CN111586139A

CN111586139A - Information transmission method and device based on echo state network

Info

Publication number: CN111586139A
Application number: CN202010361801.6A
Authority: CN
Inventors: 董亚东
Original assignee: Bank of China Ltd
Current assignee: Bank of China Ltd
Priority date: 2020-04-30
Filing date: 2020-04-30
Publication date: 2020-08-25
Anticipated expiration: 2040-04-30
Also published as: CN111586139B

Abstract

The invention provides an information transmission method and device based on an echo state network, when a server in a server cluster corresponding to an application system currently processes task information to obtain corresponding result information, a next-level application system corresponding to a target address can be selected from the application system clusters corresponding to a plurality of preset target addresses currently according to the target address of the result information, and the next-level application system corresponding to the target address is connected to transmit information. In the method provided by the embodiment of the present invention, in each time period, an application system may be determined from a plurality of next-level application systems corresponding to a preset target address based on a pre-trained echo state network, and the determined application system is used as the next-level application system corresponding to the preset target address in the time period for information transmission. Therefore, in different time periods, the utilization rate of the application system in the network is improved when the application system can be connected with different next-level application systems to transmit information.

Description

Information transmission method and device based on echo state network

Technical Field

The present invention relates to the field of computer technologies, and in particular, to an information transmission method and apparatus based on an echo state network.

Background

Many enterprises or organizations often deploy many application systems in a service network to implement various services in order to implement business needs. In the running process of the application system, the server cluster corresponding to the application system processes the task information and is connected with the next application system according to the target address in the result information to send the result information.

In the existing information transmission process, information transmission between application systems usually needs to pass through forwarding processing of multiple application systems to reach a target application system, and a next application system connected to each application system is fixed when each application system transmits information with the same target address. Each application system in the network may have a large load difference in different time periods, and the loads of the application systems in the network in the same time period may also have a large difference, and the application systems all select the next-level application system fixedly connected to transmit information, which results in a low utilization rate of the application systems in the network, and consequently, a low utilization rate of the servers in the network.

Disclosure of Invention

In view of this, embodiments of the present invention provide an information transmission method based on an echo state network, so as to solve the problem that an application system is connected to a fixed next-level application system to transmit information, which results in a low utilization rate of the application system in the network, and thus a low utilization rate of a server.

The embodiment of the invention also provides an information transmission device based on the echo state network, which is used for ensuring the actual realization and application of the method.

In order to achieve the above purpose, the embodiments of the present invention provide the following technical solutions:

an information transmission method based on an echo state network comprises the following steps:

when an allocation server of a first application system receives task information, determining a first server cluster corresponding to the first application system at present, determining a first target server in the first server cluster, and sending the task information to the first target server;

when result information corresponding to the task information sent by the first target server is received, a target address of the result information is obtained;

determining a second application system corresponding to the target address, wherein the second application system is selected from application system clusters corresponding to a plurality of preset target addresses, the application system clusters comprise a plurality of third application systems, the plurality of third application systems and the plurality of preset target addresses are in a one-to-one correspondence relationship, the preset target address is a server address corresponding to output information generated in the running process of the first application system, the third application system corresponding to each preset target address is determined from a plurality of fourth application systems corresponding to the preset target address based on a trained first echo state network model in the current time period, and the fourth application system is used for transmitting the output information to the preset target address;

and connecting the second application system and sending the result information to the second application system.

Optionally, the method for determining a third application system in a plurality of fourth application systems corresponding to the preset target address based on the trained first echo state network model includes:

when a current time period is triggered, determining all the fourth application systems corresponding to the preset target address;

acquiring system parameters of each fourth application system; the system parameters comprise the running time period, the running efficiency, the load parameters, the number of servers, the processing performance of each server and the connection parameters between the first application system and the fourth application system of the fourth application system;

constructing a first input vector according to the system parameters of all the fourth application systems and the current time;

loading the first input vector to an input node of the first echo state network model;

after idling for a first preset time, acquiring a first output vector from an output node of the first echo state network model;

and determining a fourth application system corresponding to the first output vector, and taking the fourth application system corresponding to the first output vector as a third application system corresponding to the preset target address.

Optionally, the method described above, wherein the training process of the first echo state network model includes:

loading preset sample data to an input node and an output node of the first echo state network model in sequence; the sample data is determined according to historical load data and system parameters of all fourth application systems corresponding to each preset target address in each time period;

after idling for a second preset time, updating and recording the state of the reserve pool of the first echo state network model;

and determining the output connection weight of the first echo state network model based on a linear regression algorithm.

In the foregoing method, optionally, the determining a first server cluster currently corresponding to the first application system includes:

and in the process of regularly adjusting the first server cluster, taking the first server cluster determined by the time point of adjustment before the time point of receiving the task information as the first server cluster corresponding to the first application system currently.

The method described above, optionally, the process of determining the first server cluster at the adjustment time point includes:

when the time point is triggered and adjusted, acquiring cluster parameters of an initial first server cluster; the initial first server cluster is a first server cluster corresponding to the first application system before the adjusting time point; the cluster parameters comprise the number of the servers of the initial first server cluster, the central processing unit parameter of each server, the memory parameter, the network transmission rate and the bandwidth;

constructing a second input vector according to the cluster parameters of the initial first server cluster and the current time;

loading the second input vector to an input node of a trained second echo state network model;

after idling for a third preset time, acquiring a second output vector from an output node of the second echo state network model;

determining the number of target servers corresponding to the second output vector;

and comparing the number of the servers of the initial first server cluster with the number of the target servers, and determining a first server cluster corresponding to the first application system according to a comparison result.

Optionally, the determining, according to the comparison result, the first server cluster corresponding to the first application system includes:

when the number of the servers of the initial first server cluster is larger than the number of the target servers, determining a first numerical value, wherein the first numerical value is a difference value between the number of the servers of the initial first server cluster and the number of the target servers, and determining each first server corresponding to the first numerical value in the initial first server cluster;

sending a port closing instruction to each first server to trigger the first servers to stop serving the first application systems;

and taking the servers in the initial first server cluster except all the first servers as the first server cluster corresponding to the first application system.

when the number of the servers of the initial first server cluster is smaller than the number of the target servers, determining a second value, wherein the second value is a difference value between the number of the target servers and the number of the servers of the initial first server cluster, and determining each second server corresponding to the second value in other servers deployed in a network;

sending a port opening instruction to each second server to trigger the second servers to serve the first application system;

and taking all the servers in the initial first server cluster and all the second servers as the first server cluster corresponding to the first application system.

Optionally, the method for determining the second application system corresponding to the target address includes:

comparing the target address with all the preset target addresses, and determining a first preset target address corresponding to the target address;

and acquiring a third application system corresponding to the first preset target address from the application system cluster, and taking the third application system corresponding to the first preset target address as a second application system corresponding to the target address.

An information transfer apparatus based on an echo state network, comprising:

the first determining unit is used for determining a first server cluster corresponding to a first application system at present when an allocation server of the first application system receives task information, determining a first target server in the first server cluster, and sending the task information to the first target server;

the first obtaining unit is used for obtaining a target address of result information when the result information corresponding to the task information sent by the first target server is received;

a second determining unit, configured to determine a second application system corresponding to the target address, where the second application system is obtained by selecting from application system clusters corresponding to multiple preset target addresses, the application system cluster comprises a plurality of third application systems, the third application systems and the preset target addresses are in one-to-one correspondence relationship, the preset target address is a server address corresponding to output information generated by the first application system in the operation process, the third application system corresponding to each preset target address is an application system determined in a plurality of fourth application systems corresponding to the preset target address based on a trained first echo state network model in the current time period, the fourth application system is an application system for transmitting the output information to the preset target address;

and the sending unit is used for connecting the second application system and sending the result information to the second application system.

The above apparatus, optionally, further comprises:

a third determining unit, configured to determine, when a current time period is triggered, all the fourth application systems corresponding to the preset target address;

a second obtaining unit, configured to obtain a system parameter of each of the fourth application systems; the system parameters comprise the running time period, the running efficiency, the load parameters, the number of servers, the processing performance of each server and the connection parameters between the first application system and the fourth application system of the fourth application system;

the first construction unit is used for constructing a first input vector according to the system parameters of all the fourth application systems and the current time;

a first loading unit, configured to load the first input vector to an input node of the first echo state network model;

a third obtaining unit, configured to obtain a first output vector from an output node of the first echo state network model after idling for a first preset duration;

and a fourth determining unit, configured to determine a fourth application system corresponding to the first output vector, and use the fourth application system corresponding to the first output vector as a third application system corresponding to the preset target address.

The above apparatus, optionally, further comprises:

the second loading unit is used for sequentially loading preset sample data to the input node and the output node of the first echo state network model; the sample data is determined according to historical load data and system parameters of all fourth application systems corresponding to each preset target address in each time period;

the recording unit is used for updating and recording the state of the reserve pool of the first echo state network model after idling for a second preset time length;

and the fifth determining unit is used for determining the output connection weight of the first echo state network model based on a linear regression algorithm.

The above apparatus, optionally, the first determining unit includes:

and the first determining subunit is configured to, in the process of periodically adjusting the first server cluster, use the first server cluster determined at the adjustment time point before the time point at which the task information is received as the first server cluster currently corresponding to the first application system.

The above apparatus, optionally, further comprises:

a fourth obtaining unit, configured to obtain a cluster parameter of the initial first server cluster when the adjustment time point is triggered; the initial first server cluster is a first server cluster corresponding to the first application system before the adjusting time point; the cluster parameters comprise the number of the servers of the initial first server cluster, the central processing unit parameter of each server, the memory parameter, the network transmission rate and the bandwidth;

the second construction unit is used for constructing a second input vector according to the cluster parameters of the initial first server cluster and the current time;

a third loading unit, configured to load the second input vector to an input node of a trained second echo state network model;

a fifth obtaining unit, configured to obtain a second output vector from an output node of the second echo state network model after idling for a third preset time period;

a sixth determining unit, configured to determine the number of target servers corresponding to the second output vector;

a seventh determining unit, configured to compare the number of servers of the initial first server cluster with the number of target servers, and determine, according to a comparison result, a first server cluster corresponding to the first application system.

The above apparatus, optionally, the seventh determining unit includes:

a second determining subunit, configured to determine a first value when the number of servers of the initial first server cluster is greater than the target number of servers, where the first value is a difference between the number of servers of the initial first server cluster and the target number of servers, and determine, in the initial first server cluster, each first server corresponding to the first value;

the first triggering subunit is configured to send a port closing instruction to each first server to trigger the first server to stop serving the first application system;

a third determining subunit, configured to use the servers in the initial first server cluster, except all the first servers, as the first server cluster corresponding to the first application system.

The above apparatus, optionally, the seventh determining unit includes:

a fourth determining subunit, configured to determine a second value when the number of servers of the initial first server cluster is smaller than the number of target servers, where the second value is a difference between the number of target servers and the number of servers of the initial first server cluster, and determine, in other servers deployed in the network, each second server corresponding to the second value;

the second triggering subunit is configured to send a port opening instruction to each second server, so as to trigger the second server to serve the first application system;

a fifth determining subunit, configured to use all servers in the initial first server cluster and all the second servers as the first server cluster corresponding to the first application system.

The above apparatus, optionally, the second determining unit includes:

a sixth determining subunit, configured to compare the target address with all the preset target addresses, and determine a first preset target address corresponding to the target address;

a seventh determining subunit, configured to obtain, in the application system cluster, a third application system corresponding to the first preset target address, and use the third application system corresponding to the first preset target address as the second application system corresponding to the target address.

Based on the information transmission method based on the echo state network provided by the embodiment of the invention, when the allocation server of the first application system receives the task information to be processed by the application system, the task information is sent to the first target server in the first service cluster corresponding to the first application system currently for processing, and when the corresponding result information is obtained, the target address of the result information is obtained. And selecting a second application system corresponding to the target address from application system clusters corresponding to a plurality of preset target addresses, connecting the second application system, and sending the result information to the second application system. In the method provided in the embodiment of the present invention, the preset target address corresponds to a third application system in the application system cluster one to one, the third application system is an application system determined in a plurality of fourth application systems corresponding to the preset target address based on a trained echo state network model in a current time period, and the fourth application system is an application system that can connect the first application system and transmit output information to the preset target address.

By applying the method provided by the embodiment of the invention, when the first application system needs to transmit information to the target address, the second application system determined in the plurality of application systems capable of transmitting the information to the target address based on the echo state network in the current time period can be connected, and the information is transmitted through the second application system. The second application system is an application system which is predicted by the echo state network at regular time, and the echo state network can be trained to predict different second application systems in different time periods according to a rule obtained by pre-training. Therefore, for the same target address, the application system can be connected with the corresponding next-level application system in the current time period in different time periods to transmit information, so that the utilization rate of the application system in the network is improved, and the utilization rate of the server in the network is further improved.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.

Fig. 1 is a flowchart of a method for transmitting information based on an echo state network according to an embodiment of the present invention;

fig. 2 is a schematic diagram of an echo state network according to an embodiment of the present invention;

fig. 3 is a flowchart of another method of an information transmission method based on an echo state network according to an embodiment of the present invention;

fig. 4 is a schematic structural diagram of an information transmission apparatus based on an echo state network according to an embodiment of the present invention;

fig. 5 is a schematic structural diagram of an information transmission apparatus based on an echo state network according to an embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

In this application, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.

Due to the traffic characteristics of the service network, the loads of the application systems in the network often show temporal differences, that is, a high load state is achieved for a certain period of time, and a low load state is achieved for a certain period of time, and the load characteristics of different application systems in the same period of time are different.

When an application system needs to send information to a target application system, and needs to be forwarded by other application systems, the next-level application system connected to the application system is fixed, and in a certain period of time, the next-level application system connected to the application system may be in a high-load state, and the application system capable of transmitting the information in the network is in a low-load state. Therefore, the application system is connected to the fixed next-level system to transmit information, which results in a low utilization rate of the application system in the network and is not favorable for reasonable utilization of resources.

Therefore, an embodiment of the present invention provides an information transmission method based on an echo state network, and when information to be transmitted is obtained, an application system determined based on the echo state network in a current time period is connected to perform information transmission, so that the application system can be connected to a next-level application system corresponding to the time period in different time periods, so as to improve a utilization rate of the application system in the network.

An embodiment of the present invention provides an information transmission method based on an echo state network, where the method is applicable to an application system in a service network, and an execution subject of the method may be a server of the application system, and a flowchart of the method is shown in fig. 1, and includes:

s101: when an allocation server of a first application system receives task information, determining a first server cluster corresponding to the first application system at present, determining a first target server in the first server cluster, and sending the task information to the first target server;

in the method provided by the embodiment of the invention, the application system is provided with an allocation server and a first server cluster for processing tasks. The distribution server can receive and send information to the outside, and when the distribution server receives task information sent by other application systems in the network, the first server cluster corresponding to the distribution server is determined. The task information may be task information that requires the first application system to perform response processing, or task information that requires the first application system to perform forwarding processing. The first server cluster may be a server cluster pre-configured to estimate a load of the first application system, or may automatically adjust the configured server cluster at regular time. The first target server may be determined in the first server cluster according to a load balancing algorithm, the task information may be sent to the first target server, the first target server may perform processing, and when the first target server processes the task to obtain corresponding result information, the result information may be returned to the distribution server.

S102: when result information corresponding to the task information sent by the first target server is received, a target address of the result information is obtained;

in the method provided by the embodiment of the present invention, when the distribution server receives the result information sent by the first target server, the result information may be analyzed to obtain the target address of the result information, that is, the address of the distribution server of the target application system to which the result information is to be sent.

S103: determining a second application system corresponding to the target address, wherein the second application system is selected from application system clusters corresponding to a plurality of preset target addresses, the application system clusters comprise a plurality of third application systems, the plurality of third application systems and the plurality of preset target addresses are in a one-to-one correspondence relationship, the preset target address is a server address corresponding to output information generated in the running process of the first application system, the third application system corresponding to each preset target address is determined from a plurality of fourth application systems corresponding to the preset target address based on a trained first echo state network model in the current time period, and the fourth application system is used for transmitting the output information to the preset target address;

in the method provided in the embodiment of the present invention, for a server address, that is, a preset target address, of each destination application system to which output information generated in an operation process of a first application system is to be transmitted, a third application system may be determined at regular time based on a trained echo state network model in a plurality of fourth application systems corresponding to the preset target address, that is, a next-stage application system to be connected in a transmission process when output information is to be transmitted to the preset target address in a current time period is determined. The second application system corresponding to the target address can be selected from the application system clusters corresponding to a plurality of preset target addresses, namely, the next-level application system to which the first application system is connected is sent to the target address in the transmission process of the result information. For the training of the echo state network model, the training can be carried out according to actual requirements, and the training can predict a relatively idle application system in the current time period according to historical load data. Or training the application system which can predict the condition satisfied in the current time period according to the preset condition.

S104: and connecting the second application system and sending the result information to the second application system.

In the method provided by the embodiment of the invention, after the second application system corresponding to the target address is determined, the address of the allocation server corresponding to the second application system can be obtained, the second application system is connected according to the address of the allocation server of the second application system, and the result is sent to the second application system, so that the second application system directly or indirectly transmits the result information to the target address.

Based on the method provided by the embodiment of the invention, when information needs to be transmitted to the target address, the second application system corresponding to the target address can be selected from the application system clusters corresponding to the preset target addresses for information transmission. In the application system cluster corresponding to the preset target addresses, the application system corresponding to each preset target address is an application system determined in a plurality of application systems capable of transmitting information to the preset target addresses corresponding to the preset target addresses based on the trained echo state network model in the current time period. Each preset target address can correspond to different application systems in different time periods. Therefore, when information is transmitted to the target address, the next-level application system corresponding to the target address in the current time period can be connected, and the fixed next-level application system is not connected at any time, so that the utilization rate of the application system in the network can be improved, the utilization rate of the server in the network is improved, and the resource utilization in the network is balanced.

Further, an embodiment of the present invention provides another information transmission method based on an echo state network, where on the basis of the method provided in the above embodiment, in the method provided in the embodiment of the present invention, the determining the second application system corresponding to the target address includes:

in the method provided by the embodiment of the present invention, the target address of the information to be currently transmitted may be compared with each preset target address, and the preset target address identical to the target address may be used as the first preset target address corresponding to the target address.

In the method provided by the embodiment of the present invention, the application system cluster includes the third application system corresponding to each preset target address, and the third application system corresponding to the first preset target address in the application system cluster, that is, the preset target address identical to the target address, may be used as the second application system corresponding to the target address.

Based on the method provided by the embodiment of the invention, the next-level application system corresponding to the target address can be directly obtained from the predetermined application system cluster, and the next-level application system to be connected can be conveniently determined on the basis of flexibly determining the next-level application system connected in each time period without real-time prediction according to the target address.

In order to better explain the method provided by the embodiment of the present invention, the echo state network model mentioned in the embodiment of the present invention is briefly described next.

The echo state network is a recurrent neural network, and is schematically shown in FIG. 2, and the network structure is an input layer N in turn_UStorage pool N_RAnd an output layer N_Y. Wherein the input layer N_UFor K input nodes, output layer N_YFor L output nodes, and a reserve pool N_RThere are N internal nodes, i.e., N sparsely connected internal neurons. At time n, the input to the echo state network is u (n) ═ u₁(n),u₂(n),...,u_K(n)]^TThe state of the reserve pool is x (n) ═ x₁(n),x₂(n),...,x_N(n)]^TThe output is y (n) ═ y₁(n),y₂(n),...,y_L(n)]^T。

W_inIs an input layer N_UIs connected to a reserve pool N_RThe input of (2) is connected to the weight. W is reserve tank N_RThe internal connection weights connected to the pool state at the next time may be used to retain information left over at previous times. W_outIs a reserve pool N_RConnected to the output layer N_YOutput of (2) is connected to the weight, W_backOutput layer N for the previous moment_YReserve pool N connected to the next moment_RAnd (4) outputting feedback weight values.

After inputting u (N) at each moment, reserve pool N_RThe status is updated, and the status of the reserve pool can be updated in a way of x (n +1) ═ f [ W [_inu(n)+Wx(n)]In the case of combining output feedback, the update equation for the pool state may be x (n +1) ═ f [ W [_inu(n)+Wx(n)+W_backy(n)]. f (-) is a reservoir internal neuron activation function, and a hyperbolic tangent function can be used. Echo state netThe output state equation of the complex may be

f_out(. cndot.) is an output layer activation function,

either the bias term of the output or noise.

The initialization content of the echo state network model comprises the size of a fixed reserve pool, which is usually the node number of internal nodes of the fixed reserve pool, the weight spectrum radius of internal connection weights, the input unit scale of the reserve pool, the sparsity degree of the reserve pool and other reserve pool parameters. Then randomly generating a connection matrix, setting a scaling factor, scaling the generated connection matrix to make the radius of a weight spectrum of the connection matrix smaller than one, and randomly generating an input connection weight and an output feedback weight.

To better explain the method provided by the embodiment shown in fig. 1, an embodiment of the present invention provides another information transmission method based on an echo state network, and based on the method shown in fig. 1, in the method provided by the embodiment of the present invention, a training process of the first echo state network model includes:

in the method provided by the embodiment of the invention, an echo state network model can be initialized according to actual sample size and time memory requirements, and parameters such as the size of a reserve pool, an input connection weight, an internal connection weight, an output feedback weight and the like are determined.

All preset target addresses of information transmission related to the application system, namely server addresses of a target application system to which all output information possibly generated by the first application system in the actual operation process is transmitted can be determined according to the service function of the first application system, and the server addresses are used as the preset target addresses. According to the deployment of the application systems in the network, for each preset target address, all the next-level application systems, i.e. the fourth application systems, which can be used for indirectly or directly transmitting information to the preset target address can be determined. And determining a training data set of the echo state network model according to historical system parameters and load data of all next-stage application systems corresponding to each preset target address in each time period. The sample input vector can be constructed according to each preset target address and historical system parameters of all next-level application systems in each time period, the historical system parameters comprise time corresponding to the system parameters, operation time periods, operation efficiency, server parameters and the like of the system, and a relatively idle application system is determined in each next-level application system to be used as expected output according to the historical parameters and load conditions of each next-level application system in each time period.

in the method provided by the embodiment of the invention, after sample data is sequentially input, idling for a certain time period is started, and the idling is actually the state of an initialized reserve pool, so that the noise influence of an initially input sequence on the state of the reserve pool is reduced. When idling, the reserve pool state begins to be recorded. According to the update equation of the reserve pool state, the state of the reserve pool at each moment corresponding to the input vector at each moment can be recorded, and the state of the reserve pool at the next moment can be updated.

In the method provided by the embodiment of the invention, after the reserve pool state at each moment is obtained, the prediction output of the echo state network at each moment can be determined according to the output state equation of the echo state network. The training parameter of the echo state network is the output connection weight of the echo state network, and because the state of the reserve pool and the predicted output of the echo state network are in a linear relationship, the output connection weight of the echo state network can be calculated in a manner that the predicted output is close to the expected output, and the calculation can be specifically realized by a linear regression algorithm, for example, fitting can be performed based on a least square method.

Based on the method provided by the embodiment of the invention, the sample data of the echo state network can be determined according to the historical system parameters and the load parameters of each fourth application system corresponding to each preset target address, the echo state network is trained, the output connection weight is determined, so that the echo state network can regularly predict the third application system corresponding to each preset target address according to the expected output corresponding to the system parameters of each time period memorized in the training process, in the running process of the first application system, that is, the third application system corresponding to each preset target address can be determined, that is, the information can be directly or indirectly transmitted to the relatively idle application system of the preset target address in each time period, so that when the first application system transmits the information to the outside, the relatively idle application system in the current time period can be connected to transmit the information, and the utilization rate of the application systems in the network is further improved, and is beneficial to balancing the load of each application system in the network.

To better illustrate the method provided in the embodiment shown in fig. 1, an embodiment of the present invention provides another information transmission method based on an echo state network, and based on the method provided in the above embodiment, in the method provided in the embodiment of the present invention, the process of determining a third application system in a plurality of fourth application systems corresponding to the preset target address based on the trained first echo state network model in step S103 includes:

in the method provided by the embodiment of the invention, the time period can be preset according to the actual operation condition of each application system in the network, and the time when the load of the application system in the network is relatively stable can be taken as one time period. When the current time period is triggered, the fourth application systems corresponding to the pre-stored preset target addresses are acquired, that is, the first application system can be connected with the fourth application systems, and the output information of the first application system can be indirectly or directly transmitted to the next-level application system of the preset target address.

in the method provided by the embodiment of the present invention, an information obtaining instruction may be sent to the server of each fourth application system corresponding to each preset destination address, so as to obtain the real-time system parameter of each fourth application system. The system parameters comprise the running time period, the running efficiency, the load parameters, the number of servers, the processing performance of each server and the connection parameters between the first application system and the fourth application system of the system. The operation time period is the working time period when the application system operates normally and can respond to the request. The operation efficiency may be a central processing unit parameter of the application system, the load parameter may be a concurrency amount of the system, and a connection parameter between the systems may be a bandwidth or a traffic amount between the two systems.

in the method provided by the embodiment of the present invention, after the system parameter of each fourth application system is obtained, an input vector of the echo state network model may be constructed according to the system parameter corresponding to each fourth application system and the current system time.

in the method provided by the embodiment of the present invention, the input vector constructed based on the system parameters of each current fourth application system is loaded to the input node of the trained echo state network model, so that the echo state network model can learn autonomously based on the pre-trained result, and a relatively idle system in the current time period is determined in each corresponding fourth application system in the input vector.

in the method provided by the embodiment of the invention, after the echo state network model loads the input vector, the output vector of the echo state network model can be obtained from the output node after idling for a certain time. Furthermore, the first output vector can be used as feedback, so that the echo state network can learn autonomously according to the input and the output.

In the method provided by the embodiment of the present invention, the fourth application system in the first output vector, that is, the fourth application system determined in each fourth application system by the echo state network through autonomous learning and relatively idle in the current time period, may be determined according to the identifier in the first output vector. And taking the fourth application system corresponding to the first output vector as the second application system corresponding to the preset target address, so that the first application system can be connected with the second application system for information transmission when the output information needs to be sent to the preset target address in the current time period.

According to the method provided by the embodiment of the invention, the third application system corresponding to each preset target address can be predicted at regular time according to the trained echo state network model, the input vector can be constructed according to the real-time system parameters of each fourth application system at the prediction time point, so that the echo state network is predicted based on the real-time system parameters, the prediction result is closer to the actual running state of the application systems in the network, and secondly, the multiple parameters of the system are considered in the process of predicting by using the echo state network, so that the method is beneficial to predicting the relatively idle application system by combining the overall capacity condition of the system, and the running time period of the system is considered, so that the information can be prevented from being sent to the currently inoperative application system. Furthermore, the echo state network model can automatically learn and correct the model parameters based on input and output results each time, so that the prediction result is more accurate.

Further, an embodiment of the present invention provides another information transmission method based on an echo state network, where on the basis of the method shown in fig. 1, in the method provided in the embodiment of the present invention, the determining, in step S101, a first server cluster currently corresponding to the first application system includes:

In the method provided by the embodiment of the present invention, the first server cluster configured by the first application system may be adjusted at regular time, the preset period may be determined according to the temporal difference of the load of the first application system, and the time when the load of the first application system is relatively stable may be taken as a period. The server cluster can be adjusted based on the neural network, and the load of each time can be evaluated in advance manually, so that the server cluster is configured in advance for each time period. The first server cluster determined by the current time corresponding to the previous adjustment time point may be used as the first server cluster currently corresponding to the first application system when the task information is received.

It should be noted that the time period of the server cluster of the first application system is adjusted and the time period of the third application system corresponding to the preset target address is determined based on the echo state network model, which may be the same period or different periods, and the determination of the specific time period does not affect the implementation function of the method provided by the embodiment of the present invention.

Based on the method provided by the embodiment of the invention, the server cluster corresponding to the first application system can be adjusted in a timing manner, the server cluster corresponding to the first application system in each time period can be matched with the load of the time period, when the first application system receives the task information, the server for processing the task information can be determined in the currently matched first server cluster, the utilization rate of the server is improved, and the server resources are saved.

To better illustrate the method provided by the foregoing embodiment, with reference to fig. 3, an embodiment of the present invention provides another information transmission method based on an echo state network, where, on the basis of the method provided by the foregoing embodiment, in the method provided by the embodiment of the present invention, the process of determining the first server cluster at the adjustment time point includes:

s201: when the time point is triggered and adjusted, acquiring cluster parameters of an initial first server cluster; the initial first server cluster is a first server cluster corresponding to the first application system before the adjusting time point; the cluster parameters comprise the number of the servers of the initial first server cluster, the central processing unit parameter of each server, the memory parameter, the network transmission rate and the bandwidth;

in the method provided by the embodiment of the present invention, another echo state network model may be pre-constructed, so as to predict the number of servers correspondingly configured to the first application system in a preset period, so as to adjust the server cluster correspondingly configured to the application system. When a preset adjustment time point is triggered, an information acquisition instruction may be sent to each server in a current initial first server cluster of a first application system to acquire parameters of the initial first server cluster, where the initial first server cluster is an initial server cluster in the adjustment process of the server cluster at this time, that is, a first server cluster determined at a previous adjustment time point, and does not only refer to a server cluster initially configured manually by the system. The cluster parameter of the initial first server may be the number of servers of the initial first server cluster, a central processing unit parameter, a memory parameter, a network transmission rate, a bandwidth, and other parameters of each server.

S202: constructing a second input vector according to the cluster parameters of the initial first server cluster and the current time;

in the method provided by the embodiment of the present invention, an input vector of the second echo state network model may be constructed according to the cluster parameter of the initial first server cluster and the current time.

S203: loading the second input vector to an input node of a trained second echo state network model;

in the method provided by the embodiment of the present invention, the second input vector may be loaded to the input node of the trained second echo state network model.

In the method provided in the embodiment of the present invention, the construction of the second echo state network may refer to the foregoing description about the echo state network, and is not described herein again. The training principle of the second echo state network can be described with reference to the training process of the first echo state network in the foregoing embodiments, and will not be described herein again. In the method provided by the embodiment of the invention, the training data set can be determined based on the historical load data of the first application system in each time period and the processing capacity of the server cluster, and the number of the servers matched with the load of each time period is used as expected output.

S204: after idling for a third preset time, acquiring a second output vector from an output node of the second echo state network model;

in the method provided by the embodiment of the invention, after the input vector is loaded to the input node of the echo state network model and idles for a certain time, the output vector of the echo state network model is obtained from the output node.

S205: determining the number of target servers corresponding to the second output vector;

in the method provided by the embodiment of the present invention, the number of servers corresponding to the echo state network model in the time period determined based on the autonomous learning may be obtained from the second output vector.

S206: and comparing the number of the servers of the initial first server cluster with the number of the target servers, and determining a first server cluster corresponding to the first application system according to a comparison result.

In the method provided by the embodiment of the present invention, the number of servers of the initial first server cluster may be compared with the number of target servers predicted by the echo state network model, the server cluster corresponding to the first application system is adjusted according to the comparison result, and the number of servers of the server cluster is adjusted to the number of target servers.

Based on the method provided by the embodiment of the invention, the number of the servers of the server cluster corresponding to the first application system in the next period of time can be predicted based on the echo state network at regular time, and the server cluster corresponding to the first application system is adjusted according to the predicted number of the servers. In the method provided by the embodiment of the invention, the number of the servers of the first application system, which correspond to each time period, can be determined based on the autonomous learning of the echo state network for prediction, and less servers can be configured in the time period with smaller load, and more servers can be configured in the time period with larger load, so that the application system can configure the server clusters matched with the load under different load conditions, and the utilization rate of the servers is further improved.

Further, on the basis of the foregoing embodiment, in the method provided in the embodiment of the present invention, the determining, according to a comparison result, a first server cluster corresponding to the first application system includes:

in the method provided by the embodiment of the invention, a plurality of application systems corresponding to each area can be deployed on the server in each area in the network, so that the same server can serve the plurality of application systems in the area. When the number of the servers of the initial first server cluster of the first application system is greater than the number of the target servers, the servers corresponding to the difference number may be determined in the initial first server cluster according to the difference value of the numbers, and if the number of the servers of the initial server cluster is ten and the number of the target servers is six, four first servers may be determined in the initial first server cluster according to a preset priority.

in the method provided by the embodiment of the present invention, a port closing instruction may be sent to each first server, and the first server may be triggered to close a service port corresponding to the first application system, stop serving the first application system, and release the first application system as an idle server into the area. If the first server still has the task which is not processed, the first server can be related to the server port after the first server processes the current task.

In the method provided by the embodiment of the present invention, all the first servers may be removed from the server cluster corresponding to the first application system, and the server cluster from which all the first servers are removed is used as the server cluster corresponding to the first application system.

Based on the method provided by the embodiment of the invention, when the number of the servers of the initial first server cluster is greater than that of the target servers, a certain number of servers in the initial first server cluster can be released from the cluster to become idle servers in the area, so that other application systems corresponding to the area can use the idle servers, and the utilization rate of the servers is further improved.

Further, an embodiment of the present invention provides another method for transmitting information of an echo state network, where on the basis of the foregoing embodiment, in the method provided in the embodiment of the present invention, the determining, according to a comparison result, a first server cluster corresponding to the first application system includes:

in the method provided by the embodiment of the invention, a plurality of application systems corresponding to each area can be deployed on the server in each area in the network, so that the same server can serve the plurality of application systems in the area. When the number of servers of the initial first server cluster is smaller than the number of target servers, a second server corresponding to the difference number may be determined among other servers deployed in the area corresponding to the first application system based on the difference of the numbers. If the number of the servers of the initial first server cluster is ten and the number of the target servers is thirteen, three second servers can be determined among other idle servers deployed in the area according to a preset priority, and the second servers are idle servers which do not provide services for any application system in the area.

in the method provided by the embodiment of the present invention, a port opening instruction may be sent to each second server to trigger the second server to open a service port corresponding to the first application system to serve the first application system.

In the method provided by the embodiment of the present invention, all the second servers may be incorporated into the server cluster corresponding to the first application system, that is, all the servers in all the initial first server clusters and all the second servers are used as the first server clusters corresponding to the first application system in the next time period.

Based on the method provided by the embodiment of the invention, when the number of the servers of the initial first server cluster is less than that of the target servers, other idle servers in the network can be incorporated into the server cluster of the first application system, so that the idle servers in the network can be used by the application system with load demand, and the utilization rate of the servers is further improved.

Corresponding to the information transmission method based on the echo state network shown in fig. 1, an embodiment of the present invention further provides an information transmission apparatus based on the echo state network, which is used for implementing the method shown in fig. 1 specifically, and a schematic structural diagram of the information transmission apparatus is shown in fig. 4, and includes:

a first determining unit 301, configured to determine, when a distribution server of a first application system receives task information, a first server cluster currently corresponding to the first application system, determine a first target server in the first server cluster, and send the task information to the first target server;

a first obtaining unit 302, configured to obtain a target address of result information when the result information corresponding to the task information sent by the first target server is received;

a second determining unit 303, configured to determine a second application system corresponding to the target address, where the second application system is obtained by selecting from application system clusters corresponding to multiple preset target addresses, the application system cluster comprises a plurality of third application systems, the third application systems and the preset target addresses are in one-to-one correspondence relationship, the preset target address is a server address corresponding to output information generated by the first application system in the operation process, the third application system corresponding to each preset target address is an application system determined in a plurality of fourth application systems corresponding to the preset target address based on a trained first echo state network model in the current time period, the fourth application system is an application system for transmitting the output information to the preset target address;

a sending unit 304, configured to connect to the second application system, and send the result information to the second application system.

Based on the device provided by the embodiment of the invention, when information needs to be transmitted to the target address, the second application system corresponding to the target address can be selected from the application system clusters corresponding to a plurality of preset target addresses to perform information transmission. In the application system cluster corresponding to the preset target addresses, the application system corresponding to each preset target address is an application system determined in a plurality of application systems capable of transmitting information to the preset target addresses corresponding to the preset target addresses based on the trained echo state network model in the current time period. Each preset target address can correspond to different application systems in different time periods. Therefore, when information is transmitted to the target address, the next-level application system corresponding to the target address in the current time period can be connected, and the fixed next-level application system is not connected at any time, so that the utilization rate of the application system in the network can be improved, the utilization rate of the server in the network is improved, and the resource utilization in the network is balanced.

Further, an embodiment of the present invention provides another information transmission apparatus based on an echo state network, where on the basis of the apparatus provided in the above embodiment, the apparatus provided in the embodiment of the present invention further includes:

Further, another information transmission apparatus based on an echo state network is provided in an embodiment of the present invention, and on the basis of the apparatus provided in the above embodiment, in the apparatus provided in the embodiment of the present invention, the first determining unit 301 includes:

Further, an embodiment of the present invention provides another information transmission apparatus based on an echo state network, a schematic structural diagram of which is shown in fig. 5, and on the basis of the apparatus shown in fig. 4, the apparatus provided in the embodiment of the present invention further includes:

a fourth obtaining unit 305, configured to obtain a cluster parameter of the initial first server cluster when the adjustment time point is triggered; the initial first server cluster is a first server cluster corresponding to the first application system before the adjusting time point; the cluster parameters comprise the number of the servers of the initial first server cluster, the central processing unit parameter of each server, the memory parameter, the network transmission rate and the bandwidth;

a second constructing unit 306, configured to construct a second input vector according to the cluster parameter of the initial first server cluster and the current time;

a third loading unit 307, configured to load the second input vector to an input node of a trained second echo state network model;

a fifth obtaining unit 308, configured to obtain a second output vector from an output node of the second echo state network model after idling for a third preset time;

a sixth determining unit 309, configured to determine the number of target servers corresponding to the second output vector;

a seventh determining unit 310, configured to compare the number of servers of the initial first server cluster with the number of target servers, and determine, according to a comparison result, a first server cluster corresponding to the first application system.

Further, another information transmission apparatus based on an echo state network is provided in an embodiment of the present invention, and on the basis of the apparatus provided in the above embodiment, in the apparatus provided in the embodiment of the present invention, the seventh determining unit 310 includes:

Further, an embodiment of the present invention provides another information transmission apparatus based on an echo state network, where on the basis of the apparatus provided in the above embodiment, in the apparatus provided in the embodiment of the present invention, the second determining unit 303 includes:

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the system or system embodiments are substantially similar to the method embodiments and therefore are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described system and system embodiments are only illustrative, wherein the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An information transmission method based on an echo state network is characterized by comprising the following steps:

2. The information transmission method according to claim 1, wherein the process of determining a third application system among a plurality of fourth application systems corresponding to the preset target address based on the trained first echo state network model comprises:

3. The information transmission method according to claim 1, wherein the training process of the first echo state network model comprises:

4. The information transmission method according to claim 1, wherein the determining the first server cluster to which the first application system currently corresponds includes:

5. The information transmission method according to claim 4, wherein the process of determining the first server cluster at the adjusted time point comprises:

6. The information transmission method according to claim 5, wherein the determining, according to the comparison result, the first server cluster corresponding to the first application system includes:

7. The information transmission method according to claim 5, wherein the determining, according to the comparison result, the first server cluster corresponding to the first application system includes:

8. The information transmission method according to claim 1, wherein the determining the second application system corresponding to the target address includes:

9. An information transmission apparatus based on an echo state network, comprising:

10. The information transmission apparatus according to claim 9, characterized by further comprising: