JP2019122040A - Network source reuse and routing mechanism defining multi-source by software - Google Patents

Abstract

To disclose network source reuse and a routing mechanism which defines a multi-source by software.SOLUTION: A composite network source model which can be quantified or superimposed is constructed, contribution to a decrease in time delay by the source is made to be a conversion metric of a source of a different type, influences on the time delay by cache and link are quantified respectively, quantification with respect to the cache and link is reused and are integrally expressed, and a composite source of a link between two nodes, a composite source of a route, and a composite source of a region are included. The composite source is a metric, a "dummy node" with the composite source is added on a network, the routing protocol is invited to discover a change of a network topology, and a new routing discovery step is performed.SELECTED DRAWING: Figure 1

Description

  The present invention relates to the field of cloud computing data center network technology, and more particularly to network source reuse and routing mechanism for defining multi-source by software.

Data Center Network (DCN) is an important part of cloud computing infrastructure. Studies have shown that some cache placement of network nodes can reduce the average length of data accessing paths and also avoid data hot, increasing network throughput rates. After the cache has become a kind of new network source, the DCN network environment is characterized by the coexistence of links and multiple network sources such as cache. To this end, the present invention presents and deploys a model of multi-source reuse in a kind of data center network. In the present invention, by discovering joint characteristics that network sources (links, caches) affect transmission performance parameters (time delay), they can be quantified, and a superimposable composite network source model (abbreviated as composite source) A performance comparison metric with deep matching between routing / transmission and network sources; and a source allocation metric that makes the granules of the model tunable and increases multi-source utilization.

In the present invention, all nodes stored and transferred are network coded blocks CM (Coded Message), and the idea of multi-source transfer is that a user can generate multiple data from multiple data sources (original content server or intermediate nodes). By collecting a fixed number of linearly independent CMs via a path, decryption can be realized and data transfer jobs can be completed. Among them, the link and the cache respectively affect the transmission time delay and the propagation time delay in the data transmission job, and the degree of both effects can be quantified.

Chinese Patent Application Publication No. 103685347

  The purpose of the present invention is to overcome the shortcomings and deficiencies in the current technology, to provide network source reuse, to make contribution to reduction of time delay by source into different types of source conversion metric, ie cache and so on. We quantify and express the influence on time delay by link, and then reuse quantification for cache and link.

Another object of the present invention is to provide a routing mechanism that defines multiple sources by software.

  In order to achieve the above objective, the present invention adopts the following technical plan:

Network Source Reuse includes S11-S13:

In S11, assuming that the user wants to acquire the content o, the user is located at the node s, the upstream is the node i, and the contribution to the reduction of the time delay by one link is the physical distance and the bandwidth Expressed as a function of width, further quantifying the mapping between link and time delay further on this;

The S12 expresses the degree of contribution to the reduction of time delay by one node cache as a function, and further quantifies and expresses the mapping between cache and time delay;

After obtaining the quantitative values of the link, cache and time delay mapping based on the S11 and the S12, the S13 obtains them as a composite source of one link between two nodes and a plurality of links. A unified representation is made on three types of composite sources, including a composite source of one path including, and a composite source of one area including multiple paths.

In terms of the prior art plant and quantifying the mapping between the link and the time delay in S11, it is specifically expressed as follows:

である。 In which B is the bandwidth of the link from node s to its upstream node i, d is the distance from node s to its upstream node i, and α ₁ and α ₂ are the bandwidth and distance respectively Weight parameters of the

It is.

Specifically, in the above-described S12, quantifying the mapping between cache and time delay as a priority technology plant will be expressed as follows:

Among them, M _i (o) is the quantity of CM storing the content o at node i, E (o) is the quantity of CM necessary for decoding the content o, and said CM (Coded Message) is a node Is a network coding block that stores and forwards.

As a priority technology plant, in S13, basic configuration units of three types of networks such as one link and one route and one area are all provided with metrics of composite source, and depending on the range of the network, type Superimpose the size of the composite source for each and adjust the granules; the specific steps are as follows:

The composite source of one link from node s to its upstream node i, ie the composite source of node i, is specifically expressed as follows:

In which α and β represent the weight parameters occupied by the link and the mapping of cache and time delay respectively;

If path j contains k links, then the composite source of said path is a superposition of the composite sources of k links, which is specifically expressed as follows:

When performing multi-source transmission through w paths in the area area, the composite source of the area is a superposition of w path composite sources, and specifically expressed as follows:

The routing mechanism that defines multi-source by software specifically includes S21 to S24:

In S21, in the data center based on the SDN architecture, an SDN control plane having a global network view detects composite source status and user needs, and analyzes the traffic distribution to discover idle routes and optimize traffic. Determine the goals; said SDN is a Software Defined Network and a Software Defined Network;

The step S22 determines a traffic optimization target based on the step S21, and the SDN control plane establishes an optimum plan for deploying the "dummy node" according to the optimization theory, and includes the deployment quantity and the deployment location; "Is not actually present in the network, but is a node that is falsely added to the logic;" Dummy node "deployment method broadcasts a status report of the routing protocol to the network, and the report through the" dummy node " Includes information reaching other nodes, which completes the process of falsely adding "dummy nodes";

S23 restarts the routing discovery process after the dynamic routing protocol discovers a change in network topology;

At S24, the data report is transmitted by the newly discovered routing.

As a preferential technical plan, the specific process of S22 includes S221 to S223:

に転換し、

である； The above-mentioned S221 "changes" the network topology by finding load balance as a traffic optimization target and finding a proposal to add "dummy nodes" with w composite sources in appropriate places in the network topology, for example, "dummy nodes". 'I's composite source is Cost (o) _i and the composite source Cost (o) _i is the cost metric required by the dynamic routing protocol

Switch to

Is;

The step S222 converts the specific process of the "dummy node" proposal described in the step S221 into a multi-target optimization problem as follows:

と定義し、θ_uでuつ目のリンクのトラヒック負荷を表し、θ_vでvつ目のリンクのトラヒック負荷を表し、

がリンクトラヒックの平均値である； Therein, one of the network G = assumed (V, E, C) a; E is linked collection is _{E = {e 1 ... e m} }, representing the m nd link e _m; C is a newly added “dummy node” collection, C = {c ₁ ... C _w }, c _{w represents} a w newly added “dummy node”; V is for all nodes A collection, V = {v ₁ ... V _n }, and v _n represents the n-th node; Load _G represents the Gini coefficient of link traffic throughout the network,

Define the traffic load of the u-th link by θ _u and the traffic load of the v-th link by θ _v ,

Is the average value of link traffic;

The multi-target optimization problem is constrained as follows:

であり、各リンクのトラヒック負荷が第一閾値θより小さく、クオリティオブサービスQoSを実現する；
[2] .

であり、各経路における各ノードのコストの総和が第二閾値Dより小さく、経路選択のクオリティを保証する； [1].

And the traffic load of each link is smaller than the first threshold θ to realize quality of service QoS;
[2].

And the sum of the costs of each node in each path is smaller than a second threshold D to guarantee the quality of path selection;

がノードv_iとノードv_jの間のリンクにおける各ノードのコストの総和である；ノードv_iとノードv_jとの間の経路にはzつのノードが含まれていると仮定すると、

である； Among them, v _i represents the i-th node, v _j represents the _j-th node,

Assuming the path between the node v _i and node v _j that contains z single node; but the node v _i is the sum of the cost of each node in the link between the node v _j

Is;

The step S223 solves the multi-target optimization problem using the optimization theory again to obtain the number w of "dummy nodes" satisfying the constraints and the arrangement position thereof; the optimization theory uses integer linear programming algorithm adopt.

As a priority technology plant, the specific process of S23 is as follows:

をコストメトリックとし、ルーティング発見過程を再起動し、すなわちDijkstraアルゴリズムで元-宛先の間の最短経路を改めて計算し、それに該経路における各ノードのコスト総和を最小にさせる。 When the network topology changes due to the addition of the “dummy node”, the OSPF protocol

Let C. be a cost metric, restart the routing discovery process, ie recalculate the shortest path between source and destination in the Dijkstra algorithm, and minimize the cost sum of each node in the path.

As a priority technology plant, the traffic optimization goals include any one or any combination of load balancing, failure path avoidance and minimal energy consumption.

As a priority technology plant, the traffic optimization goal of load balancing is reached by the method of moving traffic load to idle route; said idle route points to the route whose traffic load is smaller than custom threshold, by analyzing real-time traffic Find a route that meets the conditions.

The present invention has the following advantages and effects over the existing technology:

(1) In the network source reuse of the present invention, the contribution to the reduction of time delay depending on the network source is used as an intermediate conversion metric of different network sources, and this is a model based on unified description of composite source and reuse source It is a new idea to construct The model can add performance comparison metrics for routing and transmission methods, source distribution metrics to increase multi-source utilization, and be metrics of multi data source selection.

(2) The network source reuse of the present invention quantifies and expresses the influence on cache and link due to time delay respectively, and reuses quantification for cache and link, unifying them and The composite source of the link between and the composite source of the route and the composite source of the area, the sizes of the three composite sources can be superimposed, and the granules can be adjusted.

(3) The routing mechanism for defining multi sources by the software of the present invention is based on the idea of "software refactoring" and centrally arranges "dummy nodes" with composite sources + new dispatch of source arrangement "distributed routing" Present and implement traffic optimization guided by source optimization distribution. The complexity of placing centralized "dummy nodes" is not high and the routing convergence process with limited "dummy nodes" is also not complicated. This is a method to realize traffic optimization for huge Mouse flow at the expense of low complexity source arrangement.

(4) In terms of application of the present invention to routing, after using composite sources as routing metrics, sources such as links and caches are fully utilized, resulting in higher network throughput rates and lower end-to-end delays. , Energy consumption of the data center is also reduced. In terms of application to source utilization, the network COP will rise. In terms of application to multi data source selection, the load of data sources is more balanced.

FIG. 1 is an overall job flow chart of the present invention. Fig.2 (a)-FIG.2 (b) are the sketches which add a "dummy node" in this invention, and improve a routing, FIG.2 (b) is a network topology after adding a "dummy node". It is structure schematic.

The present invention will be further described by combining the attached drawings and specific examples below. As shown in FIG. 1, the method of the present invention includes two parts, the dashed line frame at the top is the network source reuse in the present invention, and the dashed line frame at the bottom is a routing mechanism that defines multi-source by the software of the present invention. is there.

Network Source Reuse includes S11-S13:

In S11, assuming that the user wants to acquire the content o, the user is located at the node s, the upstream is the node i, and the contribution to the reduction of the time delay by one link is the physical distance and the bandwidth Expressed as a function of width, and further quantifying the mapping between link and time delay, specifically expressed as follows:

In which B is the bandwidth of the link from node s to its upstream node i, d is the distance from node s to its upstream node i, and α ₁ and α ₂ are the bandwidth and distance respectively Weight parameters, and the shorter the distance of one link and the greater the bandwidth, the less time to complete the transmission job; indeed, there is not much difference in the physical distance of each link in the DCN Not, so B is the main influencer among them.

The S12 is that multi-source transmission is to collect a linearly independent CM of a specified quantity from a plurality of nodes, and a link combined from one node or one area combined from a plurality of links The transmission job can be completed when the number of CMs collected from the target reaches the target. Therefore, the more CMs of one node as the basic constituent unit of a link or area, the less time it takes to complete a transmission job. Therefore, there is a mapping between the proportion of the content (or total content) CM at one node to the target quantity and the reduction of the time delay. Depending on the overlap possibility of the ratio, in order to make the cache have the same overlap characteristic as the link, the node (one link) is further superimposed and compared with the cache in the route and the area.

Therefore, the contribution to time delay reduction by one node cache is expressed as a function, and the mapping between cache and time delay is further quantified, and specifically expressed as follows:

Among them, M _i (o) is the quantity of CM storing the content o at node i, E (o) is the quantity of CM necessary for decoding the content o, and said CM (Coded Message) is a node Is a network coding block that stores and forwards.

After obtaining the quantitative values of the link, cache and time delay mapping based on the S11 and the S12, the S13 obtains them as a composite source of one link between two nodes and a plurality of links. Unifiedly expressed in three types of composite sources including one path composite source including one path and one area composite source including multiple paths, such as one link, one path and one area The basic building blocks of all kinds of networks are equipped with metrics of composite source, and according to the range of the network, superimpose the size of composite source for each type, adjust the granules, and the concrete process is as follows:

The composite source of one link between the two nodes is specifically expressed as follows:

In which α and β represent the weight parameters occupied by the link and the mapping of cache and time delay respectively;

If path j contains k links, then the composite source of said path is a superposition of the composite sources of k links, which is specifically expressed as follows:

When performing multi-source transmission through w paths in the area area, the composite source of the area is a superposition of w path composite sources, and specifically expressed as follows:

と

はそれぞれ一つのノード（一つのリンク）と、一つの経路と、一つの領域等のネットワークの基本構成ユニットのパフォーマンスを反映し、また比較性と重畳性を備えており、従って、Cost(o)_iと、

と

はルーティングのメトリックとしても、多データソースを選択するためのメトリックとすることができる。 As you can see from the above, Cost (o) _i ,

When

Respectively reflect the performance of one basic node (one link), one route, one basic unit of the network such as one area, and also have comparability and overlap, so Cost (o) _i ,

When

Can also be a metric for selecting many data sources as a routing metric.

In general, one link includes two nodes, and in the first embodiment, the composite source includes the bandwidth and the cache, and the composite source of one node is actually the cache of one node and its downstream. And the composite of the bandwidth of one link (node), so the meaning of the composite source of one link and the composite source of one node are the same, the cache of one node and one link downstream thereof It refers to the composite of bandwidth of (node). The composite source of one link (node) defined in the first embodiment is effective in adding the "dummy node" of the composite source.

The following is one specific application example

である。その中に、ｄがノードｓからその上流のノードｉまでの距離であり、単位がKmであり；Ｂがノードｓからその上流のノードｉまでのリンクの帯域幅であり、単位がMbpsである。C_Cost_i(o)＝

。α=1またβ=1とすると、ノードｓとノードｉとの間のリンクのコンポジットソースは

である。 Assuming that the user wants to obtain content o, and the user is located at node s, and its upstream is node i, let α ₁ = 10,000, α ₂ = 100,

It is. In which d is the distance from node s to its upstream node i and the unit is Km; B is the bandwidth of the link from node s to its upstream node i and the unit is Mbps . C_Cost _i (o) =

. Assuming that α = 1 and β = 1, the composite source of the link between node s and node i is

It is.

Example 2

As described in the first embodiment, the basic constituent units of the network such as one node (one link) and one route and one area each have a composite source metric. This embodiment 2 designs distributed name routing for Mouse flow in DCN based on the concept of “software refactoring”, ie adds “dummy node” with composite source in the network, and invites to the routing protocol to add network topology A change is found, which causes traffic to be directed to a given node and path or region.

As shown in FIG. 1, the routing mechanism for defining multi-source by the software of the second embodiment includes the following S21 to S24:
In S21, in the data center based on the SDN architecture, an SDN control plane having a global network view detects composite source status and user needs, and analyzes the traffic distribution to discover idle routes and optimize traffic. Determine the goals; The SDN is a Software Defined Network, a Software Defined Network.

The step S22 determines a traffic optimization target based on the step S21, and the SDN control plane establishes an optimum plan for deploying the "dummy node" according to the optimization theory, and includes the deployment quantity and the deployment location; "Is not actually present in the network, but is a node that is falsely added to the logic;" Dummy node "deployment method broadcasts a status report of the routing protocol to the network, and the report through the" dummy node " The information to reach other nodes is included, thereby completing the process of falsely adding "dummy nodes"; S22 includes S221 to S223:

に転換し、

である； The above-mentioned S221 "changes" the network topology by finding load balance as a traffic optimization target and finding a proposal to add "dummy nodes" with w composite sources in appropriate places in the network topology, for example, "dummy nodes". 'I's composite source is Cost (o) _i and the composite source Cost (o) _i is the cost metric required by the dynamic routing protocol

Switch to

Is;

The step S222 converts the specific process of the "dummy node" proposal described in the step S221 into a multi-target optimization problem as follows:

と定義し、θ_uでuつ目のリンクのトラヒック負荷を表し、θ_vでvつ目のリンクのトラヒック負荷を表し、

がリンクトラヒックの平均値である； Therein, one of the network G = assumed (V, E, C) a; E is linked collection is _{E = {e 1 ... e m} }, representing the m nd link e _m; C is a newly added “dummy node” collection, C = {c ₁ ... C _w }, c _{w represents} a w newly added “dummy node”; V is for all nodes A collection, V = {v ₁ ... V _n }, and v _n represents the n-th node; Load _G represents the Gini coefficient of link traffic throughout the network,

Define the traffic load of the u-th link by θ _u and the traffic load of the v-th link by θ _v ,

Is the average value of link traffic;

The multi-target optimization problem is constrained as follows:

であり、各リンクのトラヒック負荷が第一閾値θより小さく、クオリティオブサービスQoSを実現する；
[2] .

であり、各経路における各ノードのコストの総和が第二閾値Dより小さく、経路選択のクオリティを保証する； [1].

And the traffic load of each link is smaller than the first threshold θ to realize quality of service QoS;
[2].

And the sum of the costs of each node in each path is smaller than a second threshold D to guarantee the quality of path selection;

がノードv_iとノードv_jの間のリンクにおける各ノードのコストの総和である；ノードv_iとノードv_jとの間の経路にはzつのノードが含まれていると仮定すると、

である； Among them, v _j represents the i-th node, v _j denotes the _j-th node,

Assuming the path between the node v _i and node v _j that contains z single node; but the node v _i is the sum of the cost of each node in the link between the node v _j

Is;

The step S223 solves the multi-target optimization problem using the optimization theory again to obtain the number w of "dummy nodes" satisfying the constraints and the arrangement position thereof; the optimization theory uses integer linear programming algorithm adopt.

S23 restarts the routing discovery process after the dynamic routing protocol discovers a change in network topology;
When a change in network topology occurs due to the addition of a "dummy node", the OSPF protocol restarts the routing discovery process with Cost (o) _i as a cost metric, ie the shortest path between source and destination with the Dijkstra algorithm. Calculate again to minimize the cost sum of each node in the path.

At S24, the data report is transmitted by the newly discovered routing.

In the second embodiment, a routing protocol is invited to discover a change in network topology by falsely adding "dummy nodes" to logic.

In the second embodiment, the traffic optimization target of the load balance is reached by the method of moving the traffic load to the idle path; the idle path represents a path with less traffic load, and the idle path traffic load is 50% of the upper limit thereof. By analyzing smaller, real-time traffic, it is possible to find a route that meets the conditions.

In the second embodiment, the traffic optimization targets include any one or any combination of load balancing, failure path avoidance, and minimum energy consumption.

The following is one specific application example

=1の「ダミーノード」Pが添加されると、自分を通じてノードFまで到着できるとネットワークに放送する。従って、OSPFがルーティング発見過程を起動し、最後、S1-D1のために新経路A-P-F（コストが6である）を見出し、またノードPが実際に存在していないため、パケット転送の実際経路はA-B-C-Fである。こういうトラヒック負荷をアイドル経路に移動する方法によって、負荷バランスの目標が達成できる。 What is shown in FIGS. 2 (a) -2 (b) is to improve routing by adding “dummy nodes”; in which FIG. 2 (a) is a schematic representation of the primitive network topology, 图 2 (b) is a schematic representation of the network topology after adding a "dummy node", the original path between S1-D1 is ADEF (with cost 12),

When = 1 “dummy node” P is added, it broadcasts to the network when it can reach node F through itself. Therefore, OSPF starts the routing discovery process and finally finds a new route APF (with cost 6) for S1-D1, and since node P does not actually exist, the actual route of packet forwarding is It is ABCF. Load balancing goals can be achieved by moving these traffic loads to the idle path.

As noted above, the present invention combines the advantages of centralized and decentralized: routing convergence processes with less "demy nodes" and less "demy nodes" in placing centralized "dummy nodes". Not too complicated. This is a method of realizing traffic optimization for a huge mouse flow at the expense of a kind of complicated source arrangement, and is a traffic optimization concept guided by source optimization distribution.

The above is only a better embodiment of the new utility model and is not used to limit the new utility model. All modifications, equivalent replacements and improvements, etc. carried out under the meaning and principle of the new utility model are included in the protection scope of the new utility model.

Claims (2)

Including S11-S13:
In S11, assuming that the user wants to acquire the content o, the user is located at the node s, the upstream is the node i, and the contribution to the reduction of the time delay by one link is the physical distance and the bandwidth Expressed as a function of width, and further quantifying the mapping between link and time delay, specifically expressed as follows:

In which B is the bandwidth of the link from node s to its upstream node i, d is the distance from node s to its upstream node i, and α ₁ and α ₂ are the bandwidth and distance respectively Weight parameters of the

Is;
The S12 expresses the contribution to the reduction of the time delay by one node cache as a function, and further quantifies the mapping between the cache and the time delay, specifically as follows:

Among them, M _i (o) is the quantity of CM storing the content o at node i, E (o) is the quantity of CM necessary for decoding the content o, and said CM (Coded Message) is a node Network coding block that stores and forwards
After obtaining the quantitative values of the link, cache and time delay mapping based on the S11 and the S12, the S13 obtains them as a composite source of one link between two nodes and a plurality of links. Unifiedly express in three types of composite sources, including a composite source of one path including, and a composite source of a region including multiple paths;
In S13, basic configuration units of three types of networks such as one link and one route and one area all have metrics of composite source, and the size of composite source for each type is determined according to the range of the network. Superimpose and prepare the granules; the specific steps are as follows:
The composite source of one link from node s to its upstream node i, ie the composite source of node i, is specifically expressed as follows:

Among them, α and β represent the weight parameters occupied by the link, mapping of cache and delay time respectively;
If path j contains k links, then the composite source of said path is a superposition of the composite sources of k links, which is specifically expressed as follows:

;
When multi-source transmission is performed by w paths in the area area, the composite source of the area is a superposition of w path composite sources, and is specifically expressed as follows:

Network source reuse characterized by being. A multi-source model is obtained by the multi-source reuse method in any one data center network according to claim 1, the composite source is arranged by the multi-source model, and specifically, S21 to 24 are included:
In S21, in the data center based on the SDN architecture, an SDN control plane having a global network view detects composite source status and user needs, and analyzes the traffic distribution to discover idle routes and optimize traffic. Determine the goals; The SDN is a Software Defined Network, which is a Software Defined Network.
The step S22 determines a traffic optimization target based on the step S21, and the SDN control plane establishes an optimum plan for deploying the "dummy node" according to the optimization theory, and includes the deployment quantity and the deployment location; "Is not actually present in the network, but is a node that is falsely added to the logic;" Dummy node "deployment method broadcasts a status report of the routing protocol to the network, and the report through the" dummy node " Includes information reaching other nodes, which completes the process of falsely adding "dummy nodes";
The S22 includes S221 to S223:
The above-mentioned S221 "changes" the network topology by finding load balance as a traffic optimization target and finding a proposal to add "dummy nodes" with w composite sources in appropriate places in the network topology, for example, "dummy nodes". The cost metric required by _i 's composite source is Cost (o) _i and composite source Cost (o) _i is required by a dynamic routing protocol

Switch to

Is;
The step S222 converts the specific process of the "dummy node" proposal described in the step S221 into a multi-target optimization problem as follows:

Therein, one of the network G = assumed (V, E, C) a; E is linked collection is _{E = {e 1 ... e m} }, representing the m nd link e _m; C is a newly added “dummy node” collection, C = {c ₁ ... C _w }, c _{w represents} a w newly added “dummy node”; V is for all nodes A collection, V = {v ₁ ... V _n }, and v _n represents the n-th node; Load _G represents the Gini coefficient of link traffic throughout the network,

Define the traffic load of the u-th link by θ _u and the traffic load of the v-th link by θ _v ,

Is the average value of link traffic;
The multi-target optimization problem is constrained as follows:
[1].

And the traffic load of each link is smaller than the first threshold θ to realize quality of service QoS;
[2].

And the sum of the costs of each node in each path is smaller than a second threshold D to guarantee the quality of path selection;
Among them, v _i represents the i-th node, v _j represents the _j-th node,

Assuming the path between the node v _i and node v _j that contains z single node; but the node v _i is the sum of the cost of each node in the link between the node v _j

Is;
The step S223 solves the multi-target optimization problem using the optimization theory again to obtain the number w of "dummy nodes" satisfying the constraints and the arrangement position thereof; the optimization theory uses integer linear programming algorithm adopt;
S23 restarts the routing discovery process after the dynamic routing protocol discovers a change in network topology;
The specific steps of S23 are as follows:
When the network topology changes due to the addition of the “dummy node”, the OSPF protocol

Let C be a cost metric, restart the routing discovery process, ie recalculate the shortest path between source and destination with the Dijkstra algorithm, and minimize the cost sum of each node in the path;
S24 transmits the data report by newly discovered routing;
The traffic optimization goals include any or all three of load balancing, failure path avoidance and minimal energy consumption;
The traffic optimization goal of load balancing is reached by the method of moving traffic load to idle route; said idle route points the route whose traffic load is smaller than custom threshold, and finds a route satisfying the condition by analyzing real-time traffic A routing mechanism that defines multiple sources by means of software.

Images