KR-20260066285-A - TRAFFIC DISTRIBUTION METHOD USING AI

KR20260066285AKR 20260066285 AKR20260066285 AKR 20260066285AKR-20260066285-A

Abstract

A method for distributing traffic to each of a plurality of servers using artificial intelligence is disclosed. The disclosed traffic distribution method using artificial intelligence includes: a step of collecting traffic information for a plurality of servers performing a preset task; a step of predicting a first traffic amount for each of the servers in a preset first prediction interval using the traffic information and a preset first traffic prediction model; and a step of distributing traffic to each of the plurality of servers according to the first traffic amount.

Inventors

박준성
이준호

Assignees

주식회사 스타랩스

Dates

Publication Date: 20260512
Application Date: 20241104

Claims (10)

A step of collecting traffic information for multiple servers performing preset tasks; A step of predicting a first traffic amount for each of the servers in a preset first prediction interval using the above traffic information and a pre-trained first traffic prediction model; and A step of distributing traffic to each of the plurality of servers according to the first traffic amount. A traffic distribution method using artificial intelligence including
In Article 1, The above traffic information including at least one of a sending IP, a receiving IP, a sending port number, a receiving port number, a protocol, and a packet length Traffic distribution method using artificial intelligence.
In Article 1, The step of distributing traffic for the plurality of servers to each of the servers A step of predicting a second traffic amount for each of the servers in a second prediction period later than the first prediction period, using the above traffic information, the above first traffic amount, and a pre-trained second traffic prediction model; and A step of distributing traffic to each of the plurality of servers using the first and second traffic amounts. A traffic distribution method using artificial intelligence including
In Paragraph 3, The step of distributing traffic for the plurality of servers to each of the servers A step of comparing the average value of the first and second traffic amounts with a preset threshold value; and A step of distributing traffic from servers whose average value is greater than or equal to the threshold to servers whose average value is less than the threshold. A traffic distribution method using artificial intelligence including
In Paragraph 3, The step of distributing traffic for the plurality of servers to each of the servers Distributing traffic to servers where the total sum of the second traffic amount is greater than the total sum of the first traffic amount to servers where the total sum of the second traffic amount is less than the total sum of the first traffic amount Traffic distribution method using artificial intelligence.
In Paragraph 5, Traffic to a server where the second traffic amount is greater than the first traffic amount is Servers where the total sum of the second traffic amount is greater than the total sum of the first traffic amount are distributed preferentially according to the magnitude of the difference between the total sums of the first and second traffic amounts. A server in which the total sum of the second traffic amount is less than the total sum of the first traffic amount Depending on the magnitude of the difference between the total sum of the first and second traffic amounts for servers where the total sum of the second traffic amount is less than the total sum of the first traffic amount, traffic is preferentially distributed to servers where the total sum of the second traffic amount is greater than the total sum of the first traffic amount. Traffic distribution method using artificial intelligence.
In Article 1, The above server is It is a server that processes requested tasks using a deep learning model, and The step of distributing traffic for the plurality of servers to each of the servers Traffic for servers where the total sum of the first traffic amount is greater than or equal to a preset first threshold is distributed to servers where the total sum of the first traffic amount is less than the first threshold. A server in which the total sum of the first traffic amount is less than the first threshold A server with available GPU resources exceeding a preset second threshold Traffic distribution method using artificial intelligence.
In Article 1, The above plurality of servers First servers that process requested tasks using a first deep learning model; and It includes second servers that process requested tasks using a second deep learning model that is lightweighted from the first deep learning model, and The step of distributing traffic for the plurality of servers to each of the servers Among the first servers, traffic for the server whose total sum of the first traffic amount is greater than or equal to a preset third threshold is distributed to the second servers. Traffic distribution method using artificial intelligence.
In Paragraph 8, The step of distributing traffic for the plurality of servers to each of the servers Traffic for servers where the total sum of the first traffic amount is greater than or equal to the third threshold is distributed among the second servers to second servers where the total sum of the first traffic amount is less than a preset fourth threshold. The above fourth threshold is, A critical amount greater than the above third critical amount Traffic distribution method using artificial intelligence.
In Article 9, The above fourth threshold amount Determined according to the lightweighting level of the second deep learning model above Traffic distribution method using artificial intelligence.

Description

Traffic Distribution Method Using Artificial Intelligence The present invention relates to a method for distributing traffic for a plurality of servers to each of the servers using artificial intelligence. Companies deploy multiple servers to effectively handle massive amounts of traffic. Traffic to these servers is then distributed to each server through various load balancing algorithms. Load balancing algorithms include round-robin, weighted round-robin, IP hash, and least connection methods. Round Robin is a method that distributes traffic to each server sequentially. Weighted Round Robin is a method that distributes traffic to servers based on weights assigned to each server. IP Hash is a method that distributes traffic by mapping a client's IP address to a specific server. Least Connection is a method that distributes traffic to the server with the least amount of traffic assigned at the time of distribution. However, these load balancing algorithms do not consider the traffic throughput of each server, or even if they do, they only consider the traffic throughput at the current time, so there is a problem in that it is difficult to provide stable service by preemptively distributing traffic in situations where a surge in traffic is expected. Relevant prior art includes Korean Published Patents No. 2023-0109117, No. 2022-0096453, No. 2022-0005064, and Korean Registered Patent No. 10-2427171. FIG. 1 is a diagram illustrating a traffic distribution system using artificial intelligence according to an embodiment of the present invention. FIG. 2 is a diagram illustrating a traffic distribution method using artificial intelligence according to an embodiment of the present invention. FIG. 3 is a diagram illustrating a traffic distribution method using artificial intelligence according to another embodiment of the present invention. The present invention is susceptible to various modifications and may have various embodiments; specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the invention to specific embodiments, and it should be understood that the invention includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention. Similar reference numerals have been used for similar components in the description of each drawing. The present invention relates to a method for distributing traffic for a plurality of servers to each of the servers using artificial intelligence. One embodiment of the present invention receives current traffic information for a plurality of servers and uses a deep learning model that predicts the amount of traffic for each server in a preset prediction interval, and distributes the traffic for the plurality of servers to each server according to the predicted amount of traffic. Accordingly, according to one embodiment of the present invention, traffic is preemptively distributed before it is concentrated on a specific server, thereby enabling the provision of a more stable and faster service. Hereinafter, embodiments according to the present invention will be described in detail with reference to the attached drawings. FIG. 1 is a diagram illustrating a traffic distribution system using artificial intelligence according to an embodiment of the present invention. Referring to FIG. 1, a traffic distribution system according to one embodiment of the present invention includes a plurality of terminals (110), a plurality of servers (120), and a traffic distribution device (130). Multiple servers (120) process tasks requested through multiple terminals (110), and as an example, the requested tasks can be processed using a deep learning model. Traffic generated according to the task request of the terminals (110) passes through a traffic distribution device (130) and is distributed to each server through the traffic distribution device (130). The traffic distribution device (130) may include a memory (131), a processor (132) and a router (133) electrically connected to the memory (131). The processor (132) collects traffic information for a plurality of servers (120) that perform a preset task. As described above, traffic between the server (120) and the terminal (110) passes through the traffic distribution device (130), and the processor (132) can collect traffic information from the traffic between the server (120) and the terminal (110). Additionally, the processor (132) can collect traffic information in a preset collection section, and the traffic information may include, as an example, at least one of a transmitting IP, a receiving IP, a transmitting port number, a receiving port number, a protocol, and a packet length. And the processor (132) predicts the amount of traffic for each server using collected traffic information and a pre-trained traffic prediction model. The traffic prediction model receives traffic information collected in the cu