CN103716402A - Selection system and method of servers - Google Patents

Abstract

The invention discloses a server selection system, the system includes a central server, an ordinary server, a landmark server, and a client, wherein the central server is used to receive service request information sent by the client, and the central server includes a network distance measurement module and a server Select the module, the network distance measurement module obtains the K common servers closest to the client and the network distance information related to the K common servers, and updates the common server map information in the central server to obtain the load information of the K common servers, The server selection module performs global optimal selection according to network distance information, load information and inter-domain traffic. The selection system of the present invention reduces the traffic transmitted between ISPs, and reduces the overhead in the server selection process by adopting a network distance prediction algorithm with less measurement overhead, thereby improving the scalability of cloud services. The invention also discloses a server selection method.

Description

Server selection system and method

technical field

The invention relates to the technical field of computer networks, in particular to a server selection system and method.

Background technique

At present, with the rapid growth of the business volume, access volume and data flow of each core department of the network, its processing power and computing intensity are also increasing correspondingly. At this time, a single server cannot independently undertake it. Although this kind of problem can be solved by upgrading the hardware, it will cause high cost investment. At this time, the bandwidth of existing network devices and servers can be expanded through load balancing technology to increase throughput, strengthen network data processing capabilities, and improve network flexibility and availability.

At present, with the rapid increase in the number of network applications and users, the scalability of network distance prediction technology is becoming more and more important. The main factors affecting scalability are: the distribution of landmarks in network coordinate calculation methods, including centralized landmarks and distributed landmarks; in non-coordinate distance prediction methods, the deployment of additional proxy nodes and the support of related servers, etc.

It can be seen that the current server selection problem mainly considers two aspects: (1) Network distance, for example, cloud service providers need to enable users to quickly obtain the required services at the "nearest" location, reducing the distance between terminals and terminals. (2) Server load balancing, for example, the load of the server needs to be within a certain tolerance range to avoid the phenomenon of being overloaded and quitting the service. However, the server selection problem not only considers the above two aspects, but also other aspects, such as inter-domain transmission traffic, network distance measurement overhead, and so on. Therefore, how to select a server more accurately and reasonably has become an urgent problem to be solved.

Contents of the invention

The present invention aims to solve at least one of the above-mentioned technical problems.

Therefore, the first object of the present invention is to propose a server selection system. The system reduces the traffic transmitted between ISPs, and by adopting a network distance prediction algorithm with less measurement overhead, it reduces the overhead in the server selection process, thereby improving the scalability of cloud services.

The second object of the present invention is to propose a server selection method.

In order to achieve the above purpose, the server selection system of the embodiment of the first aspect of the present invention includes: a central server, a common server, a landmark server, and a client, wherein the central server is used to receive the client The service request information sent, wherein the central server includes a network distance measurement module and a server selection module, and the network distance measurement module is used to obtain the K common servers and the K servers closest to the client. network distance information related to common servers, and update the common server map information in the central server to obtain the load information of the K common servers, and the server selection module is used for according to the network distance information, the load Global optimization selection for information and inter-domain traffic.

According to the server selection system of the embodiment of the present invention, after the central server receives the service request information sent by the client, the network distance measurement module in the central server obtains the K common servers and the K common servers that are closest to the client. The network distance information and the load information of K ordinary servers are obtained. The server selection module in the central server performs global optimization selection according to the network distance information, load information and inter-domain traffic, so that the optimal ordinary server can be selected to provide the client with Service, by setting the inter-domain traffic penalty coefficient in the server selection module to balance the network distance between ordinary servers and clients, and the proportion of traffic transmitted between ISPs in the server selection process, without loss of global network distance On the premise of prediction accuracy, reduce the traffic transmitted between ISPs, relieve the pressure of network service operators, and use a network distance prediction algorithm with less measurement overhead in the network distance measurement module, which reduces the overhead in the server selection process , thus improving the scalability of cloud services.

In order to achieve the above object, the server selection method of the second embodiment of the present invention includes the following steps: S1, the central server receives the service request information sent by the client, wherein the central server includes a network distance measurement module and a server selection module ; S2, the network distance measurement module acquires the K common servers closest to the client and the network distance information related to the K common servers; S3, the network distance measurement module updates the central server Common server map information, to obtain the load information of the K common servers; and S4, the server selection module performs global optimal selection according to the network distance information, the load information and inter-domain traffic.

According to the server selection method of the embodiment of the present invention, after the central server receives the service request information sent by the client, the network distance measurement module in the central server obtains the K common servers and the K common servers that are closest to the client. The network distance information and the load information of K ordinary servers are obtained. The server selection module in the central server performs global optimization selection according to the network distance information, load information and inter-domain traffic, so that the optimal ordinary server can be selected to provide the client with Service, by setting the inter-domain traffic penalty coefficient in the server selection module to balance the network distance between ordinary servers and clients, and the proportion of traffic transmitted between ISPs in the server selection process, without loss of global network distance On the premise of prediction accuracy, reduce the traffic transmitted between ISPs, relieve the pressure of network service operators, and use a network distance prediction algorithm with less measurement overhead in the network distance measurement module, which reduces the overhead in the server selection process , thus improving the scalability of cloud services.

Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Description of drawings

The above and/or additional aspects and advantages of the present invention will become apparent and easy to understand from the following description of the embodiments in conjunction with the accompanying drawings, wherein,

FIG. 1 is a block diagram of a server selection system according to an embodiment of the present invention;

FIG. 2 is a schematic structural diagram of a server selection system according to an embodiment of the present invention;

Fig. 3 is a flowchart of a method for selecting a server according to an embodiment of the present invention;

FIG. 4 is a flow chart of clustering common servers according to an embodiment of the present invention;

Fig. 5 is a flow chart of acquiring the K common servers closest to the client and their related network distance information according to one embodiment of the present invention; and

Fig. 6 is an entity relationship diagram of a method for selecting a server according to an embodiment of the present invention.

Detailed ways

Embodiments of the present invention are described in detail below, examples of which are shown in the drawings, wherein the same or similar reference numerals designate the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the figures are exemplary only for explaining the present invention and should not be construed as limiting the present invention. On the contrary, the embodiments of the present invention include all changes, modifications and equivalents coming within the spirit and scope of the appended claims.

In the description of the present invention, it should be understood that the terms "first", "second" and so on are used for descriptive purposes only, and cannot be interpreted as indicating or implying relative importance. In the description of the present invention, it should be noted that unless otherwise specified and limited, the terms "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral Ground connection; it can be mechanical connection or electrical connection; it can be direct connection or indirect connection through an intermediary. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present invention in specific situations. In addition, in the description of the present invention, unless otherwise specified, "plurality" means two or more.

Any process or method descriptions in flowcharts or otherwise described herein may be understood to represent modules, segments or portions of code comprising one or more executable instructions for implementing specific logical functions or steps of the process , and the scope of preferred embodiments of the invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including substantially concurrently or in reverse order depending on the functions involved, which shall It is understood by those skilled in the art to which the embodiments of the present invention pertain.

It should be noted that the server selection system in the embodiment of the present invention may include four network nodes: a central server, a landmark server, an ordinary server, and a client. Wherein, as shown in FIG. The server map module, the normal server and the client node respectively have a landmark server map module. The entity definition in the server selection system of the embodiment of the present invention is as follows, the network user who requests the network service can be called the client, according to the difference of function, the resource storage node can be divided into three categories: 1, carry out centralized management to various information and The server maintained can be called the central server; 2. The server serving as a reference node for network distance measurement can be called a landmark server; 3. The server providing services for network users can be called an ordinary server.

It should also be noted that in the server selection system of the embodiment of the present invention, the central server is the core part of the network application, and can be responsible for the management, maintenance, distribution and scheduling of network resources. Deploy landmark servers and common servers in different geographic regions through the landmark server map module and common server map module in the central server, and send the deployment information of the landmark server to all common servers and clients requesting services. Landmark servers can be deployed in different geographic regions with a fixed number, and are reference nodes used to measure network distances in network applications. Ordinary servers can be deployed in different geographic regions to cache network resources and provide users with content or computing services.

The system and method for selecting a server according to the embodiments of the present invention will be described below with reference to the accompanying drawings.

Fig. 2 is a schematic structural diagram of a system for selecting a server according to an embodiment of the present invention.

The server selection system of the embodiment of the present invention will be described in detail below with reference to FIG. 1 and FIG. 2 . As shown in FIG. 1 and FIG. 2 , the server selection system may include: a central server 100 , an ordinary server 200 , a landmark server 300 and a client 400 .

Specifically, the central server 100 is used to receive the service request information sent by the client 400, wherein the central server 100 may include a network distance measurement module 110 and a server selection module 120, and the network distance measurement module 110 is used to obtain the closest distance to the client 400 K common servers 200 and K common server 200-related network distance information, and update the common server map information in the central server 100, to obtain the load information of K common servers 200, the server selection module 120 is used for according to the network distance Information, load information and inter-domain traffic for global optimization selection.

Wherein, in one embodiment of the present invention, the general server map information may include the identification of the common server 200, the IP (Internet Protocol, a protocol for interconnection between networks) address of the common server 200, the first network distance vector of the common server 200 , load information of the common server 200, etc. It can be understood that the common server map information can be stored in the common server map module in the central server 100 . In addition, network distance information can be understood as network delay, which can be obtained through network distance prediction methods, and network distance prediction methods can include network coordinate calculation methods, non-coordinate distance prediction methods, and the like.

According to an embodiment of the present invention, the central server 100 is further configured to: before receiving the service request information sent by the client 400, update the map information of the first landmark server in the ordinary server 200 according to a preset frequency, and update the updated first landmark server map information A landmark server map information is synchronized to the common server map information in the central server 100; the first network distance vector of the common server 200 is graded to obtain the first network distance grade vector; The server 200 performs clustering. Wherein, in the embodiment of the present invention, the map information of the first landmark server may include one or more of the logo of the landmark server 300, the IP address of the landmark server 300, the network distance between the common server 200 and the landmark server 300, etc. kind.

That is to say, in the embodiment of the present invention, the map module of the first landmark server on the ordinary server 200 will be updated according to a certain frequency, and then the updated related information will be synchronized to the map information of the ordinary server on the central server 100 . The network distance measuring module 110 on the central server 100 obtains the first network distance vector of the common server 200 from the common server map information, and performs hierarchical processing on the first network distance vector of the common server 200 . Then, all common servers 200 are clustered by using the first network distance level vector as a cluster identifier, so that common servers 200 are divided into different clusters. Wherein, the update frequency of the cluster is the same as the update frequency of the map information of the first landmark server on the common server 200 .

的乘积，即运算γ×v，然后，对运算后的矢量中每一个元素进行取整运算，这时所获取到的第一网络距离矢量便演化成第一网络距离等级矢量。其中，a可通过对普通服务器200的等级迭代得到，使得最后的分簇大部分都包含一定数量的服务器，既不过多也不过少。Wherein, in the embodiment of the present invention, the process of classifying the first network distance vector v of the ordinary server 200 is as follows: the first network distance vector v of the ordinary server 200 is compared with the class parameter

The product of , that is, to operate γ×v, and then perform a rounding operation on each element in the calculated vector, and then the obtained first network distance vector evolves into a first network distance level vector. Wherein, a can be obtained by iterating the levels of common servers 200, so that most of the final clusters contain a certain number of servers, neither too many nor too few.

For example, if the first network distance vector from the common server 200 to the node where the landmark server 300 is located is (92, 103, 210, 321, 532), and the value of a is 100, then 1/100×(92, 103, 210, 321, 532) = (0.92, 1.03, 2.10, 3.21, 5.32), rounding to get (0, 1, 2, 3, 5), then (0, 1, 2, 3, 5) is an ordinary server 200 to the first network distance level vector of the node where the landmark server 300 is located. Finally, common servers 200 whose first network distance level vectors are the same as (0, 1, 2, 3, 5) can be classified into the same cluster.

It should be noted that, in the embodiment of the present invention, the first network distance vector can be composed of the network distances from the node where the current common server 200 is located to all the landmark servers 300, and the number of elements contained in the first network distance vector is the same as that of the landmark servers 300 For example, the network distances from the node where the current common server is located to the landmark node server 300 are d ₁ , d ₂ ... d _m , then the first network distance vector of the node where the current common server 200 is located is (d ₁ , d ₂ ... d _m ).

According to an embodiment of the present invention, the central server 100 is specifically configured to: update the second landmark server map information in the client 400, and synchronize the updated second landmark server map information to the network distance measurement module 110; The map information of the second landmark server obtains the second network distance vector of the client 400, and performs grade processing on the second network distance vector to obtain the second network distance grade vector; according to the first network distance grade vector and the second network distance The level vector obtains the cluster corresponding to the second network distance level vector; calculates and sorts the common servers 200 in the cluster according to the similarity comparison formula to obtain the distance similarity list between the client 400 and the common server 200; and measure The network distance information between the client 400 and the top K common servers 200 in the distance similarity list, and send the network distance information to the server selection module 120 . Wherein, in the embodiment of the present invention, the second landmark server map information may include one or more of the logo of the landmark server 300, the IP address of the landmark server 300, the network distance between the client 400 and the landmark server 300, etc. kind.

That is to say, in the embodiment of the present invention, after the client 400 sends the service request information, it will trigger the client 400 to update the map information of the second landmark server in the client 400, and then the updated second landmark server The map information is synchronized to the network distance measurement module 110 on the central server 100 . The network distance measurement module 110 can first obtain the network distance between the client and the landmark server in the updated map information of the second landmark server, and then obtain the second network distance vector to the client 400 according to the network distance, specifically obtain The process may refer to the process of obtaining the distance vector of the first network. Afterwards, the network distance measuring module 110 may perform level processing on the second network distance vector to obtain the second network distance level vector. Wherein, the process of performing hierarchical processing on the second network distance vector of the client 400 is the same as the process of performing hierarchical processing on the first network distance vector of the common server 200 , which will not be repeated here. Then, the second network distance level vector can be compared with the first network distance level vector in the common server 200 to find out the first network distance level vector of the common server 200 that is the same as the second network distance level vector, and then it can be based on The first network distance level vector finds the corresponding common server 200 to obtain the cluster to which the common server 200 belongs. Afterwards, all ordinary servers 200 in the cluster are calculated according to the similarity comparison formula to calculate the distance similarity list between the client 400 and the ordinary server 200, and the values in the distance similarity list can be calculated according to the similarity value. Arranged by size. Finally, directly measure the network distance information between the client 400 and the top K common servers 200 in the distance similarity list, and send it to the server selection module 120 .

Wherein, in the embodiment of the present invention, the model of the similarity comparison formula is as follows:

$\mathrm{<span\; class="notranslate">\; Sim</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; dist</span>}<span\; class="notranslate">\; (</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; ,</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; dist</span>}}_{\mathrm{<span\; class="notranslate">\; max</span>}}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{\sqrt{\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span>\frac{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}{\sqrt{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; |</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

为客户端c的第二网络距离矢量与普通服务器i的第一网络距离矢量在(m-1)维坐标空间中的欧式距离；dist_max为所有距离值中的最大值；r_c,m为客户端c与标号为m的地标服务器之间的网络距离；r_i,m为普通服务器i与标号为m的地标服务器之间的网络距离；r_cmin和r_cmax分别为客户端c与M个地标服务器之间网络距离的最小值和最大值；r_imin和r_imax分别为普通服务器i与M个地标服务器之间网络距离的最小值和最大值；Sim(c,i)∈[0,1]。Among them, Sim(c,i) is the network distance similarity between client c and common server i; M is the total number of landmark servers,

is the Euclidean distance between the second network distance vector of client c and the first network distance vector of common server i in the (m-1) dimensional coordinate space; dist _max is the maximum value among all distance values; r _c,m is The network distance between the client c and the landmark server labeled m; r _i,m is the network distance between the ordinary server i and the landmark server labeled m; r _cmin and r _cmax are the client c and M The minimum and maximum network distances between landmark servers; r _imin and r _imax are the minimum and maximum network distances between ordinary server i and M landmark servers respectively; Sim(c,i)∈[0,1 ].

According to an embodiment of the present invention, the model selected by global optimization is as follows:

$\underset{\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; C</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; I</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; ci</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; cos</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

|P _i -w _i |≤ε _i (3)

和R_ci≥0；cost(c,i)=λD(c,i)，λ为引入的域间流量传输惩罚系数，D(c,i)为客户端c与普通服务器i之间的网络距离；

为普通服务器i的负载信息在所有负载信息中所占的比例，N为普通服务器200的总数量；w_i为中心服务器100预设的普通服务器200负载信息在所有负载信息中所占的比例；ε_i为允许普通服务器200负载信息偏离w_i的额度。其中，当客户端400和普通服务器200在同一个ISP（Internet Service Provider，互联网服务提供商）内时，λ可为1，则cost(c,i)可为D(c,i)；当客户端400和普通服务器200不在同一个ISP内时，λ可为0，则cost(c,i)可为0。Among them, C and I are the sets of client 400 and common server 200 respectively, R _ci is the probability that client c is assigned to common server i, and satisfies

and R _ci ≥ 0; cost(c,i)=λD(c,i), λ is the penalty coefficient for inter-domain traffic transmission introduced, D(c,i) is the network distance between client c and common server i ;

is the proportion of the load information of the common server i in all the load information, N is the total number of common servers 200; w _i is the proportion of the common server 200 load information preset by the central server 100 in all the load information; ε _i is the amount that allows the load information of the common server 200 to deviate from w _i . Wherein, when the client 400 and the common server 200 are in the same ISP (Internet Service Provider, Internet Service Provider), λ can be 1, then cost (c, i) can be D (c, i); when the client When the end 400 and the common server 200 are not in the same ISP, λ can be 0, and cost(c,i) can be 0.

That is to say, in the embodiment of the present invention, the central server 100 can use the network distance information obtained above from the client 400 to the K ordinary servers 200 and the load information of the K ordinary servers 200 as input data, and the input data Substitute into the global optimal selection model to perform global optimal selection on the common server 200 , so that the optimal common server 200 can be obtained to provide services for the client 400 .

According to the server selection system of the embodiment of the present invention, after the central server receives the service request information sent by the client, the network distance measurement module in the central server obtains the K common servers and the K common servers that are closest to the client. The network distance information and the load information of K ordinary servers are obtained. The server selection module in the central server performs global optimization selection according to the network distance information, load information and inter-domain traffic, so that the optimal ordinary server can be selected to provide the client with Service, by setting the inter-domain traffic penalty coefficient in the server selection module to balance the network distance between ordinary servers and clients, and the proportion of traffic transmitted between ISPs in the server selection process, without loss of global network distance On the premise of prediction accuracy, reduce the traffic transmitted between ISPs, relieve the pressure of network service operators, and use a network distance prediction algorithm with less measurement overhead in the network distance measurement module, which reduces the overhead in the server selection process , thus improving the scalability of cloud services.

In order to realize the above embodiments, the present invention also proposes a method for selecting a server.

Fig. 3 is a flowchart of a method for selecting a server according to an embodiment of the present invention.

As shown in Figure 3, the server selection method may include the following steps:

S301. The central server receives service request information sent by the client, wherein the central server includes a network distance measurement module and a server selection module.

According to an embodiment of the present invention, as shown in FIG. 4, before the central server receives the service request information sent by the client, that is, before step S301, the server selection method further includes:

S401, the network distance measuring module updates the map information of the first landmark server in the common server according to a preset frequency, and synchronizes the updated map information of the first landmark server with the map information of the common server in the central server.

Wherein, in the embodiment of the present invention, the map information of the first landmark server may include one or more of the identifier of the landmark server, the IP address of the landmark server, the network distance between the common server and the landmark server, and the like. In addition, the common server map information may include one or more of common server identifiers, common server IP addresses, common server first network distance vectors, common server load information, and the like. It can be understood that the common server map information can be stored in the common server map module in the central server.

For example, the first landmark server map module on the common server will be updated according to a certain frequency, and then the updated relevant information will be synchronized to the common server map information on the central server.

S402. The network distance measurement module performs level processing on the first network distance vector of the common server to obtain the first network distance level vector.

For example, the network distance measuring module may first obtain the first network distance vector of the common server from the common server map information, and then perform hierarchical processing on the first network distance vector of the common server to obtain the first network distance level vector.

的乘积，即运算γ×v，然后，对运算后的矢量中每一个元素进行取整运算，这时所获取到的第一网络距离矢量便演化成第一网络距离等级矢量。其中，a可通过对普通服务器的等级迭代得到，使得最后的分簇大部分都包含一定数量的服务器，既不过多也不过少。Wherein, in the embodiment of the present invention, the process of classifying the first network distance vector v of the common server is as follows: the first network distance vector v of the common server is compared with the class parameter

The product of , that is, to operate γ×v, and then perform a rounding operation on each element in the calculated vector, and then the obtained first network distance vector evolves into a first network distance level vector. Among them, a can be obtained by iterating the levels of ordinary servers, so that most of the final clusters contain a certain number of servers, neither too many nor too few.

For example, if the first network distance vector from the common server to the node where the landmark server is located is (92, 103, 210, 321, 532), and the value of a is 100, then 1/100×(92, 103, 210, 321, 532) = (0.92, 1.03, 2.10, 3.21, 5.32), rounding to get (0, 1, 2, 3, 5), then (0, 1, 2, 3, 5) is the normal server to the landmark The first network distance level vector of the node where the server is located.

It should be noted that, in the embodiment of the present invention, the first network distance vector can be composed of the network distances from the node where the current common server is located to all landmark servers, and the number of elements contained in the first network distance vector is equal to the number of landmark servers , for example: the network distances from the current normal server node to the landmark node server are d ₁ , d ₂ ...d _m , then the first network distance vector of the current common server node is (d ₁ , d ₂ ...d _m ).

S403. The network distance measuring module clusters the common servers according to the first network distance level vector.

That is to say, in the embodiment of the present invention, the network distance measuring module may use the first network distance level vector as a cluster identifier to cluster all common servers, so that common servers are divided into different clusters. Wherein, the update frequency of the cluster is the same as the update frequency of the map information of the first landmark server on the common server. For example, common servers whose first network distance level vectors are both (0, 1, 2, 3, 5) may be grouped into the same cluster.

S302. The network distance measuring module acquires K common servers closest to the client and network distance information related to the K common servers.

Wherein, in the embodiment of the present invention, the network distance information can be understood as network delay, which can be obtained through a network distance prediction method, and the network distance prediction method can include a network coordinate calculation method, a non-coordinate distance prediction method, and the like.

According to an embodiment of the present invention, as shown in FIG. 5, step S302 may further include:

S501. The central server updates the map information of the second landmark server in the client, and synchronizes the updated map information of the second landmark server to the network distance measurement module.

Wherein, in the embodiment of the present invention, the map information of the second landmark server may include one or more of the identifier of the landmark server, the IP address of the landmark server, the network distance between the client and the landmark server, and the like.

For example, after the client sends the service request information, it will trigger the client to update the map information of the second landmark server in the client, and then the updated map information of the second landmark server can be synchronized to the network distance measurement module on the central server .

S502. The network distance measurement module acquires the second network distance vector of the client according to the updated map information of the second landmark server, and performs grade processing on the second network distance vector to obtain a second network distance grade vector.

For example, the network distance measurement module can first obtain the network distance between the client and the landmark server in the updated map information of the second landmark server, and then obtain the second network distance vector to the client according to the network distance. The specific acquisition process Reference may be made to the acquisition process of the first network distance vector. Afterwards, level processing may be performed on the second network distance vector to obtain the second network distance level vector. Wherein, the process of performing hierarchical processing on the second network distance vector of the client is the same as the process of performing hierarchical processing on the first network distance vector of the common server, and will not be repeated here.

S503. The network distance measurement module acquires a cluster corresponding to the second network distance level vector according to the first network distance level vector and the second network distance level vector.

For example, the network distance measurement module can compare the second network distance level vector with the first network distance level vector in the common server, find out the first network distance level vector of the common server that is the same as the second network distance level vector, and then The corresponding common server can be found according to the first network distance level vector, so as to obtain the cluster to which the common server belongs.

S504. The network distance measurement module calculates and sorts the common servers in the cluster according to the similarity comparison formula, so as to obtain a distance similarity list between the client and the common servers.

For example, the network distance measurement module can calculate the similarity comparison formula for all ordinary servers in the cluster to calculate the distance similarity list between the client and the ordinary server, and the values in the distance similarity list can be calculated according to the similarity Order of degree values.

Wherein, in the embodiment of the present invention, the model of the similarity comparison formula is as follows:

$\mathrm{<span\; class="notranslate">\; Sim</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; dist</span>}<span\; class="notranslate">\; (</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; ,</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; dist</span>}}_{\mathrm{<span\; class="notranslate">\; max</span>}}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{\sqrt{\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span>\frac{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}{\sqrt{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; |</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

为客户端c的第二网络距离矢量与普通服务器i的第一网络距离矢量在(m-1)维坐标空间中的欧式距离；dist_max为所有距离值中的最大值；r_c,m为客户端c与标号为m的地标服务器之间的网络距离；r_i,m为普通服务器i与标号为m的地标服务器之间的网络距离；r_cmin和r_cmax分别为客户端c与M个地标服务器之间网络距离的最小值和最大值；r_imin和r_imax分别为普通服务器i与M个地标服务器之间网络距离的最小值和最大值；Sim(c,i)∈[0,1]。Among them, Sim(c,i) is the network distance similarity between client c and common server i; M is the total number of landmark servers,

is the Euclidean distance between the second network distance vector of client c and the first network distance vector of common server i in the (m-1) dimensional coordinate space; dist _max is the maximum value among all distance values; r _c,m is The network distance between the client c and the landmark server labeled m; r _i,m is the network distance between the ordinary server i and the landmark server labeled m; r _cmin and r _cmax are the client c and M The minimum and maximum network distances between landmark servers; r _imin and r _imax are the minimum and maximum network distances between ordinary server i and M landmark servers respectively; Sim(c,i)∈[0,1 ].

S505. The network distance measurement module measures the network distance information between the client and the top K common servers in the distance similarity list, and sends the network distance information to the server selection module.

S303, the network distance measurement module updates the common server map information in the central server, so as to obtain the load information of K common servers.

S304. The server selection module performs global optimization selection according to network distance information, load information, and inter-domain traffic.

Wherein, in the embodiment of the present invention, the model selected by global optimization is as follows:

$\underset{\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; C</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; I</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; ci</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; cos</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

|P _i -w _i |≤ε _i (3)

和R_ci≥0；cost(c,i)=λD(c,i)，λ为引入的域间流量传输惩罚系数，D(c,i)为客户端c与普通服务器i之间的网络距离；

为普通服务器i的负载信息在所有负载信息中所占的比例，N为普通服务器的总数量；w_i为中心服务器预设的普通服务器负载信息在所有负载信息中所占的比例；ε_i为允许普通服务器负载信息偏离w_i的额度。其中，当客户端和普通服务器在同一个ISP内时，λ可为1，则cost(c,i)可为D(c,i)；当客户端和普通服务器不在同一个ISP内时，λ可为0，则cost(c,i)可为0。Among them, C and I are the sets of clients and common servers respectively, R _ci is the probability that client c is assigned to common server i, and satisfies

and R _ci ≥ 0; cost(c,i)=λD(c,i), λ is the penalty coefficient for inter-domain traffic transmission introduced, D(c,i) is the network distance between client c and common server i ;

is the proportion of the load information of the common server i in all the load information, N is the total number of common servers; w _i is the proportion of the load information of the common server preset by the central server in all the load information; ε _i is The amount by which normal server load information is allowed to deviate from _wi . Among them, when the client and the ordinary server are in the same ISP, λ can be 1, then cost(c,i) can be D(c,i); when the client and the ordinary server are not in the same ISP, λ can be 0, then cost(c,i) can be 0.

That is to say, in the embodiment of the present invention, the central server can use the network distance information obtained above from the client to the K common servers and the load information of the K common servers as input data, and substitute the input data into the global optimization selection In the model of , the global optimal selection of ordinary servers can be performed, so that the best ordinary servers can be obtained to provide services for clients.

According to the server selection method of the embodiment of the present invention, after the central server receives the service request information sent by the client, the network distance measurement module in the central server obtains the K common servers and the K common servers that are closest to the client. The network distance information and the load information of K ordinary servers are obtained. The server selection module in the central server performs global optimization selection according to the network distance information, load information and inter-domain traffic, so that the optimal ordinary server can be selected to provide the client with Service, by setting the inter-domain traffic penalty coefficient in the server selection module to balance the network distance between ordinary servers and clients, and the proportion of traffic transmitted between ISPs in the server selection process, without loss of global network distance On the premise of prediction accuracy, reduce the traffic transmitted between ISPs, relieve the pressure of network service operators, and use a network distance prediction algorithm with less measurement overhead in the network distance measurement module, which reduces the overhead in the server selection process , thus improving the scalability of cloud services.

In order to make the advantages of the present invention more obvious, the following examples will be described.

Fig. 6 is an entity relationship diagram of a method for selecting a server according to an embodiment of the present invention. For example, as shown in Figure 6, first, the client can send service request information to the central server (601), and the central server can return relevant landmark server information to the client according to the landmark server map information in the central server (602). Afterwards, the client may send detection information to the landmark server (603), so that the client obtains a detection result returned by the landmark server (604). Then, the client can send relevant network distance information to the central server according to the detection result (605), and the central server can calculate the top K common servers according to the network distance information (606). Afterwards, the client may send detection information to the normal server (607), so that the client obtains the detection result returned by the normal server (608). Then, the client can send related network distance information to the central server according to the detection result (609). Finally, the central server can make a global optimization selection based on the network distance information, the load information of the common server, and the inter-domain traffic, so that the optimal common server can be obtained to provide services for the client.

It should be understood that various parts of the present invention can be realized by hardware, software, firmware or their combination. In the embodiments described above, various steps or methods may be implemented by software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one or combination of the following techniques known in the art: Discrete logic circuits, ASICs with suitable combinational logic gates, Programmable Gate Arrays (PGAs), Field Programmable Gate Arrays (FPGAs), etc.

In the description of this specification, descriptions referring to the terms "one embodiment", "some embodiments", "example", "specific examples", or "some examples" mean that specific features described in connection with the embodiment or example , structure, material or characteristic is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

Although the embodiments of the present invention have been shown and described, those skilled in the art can understand that various changes, modifications, substitutions and modifications can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the invention is defined by the claims and their equivalents.

Claims (12)

1. A selection system for a server, comprising: a central server, a common server, a landmark server, and a client, wherein, The central server, the central server is used to receive the service request information sent by the client, wherein the central server includes a network distance measurement module and a server selection module, and the network distance measurement module is used to obtain the The client is closest to the K ordinary servers and the network distance information related to the K ordinary servers, and updates the map information of the ordinary servers in the central server to obtain the load information of the K ordinary servers, so The server selection module is used to perform global optimal selection according to the network distance information, the load information and inter-domain traffic. 2. The system according to claim 1, wherein the common server map information includes the common server's identifier, the common server's IP address, the first network distance vector of the common server, the One or more of the load information of common servers. 3. The system according to claim 2, wherein the central server is also used for: Before receiving the service request information sent by the client, updating the map information of the first landmark server in the common server according to a preset frequency, and synchronizing the updated map information of the first landmark server to the central server The general server map information in ; performing hierarchical processing on the first network distance vector of the common server to obtain a first network distance grade vector; and The common servers are clustered according to the first network distance level vector. 4. The system according to claim 3, wherein the central server is specifically used for: updating the map information of the second landmark server in the client, and synchronizing the updated map information of the second landmark server to the network distance measurement module; Acquiring a second network distance vector of the client according to the updated map information of the second landmark server, and performing grade processing on the second network distance vector to obtain a second network distance grade vector; Acquiring a cluster corresponding to the second network distance level vector according to the first network distance level vector and the second network distance level vector; calculating and sorting the common servers in the cluster according to the similarity comparison formula to obtain a distance similarity list between the client and the common servers; and Measuring the network distance information between the client and the first K common servers in the distance similarity list, and sending the network distance information to the server selection module. 5. system according to claim 4, is characterized in that, the model of described similarity comparison formula is as follows: $\mathrm{<span\; class="notranslate">\; Sim</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; dist</span>}<span\; class="notranslate">\; (</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; ,</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; dist</span>}}_{\mathrm{<span\; class="notranslate">\; max</span>}}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{\sqrt{\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span>\frac{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}{\sqrt{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; |</span>}}$ Wherein, Sim(c, i) is the network distance similarity between the client c and the common server i; M is the total number of landmark servers,

is the Euclidean distance between the second network distance vector of the client c and the first network distance vector of the common server i in the (m-1) dimensional coordinate space; dist _max is the maximum value among all distance values; r _{c, m} are the network distances between the client c and the landmark server labeled m; r _{i, m} are the network distances between the common server i and the landmark server labeled m; r _cmin and r _cmax are the minimum and maximum values of the network distances between the client c and the M landmark servers, respectively; r _imin and _rimax are the minimum and maximum values of the network distances between the common server i and the M landmark servers, respectively ; Sim(c,i)∈[0,1]. 6. The system according to claim 1, wherein the model selected by the global optimization is as follows: $\underset{\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; C</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; I</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; ci</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; cos</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}$ Among them, C and I are the sets of the client and the normal server respectively, R _ci is the probability that the client c is assigned to the normal server i, and satisfies

and R _ci ≥ 0; cost(c,i)=λD(c,i), λ is the penalty coefficient for inter-domain traffic transmission introduced, D(c,i) is the difference between the client c and the common server i network distance between

is the proportion of the load information of the common server i in all the load information, N is the total number of the common servers; w _i is the load information of the common server preset by the central server in all the load information The proportion; ε _i is the amount that allows the load information of the common server to deviate from w _i . 7. A method for selecting a server, comprising the following steps: S1. The central server receives service request information sent by the client, wherein the central server includes a network distance measurement module and a server selection module; S2, the network distance measurement module acquires K common servers closest to the client and network distance information related to the K common servers; S3, the network distance measurement module updates the common server map information in the central server, so as to obtain the load information of the K common servers; and S4. The server selection module performs global optimal selection according to the network distance information, the load information, and inter-domain traffic. 8. The method according to claim 7, wherein the common server map information comprises the common server's identifier, the common server's IP address, the first network distance vector of the common server, the One or more of the load information of common servers. 9. The method according to claim 8, further comprising: before said S1: The network distance measurement module updates the map information of the first landmark server in the common server according to a preset frequency, and synchronizes the updated map information of the first landmark server to the map of the common server in the central server information; The network distance measurement module performs grade processing on the first network distance vector of the common server to obtain a first network distance grade vector; and The network distance measurement module clusters the common servers according to the first network distance level vector. 10. The method according to claim 9, wherein said S2 further comprises: The central server updates the map information of the second landmark server in the client, and synchronizes the updated map information of the second landmark server to the network distance measurement module; The network distance measurement module obtains the second network distance vector of the client according to the updated map information of the second landmark server, and performs grade processing on the second network distance vector to obtain a second network distance grade vector ; The network distance measurement module acquires a cluster corresponding to the second network distance level vector according to the first network distance level vector and the second network distance level vector; The network distance measurement module calculates and sorts the common servers in the cluster according to a similarity comparison formula to obtain a distance similarity list between the client and the common servers; and The network distance measurement module measures network distance information between the client and the first K common servers in the distance similarity list, and sends the network distance information to the server selection module. 11. method according to claim 10, is characterized in that, the model of described similarity comparison formula is as follows: $\mathrm{<span\; class="notranslate">\; Sim</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; dist</span>}<span\; class="notranslate">\; (</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; ,</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; dist</span>}}_{\mathrm{<span\; class="notranslate">\; max</span>}}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{\sqrt{\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span>\frac{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; max</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; min</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}{\sqrt{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; |</span>}}$ Wherein, Sim(c, i) is the network distance similarity between the client c and the common server i; M is the total number of landmark servers, is the Euclidean distance between the second network distance vector of the client c and the first network distance vector of the common server i in the (m-1) dimensional coordinate space; dist _max is the maximum value among all distance values; r _{c, m} are the network distances between the client c and the landmark server labeled m; r _{i, m} are the network distances between the common server i and the landmark server labeled m; r _cmin and r _cmax are the minimum and maximum values of the network distances between the client c and the M landmark servers, respectively; r _imin and _rimax are the minimum and maximum values of the network distances between the common server i and the M landmark servers, respectively ; Sim(c,i)∈[0,1]. 12. The method according to claim 7, wherein the model selected by the global optimization is as follows: $\underset{\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; C</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; I</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; ci</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; cos</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}$ Among them, C and I are the sets of the client and the normal server respectively, R _ci is the probability that the client c is assigned to the normal server i, and satisfies

and R _ci ≥ 0; cost(c,i)=λD(c,i), λ is the penalty coefficient for inter-domain traffic transmission introduced, D(c,i) is the difference between the client c and the common server i the network distance between;

is the proportion of the load information of the common server i in all the load information, N is the total number of the common servers; w _i is the load information of the common server preset by the central server in all the load information The proportion; ε _i is the amount that allows the load information of the common server to deviate from w _i .

Images