TWI544766B - Network topology system and method - Google Patents

Network topology system and method Download PDF

Info

Publication number
TWI544766B
TWI544766B TW103138885A TW103138885A TWI544766B TW I544766 B TWI544766 B TW I544766B TW 103138885 A TW103138885 A TW 103138885A TW 103138885 A TW103138885 A TW 103138885A TW I544766 B TWI544766 B TW I544766B
Authority
TW
Taiwan
Prior art keywords
sequence
nodes
value
connection
node
Prior art date
Application number
TW103138885A
Other languages
Chinese (zh)
Other versions
TW201618504A (en
Inventor
黃吉川
梁啟修
陳昱廷
Original Assignee
國立成功大學
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 國立成功大學 filed Critical 國立成功大學
Priority to TW103138885A priority Critical patent/TWI544766B/en
Publication of TW201618504A publication Critical patent/TW201618504A/en
Application granted granted Critical
Publication of TWI544766B publication Critical patent/TWI544766B/en

Links

Landscapes

  • Data Exchanges In Wide-Area Networks (AREA)

Description

網路拓樸系統及方法 Network topology system and method

本發明有關於一種電腦網路之技術領域,特別有關於一種網路拓樸系統及方法。 The invention relates to a technical field of computer networks, and in particular to a network topology system and method.

高性能電腦或電腦系統可提供硬體和軟體的最佳化使用,並利用包含一個或多個微處理器和記憶體之處理節點來實現。這些電腦系統有時被稱作共享多處理器系統,在一個共享的多處理器電腦系統中,視處理器為網路節點以相互連接,使它們可以彼此進行通訊,以共享作業系統、資源、資料、記憶體等。 A high-performance computer or computer system provides optimal use of hardware and software and is implemented using processing nodes that include one or more microprocessors and memory. These computer systems are sometimes referred to as shared multiprocessor systems. In a shared multiprocessor computer system, the visual processors are network nodes connected to each other so that they can communicate with each other to share operating systems, resources, Information, memory, etc.

在平行運算電腦系統中,隨著電腦節點數增加,節點間網路通訊會大幅度增加,為了使兩兩節點之間的通訊速度增快,會使用交換器來作為各節點間通訊之用。然而,節點的數量愈大,所使用的交換器的數量就愈多,而使用例如胖樹(Fat Tree)結構之交換器的層數亦會增加,因而造成整個網路系統的結構相當複雜且耗電,且網路系統的成本亦會增加。 In the parallel computing computer system, as the number of computer nodes increases, the network communication between nodes will increase greatly. In order to increase the communication speed between the two nodes, the switch is used as the communication between the nodes. However, the larger the number of nodes, the more switches are used, and the number of layers using switches such as the Fat Tree structure is also increased, resulting in a complicated structure of the entire network system. Power is consumed and the cost of the network system will increase.

有鑑於上述問題,本發明之目的係提供一種網路拓樸系統及方法,其拓樸數學計算方式是在具有多個節點的系統中為各節點尋找一組節點連線規則(connection rules)以及各節點與其他 節點連接的數目,使其可建構出一個讓所有節點根據其連線規則和其連接數所構成的網路拓樸,而無需使用交換器,使整個網路系統的結構變得簡單且無交換器之耗電,所以網路系統的成本不會增加,再者每個節點透過相同形式的通訊自由(使用同一組連線規則),並使得所有節點之連接數非常接近者,而達成所謂民主化維度(Democratized Dimension)。 In view of the above problems, an object of the present invention is to provide a network topology system and method, wherein the topology mathematical calculation method is to find a set of node connection rules for each node in a system with multiple nodes. Each node and other The number of nodes connected, so that it can construct a network topology that allows all nodes to be based on their connection rules and their connections without using a switch, making the structure of the entire network system simple and non-exchangeable. The power consumption of the network, so the cost of the network system will not increase, and each node will pass the same form of communication freedom (using the same set of connection rules), and the connection number of all nodes is very close, and the so-called democracy is achieved. Democratized Dimension.

本發明之第一態樣係提供一種網路拓樸系統,其包括:複數個節點,每一節點係一單核心晶片、一處理器、一電腦、一組電腦之內部最佳化組及一計算機中心之其中一者,每一節點根據連線規則與該等節點中之複數目的節點進行連接;其中,該連線規則之建立包括下列步驟:計算一質數數列中相鄰質數之差值,以產生一質數數值差數列;計算該質數數值差數列中目前位置之數值與所有之前位置之數值之和,以產生一通訊策略數值數列;取出一遞迴數列中個數為一連接節點數之數值作為一索引數列,取出該通訊策略數值數列中個數為該連接節點數之數值,以產生一連線策略數值數列,其中該索引數列中之數值係為該通訊策略數值數列之數列位置,其數列位置之數值為該連線策略數值數列之數值;以及計算每一奇數與該連線策略數值數列中之數值之和,以產生複數個連線節點編號數列,以建立為每一奇數編號之節點連結相應之連線節點編號數列中之數值編號之節點之該連線規則。 A first aspect of the present invention provides a network topology system including: a plurality of nodes, each node being a single core chip, a processor, a computer, an internal optimization group of a set of computers, and a One of the computer centers, each node is connected to a plurality of nodes in the nodes according to a connection rule; wherein the establishing of the connection rule includes the following steps: calculating a difference between adjacent prime numbers in a prime number column, To generate a prime numerical difference series; calculate the sum of the current position value of the prime numerical difference series and the values of all previous positions to generate a communication strategy numerical sequence; take out a number of reciprocal series as a number of connected nodes The value is used as an index sequence, and the number in the number sequence of the communication strategy is taken as the value of the number of connected nodes to generate a connection strategy number sequence, wherein the value in the index sequence is the sequence position of the number sequence of the communication strategy. The value of the sequence position is the value of the connection strategy numerical sequence; and the sum of each odd number and the value in the connection strategy numerical sequence Generating a plurality of connections to the number of nodes numbered columns, linked to the connection node of rule values corresponding to the node number of the connection in series of the node number of each of odd-numbered.

本發明之第二態樣係提供一種網路拓樸方法,其應用於一網路拓樸系統,該網路拓樸系統包括複數個節點,每一節點係一單核心晶片、一處理器、一電腦、一組電腦之內部最佳化組及一計算機中心之其中一者,該方法包括下列步驟:由該網路拓樸系統根據一組連線規則將該等節點中之任一起始節點與複數節點進行連接;其中,建立該連線規則包括下列步驟:計算一質數數列中相鄰質數之差值,以產生一質數數值差數列;計算該質數數值差數列中目前位置之數值與所有之前位置之數值之和,以產生一通訊策略數值數列;取出一遞迴數列中個數為一連接節點數之數值作為一索引數列,取出該通訊策略數值數列中個數為該連接節點數之數值,以產生一連線策略數值數列,其中該索引數列中之數值係為該通訊策略數值數列之數列位置,其數列位置之數值為該連線策略數值數列之數值;以及計算每一奇數與該連線策略數值數列中之數值之和,以產生複數個連線節點編號數列,以建立為每一奇數編號之節點連結相應之連線節點編號數列中之數值編號之節點之該連線規則。 A second aspect of the present invention provides a network topology method for a network topology system, the network topology system including a plurality of nodes, each node being a single core chip, a processor, A computer, an internal optimization group of a set of computers, and a computer center, the method comprising the steps of: starting, by the network topology system, any one of the nodes according to a set of connection rules Connecting with the plurality of nodes; wherein establishing the connection rule comprises the steps of: calculating a difference between adjacent prime numbers in a prime number column to generate a prime numerical difference series; calculating a value of the current position in the prime value difference series and all The sum of the values of the previous positions to generate a communication strategy numerical sequence; taking the number of the number of connected nodes in a recursive sequence as an index sequence, and extracting the number of the communication strategy numerical sequence as the number of the connected nodes The value is used to generate a connection strategy value sequence, wherein the value in the index sequence is the sequence position of the number sequence of the communication strategy, and the value of the sequence position is The value of the connection strategy value sequence; and the sum of each odd number and the value in the connection strategy value sequence to generate a plurality of connection node number sequence columns to establish a connection for each odd numbered node link The connection rule for the node of the numerical number in the node number column.

圖1為本發明之網路拓樸系統之節點架構示意圖;圖2為本發明之網路拓樸方法之流程圖;圖3為本發明之建立連線規則之方法之流程圖;以及 圖4A至4D為本發明之網路拓樸系統與各種網路拓樸系統之效能比較之示意圖。 1 is a schematic diagram of a node architecture of a network topology system according to the present invention; FIG. 2 is a flowchart of a network topology method of the present invention; FIG. 3 is a flowchart of a method for establishing a connection rule according to the present invention; 4A to 4D are diagrams showing the performance comparison between the network topology system of the present invention and various network topology systems.

為使熟習本發明所屬技術領域之一般技藝者能更進一步了解本發明,下文特列舉本發明之較佳實施例,並配合所附圖式,詳細說明本發明的構成內容及所欲達成之功效。 The present invention will be further understood by those of ordinary skill in the art to which the present invention pertains. .

圖1為本發明之網路拓樸系統之節點架構示意圖。在圖1中,網路拓樸系統包括複數個節點(節點編號1、2、...、16),每一節點係為一單核心晶片(single core on chip)、一處理器、一電腦、一組電腦之內部最佳化組(a group of internally optimized group of computers)或一計算機中心(a computational center),網路拓樸系統中之每一節點依據一組連線規則(connection rule)於該等節點中之任一起始節點與複數節點進行連接。 FIG. 1 is a schematic diagram of a node architecture of a network topology system according to the present invention. In FIG. 1, the network topology system includes a plurality of nodes (node numbers 1, 2, ..., 16), each node being a single core on chip, a processor, a computer. , a group of internally optimized group of computers or a computational center, each node in the network topology system is based on a set of connection rules Any of the starting nodes in the nodes are connected to a plurality of nodes.

本發明之實施例以16個節點為例說明網路拓樸系統之通訊操作,但該節點數量並非侷限本發明,本發明之網路拓樸系統之節點數量可高達數千個或甚至百萬個節點數量。 The embodiment of the present invention uses 16 nodes as an example to illustrate the communication operation of the network topology system, but the number of nodes is not limited to the present invention, and the number of nodes of the network topology system of the present invention can be up to thousands or even millions. The number of nodes.

連線規則之建立方式說明如下。首先輸入一總節點數N及一連接節點數K,其中總節點數N表示網路拓樸系統之節點數量,連接節點數K表示每一節點可平行連接其它節點的數量。在本實施中,總節點數N為16,連接節點數K為4。 The way to establish the connection rules is as follows. First, a total number of nodes N and a number of connected nodes K are input, wherein the total number of nodes N represents the number of nodes of the network topology system, and the number K of connected nodes indicates the number of other nodes that can be connected in parallel to each node. In the present embodiment, the total number of nodes N is 16, and the number K of connection nodes is 4.

接著,列出數值小於總節點數N之一遞迴數列,其中遞迴數列可採用熟知之遞迴數列。列出數值為質數之一質數數列,如2、3、5、7、11、13、17、19、...。計算質數數列中相鄰質數之差值,如3-2=1、5-3=2、7-5=2、11-7=4、13-11=2、17-13=4、...以 產生數值為1、2、2、4、2、4、...之一質數數值差數列。 Next, a recursive sequence in which the value is less than one of the total number of nodes N is listed, wherein the recursive sequence can use a well-known recursive sequence. List the prime numbers as a prime number, such as 2, 3, 5, 7, 11, 13, 17, 19, .... Calculate the difference between adjacent prime numbers in the prime number series, such as 3-2=1, 5-3=2, 7-5=2, 11-7=4, 13-11=2, 17-13=4, .. To A numerical difference series of values of 1, 2, 2, 4, 2, 4, ... is generated.

計算質數數值差數列中目前位置之數值與所有之前位置之數值之和,亦即第1位置之數值和為1、第2位置之數值和為1+2=3、第3位置之數值和為1+2+2=5、第4位置之數值和為1+2+2+4=9、第5位置之數值和為1+2+2+4+2=11、第6位置之數值和為1+2+2+4+2+4=15、...,經上述計算以產生數值為1(1)、3(2)、5(3)、9(4)、11(5)、15(6)、...之一通訊策略數值數列,其中括弧中之數字表示該數值在通訊策略數值數列中之位置之索引。 Calculate the sum of the value of the current position in the numerical difference series and the value of all previous positions, that is, the value of the first position is 1, the value of the second position is 1 + 2 = 3, and the value of the third position is 1+2+2=5, the value of the 4th position is 1+2+2+4=9, the value of the 5th position, and the value of 1+2+2+4+2=11, the 6th position and Is 1+2+2+4+2+4=15,..., calculated by the above to produce values of 1(1), 3(2), 5(3), 9(4), 11(5) , 15(6), ... A communication strategy numerical sequence, where the number in parentheses indicates the index of the position of the value in the number sequence of the communication strategy.

取出遞迴數列中位置為第3個至第K+2個(K為4)之數值作為一索引數列,在本實施例中,索引數列之數值為1、2、3、5。取出通訊策略數值數列中K個數值,以產生一連線策略數值數列,其中K個數值係索引數列中之數值對應通訊策略數值數列之數列位置之數值,在本實施例中,連線策略數值數列之數值為1(1)、3(2)、5(3)、11(5)。 The value of the third to K+2 (K is 4) in the recursive sequence is taken as an index sequence. In the present embodiment, the value of the index sequence is 1, 2, 3, and 5. The K values in the number sequence of the communication strategy are retrieved to generate a connection strategy number sequence, wherein the values in the K number index series correspond to the values of the sequence positions of the communication strategy numerical sequence. In this embodiment, the connection strategy value The values of the series are 1 (1), 3 (2), 5 (3), and 11 (5).

計算每一奇數與連線策略數值數列中之數值之和,以產生複數個連線節點編號數列,以建立為每一奇數編號之節點連結相應之連線節點編號數列中之數值編號之節點之連線規則,在本實施例中,奇數為1之節點平行連結編號為1+1=2、1+3=4、1+5=6、1+11=12等節點、奇數為3之節點平行連結編號為3+1=4、3+3=6、3+5=8、3+11=14等節點、...(如圖1所示)。 Calculating the sum of the values in each of the odd number and the connection strategy value series to generate a plurality of connection node number sequence columns to establish a node for each odd numbered node to link the corresponding numerical number in the connection node number sequence The connection rule, in this embodiment, the node with the odd number of 1 is a node with a parallel connection number of 1+1=2, 1+3=4, 1+5=6, 1+11=12, and an odd number of 3 nodes. The parallel link numbers are 3+1=4, 3+3=6, 3+5=8, 3+11=14, etc., as shown in Fig. 1.

在一總節點數為N的網路拓樸系統中找出每一個節點之連接數接近一特定值,即K i K,i=1,2,.....,N,其中K=log2 N,且具有最短之平均距離者,稱為拓樸優化(topology optimization)。 In a network topology system with a total number of nodes N, find that the number of connections of each node is close to a specific value, that is, K i K, i =1 , 2 , ..... , N, where K = log 2 N, and with the shortest average distance, known topology optimization (topology optimization).

在一個總節點數為N的網路拓樸中,所有節點彼此之 間的最長距離可以表示成以下的距離矩陣(Distance Matrix): 其中d ij 為節點i與j之間的最長距離,且d ii =0,即對角線元素皆為0。 In a network topology with a total number of nodes of N, the longest distance between all nodes can be expressed as the following distance matrix (Distance Matrix): Where d ij is the longest distance between nodes i and j, and d ii =0, that is, the diagonal elements are all 0.

一個拓樸的平均距離A的定義如下: ,i,j=1,2,....,N其中d ij 是指任意兩節點i與j之間的最長距離。例如總節點數N為4,連接節點數K為2,距離矩陣表示成 則A=(1+2+1+…+2+1)/(4×3)=1.33 The definition of the average distance A of a topology is as follows: , i, j =1 , 2 , .... , N where d ij is the longest distance between any two nodes i and j. For example, the total number of nodes N is 4, the number of connected nodes K is 2, and the distance matrix is expressed as Then A=(1+2+1+...+2+1)/(4×3)=1.33

讓每個節點透過相同形式的通訊自由(使用同一組節點連線規則),並使得所有節點之連接數非常接近者,即K i K,i=1,2,.....,N,便稱為民主化維度環境。 Let each node pass the same form of communication freedom (using the same set of node connection rules) and make the number of connections of all nodes very close, ie K i K, i =1 , 2 , ..... , N It is called the democratization dimension environment.

圖2為本發明之網路拓樸方法之流程圖。在說明圖2之流程步驟時參考圖1之節點。 2 is a flow chart of a method for network topology of the present invention. Reference is made to the node of Figure 1 in illustrating the flow steps of Figure 2.

在圖2中,建立好的連線規則首先用以連接於圖1之各個節點(步驟S20)。 In FIG. 2, the established connection rules are first used to connect to the respective nodes of FIG. 1 (step S20).

圖3為本發明之建立連線規則之方法之流程圖,建立連線規則之資料的步驟如圖3之流程圖所示。在圖3中,輸入一總節點數N及一連接節點數K,在本實施中,網路拓樸系統之總節點數N為16,每一節點可平行連接其它節點的連接節點數K為4(步 驟S30)。 3 is a flow chart of a method for establishing a connection rule according to the present invention. The steps of establishing the data of the connection rule are as shown in the flow chart of FIG. 3. In FIG. 3, a total number of nodes N and a number K of connection nodes are input. In this implementation, the total number of nodes N of the network topology system is 16, and the number K of connection nodes of each node that can be connected in parallel to other nodes is 4 (step Step S30).

列出數值小於總節點數N之一遞迴數列,在本實施例中,遞迴數列之數值為0、1、1、2、3、5、8、13(步驟S32),其中遞迴數列可採用熟知之遞迴數列。 The recursive sequence is set to 0, 1, 1, 2, 3, 5, 8, and 13 (step S32), wherein the value is less than one of the total number of nodes N. In this embodiment, the value is 0, 1, 1, 2, 3, 5, 8, and 13 (step S32). A well-known recursive sequence can be used.

列出數值為質數之一質數數列,其數值為2、3、5、7、11、13、17、19、...(步驟S34)。計算質數數列中相鄰質數之差值,如3-2=1、5-3=2、7-5=2、11-7=4、13-11=2、17-13=4、...以產生數值為1、2、2、4、2、4、...之一質數數值差數列(步驟S36)。 The numerical value is listed as a prime number of prime numbers, and the values are 2, 3, 5, 7, 11, 13, 17, 19, ... (step S34). Calculate the difference between adjacent prime numbers in the prime number series, such as 3-2=1, 5-3=2, 7-5=2, 11-7=4, 13-11=2, 17-13=4, .. A numerical difference series of values of 1, 2, 2, 4, 2, 4, ... is generated (step S36).

計算質數數值差數列中目前位置之數值與所有之前位置之數值之和,以產生一通訊策略數值數列,如在本實施例中,第1位置之數值和為1、第2位置之數值和為1+2=3、第3位置之數值和為1+2+2=5、第4位置之數值和為1+2+2+4=9、第5位置之數值和為1+2+2+4+2=11、第6位置之數值和為1+2+2+4+2+4=15、...,經上述計算以產生數值為1(1)、3(2)、5(3)、9(4)、11(5)、15(6)、...之通訊策略數值數列,其中括弧中之數字表示該數值在通訊策略數值數列中之位置之索引(步驟S38)。 Calculating the sum of the value of the current position in the prime value difference series and the value of all previous positions to generate a communication strategy value sequence. As in this embodiment, the value of the first position is the sum of the values of the first and second positions. 1+2=3, the value of the third position is 1+2+2=5, the value of the 4th position is 1+2+2+4=9, the value of the 5th position is 1+2+2 +4+2=11, the value of the sixth position is 1+2+2+4+2+4=15,..., and the above calculation is performed to generate values of 1(1), 3(2), 5 (3), 9(4), 11(5), 15(6), ..., a communication strategy numerical sequence, wherein the number in parentheses indicates an index of the position of the value in the communication strategy numerical sequence (step S38) .

取出遞迴數列(0、1、1、2、3、5、8、13)中位置為第3個至第K+2個(本實施例之K為4)之數值作為一索引數列,在本實施例中,索引數列位置為第3個至第6個之數值為1、2、3、5。接著,取出通訊策略數值數列中K個數值,以產生一連線策略數值數列,其中K個數值係索引數列中之數值對應通訊策略數值數列之數列位置之數值,在本實施例中,連線策略數值數列之數值為1(1)、3(2)、5(3)、11(5)(步驟S40)。 The value of the third to K+2 (the K of the embodiment is 4) in the recursive sequence (0, 1, 1, 2, 3, 5, 8, 13) is taken as an index sequence. In this embodiment, the values of the index number column positions from the third to the sixth are 1, 2, 3, and 5. Then, the K values in the communication strategy numerical sequence are extracted to generate a connection strategy numerical sequence, wherein the values in the K numerical index series correspond to the values of the serial positions of the communication strategy numerical sequence. In this embodiment, the connection is performed. The values of the policy numerical sequence are 1 (1), 3 (2), 5 (3), and 11 (5) (step S40).

計算每一奇數與連線策略數值數列中之數值之和,以 產生複數個連線節點編號數列,以建立為每一奇數編號之節點連結相應之連線節點編號數列中之數值編號之節點之連線規則(步驟S42),如在本實施例中,奇數為1之節點平行連結編號為1+1=2、1+3=4、1+5=6、1+11=12等節點、奇數為3之節點平行連結編號為3+1=4、3+3=6、3+5=8、3+11=14等節點、...(如圖1所示),而將所建立之連線規則用以連接圖1之網路節點。 Calculate the sum of the values in each of the odd and connection strategy numeric columns to Generating a plurality of connection node number sequence columns to establish a connection rule for a node of each odd numbered node to link the numerical number in the corresponding connection node number sequence (step S42), as in the embodiment, the odd number is The node parallel connection number of 1 is 1+1=2, 1+3=4, 1+5=6, 1+11=12, etc., and the node with odd number is 3, the parallel connection number is 3+1=4, 3+ 3=6, 3+5=8, 3+11=14 nodes, etc. (as shown in Figure 1), and the established connection rules are used to connect the network nodes of Figure 1.

再次參考圖2,由網路拓樸系統於圖1之該等節點中之任一起始節點及任一目的節點間進行通訊傳輸(步驟S22)。 Referring again to FIG. 2, the network topology system performs communication transmission between any of the nodes and any of the destination nodes of FIG. 1 (step S22).

上述之網路拓樸系統與已知之網路拓樸系統的比較說明如下。在圖4A中,本發明之總節點數N為324及連接節點數K為8之網路拓樸系統與Fijitsu之TOFU-3x3x3 N324K7-10之網路拓樸系統比較發現,本發明之網路拓樸系統之平均距離A=3.41、最長距離D=5,而Fijitsu之網路拓樸系統之平均距離A=4.35、最長距離D=9,由於本發明之網路拓樸系統之平均距離A與最長距離D之數值皆小於Fijitsu之網路拓樸系統之平均距離A與最長距離D之數值,所以本發明之網路拓樸系統之通訊傳輸效能優於Fijitsu之網路拓樸系統之通訊傳輸效能。 A comparison of the above network topology system with a known network topology system is described below. In FIG. 4A, the network topology system of the present invention having a total node number N of 324 and a connection node number K of 8 is compared with the network topology system of Fijitsu's TOFU-3x3x3 N324K7-10, and the network of the present invention is found. The average distance of the topology system is A= 3.41, and the longest distance is D=5, while the average distance of Fijitsu's network topology system is A=4.35 and the longest distance D=9, due to the average distance A of the network topology system of the present invention. The value of the longest distance D is smaller than the average distance A and the longest distance D of the network topology system of Fijitsu, so the communication transmission performance of the network topology system of the present invention is superior to that of the network topology system of Fijitsu. Transmission efficiency.

同樣地,在圖4B中,本發明之總節點數N為1500及連接節點數K為9之網路拓樸系統之平均距離A=4.23、最長距離D=6,Fijitsu之TOFU-5x5x5 N1500K7-10之網路拓樸系統之平均距離A=6.47、最長距離D=15i在圖4C中,本發明之總節點數N為1024及連接節點數K為10之網路拓樸系統之平均距離A=3.63、最長距離D=5,IBM之BlueGene/Q N1024K10之網路拓樸系統之平均距離A=5.00、最長距離D=10;在圖4D中,本發明之總節點數N 為80及連接節點數K為4之網路拓樸系統之平均距離A=3.59、最長距離D=6,Motorola's之MC68xxx processor N80K2-4之網路拓樸系統之平均距離A=4.73、最長距離D=8。 Similarly, in FIG. 4B, the total node number N of the present invention is 1500 and the average distance of the network topology system with the number of connected nodes K is 9 is A=4.23, the longest distance D=6, and the TOFU-5x5x5 N1500K7 of Fijitsu- The average distance of the network topology system of 10 is A=6.47, and the longest distance D=15i. In FIG. 4C, the total number of nodes N of the present invention is 1024 and the average distance A of the network topology system with the number of connected nodes K is 10. =3.63, the longest distance D=5, the average distance of the network topology system of IBM's BlueGene/Q N1024K10 is A=5.00, the longest distance D=10; in Fig. 4D, the total number of nodes of the invention N The average distance of the network topology system with 80 and the number of nodes K is 4 is A=3.59, and the longest distance is D=6. The average distance of the network topology system of Motorola's MC68xxx processor N80K2-4 is A=4.73, the longest distance. D=8.

由圖4B至4D中發現本發明之網路拓樸系統之通訊傳輸效能係優於Fijitsu之TOFU-5x5x5 N1500K7-10之網路拓樸系統、IBM之BlueGene/Q N1024K10之網路拓樸系統及Motorola's之MC68xxx processor N80K2-4之網路拓樸系統等之通訊傳輸效能。 It is found from FIGS. 4B to 4D that the communication transmission performance of the network topology system of the present invention is superior to that of Fijitsu's TOFU-5x5x5 N1500K7-10 network topology system, IBM's BlueGene/Q N1024K10 network topology system and Motorola's MC68xxx processor N80K2-4 network topology system and other communication transmission performance.

本發明係提供一種網路拓樸系統及方法,其優點在於拓樸數學計算方式是在具有多個節點的系統中為各節點尋找一組節點連線規則以及各節點與其他節點連接的數目,使其可建構出一個讓所有節點根據其連線規則和其連接數所構成的網路拓樸,而無需使用交換器,使整個網路系統的結構變的簡單且無交換器之耗電,所以網路系統的成本不會增加,再者每個節點透過相同形式的通訊自由(使用同一組連線規則),並使得所有節點之連接數非常接近者,而達成所謂民主化維度。 The present invention provides a network topology system and method, which has the advantage that the topology mathematical calculation method is to find a set of node connection rules for each node in a system with multiple nodes and the number of connections between each node and other nodes. It makes it possible to construct a network topology that allows all nodes to be based on their connection rules and their connections without using a switch, making the structure of the entire network system simple and without the power consumption of the switch. Therefore, the cost of the network system will not increase, and each node will pass the same form of communication freedom (using the same set of connection rules), and the connection number of all nodes is very close to the so-called democratization dimension.

雖然本發明已參照較佳具體例及舉例性附圖敘述如上,惟其應不被視為係限制性者。熟悉本技藝者對其形態及具體例之內容做各種修改、省略及變化,均不離開本發明之申請專利範圍之所主張範圍。 The present invention has been described above with reference to the preferred embodiments and the accompanying drawings, and should not be considered as limiting. Various modifications, omissions and changes may be made without departing from the scope of the invention.

Claims (6)

一種網路拓樸系統,其包括:複數個節點,每一節點係一單核心晶片、一處理器、一電腦、一組電腦之內部最佳化組及一計算機中心之其中一者,根據一連線規則於該等節點中之任一起始節點及任一目的節點間進行連接;其中,該連線規則之建立包括下列步驟:計算一質數數列中相鄰質數之差值,以產生一質數數值差數列;計算該質數數值差數列中目前位置之數值與所有之前位置之數值之和,以產生一通訊策略數值數列;取出一遞迴數列中個數為一連接節點數之數值作為一索引數列,取出該通訊策略數值數列中個數為該連接節點數之數值,以產生一連線策略數值數列,其中該索引數列中之數值係為該通訊策略數值數列之數列位置,其數列位置之數值為該連線策略數值數列之數值;以及計算每一奇數與該連線策略數值數列中之數值之和,以產生複數個連線節點編號數列,建立為每一奇數編號之節點連結相應之連線節點編號數列中之數值編號之節點之該連線規則。 A network topology system comprising: a plurality of nodes, each node being a single core chip, a processor, a computer, an internal optimization group of a set of computers, and a computer center, according to one The connection rule is connected between any one of the starting nodes and any one of the destination nodes; wherein the establishing of the connection rule comprises the following steps: calculating a difference between adjacent prime numbers in a prime number column to generate a prime number a numerical difference series; calculating a sum of a value of a current position in the numerical difference series and a value of all previous positions to generate a communication strategy numerical sequence; taking a number in a recursive sequence as a number of connected nodes as an index a number column, the number of the number of the communication strategy is taken as the number of the number of connected nodes, to generate a connection strategy number sequence, wherein the value in the index sequence is the sequence position of the number sequence of the communication strategy, and the sequence position thereof The value is the value of the number of values in the connection strategy; and the sum of each odd number and the value in the number sequence of the connection strategy is calculated to generate a complex number Connection node ID number of columns, the connection rule to link the corresponding value of the connection node of the node number of the column number of the node number of each odd-numbered. 如申請專利範圍第1項之網路拓樸系統,其中,該連線規則之建立步驟進一步包括:輸入一總節點數及該連接節點數;列出數值小於該總節點數之該遞迴數列;以及列出數值為質數之該質數數列。 The network topology system of claim 1, wherein the establishing step of the connection rule further comprises: inputting a total number of nodes and the number of the connected nodes; and listing the recursive sequence whose value is less than the total number of nodes ; and list the prime number of the number as a prime number. 如申請專利範圍第1或2項之網路拓樸系統,其中,從該遞迴 數列中第3位置取出個數為該連接節點數之數值作為該索引數列。 For example, the network topology system of claim 1 or 2, wherein the recursive The third position in the sequence lists the number of the number of connected nodes as the index sequence. 一種網路拓樸方法,其應用於一網路拓樸系統,該網路拓樸系統包括複數個節點,每一節點係一單核心晶片、一處理器、一電腦、一組電腦之內部最佳化組及一計算機中心之其中一者,該方法包括下列步驟:由該網路拓樸系統根據一連線規則於該等節點中之任一起始節點及任一目的節點間進行連接;其中,建立該連線規則包括下列步驟:計算一質數數列中相鄰質數之差值,以產生一質數數值差數列;計算該質數數值差數列中目前位置之數值與所有之前位置之數值之和,以產生一通訊策略數值數列;取出一遞迴數列中個數為一連接節點數之數值作為一索引數列,取出該通訊策略數值數列中個數為該連接節點數之數值,以產生一連線策略數值數列,其中該索引數列中之數值係為該通訊策略數值數列之數列位置,其數列位置之數值為該連線策略數值數列之數值;以及計算每一奇數與該連線策略數值數列中之數值之和,以產生複數個連線節點編號數列,建立為每一奇數編號之節點連結相應之連線節點編號數列中之數值編號之節點之該連線規則。 A network topology method is applied to a network topology system. The network topology system includes a plurality of nodes, each node being a single core chip, a processor, a computer, and a group of computers. In one of the Jiahua group and a computer center, the method includes the following steps: connecting, by the network topology system, any one of the starting nodes and any one of the destination nodes according to a connection rule; The establishing the connection rule comprises the steps of: calculating a difference between adjacent prime numbers in a prime number series to generate a prime numerical difference series; calculating a sum of a value of a current position in the numerical difference series and a value of all previous positions, To generate a communication strategy numerical sequence; take the number of the number of connected nodes in a recursive sequence as an index sequence, and take out the number in the numerical sequence of the communication strategy as the number of the connected nodes to generate a connection. The value sequence of the strategy, wherein the value in the index sequence is the position of the number column of the communication strategy, and the value of the sequence position is the numerical sequence of the connection strategy a value; and calculating a sum of each odd number and a value in the number of values of the connection strategy to generate a plurality of connection node number series, establishing a numerical number in the number of connected node number numbers for each odd numbered node The connection rule of the node. 如申請專利範圍第4項之網路拓樸方法,其中,建立該連線規則之步驟進一步包括:輸入一總節點數及該連接節點數;列出數值小於該總節點數之該遞迴數列;以及 列出數值為質數之該質數數列。 The network topology method of claim 4, wherein the step of establishing the connection rule further comprises: inputting a total number of nodes and the number of the connection nodes; and listing the recursive sequence whose value is less than the total number of nodes ;as well as List the prime numbers whose values are prime. 如申請專利範圍第4或5項之網路拓樸方法,其中,從該遞迴數列中第3位置取出個數為該連接節點數之數值作為該索引數列。 The network topology method of claim 4, wherein the number of the number of connected nodes is taken as the index sequence from the third position in the recursive sequence.
TW103138885A 2014-11-10 2014-11-10 Network topology system and method TWI544766B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW103138885A TWI544766B (en) 2014-11-10 2014-11-10 Network topology system and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW103138885A TWI544766B (en) 2014-11-10 2014-11-10 Network topology system and method

Publications (2)

Publication Number Publication Date
TW201618504A TW201618504A (en) 2016-05-16
TWI544766B true TWI544766B (en) 2016-08-01

Family

ID=56509114

Family Applications (1)

Application Number Title Priority Date Filing Date
TW103138885A TWI544766B (en) 2014-11-10 2014-11-10 Network topology system and method

Country Status (1)

Country Link
TW (1) TWI544766B (en)

Also Published As

Publication number Publication date
TW201618504A (en) 2016-05-16

Similar Documents

Publication Publication Date Title
US20210117810A1 (en) On-chip code breakpoint debugging method, on-chip processor, and chip breakpoint debugging system
Wang et al. Membrane computing model for IIR filter design
US8694979B2 (en) Efficient egonet computation in a weighted directed graph
WO2016123808A1 (en) Data processing system, calculation node and data processing method
Rashidi et al. Themis: A network bandwidth-aware collective scheduling policy for distributed training of dl models
CN110034920B (en) Mapping method and device for coarse-grained reconfigurable cipher logic array
CN111478811A (en) Network key point analysis method based on double-layer information flow transmission
CN110766089A (en) Model structure sampling method and device of hyper network and electronic equipment
JP7063274B2 (en) Information processing equipment, neural network design method and program
JP6728227B2 (en) Device and/or method for providing query response based on ephemeral data
Bossard et al. The set-to-set disjoint-path problem in perfect hierarchical hypercubes
Fan et al. Reconfigurable Fault‐tolerance mapping of ternary N‐cubes onto chips
Liu et al. Dynamic threshold models of collective action in social networks
Park et al. On the power of gradual network alignment using dual-perception similarities
Rumley et al. Design methodology for optimizing optical interconnection networks in high performance systems
US10193757B2 (en) Network topology system and method
TWI544766B (en) Network topology system and method
Chang et al. Fault-tolerant bipancyclicity of faulty hypercubes under the generalized conditional-fault model
Kocik et al. Node-to-node disjoint paths problem in Möbius cubes
Li et al. A parallel algorithm for constructing two edge-disjoint hamiltonian cycles in locally twisted cubes
JP7400833B2 (en) Topology design device, topology design method, and program
CN107040394B (en) Network topological system and method
Habibi et al. High-performance 3D mesh-based NOC architecture using node-layer clustering
Chen et al. The fault-tolerant Hamiltonian problems of crossed cubes with path faults
Kocík et al. An algorithm for node-to-node disjoint paths problem in a Möbius cube