CN103957163A - Network topology structure based on fat tree high scalability hypercube - Google Patents
Network topology structure based on fat tree high scalability hypercube Download PDFInfo
- Publication number
- CN103957163A CN103957163A CN201410084035.8A CN201410084035A CN103957163A CN 103957163 A CN103957163 A CN 103957163A CN 201410084035 A CN201410084035 A CN 201410084035A CN 103957163 A CN103957163 A CN 103957163A
- Authority
- CN
- China
- Prior art keywords
- network
- network topology
- topology structure
- fat tree
- switch
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention is applied to the field of network communication, and provides a network topology structure based on a fat tree high scalability hypercube. The network topology structure comprises multiple switches and multiple servers. The switches employ a recursion unit hierarchical mode to form a topology network. A lowest recursion unit employs an m-port n- tree fat tree network structure, and the number of servers supported by the lower recursion unit is: g0=2*(m/2)n, wherein m is the number of ports of the switches in a fat tree, and n is the number of layers of the fat tree structure. In a network topology, the k-th port of the i-th switch at a highest layer is interconnected with the k-th server of the i-th subunit for realizing a multipath network topology structure, wherein i and k belong to a set of {1,2, ..., gk-1}. Through fusing a fat tree network topology in hierarchical recursion, the high scalability of the network topology is realized, at the same time, the characteristics of equal bandwidth and multipath performance of a fat tree network are also inherited, the bandwidth throughput is quite large, the fault tolerance is quite good, and the average time delay is quite small.
Description
Technical field
The invention belongs to network communication field, relate in particular to a kind of based on the fat tree-like high network topology structure of expanding hypercube.
Background technology
Along with further developing of mobile communication, the flow of mobile Internet rises on a large scale, when the large service traffics of reply, there is very serious link bottleneck and " hot spot phenomenon " in traditional data center network topological structure such as tree-like, and along with the expansion of business scale, autgmentability deficiency increases network construction cost greatly, therefore needs the strong high speed internet data central site network structure of a kind of autgmentability.
Data center network structure be computer with communicate by letter in popular research field, it is being connected a large amount of terminal nodes by a high speed wired or wireless link with switch, the topology of network configuration has provided the connected mode between the whole network switch and node, directly determining handling up of whole network, the performance such as fault-tolerant and congested.
At present, comprise that the various new data center network structures such as fat tree, BCube, DCell, VL2 are extensively proposed and dispose.The main designing requirement of new types of data central site network structure has:
1. to possess stronger expansibility, not only want to hold a large amount of servers, also want the sustainable growth of back-level server quantity.
2. to possess higher reliability, to tackle various server failures and link failure.
3. to possess good network performance, with the service of supporting that bandwidth demand is large.
(1) fat tree (Fat-Tree)
Fat tree is a kind of improvement structure of traditional tree structure, wherein any one intermediate node can have a plurality of father nodes, increase the number of links between levels aggregation switch and between aggregation switch and core switch, thereby increased the connectedness of network.But this structure cannot fundamentally solve expansion and the upgrade problem of large-scale data central site network structure, increase along with number of servers, need to change original topological structure, increase the level of aggregation switch, cause greatly having increased the construction cost of network.
(2)BCube
BCube is a kind of modularization solution that is beneficial to large scale network, by Microsoft Research Asia, in 2009, is proposed.BCube is also interconnected by using low side, cheap switch that server is carried out, and forms hypercube structure.BCube utilizes the node annexation of hypercube as recurrence rule, and the Servers-all in the different recurrence of same level unit in same position all interconnects by a switch.Each server of this structural requirement is equipped with a plurality of ports, has possessed the advantage that hypercube connectedness is high, diameter is little, reliability is high.
(3)DCell
DCell utilizes the node annexation of complete graph as recurrence rule, between any two recurrence unit of same level, there is a pair of server to be directly connected, connectivity between its recurrence unit has improved the reliability of network greatly, but each server has multilink, Routing Protocol is comparatively complicated, simultaneously because DCell expands by bottom server completely, be not suitable for the service large to bandwidth demand amount, the performance of server is had to higher requirement, and the construction cost of network is higher.
(4)VL2
VL2 structure is the improvement based on traditional tree structure, has realized the dynamic assignment of Service Source.This structure utilizes virtual machine technique to turn to unified territory by the set layer of traditional tree structure is virtual, make all servers be connected to seemingly same local area network (LAN), and according to the IP address of the demand dynamic assignment server of server, network performance and efficiency of service have effectively been improved.VL2 itself does not change the node connected mode of traditional tree structure, and being conducive to, for the transformation to legacy data division center, has good effect to traffic assignments and flow control etc., but the reliability of network does not have greatly improved.
(5) other structures
Under the driving of the factors such as cost control and resource utilization, the data center network that industrial quarters also accords with the demands of the market in research always, if company of Boke is in data center's framework (DCF) of issue in 2007, by integrating storage networking and server, build server cluster, become data center's infrastructure of single fusion, simplified the connection of data center and reduced cost.The researcher of Google has designed a kind of energy proportion structure, node connected mode according to flattening butterfly topology connects each server, to guarantee that the power consumption of data center network and the utilance of server match, thereby reduce to greatest extent the operating cost of data center.
Development along with data center, although traditional tree data central site network structure builds fairly simple convenience, but be not easy to expand and upgrading, and core switch breaks down and may cause thousands of station servers to lose efficacy, therefore the performance needs such as the autgmentability of network topology, reliability are had to higher requirement.
An intermediate node in fat tree structure can have a plurality of father nodes, has increased the number of links between levels aggregation switch and between aggregation switch and core switch, has increased the connectedness of network.But fat tree can't fundamentally solve, data center network is expanded and the problem of upgrading, when switch quantity is increased to a certain degree, need to change the structure of legacy data central site network, increase the level quantity of aggregation switch, be unfavorable for the structure of large-scale data central site network.
DCell adopts the node annexation of complete graph as recurrence rule, between any two recurrence unit of same level, there is a pair of server to be directly connected, but its each terminal node is connected with multilink, therefore each node needs to have a plurality of ports, expansion along with network size, node communication path between different subelements is longer, and communication delay increases.Meanwhile, this structure is comparatively complicated, and Routing Protocol is difficult for design.
BCube utilizes the node annexation of hypercube as recurrence rule, server in the different recurrence of same level unit in same position all interconnects by a switch, it is good that this structure has possessed hypercube connectedness, the advantage that diameter is little, its Path diversity does not have fat tree topology abundant, is difficult to accomplish the unification of upstream and downstream bandwidth.
VL2 is by the upgrading of original fat tree network is transformed to network, by amateurish demand dynamic assignment resource, has improved the utilance of network, but cannot change the poor fact of the autgmentability of network own, and does not improve the reliability of network.
Summary of the invention
The invention provides a kind of network topology structure based on fat tree-like high expansion hypercube, be intended to solve the problem that existing network topological structure expansion is poor, reliability is low, network performance is low.
The present invention realizes like this, a kind of based on the fat tree-like high network topology structure of expanding hypercube, described network topology structure comprises a plurality of switches and a plurality of server, described switch adopts recurrence unit level mode to form topological network, wherein minimum recurrence unit adopts the fat tree network configuration of m-port n-tree, and the number of servers of minimum recurrence unit support is: g
0=2 * (m/2)
n, wherein m is the port number of switch in fat tree, the level that n is fat tree structure, in network topology, k port of top i switch and k server of i subelement are interconnected, realize multi-path network topological structure, i wherein, k ∈ { 1,2...g
k-1.
Further technical scheme of the present invention is:, it is characterized in that: in described network topology structure, when carrying out high-rise recurrence, still kept decile bandwidth and multipath performance.
Further technical scheme of the present invention is: the decline along with the increase bandwidth throughput of link failure ratio in described network topology structure comparatively relaxes.
Further technical scheme of the present invention is: the described switch of the recurrence unit of each level in described network topology structure adopts unified model.
Further technical scheme of the present invention is: the switch high-level port number in the unit of recurrence described in network is greater than low-level port number.
Further technical scheme of the present invention is: being denoted as of switch described in network
SW < I, C=C
n-2c
n-3... C
1c
0>, wherein 1 is switch place level number, { 0,1...n-1}, the span of C is I ∈
Further technical scheme of the present invention is: at server described in networking, be labeled as P (P
n-1p
n-2... P
1p
0), P ∈ { 0,1...m-1} * { 0,1... (m/2)-1} wherein
n-1.
Further technical scheme of the present invention is: in network topology structure, route adopts divide-and-conquer strategy, successively obtains subpath and merge to draw routed path.
The invention has the beneficial effects as follows: by incorporate fat tree network topology in level recurrence, realized the high expansion of network topology, also inherited the decile bandwidth at fat tree networking and the feature of multipath performance, bandwidth throughput is large, fault-tolerance is better simultaneously, and average delay is less.
Accompanying drawing explanation
Fig. 1 be the embodiment of the present invention provide k-1 time hypercube fat tree network topology structure;
Fig. 2 is the equivalent hypercube mesh that the embodiment of the present invention provides;
Fig. 3 is the CLOS structural network that the embodiment of the present invention provides;
Fig. 4 be BFT and DCell throughput ratio;
Fig. 5 is the communication delay comparison of BFT and DCell;
Fig. 6 is the point-to-points performance comparison figure mono-of BFT network;
Fig. 7 is the point-to-points performance comparison figure bis-of BFT network.
Embodiment
Fig. 1-3 show network topology structure described in the network topology structure based on fat tree-like high expansion hypercube provided by the invention and comprise a plurality of switches and a plurality of server, described switch adopts recurrence unit level mode to form topological network, wherein minimum recurrence unit adopts the fat tree network configuration of m-port n-tree, and the number of servers of minimum recurrence unit support is: g
0=2 * (m/2)
n, wherein m is the port number of switch in fat tree, the level that n is fat tree structure, in network topology, k port of top i switch and k server of i subelement are interconnected, realize multi-path network topological structure, i wherein, k ∈ { 1,2...g
k-1.By incorporate fat tree network topology in level recurrence, realized the high expansion of network topology, also inherited the decile bandwidth at fat tree networking and the feature of multipath performance simultaneously, bandwidth throughput is large, fault-tolerance is better, and average delay is less.The composition of the fat tree network of m-port n-tree comprises 2 * (m/2)
nindividual server and (n-1) * (m/2)
n-1individual m port switch forms, server be labeled as P (P
n-1p
n-2... P
1p
0), wherein
P ∈ { 0,1...m-1} * { 0,1... (m/2)-1}
n-1, switch identification is SW < I, C=C
n-2c
n-3... C
1c
0>, wherein 1 is switch place level number, { 0,1...n-1}, the span of C is I ∈
BFT (k) (BFT: be the fat tree network configuration of hypercube) the fat tree network of hypercube of a k level, the number of servers of its support meets g
k=g
k-1 2, server identification is < q, p >, and switch identification is < q, I, C=C
n-2c
n-3... C
1c
0>, the span of q is
Switch identification is < q, I, C=C
n-2c
n-3... C
1c
0>.
Decile bandwidth and multipath performance in described network topology structure, when carrying out high-rise recurrence, have still been kept.
In BFT (k), if each BFT (k-1) is equivalent to a node, BFT (k) can be reduced to model as shown in Figure 3, as seen from Figure 3, if total m Higher layer switch machine, between each node, have m communication link, visible whole communication network is not only fat tree structure at bottom, has still kept the performance characteristics such as the decile bandwidth of fat tree network and multipath when carrying out high-rise recurrence.
In tree network, there is serious hot spot phenomenon in the communication between node, and the communications burden of Higher layer switch machine is serious, understand the while with serious communication delay, and BFT preferably resolves this problem when load is larger.Between the inner node of BFT (0), due to the architectural characteristic of fat tree, communication has extraordinary performance mutually.When different subelements communicates; the node recurrence that makes progress is found nearest public ancestors; by this public branch exchange, connect different subelements; for example this public branch exchange is positioned at top; so lower than the switch of this level still in idle condition, be conducive to interconnecting on a large scale of whole network.
Decline along with the increase bandwidth throughput of link failure ratio in described network topology structure comparatively relaxes.Due to the Path diversity of BFT, its bandwidth throughput is larger, and fault-tolerance is also better, and along with the increase of link failure ratio, the decline of bandwidth throughput also can comparatively relax.In the structure of Fig. 3, suppose total k Higher layer switch machine, to there being k equivalent unit node, between any two equivalent nodes, have k communication link, suppose that a certain node sends or receive data to other k-1 nodes, all can be undertaken by non-interfering link, reach maximum network throughput, realize the optimal performance of fat tree network.And the fat tree structure of hypercube do not exist link bottleneck, as long as have one can communicate by letter in the k bar link in theory between arbitrary node, two subelements just can keep connection.Time delay represents from transmitting terminal sends data, to transmitting terminal, receive the confirmation from receiving terminal, altogether the time difference of experience, delay performance is directly proportional to the shortest path between 2 o'clock, the average shortest path of BFT is less than the DCell structure of same size, and average delay is also less.
The described switch of the recurrence unit of each level in described network topology structure adopts unified model.
Switch high-level port number in the unit of recurrence described in network is greater than low-level port number.
At being denoted as SW<I of switch described in network, C=C
n-2c
n-3... C
1c
0>, wherein 1 is switch place level number, { 0,1...n-1}, the span of C is I ∈
(the P for P that is labeled as at server described in networking
n-1p
n-2... P
1p
0), P ∈ { 0,1...m-1} * { 0,1... (m/2)-1} wherein
n-1.
The general routing algorithm adopting in the fat tree topology network configuration of hypercube is: for the network that adopts recursive structure design, general routing algorithm can adopt divide-and-conquer strategy, successively obtains subpath and merge to show that routed path src is source node src=[q
n-1... q
0, p
n-1... p
0], dst is receiving node dst=[u
n-1... u
0, v
n-1... v
0], the marker method of src and dst is all deferred to the labeling method of BFT to server node, q wherein,
p, v ∈ { 0,1...m-1} * { 0,1... (m/2)-1}
n-1, the false code of divide-and-conquer strategy is:
First contrast the prefix of src and dst, if prefix is different, illustrate that src and dst are in different BFT (k-1), can pass through top routing to communicate, if prefix is the same, illustrate that src and dst, in identical BFT (k-1), continue downward recurrence.
For tree-shaped topological structure, that often adopt is nearest public ancestors (least common ancestor, LCA) algorithm, when communicating by letter with destination node, first in father's switching node, search source node, if searching in oneself child node, father node there is no destination node, will inquire about to upper strata, while knowing the common ancestor's node that finds source node and destination node, then search destination node downwards.
In the fat tree network of hypercube, under no-failure link condition, first judge whether source processing node and destination node belong to the coverage of source node father node, if belonged to, be forwarded to destination node, otherwise be forwarded to any one in upper layer node, then judge whether to belong to the coverage of current parent's node, until be forwarded to the father node that covers destination node, then be forwarded to destination node downwards.In the process of route that makes progress, can there is mulitpath to select, thereby have a plurality of public ancestors.
At network, exist in certain mistake (path failure, node damages) situation, can adopt recurrence strategy to realize fault-tolerant route.In a BFT (k), suppose that 2 nodes are in different sub-BFT (k-1), all the time exist 2 communication paths the shortest (not comprising that two nodes, directly by the connected situation of high-rise router, now only have the shortest path that a paths length is 1).Path one: from source node to the high-rise router being directly connected with it, this high level router, to the sub-BFT (k-1) with destination node place, supposes that the via node that the sub-BFT of high-rise router and this (k-1) is directly connected is n.Continue judge that n and source node are whether in different BFT (k-2), the if so, continuation repetition first step, until in via node and the destination node subelement in the bottom, be also that n and destination node are in the fat tree of same height.Path two: find shortest path from reverse link, determine that destination node arrives the high-rise router being directly connected with it, this high level router calculates forward and the via node m of source node at same BFT (k-1), and residue routing mechanism is identical with path one.
The BFT structure that is 1 by the DCell structure that is 2 to Degree and Degree is below carried out Performance Ratio.
Simulated environment: communication simulation software NS2 and Opnet.
Emulation content:
1, for the ease of the performance of contrast DCell network and BFT network, in NS2, set up respectively two kinds of networks that comprise 20 servers, choose four random nodes and to a certain node of different subelements, send data simultaneously, difference test throughput, delay performance.
The link bandwidth of setting between switch is 1Gbps, bandwidth between switch-to-server is 100Mbps, as seen from Figure 6, the link paths limit transmission speed of two kinds of networks is all less than 100Mb/s, be limited to the bandwidth between switch-to-server, along with the carrying out of emulation, the point-to-point communication performance of two kinds of structures can reach the limiting value of bandwidth, green curve is BFT topology, and the time of the value of reaching capacity will be less than DCell network slightly.
2, the delay performance of test point point to-point communication.
Grey Point is BFT network, and redness is DCell network, and under equal conditions, the conventional time delay of BFT network is less than 0.05s, and the conventional time delay of DCell network concentrates on the time period of 0.05-0.10s, and the maximum delay of DCell has reached 0.3s, and the performance of BFT is slightly good.
Utilize Opnet software to set up the complete BFT network of supporting 64 servers, minimum recursive structure is the fat tree structure of 4-port2-tree, choose at random a server and send data to 16 servers, and check the performance situation of BFT network under certain link failure ratio (5%).
In Fig. 6,7, black curve is that a node of choosing at random sends load and the time delay situation of data to a plurality of (16) node, Grey curves is load and the time delay situation under 5% link failure ratio, as can be seen from the figure, under the link failure rate of existence 5%, the load of network is not affected, the robustness that has fully reflected BFT network, transfer of data can rationally be avoided wrong path, and from delay performance, path failure can cause the time delay of network to increase a little, but the amplitude increasing is not obvious, has still kept good delay performance
The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, all any modifications of doing within the spirit and principles in the present invention, be equal to and replace and improvement etc., within all should being included in protection scope of the present invention.
Claims (8)
1. the network topology structure based on fat tree-like high expansion hypercube, it is characterized in that: described network topology structure comprises a plurality of switches and a plurality of server, described switch adopts recurrence unit level mode to form topological network, wherein minimum recurrence unit adopts the fat tree network configuration of m-port n-tree, and the number of servers of minimum recurrence unit support is: g
0=2 * (m/2)
n, wherein m is the port number of switch in fat tree, the level that n is fat tree structure, in network topology, k port of top i switch and k server of i subelement are interconnected, realize multi-path network topological structure, i wherein, k ∈ { 1,2...g
k-1.
2. network topology structure according to claim 1, is characterized in that: in described network topology structure, when carrying out high-rise recurrence, still kept decile bandwidth and multipath performance.
3. network topology structure according to claim 2, is characterized in that: the decline along with the increase bandwidth throughput of link failure ratio in described network topology structure comparatively relaxes.
4. according to the network topology structure described in claim 1-3 any one, it is characterized in that: the described switch of the recurrence unit of each level in described network topology structure adopts unified model.
5. network topology structure according to claim 4, is characterized in that: the switch high-level port number in the unit of recurrence described in network is greater than low-level port number.
6. network topology structure according to claim 4, is characterized in that: at switch described in network, be denoted as SW < I, C=C
n-2c
n-3... C
1c
0>, wherein 1 is switch place level number, { 0,1...n-1}, the span of C is I ∈
7. network topology structure according to claim 4, is characterized in that: at server described in networking, be labeled as P (P
n-1p
n-2... P
1p
0), P ∈ { 0,1...m-1} * { 0,1... (m/2)-1} wherein
n-1.
8. network topology structure according to claim 4, is characterized in that: in network topology structure, route adopts divide-and-conquer strategy, successively obtains subpath and merge to draw routed path.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410084035.8A CN103957163A (en) | 2014-03-07 | 2014-03-07 | Network topology structure based on fat tree high scalability hypercube |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410084035.8A CN103957163A (en) | 2014-03-07 | 2014-03-07 | Network topology structure based on fat tree high scalability hypercube |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103957163A true CN103957163A (en) | 2014-07-30 |
Family
ID=51334390
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410084035.8A Pending CN103957163A (en) | 2014-03-07 | 2014-03-07 | Network topology structure based on fat tree high scalability hypercube |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103957163A (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104717111A (en) * | 2015-03-17 | 2015-06-17 | 中国科学技术大学苏州研究院 | Internet structure of extensible exchange cube |
| CN107623644A (en) * | 2017-09-30 | 2018-01-23 | 新华三技术有限公司 | A kind of bandwidth speed limiting method, apparatus and equipment |
| CN107819884A (en) * | 2017-12-16 | 2018-03-20 | 长沙志鸿文化传媒有限公司 | A kind of network topology structure based on Internet of Things |
| CN108243103A (en) * | 2016-12-23 | 2018-07-03 | 瞻博网络公司 | The devices, systems, and methods of routing protocol information are distributed in this fabric network of Crow |
| CN108768864A (en) * | 2018-07-16 | 2018-11-06 | 南京邮电大学 | A kind of high fault-tolerant data center network topology structure of easy expansion |
| CN109039672A (en) * | 2018-07-20 | 2018-12-18 | 苏州大学张家港工业技术研究院 | The information broadcast method and system of hypercube line chart structured data central site network |
| CN109617800A (en) * | 2019-01-18 | 2019-04-12 | 福建师范大学 | A kind of data center network fault tolerant secure method for routing based on balance hypercube |
| CN110535791A (en) * | 2019-06-25 | 2019-12-03 | 南京邮电大学 | A kind of data center network based on prism structure |
| CN110719193A (en) * | 2019-09-12 | 2020-01-21 | 无锡江南计算技术研究所 | High-performance computing-oriented high-reliability universal tree network topology method and structure |
| CN111245655A (en) * | 2020-01-10 | 2020-06-05 | 长春理工大学 | Intra-network topology structure, construction method and system |
| CN111262913A (en) * | 2020-01-10 | 2020-06-09 | 长春理工大学 | Data center network topology structure, determination method and system |
| CN111582453A (en) * | 2020-05-09 | 2020-08-25 | 北京百度网讯科技有限公司 | Method and device for generating neural network model |
| CN112165405A (en) * | 2020-10-13 | 2021-01-01 | 中国人民解放军国防科技大学 | Method for testing big data processing capacity of supercomputer based on network topological structure |
| CN113595896A (en) * | 2021-08-05 | 2021-11-02 | 河南大学 | Multi-route communication method of mixed data center network based on VLC link |
| CN113708963A (en) * | 2021-08-20 | 2021-11-26 | 北京大学深圳研究生院 | Optimized topology recursion method and system meeting alliance chain partition tolerance |
| CN114584868A (en) * | 2022-02-12 | 2022-06-03 | 国网宁夏电力有限公司电力科学研究院 | Data center photoelectric hybrid switching architecture upgrading method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100020806A1 (en) * | 2008-07-22 | 2010-01-28 | Amin Vahdat | Scalable Commodity Data Center Network Architecture |
| US20100180048A1 (en) * | 2009-01-09 | 2010-07-15 | Microsoft Corporation | Server-Centric High Performance Network Architecture for Modular Data Centers |
| CN101945050A (en) * | 2010-09-25 | 2011-01-12 | 中国科学院计算技术研究所 | Dynamic fault tolerance method and system based on fat tree structure |
| CN102130810A (en) * | 2011-01-27 | 2011-07-20 | 电子科技大学 | Method for realizing interconnection structure in same layer domain of tree topology |
-
2014
- 2014-03-07 CN CN201410084035.8A patent/CN103957163A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100020806A1 (en) * | 2008-07-22 | 2010-01-28 | Amin Vahdat | Scalable Commodity Data Center Network Architecture |
| US20100180048A1 (en) * | 2009-01-09 | 2010-07-15 | Microsoft Corporation | Server-Centric High Performance Network Architecture for Modular Data Centers |
| CN101945050A (en) * | 2010-09-25 | 2011-01-12 | 中国科学院计算技术研究所 | Dynamic fault tolerance method and system based on fat tree structure |
| CN102130810A (en) * | 2011-01-27 | 2011-07-20 | 电子科技大学 | Method for realizing interconnection structure in same layer domain of tree topology |
Non-Patent Citations (2)
| Title |
|---|
| CHUANXIONG GUO等: "BCube:A High Performance,Server-Centric Network Architecture for Modular Data Centers", 《ACM SIGCOMM COMPUTER COMMUNICATION REVIEW-SIGMOCC’09》 * |
| MOHAMMAD AI-FARES等: "A Scalabel,Commodity Data Center Network Architecture", 《ACM SIGCOMM COMPUTER COMMUNICATION REVIEW》 * |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104717111B (en) * | 2015-03-17 | 2018-07-06 | 中国科学技术大学苏州研究院 | A kind of extension exchanges cubical internet system |
| CN104717111A (en) * | 2015-03-17 | 2015-06-17 | 中国科学技术大学苏州研究院 | Internet structure of extensible exchange cube |
| CN108243103B (en) * | 2016-12-23 | 2020-12-04 | 瞻博网络公司 | Apparatus, system and method for distributing routing protocol information in a clos network |
| CN108243103A (en) * | 2016-12-23 | 2018-07-03 | 瞻博网络公司 | The devices, systems, and methods of routing protocol information are distributed in this fabric network of Crow |
| CN107623644A (en) * | 2017-09-30 | 2018-01-23 | 新华三技术有限公司 | A kind of bandwidth speed limiting method, apparatus and equipment |
| CN107819884A (en) * | 2017-12-16 | 2018-03-20 | 长沙志鸿文化传媒有限公司 | A kind of network topology structure based on Internet of Things |
| CN108768864A (en) * | 2018-07-16 | 2018-11-06 | 南京邮电大学 | A kind of high fault-tolerant data center network topology structure of easy expansion |
| CN108768864B (en) * | 2018-07-16 | 2021-05-28 | 南京邮电大学 | Data center network topology system easy to expand and high in fault tolerance |
| CN109039672A (en) * | 2018-07-20 | 2018-12-18 | 苏州大学张家港工业技术研究院 | The information broadcast method and system of hypercube line chart structured data central site network |
| CN109039672B (en) * | 2018-07-20 | 2021-04-20 | 苏州大学张家港工业技术研究院 | Information broadcasting method and system for hypercube line graph structure data center network |
| CN109617800A (en) * | 2019-01-18 | 2019-04-12 | 福建师范大学 | A kind of data center network fault tolerant secure method for routing based on balance hypercube |
| CN109617800B (en) * | 2019-01-18 | 2021-06-01 | 福建师范大学 | Data center network fault-tolerant safe routing method based on balance hypercube |
| CN110535791B (en) * | 2019-06-25 | 2022-03-08 | 南京邮电大学 | Data center network based on prism structure |
| CN110535791A (en) * | 2019-06-25 | 2019-12-03 | 南京邮电大学 | A kind of data center network based on prism structure |
| CN110719193B (en) * | 2019-09-12 | 2021-02-02 | 无锡江南计算技术研究所 | High-performance computing-oriented high-reliability universal tree network topology method and structure |
| CN110719193A (en) * | 2019-09-12 | 2020-01-21 | 无锡江南计算技术研究所 | High-performance computing-oriented high-reliability universal tree network topology method and structure |
| CN111262913A (en) * | 2020-01-10 | 2020-06-09 | 长春理工大学 | Data center network topology structure, determination method and system |
| CN111245655A (en) * | 2020-01-10 | 2020-06-05 | 长春理工大学 | Intra-network topology structure, construction method and system |
| CN111262913B (en) * | 2020-01-10 | 2022-08-19 | 长春理工大学 | Data center network topology structure, determination method and system |
| CN111582453B (en) * | 2020-05-09 | 2023-10-27 | 北京百度网讯科技有限公司 | Method and device for generating neural network model |
| CN111582453A (en) * | 2020-05-09 | 2020-08-25 | 北京百度网讯科技有限公司 | Method and device for generating neural network model |
| CN112165405B (en) * | 2020-10-13 | 2022-04-22 | 中国人民解放军国防科技大学 | Method for testing big data processing capacity of supercomputer based on network topological structure |
| CN112165405A (en) * | 2020-10-13 | 2021-01-01 | 中国人民解放军国防科技大学 | Method for testing big data processing capacity of supercomputer based on network topological structure |
| CN113595896A (en) * | 2021-08-05 | 2021-11-02 | 河南大学 | Multi-route communication method of mixed data center network based on VLC link |
| CN113708963A (en) * | 2021-08-20 | 2021-11-26 | 北京大学深圳研究生院 | Optimized topology recursion method and system meeting alliance chain partition tolerance |
| CN113708963B (en) * | 2021-08-20 | 2022-08-16 | 北京大学深圳研究生院 | Optimized topology recursion method and system meeting alliance chain partition tolerance |
| CN114584868A (en) * | 2022-02-12 | 2022-06-03 | 国网宁夏电力有限公司电力科学研究院 | Data center photoelectric hybrid switching architecture upgrading method |
| CN114584868B (en) * | 2022-02-12 | 2023-07-18 | 国网宁夏电力有限公司电力科学研究院 | Data center photoelectric hybrid architecture upgrading method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103957163A (en) | Network topology structure based on fat tree high scalability hypercube | |
| US10673741B2 (en) | Control device discovery in networks having separate control and forwarding devices | |
| CN104158747B (en) | Method for discovering network topology and system | |
| CN102150148B (en) | Differentiated services for unicast multicast frames in layer 2 topologies | |
| CN102037685B (en) | IP forwarding across a link state protocol controlled Ethernet network | |
| CN103428031B (en) | A kind of inter-domain link fast failure recovery method based on software defined network | |
| KR101208230B1 (en) | Node device, executing method for node device and computer-readable storage medium having program | |
| JP6117911B2 (en) | Optimization of 3-stage folded CLOS for 802.1AQ | |
| CN103561445B (en) | A kind of wireless sense network multi-path routing method based on network code | |
| CN103281247B (en) | The general method for routing of a kind of data center network and system | |
| CN102368727B (en) | Crossed IP network TRILL network communication method, system thereof and devices | |
| CN102510404B (en) | Nondestructive continuous extensible interconnection structure for data center | |
| CN103179046A (en) | Data center flow control method and data center flow control system based on openflow | |
| JP2015512584A (en) | Packet flow interconnect fabric | |
| CN104378297A (en) | Message forwarding method and device | |
| CN102546351A (en) | System and method for interconnecting openflow network and conventional Internet protocol (IP) network | |
| CN102185772B (en) | Method for routing data centre network system | |
| CN104243323A (en) | Exchange network multicast route method and system | |
| CN105763439B (en) | The interconnected method and device of data center in a kind of multilink transparent internet | |
| CN101815040A (en) | Method for forwarding packet in hybrid switch network and switch | |
| CN102098729B (en) | Construction method and related packet forwarding method for tree backbone structure in wireless network | |
| CN103095579B (en) | TRILL network interconnected method, Apparatus and system | |
| CN104469874A (en) | Message forwarding method of opportunistic network based on probability centrality | |
| Zhu et al. | Efficient hybrid multicast approach in wireless data center network | |
| CN110290069A (en) | A kind of reliable multicast method for routing under SDN scene |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20140730 |