TWI672018B - System and method of proactive multi-path routing with a predictive mechanism - Google Patents
System and method of proactive multi-path routing with a predictive mechanism Download PDFInfo
- Publication number
- TWI672018B TWI672018B TW107126494A TW107126494A TWI672018B TW I672018 B TWI672018 B TW I672018B TW 107126494 A TW107126494 A TW 107126494A TW 107126494 A TW107126494 A TW 107126494A TW I672018 B TWI672018 B TW I672018B
- Authority
- TW
- Taiwan
- Prior art keywords
- path
- traffic
- subnets
- routing
- multipath
- Prior art date
Links
Abstract
本發明之具預測機制之主動式多路徑路由系統及方法，包括：多個子網路；以及透過多個交換器連接於該多個子網路的控制器，該控制器在該多個子網路的出口端交換器監測該多個子網路至其他子網路之流量並包括流量統計及預測模組、路徑計算及選擇模組及多路徑路由模組，該流量統計及預測模組透過該出口端交換器監測該流量，以使用指數平滑法預測下一時間點的流量；該路徑計算及選擇模組計算該多個子網路至其他子網路的可能多路徑，以根據該可能多個路徑的相依度選擇多路徑候選；該多路徑路由模組依據該路徑計算及選擇模組所選擇之該多路徑候選，藉由重複組合產生該多路徑候選經過權重的組合，根據該流量統計及預測模組預測的該流量，模擬該多路徑候選經過權重的組合的加權延遲，以選擇最佳路由配置並設定於該多個交換器的實體層。 The active multi-path routing system and method with a prediction mechanism of the present invention includes: multiple subnets; and a controller connected to the multiple subnets through multiple switches. The controllers are located in the multiple subnets. The egress switch monitors the traffic from the multiple subnets to other subnets and includes a traffic statistics and prediction module, a path calculation and selection module, and a multipath routing module. The traffic statistics and prediction module passes the egress The switch monitors the traffic to predict the traffic at the next time point using the exponential smoothing method; the path calculation and selection module calculates the possible multi-path from the multiple subnets to other sub-networks, based on the Dependency selects multipath candidates; the multipath routing module calculates and selects the multipath candidates based on the path, and generates a weighted combination of the multipath candidates by repeating the combination, according to the traffic statistics and prediction model The predicted traffic of the group simulates the weighted delay of the multipath candidate through a combination of weights to select the optimal routing configuration and set the actual routing of the multiple switches. Floor.
Description
本發明係關於一種應用於網路環境之多子網路多路徑路由安排，特別是一種具預測機制之主動式多路徑路由系統及方法。 The invention relates to a multi-subnet multi-path routing arrangement applied to a network environment, in particular to an active multi-path routing system and method with a prediction mechanism.
傳統網路的路由協定，因其使用單一最短路徑之路由演算法，且根據每30秒啟動一次的鏈結狀態通告來改變路徑權重，造成路由協定無法有效解決網路壅塞之問題，在傳統網路下，若要改善此現象需改動到標準及網路環境的所有路由器，成本過高實難達成。SDN是近年來的新興網路架構，它的概念是將網路的控制層(control-plane)功能與資料層(data-plane)功能分離，並透過OpenFlow協定來設定或取得交換器之資訊，相對於傳統網路架構較為彈性，可輕易解決上述之缺點。 The routing protocol of the traditional network, because it uses a single shortest path routing algorithm, and changes the weight of the path according to the link status announcement started every 30 seconds, resulting in the routing protocol cannot effectively solve the problem of network congestion. Down the road, if you want to improve this phenomenon, you need to change all routers to the standard and network environment, which is too expensive to achieve. SDN is an emerging network architecture in recent years. Its concept is to separate the control-plane function and data-plane function of the network, and set or obtain the information of the switch through the OpenFlow protocol. Compared with the traditional network architecture, it is more flexible and can easily solve the above disadvantages.
本發明針對多SDN網路設計多子網路多路徑路由安排方法。先前應用於普通SDN網路環境之路由方法大都以等價多路徑路由(Equal-cost multi-path routing)來分配多路徑路由之流量，如：U.S.Pat.No.9,369,387「Segment routing based wide area network orchestration in a network environment」專利。另一方面如：U.S.2015/0327135「Apparatus and method for dynamic hybrid routing in sdn networks to avoid congestion and balance loads under changing traffic load」、U.S.Pat.No.9,608,858「Reliable multipath forwarding for encapsulation protocols」及U.S.2014/0192645「Method for Internet Traffic Management Using a Central Traffic Controller」等專利雖皆聲明路由安排會依據流量大小來做多路徑之分流，但係使用目前所監測之流量大小來做分流，一方面當網路拓樸增大，控制器實難處理及監測流量數據，另外亦無法克服流量大小變化所造成之錯誤的路由安排。 The invention designs a multi-subnet and multi-path routing arrangement method for a multi-SDN network. Most of the routing methods previously applied to the ordinary SDN network environment use Equal-cost multi-path routing to distribute the traffic of multi-path routing, such as: U.S.Pat.No.9,369,387 "Segment routing based wide area network orchestration in a network environment "patent. On the other hand: US2015 / 0327135 "Apparatus and method for dynamic hybrid routing in sdn networks to avoid congestion and balance loads under changing traffic load", USPat. No. 9,608,858 "Reliable multipath forwarding for encapsulation protocols" and US2014 / 0192645 "Method for Internet Traffic Management Using a Central Traffic Controller" and other patents all state that routing arrangements will perform multi-path offloading based on the size of the traffic, but use the currently monitored traffic size for offloading. As the size increases, the controller can hardly process and monitor the traffic data. In addition, it cannot overcome the incorrect routing arrangement caused by the change in traffic size.
因傳統網路使用單一最短路徑來安排路由繞徑，及藉由每30秒的鏈路狀態通告(Link-state advertisement)來改變路徑權重導致路由協定缺乏全域意識，因而路由協定無法有效解決網路壅塞問題。 Because traditional networks use a single shortest path to arrange routing detours, and change the weight of a path through Link-state advertisements every 30 seconds, the routing protocol lacks global awareness, so routing protocols cannot effectively solve the network Congestion.
由此可見，上述習用方式仍有諸多缺失，實非一良善之設計，而亟待加以改良。 It can be seen that there are still many shortcomings in the above-mentioned customary methods. It is not a good design, and it needs to be improved.
本案發明人鑑於上述習用方式所衍生的各項缺點，乃亟思加以改良創新，並經多年苦心孤詣潛心研究後，終於成功研發完成本件應用於SDN網路環境之多子網路多路徑路由安排方法。 In view of the various shortcomings derived from the above-mentioned conventional methods, the inventor of this case was eager to improve and innovate. After years of painstaking research, he finally successfully developed a multi-subnet multi-path routing method for SDN network environment. .
為解決上述之問題，本發明提出流量預測方法及主動 式多路徑路由之目的即基於軟體定義網路之路由方法，解決網路壅塞問題，使網路服務供應商有效提升路由安排的效能。使用流量統計資料來預測各子網路間之流量變化來預測下一時間點之流量大小並使用排隊理論，依據預測之流量大小來模擬各種路徑權重安排的網路延遲，並藉由基因演算法來加速最佳解的搜尋速度，達到路由安排最佳化。 In order to solve the above problems, the present invention proposes a traffic prediction method and an initiative The purpose of the multi-path routing is to solve the problem of network congestion based on the routing method of software-defined networks, so that network service providers can effectively improve the efficiency of routing arrangements. Use traffic statistics to predict changes in traffic between subnets to predict the size of traffic at the next point in time and use queuing theory to simulate network delays for various path weighting arrangements based on the predicted traffic size and use genetic algorithms To speed up the search for the best solution and optimize routing arrangements.
本發明係提供一種具預測機制之主動式多路徑路由系統，包括：多個子網路；以及控制器，其透過多個交換器連接於該多個子網路，並在該多個子網路的出口端交換器監測該多個子網路至其他子網路之流量，該控制器包括：流量統計及預測模組，其透過該出口端交換器監測該流量，以使用指數平滑法預測下一時間點的流量；路徑計算及選擇模組，其計算該多個子網路至其他子網路的可能多路徑，以根據該可能多個路徑的相依度選擇多路徑候選；以及多路徑路由模組，其依據該路徑計算及選擇模組所選擇之該多路徑候選，藉由重複組合產生該多路徑候選經過權重的組合，並根據該流量統計及預測模組預測的該流量，模擬該多路徑候選經過權重的組合的加權延遲，以選擇最佳路由配置並設定於該多個交換器的實體層。 The invention provides an active multi-path routing system with a prediction mechanism, including: a plurality of subnets; and a controller connected to the plurality of subnets through a plurality of switches and exiting the plurality of subnets. The end switch monitors the traffic from the multiple subnets to other subnets. The controller includes: a traffic statistics and prediction module that monitors the traffic through the egress switch to predict the next point in time using exponential smoothing. Path calculation and selection module that calculates possible multipaths from the multiple subnets to other subnets to select multipath candidates based on the degree of dependency of the possible multiple paths; and a multipath routing module that According to the path calculation and selection module, the multi-path candidate is selected, a combination of weights of the multi-path candidate is generated through repeated combination, and the multi-path candidate is simulated according to the traffic predicted by the traffic statistics and prediction module. A combination of weighted delays to select the optimal routing configuration and set it at the physical layer of the multiple switches.
在前述之具預測機制之主動式多路徑路由系統中，該路徑計算及選擇模組透過廣度優先搜尋法計算該多個子網路至其他子網路的該可能多路徑。 In the aforementioned active multi-path routing system with a prediction mechanism, the path calculation and selection module calculates the possible multi-path from the multiple sub-networks to other sub-networks using a breadth-first search method.
在前述之具預測機制之主動式多路徑路由系統中，該 路徑計算及選擇模組根據該可能多個路徑的相依度中，選擇相對低相依度的多路徑候選。 In the aforementioned active multi-path routing system with a prediction mechanism, the The path calculation and selection module selects a multipath candidate with a relatively low degree of dependence from among the degrees of dependence of the possible multiple paths.
在前述之具預測機制之主動式多路徑路由系統中，該多路徑路由模組使用排隊模型模擬該多路徑候選經過權重的組合的加權延遲。 In the aforementioned active multi-path routing system with a prediction mechanism, the multi-path routing module uses a queuing model to simulate the weighted delay of the multi-path candidate through a combination of weights.
在前述之具預測機制之主動式多路徑路由系統中，該多路徑路由模組使用基因演算法選擇最佳路由配置。 In the aforementioned active multi-path routing system with a prediction mechanism, the multi-path routing module uses a genetic algorithm to select the optimal routing configuration.
本發明另提供一種具預測機制之主動式多路徑路由方法，包括：連接多個交換器於多個子網路；監測該多個子網路至其他子網路之流量；透過該多個子網路的出口端交換器監測該流量，以使用指數平滑法預測下一時間點的流量；計算該多個子網路至其他子網路的可能多路徑，以根據該可能多個路徑的相依度選擇多路徑候選；依據所選擇之該多路徑候選，產生該多路徑候選經過權重的組合；依據所預測的該流量，模擬該多路徑候選經過權重的組合的加權延遲；選擇最佳路由配置；以及設定該最佳路由配置於該多個交換器的實體層。 The invention also provides an active multi-path routing method with a prediction mechanism, which includes: connecting multiple switches to multiple subnets; monitoring traffic from the multiple subnets to other subnets; The egress switch monitors the traffic to predict the traffic at the next time point using the exponential smoothing method; calculates the possible multipath from the multiple subnets to other subnets to select the multipath based on the degree of dependency of the possible multiple paths Candidate; generating a weighted combination of the multipath candidate according to the selected multipath candidate; simulating the weighted delay of the combination of the multipath candidate through the weight according to the predicted traffic; selecting an optimal routing configuration; and setting the The optimal route is configured at the physical layer of the plurality of switches.
在前述之具預測機制之主動式多路徑路由方法中，該計算該多個子網路至其他子網路的可能多路徑，是透過廣度優先搜尋法計算該多個子網路至其他子網路的該可能多路徑。 In the aforementioned active multi-path routing method with a prediction mechanism, the calculation of possible multi-paths of the multiple sub-networks to other sub-networks is performed by a breadth-first search method to calculate the multiple sub-networks to other sub-networks. This may be multipathing.
在前述之具預測機制之主動式多路徑路由方法中，該選擇多路徑候選，是根據該可能多個路徑的相依度中，選擇相對低相依度的多路徑候選。 In the aforementioned active multi-path routing method with a prediction mechanism, the selection of a multi-path candidate is to select a multi-path candidate with a relatively low degree of dependency based on the degree of dependency of the possible multiple paths.
在前述之具預測機制之主動式多路徑路由方法中，該 模擬該多路徑候選經過權重的組合的加權延遲，是使用排隊模型模擬該多路徑候選經過權重的組合的加權延遲。 In the aforementioned active multi-path routing method with a prediction mechanism, the The weighted delay that simulates the weighted combination of the multipath candidate is a weighted delay that simulates the weighted combination of the multipath candidate using a queuing model.
在前述之具預測機制之主動式多路徑路由方法中，該多路徑路由模組使用基因演算法選擇最佳路由配置。 In the aforementioned active multi-path routing method with a prediction mechanism, the multi-path routing module uses a genetic algorithm to select the optimal routing configuration.
為達成上述發明目的，本發明根據各個子網路間之流量統計資料及使用指數平滑法來預測下一時間點之流量大小，並使用排隊模型及流量預測資料來模擬各種路徑權重安排的網路延遲，並藉由基因演算法來加速最佳解的搜尋速度，可有效降低總體的網路延遲，提升網路效能，在各個子網路之出口端交換器監測該子網路到所有其他子網路之流量大小，再使用指數平滑法預測下一時間點之流量大小，先計算所有子網路到所有其他子網路之所有路徑，並使用重複組合(Combination with repetition)來產生所有路徑權重值的組合，之後以先前預測之流量數據及基因演算法來搜尋最佳路徑權重值之安排。 In order to achieve the above-mentioned object of the invention, the present invention uses the exponential smoothing method to predict the traffic volume at the next point in time based on the traffic statistics between each subnet and uses a queuing model and traffic prediction data to simulate a network with various path weights Delay, and use genetic algorithms to speed up the search for the best solution, which can effectively reduce the overall network delay, improve network performance, and monitor the subnet to all other subnets at the egress switch of each subnet The amount of network traffic, then use the exponential smoothing method to predict the amount of traffic at the next point in time. First calculate all paths from all subnets to all other subnets, and use Combination with repetition to generate all path weights. The combination of values is then used to search for the best path weights based on previously predicted traffic data and genetic algorithms.
10‧‧‧控制器 10‧‧‧ Controller
101‧‧‧流量統計及預測模組 101‧‧‧Traffic Statistics and Forecasting Module
102‧‧‧路徑計算及選擇模組 102‧‧‧ Path calculation and selection module
103‧‧‧多路徑路由模組 103‧‧‧Multi-path routing module
201‧‧‧子網路A 201‧‧‧Subnet A
202‧‧‧子網路B 202‧‧‧Subnet B
30‧‧‧出口端交換器 30‧‧‧Export-side switch
31‧‧‧交換器 31‧‧‧exchanger
1011‧‧‧步驟 1011‧‧‧step
1012‧‧‧步驟 1012‧‧‧step
10121‧‧‧步驟 10121‧‧‧step
1013‧‧‧步驟 1013‧‧‧step
1014‧‧‧步驟 1014‧‧‧step
1015‧‧‧步驟 1015‧‧‧step
10151‧‧‧步驟 10151‧‧‧step
1016‧‧‧步驟 1016‧‧‧step
10161‧‧‧步驟 10161‧‧‧step
10162‧‧‧步驟 10162‧‧‧step
1017‧‧‧步驟 1017‧‧‧step
第1圖為本發明具預測機制之主動式多路徑路由系統之系統架構圖。 FIG. 1 is a system architecture diagram of an active multi-path routing system with a prediction mechanism according to the present invention.
第2圖為本發明具預測機制之主動式多路徑路由方法之運作流程圖。 FIG. 2 is a flowchart of the operation of the active multi-path routing method with a prediction mechanism of the present invention.
第3圖為本發明具預測機制之主動式多路徑路由方法流量統計及預測模組之動作流程圖。 FIG. 3 is an operation flowchart of the traffic statistics and prediction module of the active multi-path routing method with a prediction mechanism of the present invention.
第4圖為本發明具預測機制之主動式多路徑路由方法 路徑計算及選擇模組之動作流程圖。 FIG. 4 is an active multi-path routing method with a prediction mechanism according to the present invention Flow chart of path calculation and selection module.
第5圖為本發明具預測機制之主動式多路徑路由方法之多路徑路由模組動作流程圖。 FIG. 5 is a flowchart of the operation of the multi-path routing module of the active multi-path routing method with a prediction mechanism of the present invention.
在網路環境下管理多個子網路間之多路徑路由安排，所謂具預測機制之主動式多路徑路由(Proactive Multi-path Routing with a Predictive mechanism，PMRP)主要想達到高效能壅塞處理及具flow table使用量意識的路由演算法。PMRP主要有兩點概念，(1)預測性：PMRP使用指數平滑法(Exponential smoothing)來預測各子網路間的流量大小以克服流量變化；(2)主動性：PMRP定時使用排隊模型及流量預測資料來模擬各種路徑權重安排的網路延遲，並藉由基因演算法(Genetic algorithm)來加速最佳解的搜尋速度。 Manage multi-path routing arrangements between multiple sub-networks in a network environment. The so-called Proactive Multi-path Routing with a Predictive mechanism (PMRP) mainly aims to achieve high-efficiency congestion processing and flow control. Table uses volume-aware routing algorithms. PMRP mainly has two concepts: (1) predictive: PMRP uses exponential smoothing to predict the size of traffic between subnets to overcome traffic changes; (2) initiative: PMRP uses a queuing model and traffic regularly The prediction data is used to simulate the network delay of various path weights, and the genetic algorithm is used to accelerate the search speed of the best solution.
本發明為一種具預測機制之主動式多路徑路由系統及方法，其在網路環境中，提供多子網路之多路徑路由功能，利用各子網路之出口端流量監測與指數平滑法來預測下時間點之流量大小，計算所有子網路到所有其他子網路之所有路徑，並利用重複組合產生所有權重值，再以基因演算法及排隊理論模擬各種權重值安排，搜尋最佳路徑權重值之安排。 The present invention is an active multi-path routing system and method with a prediction mechanism. In a network environment, it provides a multi-path routing function for multiple sub-networks. Predict the amount of traffic at the time point, calculate all the paths from all subnets to all other subnets, and use the repeated combination to generate the weight value, and then use genetic algorithms and queuing theory to simulate various weight value arrangements to search for the best path Arrangement of weight values.
請參閱第1圖，第1圖為本發明具預測機制之主動式多路徑路由系統之系統架構圖，其組成包括控制器10、流量統計及預測模組101、路徑計算及選擇模組102、多路徑 路由模組103、範例子網路A201、範例子網路B202、出口端交換器30。 Please refer to FIG. 1. FIG. 1 is a system architecture diagram of an active multi-path routing system with a prediction mechanism according to the present invention, which includes a controller 10, a traffic statistics and prediction module 101, a path calculation and selection module 102, Multipath The routing module 103, the example subnet A201, the example subnet B202, and the egress switch 30.
請一併參閱第2圖，第2圖為本發明具預測機制之主動式多路徑路由方法之運作流程圖。步驟1014，控制器10向實體層提出拓樸資料要求；步驟1015，待實體層回報給路徑計算及選擇模組102來運算；步驟1016，路徑計算及選擇模組102將所有子網路到所有其他子網路候選路徑之計算結果提供予多路徑路由模組103進行路由安排模擬；步驟1011，由控制器10向所有子網路之出口端交換器30要求其流量之監測資料；步驟1012，等待交換器回報數據；步驟1013，由流量統計及預測模組101進行預測後提供多路徑路由模組103運算；步驟1017，最後多路徑路由模組103綜合步驟1013及1016之結果進行路由安排模擬，最後將計算之路由安排設定於實體層。 Please refer to FIG. 2 together, which is a flowchart of the operation of the active multi-path routing method with a prediction mechanism of the present invention. In step 1014, the controller 10 submits a topology data request to the physical layer; in step 1015, the physical layer reports to the path calculation and selection module 102 for calculation; in step 1016, the path calculation and selection module 102 sends all subnets to all The calculation results of other subnet candidate paths are provided to the multipath routing module 103 for routing arrangement simulation; step 1011, the controller 10 requests monitoring data of its traffic to the egress switches 30 of all subnets; step 1012, Wait for the switch to report the data; Step 1013, the traffic statistics and prediction module 101 performs the prediction to provide the multi-path routing module 103 operation; Step 1017, the final multi-path routing module 103 combines the results of steps 1013 and 1016 to simulate the routing arrangement , And finally set the calculated routing arrangement at the physical layer.
流量統計及預測模組101會依據流量大小及變化來安排路由，所以會監測各子網路到所有其他子網路的流量統計資料，在各個子網路之出口端交換器30監測其流量，將各流量數據記錄下來，並使用指數平滑法做下一時間點之流量預測。 The traffic statistics and prediction module 101 arranges routes according to the size and change of traffic, so it will monitor the traffic statistics of each subnet to all other subnets, and monitor its traffic at the egress switch 30 of each subnet. Record each flow of data, and use the exponential smoothing method to make the flow prediction at the next time point.
本發明流量統計及預測模組101係藉由控制器10來取得各子網路之出***換器的流量統計資料；將收集的流量統計資料使用指數平滑法來預測各子網路間下一時間之流量大小，並交由多路徑路由模組103來作流量分配，請參閱第3圖。 The traffic statistics and prediction module 101 of the present invention uses the controller 10 to obtain the traffic statistics of the egress switches of each subnet; the collected traffic statistics are used to predict the next time between the subnets by using the exponential smoothing method. The amount of traffic is transferred to the multi-path routing module 103 for traffic distribution, see FIG. 3.
如第3圖所示，步驟1011，由控制器10對各子網路出口端交換器30要求流量監測之數據。步驟1012，出口端交換器30回報流量統計資料；步驟10121，控制器10以指數平滑法模型對流量進行下一時間點流量大小預測；步驟1013，提供流量預測結果，以備多路徑路由模組103模擬運算。 As shown in FIG. 3, in step 1011, the controller 10 requests the data of the traffic monitoring of the sub-network egress switch 30. In step 1012, the egress switch 30 reports traffic statistics; in step 10121, the controller 10 performs an exponential smoothing model on the traffic at the next time point to predict the traffic size; in step 1013, the traffic prediction result is provided for the multi-path routing module. 103 analog operations.
路徑計算及選擇模組102，計算所有子網路到其他所有子網路之多路徑候選。首先使用廣度優先搜尋法(Breadth-first Search)計算所有子網路到所有其他子網路的所有路徑，但若網路拓樸稍大的情況下，過多的路徑參與路徑安排會造成時間複雜度提升，所有多路徑候選再交由多路徑路由模組103安排路由前，會先計算各路徑之相依度並挑選相依度較低之路徑。 The path calculation and selection module 102 calculates multipath candidates from all subnets to all other subnets. First use Breadth-first Search to calculate all paths from all subnets to all other subnets, but if the network topology is slightly larger, too many paths participating in the path arrangement will cause time complexity Upgrading, before all multipath candidates are handed over to the multipath routing module 103 to arrange routing, the degree of dependency of each path is calculated and the path with a lower degree of dependency is selected.
路徑計算及選擇模組102係藉由控制器10來取得實體資源之網路拓樸，並使用廣度優先搜尋法來計算各子網路到所有其他子網路之所有可能路徑；將各子網路到所有其他子網路之所有可能路徑，計算各路徑間的相依性，並從中挑選出相對低相依性之多路徑，提供多路徑路由模組103作流量分配的多路徑候選，請參閱第4圖。 The path calculation and selection module 102 obtains the network topology of the physical resources by the controller 10 and uses the breadth-first search method to calculate all possible paths from each subnet to all other subnets; All possible paths to all other subnets, calculate the dependencies between each path, and select the multipath with relatively low dependency from it, and provide the multipath routing module 103 as a multipath candidate for traffic distribution. 4 Figure.
如第4圖所示，步驟1014，控制器10向實體層提出拓樸資料要求；步驟1015，並待實體層回報給路徑計算及選擇模組102來運算，由控制器10取得整體網路拓樸；步驟10151，以廣度優先搜尋法計算各子網路到所有其他子網路的所有路徑；步驟1016，計算各路徑間相依度，並挑 選相依度較低者，交由多路徑路由模組103模擬。 As shown in FIG. 4, in step 1014, the controller 10 makes a request for topology information to the physical layer; in step 1015, the physical layer returns to the path calculation and selection module 102 for calculation, and the controller 10 obtains the overall network topology. Step 10151: Calculate all paths from each subnet to all other subnets using the breadth-first search method; Step 1016, calculate the degree of dependency between each path, and select The one with the lower dependency is selected and simulated by the multi-path routing module 103.
多路徑路由模組103，依照路徑計算及選擇模組102所挑選之多路徑候選，產生所有路徑權重值之所有組合，並依照流量統計及預測模組101之預測流量，以排隊理論來模擬總體網路延遲，因無法以暴力法(Brute force algorithm)來計算所有組合，因此使用基因演算法來加速搜尋最佳解。 The multi-path routing module 103 generates all combinations of all path weight values according to the multi-path candidates selected by the path calculation and selection module 102, and uses the queuing theory to simulate the overall population in accordance with the predicted traffic of the traffic statistics and prediction module 101. Network latency, because it is not possible to calculate all combinations using the brute force algorithm, so genetic algorithms are used to speed up the search for the best solution.
本發明係藉由路徑計算及選擇模組102所挑選出的多路徑候選以重複組合來產生所有子網路到另一其他子網路之路徑權重的所有可能組合，使用排隊模型及流量統計及預測模組101之預測的流量大小，來模擬各子網路到所有其他子網路之路徑權重的整體網路加權延遲，並使用基因演算法來加快挑選出最佳解，再設定到實體資源，請參閱第5圖。 The present invention uses the multipath candidates selected by the path calculation and selection module 102 to repeatedly combine to generate all possible combinations of path weights from all subnets to another subnet, using a queuing model and traffic statistics, and The predicted traffic size of the prediction module 101 is used to simulate the overall network weighted delay of the path weight of each subnet to all other subnets, and the genetic algorithm is used to accelerate the selection of the best solution and then set it to the physical resources See Figure 5.
如第5圖所示，步驟10161，將路徑計算及選擇模組102所挑選之候選路徑，以重複組合來產生所有子網路到其他子網路之所有路徑權重值的所有組合；步驟10162，以基因演算法來模擬各種組合並以排隊理論來驗證，搜尋最佳路由安排；步驟1017，將計算出之最佳之路由安排設定於實體資源。 As shown in FIG. 5, step 10161, the candidate paths selected by the path calculation and selection module 102 are repeatedly combined to generate all combinations of all path weight values of all subnets to other subnets; step 10162, A genetic algorithm is used to simulate various combinations and verified by queuing theory to search for the best routing arrangement. Step 1017, the calculated optimal routing arrangement is set to the physical resource.
其中，步驟10161中，以重複組合產生所有子網路到另一其他子網路之路徑權重的所有可能組合，為了減少過多組合數造成流量分配之計算的時間複雜度急遽提升，係基於可用之多路徑候選以某一百分率來在多路徑間分配一 子網路到另一其他子網路的流量。 Among them, in step 10161, all possible combinations of the path weights of all subnets to another other subnet are generated by repeated combinations. In order to reduce the excessive number of combinations, the time complexity of the traffic allocation calculation is rapidly increased. Multipath candidates assign a certain percentage among multiple paths Traffic from one subnet to another.
其中，步驟10162中，以基因演算法來挑選最佳解，係先產生出多組所有子網路到另一其他子網路之路徑權重的組合，以排隊模型及流量統計及預測模組之預測的流量大小來模擬上述之組合的整體網路加權延遲，並依照基因演算法的選擇、交配、突變之邏輯，逐步搜尋出最佳解。 Among them, in step 10162, a genetic algorithm is used to select the best solution. First, a combination of path weights of all sets of all subnets to another other subnet is generated, and a queuing model and traffic statistics and prediction module are used. The predicted traffic size simulates the overall network weighted delay of the above combination, and gradually searches for the best solution according to the logic of the selection, mating, and mutation of the genetic algorithm.
先前應用於軟體定義網路路由安排之技術，大多以目前監測網路流量之數據來做路由安排，並沒有考慮流量變化之因素，本發明能依據流量變化對下一時間點流量大小進行預測，並做動態多路徑之更新，與其他習用技術相互比較時，更具備下列優點：本發明利用指數平滑法做流量大小預測，可更貼近實體流量變化；且本發明利用基因演算法及排隊理論來模擬總體網路延遲，可提升網路效能及加快路由計算時間。 The technologies previously applied to software-defined network routing arrangements mostly use current monitoring network traffic data to make routing arrangements, and do not consider the factors of traffic changes. The present invention can predict the size of traffic at the next point in time based on traffic changes. It also updates the dynamic multi-path. Compared with other conventional technologies, it also has the following advantages: The present invention uses the exponential smoothing method to make traffic size prediction, which can be closer to the change of physical traffic; and the present invention uses genetic algorithms and queuing theory to Simulate the overall network latency, which can improve network performance and speed up route calculation time.
上列詳細說明乃針對本發明之一可行實施例進行具體說明，惟該實施例並非用以限制本發明之專利範圍，凡未脫離本發明技藝精神所為之等效實施或變更，均應包含於本案之專利範圍中。 The above detailed description is a specific description of a feasible embodiment of the present invention, but this embodiment is not intended to limit the patent scope of the present invention. Any equivalent implementation or change that does not depart from the technical spirit of the present invention should be included in Within the scope of the patent in this case.
綜上所述，本案不僅於技術思想上確屬創新，並具備習用之傳統方法所不及之上述多項功效，已充分符合新穎性及進步性之法定發明專利要件，爰依法提出申請，懇請貴局核准本件發明專利申請案，以勵發明，至感德便。 To sum up, this case is not only innovative in terms of technical ideas, but also has many of the above-mentioned effects that are not used by traditional methods. It has fully met the requirements of statutory invention patents that are novel and progressive. To approve this invention patent application, to encourage invention, to the utmost convenience.
Claims (8)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW107126494A TWI672018B (en) | 2018-07-31 | 2018-07-31 | System and method of proactive multi-path routing with a predictive mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW107126494A TWI672018B (en) | 2018-07-31 | 2018-07-31 | System and method of proactive multi-path routing with a predictive mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| TWI672018B true TWI672018B (en) | 2019-09-11 |
| TW202008751A TW202008751A (en) | 2020-02-16 |
Family
ID=68618992
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| TW107126494A TWI672018B (en) | 2018-07-31 | 2018-07-31 | System and method of proactive multi-path routing with a predictive mechanism |
Country Status (1)
| Country | Link |
|---|---|
| TW (1) | TWI672018B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2002017110A1 (en) * | 2000-08-22 | 2002-02-28 | Internap Network Services Corporation | Method and system for optimizing routing through multiple available internet route providers |
| CN103414586A (en) * | 2013-08-05 | 2013-11-27 | 浙江工商大学 | Novel network view generating and updating method |
| WO2015168611A1 (en) * | 2014-05-01 | 2015-11-05 | Netflow Logic Corporation | Method and system for confident anomaly detection in computer network traffic |
-
2018
- 2018-07-31 TW TW107126494A patent/TWI672018B/en active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2002017110A1 (en) * | 2000-08-22 | 2002-02-28 | Internap Network Services Corporation | Method and system for optimizing routing through multiple available internet route providers |
| CN103414586A (en) * | 2013-08-05 | 2013-11-27 | 浙江工商大学 | Novel network view generating and updating method |
| WO2015168611A1 (en) * | 2014-05-01 | 2015-11-05 | Netflow Logic Corporation | Method and system for confident anomaly detection in computer network traffic |
Also Published As
| Publication number | Publication date |
|---|---|
| TW202008751A (en) | 2020-02-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20240129238A1 (en) | Flow-Based Load Balancing | |
| Xiong et al. | Performance evaluation of OpenFlow-based software-defined networks based on queueing model | |
| Li et al. | OpenFlow based load balancing for fat-tree networks with multipath support | |
| Li et al. | An effective path load balancing mechanism based on SDN | |
| Shafiee et al. | A simple congestion-aware algorithm for load balancing in datacenter networks | |
| CN105471764B (en) | A kind of method of end-to-end QoS guarantee in SDN network | |
| CN112600759B (en) | Multipath traffic scheduling method and system based on deep reinforcement learning under Overlay network | |
| CN102112981A (en) | Distribution of virtual machines in a communication network | |
| CN110891019B (en) | Data center flow scheduling method based on load balancing | |
| Tajiki et al. | CECT: computationally efficient congestion-avoidance and traffic engineering in software-defined cloud data centers | |
| Otokura et al. | Application of evolutionary mechanism to dynamic virtual network function placement | |
| Liu | Intelligent routing based on deep reinforcement learning in software-defined data-center networks | |
| CN108028805A (en) | A kind of system and method for control flow equalization in band in software defined network | |
| CN106161270B (en) | A kind of network deployment method | |
| Babbar et al. | A genetic load balancing algorithm to improve the QoS metrics for software defined networking for multimedia applications | |
| Rikhtegar et al. | DeepRLB: A deep reinforcement learning‐based load balancing in data center networks | |
| CN116233133A (en) | Method for balancing network load of digital twin auxiliary data center | |
| TWI672018B (en) | System and method of proactive multi-path routing with a predictive mechanism | |
| Cheng et al. | Application-aware routing scheme for sdn-based cloud datacenters | |
| Chekired et al. | A hybrid SDN path computation for scaling data centers networks | |
| Farhoudi et al. | Server load balancing in software-defined networks | |
| CN105306367A (en) | A method of communicating information over a network and a system for communicating information over a network | |
| Yuan et al. | Workload-aware request routing in cloud data center using software-defined networking | |
| Lokare et al. | QoS based Routing using the Fuzzy TOPSIS MCDM method to enhance the performance of the MANET | |
| Pang et al. | Horizon: a QoS management framework for SDN-based data center networks |