and the nodes are a residual tree. , assuming c _{c , d} = 0.5, c _{d , c} = 1.2, =1.0 and =0.8, available from equation (3)

b(c,d) = c _{d , c} ‧ ( S ( d , c ) - R ( c , d )) =1.2×(20-18)=3.6,

b(d,c) = c _{c , d} ‧ ( S ( c , d ) - R ( d , c )) =0.5×(21-10)=5.5

Therefore, the communication cost that node S ₁ can reduce is [ S ₁ ]= b(d,c) + b(c, S ₁ )=6.5, and the communication cost that node c can reduce is [ c ]= b(S ₁ , c) + b(d,c) = 15.1, the communication cost that can be reduced by node d is [ d ]= b(S ₁ ,c) + b(c,d) =12.0. It can be seen from the above that the node c is a node that can achieve the lowest overall communication cost, and therefore, the node c is a storage node. See FIG. Ninth, a node where data may be _an inquiry, the node g, h may query information to node C, node d to the node may be c, S ₂ inquiry data, the node f to a node S, i can ask Node S ₂ data.

In another embodiment of the present invention, a method for determining an overall minimum communication cost in a tree network structure T having one or more data source nodes is disclosed. If the method determines a plurality of storage nodes, if any of the storage nodes are not queried by other nodes, then the storage node will not be a storage node. The measurement method of the storage node of this embodiment is disclosed in the foregoing embodiments. Suppose there is a tree network T , which contains a data source node and a plurality of storage nodes, one of which has a storage node j , and any node adjacent to the node j is a node x (but x ≠ sn(j) ), if The sum of the interrogation rates of the nodes x is R _x =0, which means that the storage node j is not interrogated by any of the subtrees T(x, j) , so the node j is not a storage node. However, if the embodiment finally includes only one storage node, the storage node is reserved in the tree network, and the storage node is retained even if the storage node is not queried by any node. In the case of multiple data source nodes and a plurality of storage nodes, one of the storage nodes j and the neighbor nodes of the node j are x , if r _i =0 and all neighbor nodes x are completely covered nodes, or R(x,j) =0, it means that the storage node j is not queried by any node in the subtree T(x, j) , so the node j is not a storage node.

The above description is a preferred embodiment of the invention. Those skilled in the art should be able to understand the invention and not to limit the scope of the patent claims claimed herein. The scope of patent protection is subject to the scope of the patent application and its equivalent fields. Any modification or refinement made by those skilled in the art without departing from the spirit or scope of the present invention is equivalent to the equivalent change or design made in the spirit of the present disclosure, and should be included in the following patent application scope. Inside.

T, T’. . . network

a, b, g, s, s ₁ , s ₂ , h, d, e, f, i. . . node

The first diagram shows a tree network of single source nodes.

The second figure shows the path removal diagram of node a and node b of the tree network of a single data source node.

The third figure shows the path removal of the node s and node d of the tree network of a single source node.

The fourth graph shows the results of determining the best storage node in a tree network of single data source nodes.

The fifth picture shows a tree network of multiple sources.

The sixth figure shows a schematic diagram of path removal of node S ₁ and node c of a tree network of a plurality of data sources.

The seventh figure shows a schematic diagram of node c and node d path removal of a tree network of a plurality of data sources.

The eighth diagram shows the best storage node results for a tree network of multiple data sources.

The ninth figure shows the best storage node results for a tree network that cannot determine the multiple data sources of the storage node.

T’. . . network

a, c, g, s ₁ , s ₂ , h, d, f, i. . . node

Claims (23)

A method for optimizing a storage node in a network, the method comprising the steps of: (a) first obtaining structural information of a network; (b) the network includes a data source node, and the data source node generates data; (c) finding a node with a single path as an edge node; (d) starting from any node in the network, removing the path between the node and one of the neighbor nodes of the node to form two sub-trees, if one If the sum of the data source node generation rates of the subtree is smaller than the sum of all node data query rates of the other subtree, the node is a storage node; (e) recovering the removed path, repeating step (d) until all the networks All of the paths perform step (d); and (f) if there is at least one storage node, the at least one storage node and the data source node form a set of storage nodes. The method of claim 1, further comprising the step of: (g) if the storage node set includes more than one node, and the storage node set has nodes that are not inquired by any node, The node moves out of the set of storage nodes and makes the node not a storage node. The method of claim 1 or 2 further includes a step of: (h) all nodes in the network can query data on any node in the set of storage nodes. The method of claim 1 or 2, wherein the channel communication cost of all adjacent nodes in the network can be different. The method of claim 1 or 2, wherein the communication cost comprises a communication amount when the node in the collection of storage nodes returns data. The method of claim 4, wherein the communication cost comprises a communication amount when the node in the collection of storage nodes returns data. The method of claim 1 or 2, wherein the structural information includes the network structure, the data source node generation rate, and the node query rate. The method of claim 4, wherein the structural information includes the network structure, the data source node generation rate, and the node query rate. The method of claim 1 or 2, wherein the data generated by the data source node is actively pushed by the data source node to the at least one storage node. The method of claim 4, wherein the data generated by the data source node is actively pushed by the data source node to the at least one storage node. The method of claim 3, wherein all nodes in the network first transmit an inquiry message to the closest storage node or the closest source node, and then the storage node or the data source node The stored data is transferred to this node. A method for optimizing a storage node in a network, the method comprising the steps of: (a) first obtaining structural information of a network; (b) the network comprises a plurality of data source nodes, and the plurality of data source nodes are Generating data; (c) finding a node with a single path as an edge node; (d) removing a path from any node in the network and one of the neighbor nodes of the node to form two subtrees, if one If the sum of the data source node generation rates of the tree is less than the sum of all node data query rates of the other subtree, then the node is the covered node; (e) recovering the removed path, repeating step (d) until all the networks All of the paths perform step (d); (f) if one node is completely covered by the neighbor node, the node is a storage node; and (g) finally if at least one storage node is included, the at least one storage node A set of storage nodes is formed. The method of claim 12, further comprising the step of: (h) if at least one of the storage nodes cannot be found in the network, finding a node that can achieve the lowest overall communication cost is the at least one storage. node. The method of claim 12, further comprising the step of: (i) if the storage node set includes more than one node, and the storage node set has nodes that are not inquired by any node, The node moves out of the set of storage nodes and makes the node not a storage node. The method of claim 12, 13 or 14 further comprises a step of: (j) all nodes in the network can query data on any node in the set of storage nodes. The method of claim 12, 13 or 14, wherein the cost of channel communication for all adjacent nodes in the network can be different. The method of claim 12, 13 or 14, wherein the structural information comprises the network structure, the data source node generation rate, and the node query rate. The method of claim 16, wherein the structural information includes the network structure, the data source node generation rate, and the node query rate. The method of claim 12, 13 or 14, wherein the communication cost comprises a communication amount when the node in the collection of storage nodes returns data. The method of claim 16, wherein the communication cost comprises a communication amount when the node in the collection of storage nodes returns data. The method of claim 12, 13 or 14, wherein the data generated by the plurality of data source nodes is actively pushed to the at least one storage node by the plurality of data source nodes. The method of claim 16, wherein the data generated by the plurality of data source nodes is actively pushed to the at least one storage node by the plurality of data source nodes. The method of claim 15, wherein all nodes in the network first transmit an inquiry message to the closest storage node, and the storage node transmits the stored data to the node.