Network twin multi-sink transmission method and system for multi-cloud native application

Abstract

Disclosed in the present invention is a network twin multi-sink transmission method and system for a multi-cloud native application, and the present invention provides a solution for fully utilizing multi-sink network resources and improving user communication experience. First, a network twin service is introduced to serve as a communication agent of all cloud native applications of a user, and traditional end-to-end one segment of communication is divided into three segments of communication to realize decoupling of a cloud native application client and server; second, the network twin service establishes an all-weather normally-open communication connection using a multi-sink network, and data of the cloud native applications of the user is transmitted by using the connection under the scheduling of network twinning; and finally, multiple data streams of the cloud native applications are multiplexed into the normally-open connection established by the network twin service by using QUIC streaming technology for communication transmission. Compared with existing application end-to-end communication transmission methods, according to the present invention, a multi-sink network can be fully utilized to realize larger-bandwidth and more stable communication, and blockage caused by a traditional multi-path communication protocol to multiple applications is also avoided.

Description

A network twin multi-host transmission method and system for multi-cloud native applications Technical Field

[0001] The present invention belongs to the field of computer network transmission protocol technology, relates to transport layer protocol theory and efficient multi-host network transmission solution design, and specifically relates to a network twin multi-host transmission method and system for multi-cloud native applications. Background Art

[0002] With the advancement of various wireless communication technologies, multi-homed devices, capable of establishing connections simultaneously through multiple network interfaces, are becoming increasingly popular. Users of multi-homed devices should enjoy the many benefits of multi-homed networks, including greater aggregate bandwidth and more stable communication quality, thereby further satisfying the growing demand for high-bandwidth, low-latency, and reliable communication services. However, because most traditional Internet applications do not support multipath outbound protocols, users with multi-homed devices cannot fully utilize the network resources provided by multi-homed networks, resulting in resource waste. Cloud-native applications, which are decoupled from underlying hardware devices, are becoming increasingly popular. Their unique microservices architecture and high scalability enable faster updates than traditional applications. To meet the growing popularity of cloud-native applications, it is crucial to design a transmission method for cloud-native applications that fully utilizes multi-homed network resources.

[0003] After searching the existing literature, it was found that in order to avoid the head-of-line blocking problem and improve the communication quality to a certain extent, A. Langley et al. and J. Iyengar et al. published articles entitled "The QUIC transport protocol: Design and Internet-scale deployment" and "QUIC: A UDP-based multiplexed and secure transport" in 2017 and 2021, respectively. These two articles proposed and standardized a new transport protocol implemented in user space. The Quick UDP Internet Connection (QUIC) protocol developed the concept of stream and implemented stream-level scheduling of transmitted data, thereby avoiding the head-of-line blocking problem to a great extent. In 2021, K. Hou et al. published an article entitled "Accelerate and secure serverless networks with QUIC" to prove that the use of the QUIC communication protocol can achieve better transmission performance under certain conditions. However, the QUIC protocol is only a single-path communication transmission protocol and does not support establishing connections using multiple network interfaces at the same time.

[0004] In order to make full use of the network resources provided by multi-homed networks, a search revealed that Q. De Coninck et al. published an article titled "Multipath QUIC: Design and evaluation" in 2017. This article proposed and proved the feasibility of the multipath QUIC protocol. Multipath QUIC aggregates the bandwidth of multiple networks to achieve more efficient transmission. In 2021, W. Yang et al. published an article titled "MM-QUIC: Mobility-aware multipath QUIC for satellite networks", which demonstrated the improvement of the multipath QUIC protocol on the performance of satellite communication systems. However, the multipath QUIC protocol is similar to the traditional multipath transmission protocol. It splits the data of the same application into different paths for transmission. Due to the different network conditions of different paths, the data on different paths will be blocked. Therefore, the packet scheduler largely determines the performance of the multipath transmission protocol. In 2018, A. Rabitsch et al. published a paper titled "A stream-aware multipath QUIC scheduler," proposing a stream-aware scheduler. However, different multipath schedulers only perform well in specific scenarios. In summary, existing transport protocols cannot meet the growing user demand for high-bandwidth, low-latency, and reliable communication services.

[0005] Summary of the Invention

[0006] Purpose of the invention: The purpose of the present invention is to provide a network twin multi-host transmission method and system for multi-cloud native applications, which can fully utilize multi-host network resources and thus improve the user communication experience.

[0007] Technical solution: To achieve the above-mentioned purpose, the present invention adopts the following technical solution:

[0008] A network twin multi-homed transmission method for multi-cloud native applications includes the following steps:

[0009] The network twin service is introduced as the communication agent for all user cloud-native applications, dividing the traditional one-stage end-to-end communication of cloud-native applications into three-stage communication. The network twin service consists of a network twin client and a network twin server running on the user terminal device and the cloud server respectively. The three-stage communication is the communication between the cloud-native application client and the network twin client, the communication between the network twin client and the network twin server, and the communication between the network twin server and the cloud-native application server.

[0010] The network twin service establishes an all-weather always-open communication connection using a multi-homed network. The data of each user's cloud native application is transmitted using the always-open communication connection under the scheduling of the network twin service;

[0011] The streaming technology of the Quick UDP Internet Connection Protocol (QUIC) is used to multiplex multiple streams of cloud native application data onto the always-open connection established by the network twin service for communication and transmission.

[0012] Furthermore, cloud-native applications are designed for cloud computing architectures and deployed on cloud infrastructure, leveraging the cloud's flexibility, scalability, and elasticity. They can run and scale seamlessly on cloud infrastructure. Cloud-native applications utilize a microservices architecture, where each service is responsible for a specific area and is no longer tightly coupled to hardware devices, resulting in strong scalability.

[0013] Furthermore, the multi-cloud native application is based on the overall user experience and does not optimize the transmission of a single cloud native application. Multiple cloud native applications adopt the multi-host transmission method to avoid the situation where a single application greedily seizes device and network resources.

[0014] Furthermore, the network twin's server and the cloud native application's server run on the same cloud service provider's cloud server. Unlike conventional networks, the network twin's communication proxy service is opaque to cloud native applications.

[0015] Furthermore, communication between the cloud-native application client and the network twin client, and between the network twin server and the cloud-native application server, all occurs on the same device or cloud, minimizing the latency introduced by the additional communication between the two ends. This three-stage communication decouples the cloud-native application client and server, eliminating the need for cloud-native applications to consider communication speed and stability.

[0016] Furthermore, the all-weather always-open connection is a connection between the network twin client and the network twin server established using the recently improved Quick UDP Internet Connection Protocol (QUIC). This connection is established using a multi-homed network, and the sub-connections established through different network interfaces are independent of each other and have independent connectionIDs but are uniformly subject to the global scheduling of the network twin service.

[0017] Furthermore, the network twin service schedules cloud-native data based on the current state of different networks, seamlessly dispatching data from multiple cloud-native applications to sub-connections established across different networks. This aggregates multi-homed network bandwidth while avoiding the increased complexity and congestion associated with traditional multipath communication protocols, which often split the same application data across different sub-connections.

[0018] Furthermore, the multiplexing of cloud-native application data into multiple streams is to use the Quick UDP Internet Connection Protocol (QUIC) to schedule the data at the stream level. The network twin client carries the data of a single cloud-native application into multiple streams in a sub-link, so that multiple cloud-native applications share a sub-link without interfering with each other.

[0019] Furthermore, the network twin service can perform global scheduling and management of normally open communication connections. When the network fluctuation or interference of a sub-connection exceeds expectations, the network twin service can use the flow migration function to migrate the cloud-native application data flow on the sub-connection to the remaining sub-connections with normal network status for transmission, thereby improving the stability of transmission.

[0020] A network twin multi-host transmission system for multi-cloud native applications, including a network twin client running on a user terminal device, and a network twin server running on a cloud server; the network twin client and the network twin server constitute a network twin service, and the network twin service acts as a communication agent for all the user's cloud native applications, dividing the end-to-end one-stage communication of the cloud native application into three-stage communication; the three-stage communication is the communication between the cloud native application client and the network twin client, the communication between the network twin client and the network twin server, and the communication between the network twin server and the cloud native application server; the network twin service establishes an all-weather always-open communication connection using a multi-host network, and the data of each user's cloud native application is transmitted using the always-open communication connection under the scheduling of the network twin service; the network twin service uses the streaming technology of the fast UDP Internet connection protocol QUIC to multiplex multiple streams of cloud native application data onto the established always-open connection for communication transmission.

[0021] Beneficial effects: Compared with the existing technology, the network twin multi-host transmission method for multi-cloud native applications provided by the present invention has the following advantages: the method introduces the network twin service as a communication agent for cloud native applications, which can avoid the defects of traditional multi-path transmission protocols while making full use of the multi-host network to achieve larger bandwidth and more stable transmission. In addition, the decoupled three-stage transmission proposed in the present invention combines the Quick UDP Internet Connection Protocol (QUIC) to schedule data at the stream level, solving the head-of-line blocking problem between different cloud native applications. In addition, by comparing the traditional end-to-end transmission without the introduction of network twins in a real environment, the effectiveness and superiority of the present invention are verified, providing an efficient solution for end-to-end multi-host network communication. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a scenario diagram of a three-stage transmission architecture after network twins are introduced in an embodiment of the present invention.

[0023] FIG2 is a schematic diagram showing a comparison of round-trip delays when establishing a connection between an embodiment of the present invention and traditional end-to-end transmission.

[0024] FIG3 is a schematic diagram showing a comparison of transmission completion time in a multi-host network between an embodiment of the present invention and traditional end-to-end transmission.

[0025] FIG4 is a schematic diagram showing a comparison of transmission completion time boxes between an embodiment of the present invention and traditional end-to-end transmission under a 3% packet loss rate.

[0026] FIG5 is a schematic diagram showing a comparison between the transmission completion time using the stream migration function in the case of network interference according to an embodiment of the present invention and the traditional end-to-end transmission completion time.

[0027] FIG6 is a schematic diagram showing a comparison between the download speed using the stream migration function in the event of network interference according to an embodiment of the present invention and the download speed using traditional end-to-end transmission. DETAILED DESCRIPTION

[0028] In order to make the objectives, technical solutions, and advantages of the present invention more clearly understood, the following detailed description of an embodiment of the present invention is given in conjunction with the accompanying drawings. This embodiment is implemented based on the technical solutions of the present invention, and provides a detailed implementation method and specific operation process. It should be understood that the specific examples described herein are only used to illustrate the present invention, and the scope of protection of the present invention is not limited to the following embodiments.

[0029] This embodiment, based on the three-stage multi-host transmission scenario of Figure 1, proposes a network twin multi-host transmission method for multi-cloud native applications. In this scenario, the traditional communication from the cloud native application client to the cloud native application server is split into three stages with the introduction of the network twin proxy service, where the network twin service is jointly provided by the network twin client and the network twin server. We will further explain each communication stage from the client to the server. The first stage of communication occurs between all cloud native application clients and the network twin client. These communication transmissions all occur on the user's local terminal device, so the communication efficiency is extremely high and the delay is negligible. In the first stage of communication, the simple and mature Transmission Control Protocol (TCP) can be used. The second stage of communication occurs between the network twin client running on the local terminal device and the network twin server deployed in the cloud. Communication transmission established using a multi-host network including the fifth generation communication (5G) network or wireless fidelity network (Wi-Fi) occurs in this stage. The transmission distance from the terminal device to the cloud server is relatively long and the environment is complex. Therefore, the communication transmission in the second stage is relatively unstable, and the transmission method of the second stage of communication needs to be given special consideration. The third communication segment occurs between the network twin server and the cloud-native application server running on the same cloud server. Similar to the first segment, this segment occurs within the same cloud server, offering extremely high efficiency and negligible latency, and can also utilize the existing Transmission Control Protocol (TCP).

[0030] In order to overcome the problem of unstable communication quality in the second segment of communication, this embodiment adopts a suitably improved Quick UDP Internet Connection Protocol (QUIC). As shown in Figure 1, the network twin client and the network twin server play a gateway-like role for all the user's cloud native applications, so the QUIC connection established through the network twin can be set to always open and carry the data stream of all cloud native applications. In this embodiment, by adopting the above-mentioned QUIC connection that is always open and carries the data of multiple cloud native applications, it is possible to reduce the overhead of establishing multiple connections for multiple different cloud native applications while saving resources for maintaining corresponding connections for multiple different applications. Multiple QUIC connections will be established through heterogeneous physical paths of multi-host networks. Unlike traditional multi-path QUIC protocols, the established QUIC connections have independent connectionIDs and are independent of each other, but are subject to unified scheduling by the network twin service. The data of the same cloud native application is not split and transmitted through an independent QUIC connection. The data streams of different cloud-native applications will be multiplexed onto the established QUIC connection through the stream-level control multi-stream of the fast UDP Internet connection protocol. The data of different cloud-native applications will be transmitted using independent streams without affecting each other's transmission.

[0031] In this embodiment, the normal request-reply process of a cloud native application is described as follows. In the first communication segment, the network twin client acts as an opaque communication proxy server responsible for receiving requests from the local cloud native application client. Next, the path scheduler of the network twin service selects a suitable one from the multiple established QUIC connections for transmission. The scheduling decision of the path scheduler is based on the implemented network conditions and the user's priority for the cloud native application data. Unlike traditional multi-path transmission protocols, the network twin service does not split the data of the same application into different paths. Instead, it carries the cloud native application data through multiple streams, thereby avoiding the head-of-line blocking problem caused by different path network conditions and realizing one-level scheduling of application data streams. Finally, in the third communication segment, this request will be forwarded by the network twin server deployed on the cloud server to the corresponding cloud native application server. The corresponding reply to this request will be transmitted back to the cloud native application client on the terminal through the same path. After introducing the network twin, we have achieved the decoupling of the cloud native application client and server.

[0032] As can be seen from the above description, the Network Twin service is not transparent to cloud-native applications, unlike traditional communication proxies. Under the global supervision of the Network Twin, the stream migration function is implemented, which can seamlessly migrate specific data streams from one QUIC connection to another QUIC connection managed by the Network Twin. When the network conditions of a connection fluctuate or are interrupted, the stream migration function can be used to migrate the stream transmitted on that connection to a connection with better network conditions, greatly improving the stability and flexibility of multi-host network transmission.

[0033] The key functions and main parts of this embodiment are implemented based on the open source database quic-go and have been improved to a certain extent. The code base quic-go is an open source method for implementing the Quick UDP Internet Connection Protocol (QUIC) developed based on the Golang language, with extremely high cross-platform portability and rich community support. In order to maintain a constantly open QUIC connection between the network twin client and the network twin server, we utilize the two parameters MaxIdleTimeout and KeepAlivePeriod. In order to maintain the QUIC connection in a constantly open state and ensure that data from cloud-native applications can use the QUIC connection at any time, the KeepAlivePeriod parameter is set to one-quarter of MaxIdleTimeout. Then, keep-alive packets are sent at fixed intervals to keep the connection active and prevent the QUIC connection from being prematurely interrupted due to inactivity. Because the Quick UDP Internet Connection Protocol (QUIC) is built on top of the User Datagram Protocol (UDP), we specify the underlying udpConnection parameter of the QUIC connection as the IP addresses of different network interfaces. This allows multiple QUIC connections to be established over different physical paths, fully utilizing multi-homed network resources. To enable multiplexing of streams across QUIC connections for different cloud-native applications, we developed dedicated HTTP / 3 clients and servers to serve as the Network Twin client and server, jointly providing the Network Twin service. Within the dedicated HTTP / 3 client, the improved QUIC protocol uses streams to carry data for a single cloud-native application, but the number of streams a single application can use is not limited to one. When the Network Twin service is started, a QUIC connection is established over the multi-homed network and prepared for future data transmission. To facilitate request-reply transmission for cloud-native applications in three-phase communication, the Network Twin client also acts as a server for the cloud-native application client. After receiving a request, it sends it to the Network Twin server over a dedicated QUIC connection under the management of the path scheduler. When the network encounters jitter or interruption, the network twin service will make a decision to migrate the stream. At this time, the network twin client will check the data that has been transmitted through the localSize parameter and send a retransmission request to the network twin server through the migration target QUIC connection. The latter will continue to send data in the range of (localSize, fileSize].

[0034] We conducted experiments on this embodiment in the real world. We used a laptop with multi-homed network connectivity as the terminal device. In addition, we used a commercial cloud server from a major cloud service provider to deploy the cloud native application server and the network twin server. The distance between the cloud server and the terminal device is approximately 1,175 kilometers, and the outbound bandwidth of the cloud server is 5 Mbps. Using the traffic controller (tc), we achieved control over the bandwidth and packet loss rate of different paths. The traffic controller is a powerful network traffic control tool in the Linux operating system that provides the ability to deeply control and optimize network traffic. Ordinary cloud native application request-reply requests were tested in the above environment. We set two different sizes of reply data streams, 10 MB and 2.5 MB respectively. The proposed network twin multi-homed transmission method for multi-cloud native applications will be compared with the traditional end-to-end transmission without the introduction of the network twin service.

[0035] In this embodiment, we first use the process of connection establishment to prove the feasibility of three-stage transmission with network twins. Figure 2 shows the round-trip time (Round-trip time) of connection establishment in two cases, including the handshake process and excluding the handshake process. When considering the connection establishment without the handshake process, the round-trip delay of the three-stage communication with the introduction of network twins is only slightly longer than the traditional end-to-end round-trip delay, and the delay of the second segment of communication accounts for most of the overall delay. This shows that the delay of the additional two segments of communication formed by the introduction of network twin services is insignificant to the overall delay and will not cause a decline in user experience. When considering the handshake process, due to the unique 0-RTT handshake mechanism of the QUIC connection, a lot of time can be saved in the handshake stage, resulting in the overall round-trip delay of the three-stage transmission with the introduction of network twins being much lower than the traditional end-to-end transmission delay.

[0036] We illustrate the advantages of this embodiment compared to the traditional end-to-end transmission method in two different scenarios.

[0037] In Scenario 1, a multi-homed network environment provides two paths, each with a different packet loss rate, but both offering a bandwidth of 2.5 Mbps. On the terminal device, two cloud-native applications request data of 10 MB and 2.5 MB, respectively. Using a three-stage transmission method that incorporates network twinning, the user's terminal device can fully utilize the aggregated bandwidth of the two paths. Specifically, each path is used to transmit the data of a single cloud-native application, thus avoiding the increased complexity and congestion caused by traditional multi-path transmission protocols that split the data of the same cloud-native application across different paths. As shown in Figure 3, the three-stage transmission method, after introducing the network twin service, reduces the overall transmission time of the two cloud-native applications' data at various packet loss rates. Furthermore, Figure 4 shows that the proposed three-stage communication transmission method based on network twinning not only reduces transmission time but also improves transmission stability even under high packet loss rates. In contrast, terminal devices using traditional end-to-end transmission methods cannot utilize the aggregated bandwidth of multiple paths, even in a multi-homed network environment, resulting in a significant increase in overall transmission time.

[0038] In scenario two, we consider a cloud-native application transmitting 10MB of data along a specific path. After a certain period of transmission, the network environment changes, causing a significant increase in packet loss along this path. In a three-stage transmission using the Network Twin service, network performance can be monitored with minimal effort, as the QUIC connections on the remaining paths can be leveraged to synchronize the path status between the current Network Twin client and server. Figure 5 shows the data transmission completion time after encountering network interference at different stages of the transmission. It can be observed that in this three-stage transmission using the Network Twin, the unique stream migration feature minimizes the impact of network interference on transmission. When the Network Twin service detects a degradation in transmission performance on a particular path, it initiates a stream migration, relocating the data stream experiencing the interference to a path with better network conditions. In traditional end-to-end transmission, even if network interference occurs, without the intervention of the Network Twin service, transmission continues along the path with the degraded performance, resulting in a sharp increase in transmission completion time. Figure 6 clearly shows that with the intervention of the stream migration feature, the transmission rate does not significantly decrease in the face of network interference.

[0039] Based on the same inventive concept, an embodiment of the present invention also discloses a network twin multi-host transmission system for multi-cloud native applications, including a network twin client running on a user terminal device, and a network twin server running on a cloud server; the network twin client and the network twin server constitute a network twin service, and the network twin service acts as a communication agent for all the user's cloud native applications, dividing the end-to-end one-stage communication of the cloud native application into three-stage communication; the three-stage communication is the communication between the cloud native application client and the network twin client, the communication between the network twin client and the network twin server, and the communication between the network twin server and the cloud native application server; the network twin service establishes an all-weather always-open communication connection using a multi-host network, and the data of each user's cloud native application is transmitted using the always-open communication connection under the scheduling of the network twin service; the network twin service uses the stream technology of the fast UDP Internet connection protocol QUIC to multiplex multiple streams of cloud native application data onto the established always-open connection for communication transmission.

[0040] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A network twin multi-host transmission method for multi-cloud native applications, characterized by: The following steps are involved: The network twin service is introduced as the communication agent for all user cloud-native applications, dividing the end-to-end one-stage communication of cloud-native applications into three-stage communication; the network twin service consists of a network twin client and a network twin server running on the user terminal device and the cloud server respectively; the three-stage communication is the communication between the cloud-native application client and the network twin client, the communication between the network twin client and the network twin server, and the communication between the network twin server and the cloud-native application server; The network twin service establishes an all-weather always-open communication connection using a multi-homed network. The data of each user's cloud native application is transmitted using the always-open communication connection under the scheduling of the network twin service; The streaming technology of the fast UDP Internet connection protocol QUIC is used to multiplex multiple streams of cloud native application data onto the always-open connection established by the network twin service for communication and transmission.

2. A network twin multi-host transmission method for multi-cloud native applications according to claim 1, characterized in that: The cloud-native applications are applications designed for cloud computing architecture and deployed on cloud infrastructure.

3. A network twin multi-host transmission method for multi-cloud native applications according to claim 1, characterized in that: The multi-cloud native application approach is based on the overall user experience and does not optimize the transmission of a single cloud native application. Multiple cloud native applications adopt the multi-host transmission method to avoid a single application greedily occupying device and network resources.

4. A network twin multi-host transmission method for multi-cloud native applications according to claim 1, characterized in that: The network twin server and the cloud native application server run on the cloud server of the same cloud service operator.

5. The network twin multi-host transmission method for multi-cloud native applications according to claim 1 is characterized by: The communication between the cloud-native application client and the network twin client, and the communication between the network twin server and the cloud-native application server occur in the same device or cloud.

6. A network twin multi-host transmission method for multi-cloud native applications according to claim 1, characterized in that: The all-weather always-open communication connection is a connection between the network twin client and the network twin server established using the improved fast UDP Internet connection protocol QUIC; the connection is established using a multi-host network, and the sub-connections established through different network interfaces are independent of each other and have separate connectionIDs but are uniformly subject to the global scheduling of the network twin service.

7. A network twin multi-homed transmission method for multi-cloud native applications according to claim 1, characterized in that: The network twin service schedules cloud native data based on the current status of different networks, and schedules multiple different cloud native application data to sub-connections established using different networks for transmission without splitting.

8. The network twin multi-host transmission method for multi-cloud native applications according to claim 1, characterized in that: The multiplexing of cloud-native application data into multiple streams is to use the fast UDP Internet connection protocol QUIC to schedule the data at the stream level. The network twin client carries the data of a single cloud-native application into multiple streams in a sub-link, so that multiple cloud-native applications share a sub-link without interfering with each other.

9. A network twin multi-host transmission method for multi-cloud native applications according to claim 1, characterized in that: The network twin service performs global scheduling and management of normally open communication connections. When the network fluctuation or interference of a sub-connection exceeds expectations, the network twin service uses the flow migration function to migrate the cloud-native application data flow on the sub-connection to the remaining sub-connections with normal network status for transmission.

10. A network twin multi-host transmission system for multi-cloud native applications, characterized by: It includes a network twin client running on the user terminal device and a network twin server running on the cloud server; the network twin client and the network twin server constitute a network twin service, and the network twin service acts as a communication agent for all the user's cloud native applications, dividing the end-to-end one-stage communication of the cloud native application into three-stage communication; the three-stage communication is the communication between the cloud native application client and the network twin client, the communication between the network twin client and the network twin server, and the communication between the network twin server and the cloud native application server; The network twin service establishes an all-weather always-open communication connection using a multi-host network. The data of each user's cloud native application is transmitted using the always-open communication connection under the scheduling of the network twin service; The network twin service uses the streaming technology of the fast UDP Internet connection protocol QUIC to multiplex multiple streams of cloud native application data onto the established open connection for communication transmission.

Images