KR20210129872A - Method and apparatus for high precision data communication - Google Patents

Abstract

Provided are a method and an apparatus for super-precision data communication. The method of the present invention is a method for super-precision data communication performed on a terminal linked to a multi network through a plurality of network interfaces, which comprises the following steps of: determining a transmission mode matched to information on an application ID or a destination, which represents service properties of a transmission packet when a transmission packet occurs; combining the network interfaces when the determined transmission mode is redundancy transmission to select a combination of the network interfaces where an average delay time of the network interfaces included in each combination is minimal; combining the network interfaces when the determined transmission mode is aggregation transmission to select a combination of the network interfaces where a deviation in the delay times of the network interfaces included in each combination is minimal; using the selected combination of the network interfaces to perform redundancy transmission or aggregation transmission of the transmission packet. Accordingly, the redundancy transmission or aggregation transmission is selectively applied to allow optimized service to be provided in terms of a delay time and a transmission speed.

Description

Ultra-precise data communication method and apparatus

The present invention relates to an ultra-precise data communication method and an apparatus therefor.

In general, the propagation latency of a packet that a mobile communication terminal receives from a content server through an access network such as 5G network, LTE (Long Term Evolution) network, Wi-Fi network, etc. depends on the situation of each access network. It is variable according to each access network, and there is a difference between each access network. This is because each access network and the content server are composed of various network elements (NEs).

Each network element (NE) allocates a network resource shared and used by terminals simultaneously connected to a plurality of terminals instead of one terminal. Each network element NE allocates network resources to terminals according to a packet forwarding priority control method, a packet scheduling method, and an applied connection load and performance. As such, it is practically very difficult or impossible to implement transmission that guarantees always uniform packet propagation delay in a network path composed of various network elements NE. This is because the network resource allocation of each network element NE is variable depending on the queuing or buffering structure and the factors constituting the entire section of the traffic flow.

In addition, when transmitting a packet through one access network, the packet forwarding delay that can be provided by the network element (NE) nodes of the access network and the associated core network may be outside the range that can smoothly perform the service. In this case, it may be difficult to smoothly proceed with the service, such as when the delivery delay of a packet transferred from the content server to the terminal is changed and the service is stopped. In particular, in order for a service used in a smart city or a smart factory having hard real-time or mission critical characteristics to operate smoothly, the transmission delay of a transmitted packet or signaling packet must be short. Of course, the data transmitted from the service server must reach the terminal with 100% reliability within the given time.

However, when a problem such as network congestion occurs during packet transmission through a single access network or the arrival time of a transmitted packet is delayed, the entire service itself may be stopped or the quality of experience (QoE) may deteriorate due to deterioration of quality. . In addition, in the case of a remote control service, a fatal service failure may occur due to a disruption in precise machine control.

As such, in order to solve the problem of using only a single access network, a packet transmission service using a multiple access network has been proposed. A typical example is aggregated transmission. Merge transmission is a method of transmitting data through a plurality of access networks, sending some packets out of all packets to a random access network, transmitting some other packets to another access network, merging them and restoring the original packet. , it is possible to increase the efficiency compared to single network transmission. However, a relatively fast access network among a plurality of access networks always exists. Therefore, there is a limit in reducing the overall delay time for all packets to be transmitted in terms of restoring the original packet.

Also, a particular service/application may not always require only the lowest latency. For example, when a specific service/application requires packets to be delivered within an average of 10 msec, a stable service is possible when all packets are delivered within a 10 msec delay time. For example, if the five packet delivery delays are 6ms, 7ms, 8ms, 9ms, and 10ms, respectively, the average packet delivery delay is 8ms, and therefore the average packet delivery delay is within 10msec due to network effects. However, when the distribution of packet delivery delay is large due to severe fluctuations, for example, when the packet delivery delay is 8ms, 8ms, 10ms, 12ms, or 12ms, the average packet delivery delay is 8ms, but it is less than the service request delay time of 10ms. The stability of the service may not be guaranteed by slower incoming packets (12ms). Therefore, it is possible to smoothly perform the ultra-low delay service by considering the minimization of packet delivery delay and the arrival distribution (jitter) of the transmitted packets at the same time.

However, up to now, packet transmission techniques have been mainly focused on minimizing the delay time. Even in the case of multi-path transmission, there is a limitation in that the purpose of unified transmission is only through a path with a minimized delay time.

The problem to be solved by the present invention is that precise control of network performance is possible in consideration of the maximum packet delivery delay, deviation of delivery delay, and availability (reliability/survivability) according to service requirements. An object of the present invention is to provide an ultra-precise data communication method and an apparatus therefor.

According to one feature of the present invention, the ultra-precise data communication method is an ultra-precision data communication method of a terminal connected to a multi-network through a plurality of network interfaces. When a transport packet is generated, an application ID or destination indicating the service characteristics of the transport packet determining a transmission mode matching the information; if the determined transmission mode is redundant transmission, combining the plurality of network interfaces to select a network interface combination having the minimum average delay time of network interfaces included in each combination , when the determined transmission mode is merged transmission, combining the plurality of network interfaces to select a network interface combination having a minimum delay time deviation of network interfaces included in each combination, and using the selected network interface combination and transmitting the transport packets in duplicate or merged transmission.

The step of selecting a network interface combination having the minimum average delay time includes measuring each delay time and link capacity for the plurality of network interfaces, and an objective function using each delay time and link capacity is the minimum. and selecting a network interface combination that becomes The total delay time including time deviations may be calculated to have a weight higher than the average link capacity of the network interfaces included in each combination.

The step of selecting a network interface combination having the minimum delay time deviation includes measuring each delay time and link capacity for the plurality of network interfaces, and an objective function using each delay time and link capacity is the minimum. and selecting a network interface combination that becomes The average link capacity of the network interfaces included in each combination may be calculated to have a higher weight than the total delay time including time deviations.

In the redundant transmission, a portion of the transport packet is transmitted over all network interfaces included in the selected combination, and another portion of the transport packet is transmitted without overlapping through some network interfaces included in the selected combination. .

As for the some network interfaces, a network interface having the maximum network performance based on a transmission speed or a delay time may be selected from among a plurality of network interfaces included in the selected combination.

After the transmitting step, receiving a duplicate transport packet using the selected network interface combination, and excluding the duplicate transport packet received from a network interface having a minimum delay time among network interfaces included in the selected network interface combination and discarding duplicate transport packets received from the remaining network interfaces.

After the step of transmitting, storing a transmission packet received using the selected network interface combination in a reception buffer, and adjusting the speed of outputting the transmission packet stored in the reception buffer to the application layer according to the transmission mode may include more.

In the step of adjusting, in the case of merge transmission, the greater the delay time difference between the interfaces in the selected network interface combination, the slower the output speed, and the smaller the delay time difference, the higher the output speed.

In the adjusting step, in the case of redundant transmission, when packet loss occurs, the output speed is lowered as the delay time deviation between interfaces in the selected network interface combination is larger, and the output speed is increased as the delay time deviation is smaller, and the packet loss is reduced. If not, the output speed may be adjusted based on the packet reception speed of the network interface having the largest link capacity among the interfaces in the selected network interface combination.

According to another feature of the present invention, the ultra-precise data communication method is an ultra-precision data communication method of a terminal connected to a multi-network through a plurality of network interfaces, and a redundant transmission mode matching an application ID or destination information indicating a service characteristic of a transport packet. determining, combining the plurality of network interfaces to determine a network interface combination having a minimum average delay time, and repeatedly transmitting the transport packet using the determined network interface combination; In the step of, a part of the transport packet may be repeatedly transmitted through all network interfaces included in the selected combination, and another part of the transport packet may be transmitted without overlapping through some network interfaces included in the selected combination.

In the transmitting step, the available link capacity of the network interfaces included in the determined network interface combination is checked, and the network interface whose available link capacity does not meet the minimum link capacity according to the service characteristic of the transport packet is a redundant transmission path. can be excluded.

The determining may include measuring a delay time and a link capacity for each of a plurality of network interfaces, combining the plurality of network interfaces, and using the delay time and the link capacity, a delay time of a network interface for each combination calculating the deviation, average delay time, and average link capacity of , and weighting the average delay time higher than the deviation of the delay time and calculating the average total delay time obtained by adding the deviation of the average delay time and the delay time to the average The method may include determining a network interface combination in which an objective function given a weight higher than the link capacity is minimized.

According to the embodiment, when transmitting a packet, by selectively applying either overlapping transmission or merged transmission in a service application unit or a packet flow unit, packet delivery delay is minimized and at the same time, it is ultra-low in consideration of various performance measures such as packet arrival distribution (jitter) and transmission speed. Delayed service can be provided smoothly. In addition, by selectively applying redundant transmission and merged transmission, it is possible to provide a service optimized in terms of two performance measures, that is, delay time and transmission speed.

In addition, duplicate packets are transmitted through multiple paths, but some of the packets are duplicated through multiple paths and the remaining packets are transmitted simultaneously through a single path with a large transmission capacity, thereby improving packet transmission delay and reliability as well as transmission speed. can do it

In addition, when a packet error or packet omission occurs, transmission reliability can be improved through a packet transmitted through another available path. Accordingly, it is possible to ensure transmission delay, transmission speed, and reliability when delivering applications requiring real-time performance or mission-critical applications required by smart factories, smart cities, and the like.

In addition, by applying the terminal's own policy, network policy, and policy based on service request, it can be applied to various products or services by grafting it to the corporate market as well as the mass market.

1 is a block diagram of an ultra-precision data communication system according to an embodiment of the present invention.
2 is a block diagram of an ultra-precision data communication system according to another embodiment of the present invention.
3 is a block diagram of a terminal according to an embodiment of the present invention.
4 is an exemplary diagram illustrating redundant transmission according to an embodiment of the present invention.
5 is a diagram for explaining a delay time for each radio path according to an embodiment of the present invention.
6 is a block diagram of a gateway according to an embodiment of the present invention.
7 is a flowchart illustrating an ultra-precision data communication method according to an embodiment of the present invention.
8 is a diagram illustrating redundant transmission through multiple paths according to an embodiment of the present invention.
9 is a diagram illustrating redundant transmission through multiple paths according to another embodiment of the present invention.
10 is a diagram for explaining redundant transmission through multiple paths according to another embodiment of the present invention.

Hereinafter, with reference to the accompanying drawings, the embodiments of the present invention will be described in detail so that those of ordinary skill in the art can easily implement them. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout the specification, when a part "includes" a certain element, it means that other elements may be further included, rather than excluding other elements, unless otherwise stated.

In addition, terms such as “…unit”, “…group”, and “…module” described in the specification mean a unit that processes at least one function or operation, which may be implemented by hardware or software or a combination of hardware and software. can

The devices described in the present invention are composed of hardware including at least one processor, a memory device, a communication device, and the like, and a program executed in combination with the hardware is stored in a designated place. The hardware has the configuration and capability to implement the method of the present invention. The program includes instructions for implementing the method of operation of the present invention described with reference to the drawings, and is combined with hardware such as a processor and a memory device to execute the present invention.

As used herein, "transmission or provision" may include not only direct transmission or provision, but also transmission or provision indirectly through another device or using a detour path.

In the present specification, expressions described in the singular may be construed in the singular or plural unless an explicit expression such as “a” or “single” is used.

In this specification, the same reference numbers refer to the same components regardless of the drawings, and "and/or" includes each and every combination of one or more of the referenced components.

In this specification, terms including an ordinal number such as first, second, etc. may be used to describe various elements, but the elements are not limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present disclosure, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component.

In the flowchart described with reference to the drawings in this specification, the order of operations may be changed, several operations may be merged, some operations may be divided, and specific operations may not be performed.

In the specification, a terminal/user equipment accesses a base station of an access network/radio access network (RAN) and uses network functions (NFs) of a core network. Here, the base station may include an eNB (or eNodeB), a gNB (or a next generation Node B), an access point (AP), etc. according to an access network.

The terminal may be a terminal of various types and uses, such as a portable terminal, an IoT terminal, a vehicle terminal, a display terminal, a broadcasting terminal, and a game terminal.

The communication interface may vary. For example, the communication interface is a short-range wireless network interface such as Wi-Fi/Wireless Local Area Network (WLAN)/Bluetooth, and 3G/LTE (Long Term Evolution)/LTE-A (Long Term) It may include a mobile communication network interface such as Evolution-Advanced)/5G, and a terminal manufacturer may add various communication interfaces.

In this specification, a WiFi interface, an LTE interface, and a 5G interface will be described as examples, but the communication interface is not limited thereto.

Embodiments of the present invention are 3GPP (3rd Generation Partnership Project) Long Term Evolution (LTE), LTE-A, 3GPP2, IEEE (Institute of Electrical and Electronics Engineers) system, 3GPP such as FS_NextGen (Study on Architecture for Next Generation System). It may be supported by standard documents related to at least one of the 5G systems. That is, steps or parts not described in order to clearly reveal the technical spirit of the present invention among the embodiments of the present invention may be supported by the document. In addition, all terms disclosed in this document may be explained by the standard document.

Multi-path transmission is a transmission method that simultaneously uses heterogeneous network interfaces in mobile devices equipped with various network interfaces.

Aggregation transmission among multi-path transmissions is a technology for transmitting data using at least two radio paths simultaneously, and processes data transmitted through each radio path as one session. In the merge transmission, data transmitted through a single path may be divided into paths of a plurality of homogeneous networks or a plurality of heterogeneous networks and transmitted, or data transmitted through a plurality of paths may be bundled into one path and transmitted. Such aggregation transmission may be called MultiNet Aggregation in the sense of merging a plurality of networks.

Since merged transmission transmits data through a plurality of paths, in order to improve transmission speed compared to when transmitting through a single network, it is possible to divide the entire data and transmit it through each wireless path. For example, after the transmitting side transmits some data to the LTE network and the other partial data to the WiFi network, the receiving side merges the partial data received through each wireless path to restore the original whole data.

However, in the case of merge transmission, a relatively fast radio path is always present among a plurality of radio paths. That is, since the transmission speed of each wireless path is different, there is a limit to reducing the overall delay time for the data to be transmitted in terms of restoring the entire original data.

Among multi-path transmissions, redundant transmission is a technique for transmitting the same data through at least two radio paths.

There are two methods of redundant transmission. One method is to use Internet Protocol (IP)-in-IP encapsulation and decapsulation (encapsulation and decapsulation) technology. In this method, the same data, not partial data, is simultaneously transmitted through at least two radio paths, and data of a radio path having a relatively short transmission delay is first acquired at the receiving end. Data generated by the application or service on the sending side is located in the inner IP payload of the IP packet, and this data is encapsulated with an external IP address and transmitted. The external IP address is an address used for transmission and routing in a data transmission network, and may be, for example, IP addresses of network devices of a gateway or a core network. On the receiving side, the original data is obtained by decapsulating the encapsulated data.

In this case, the upper limit of the Path Maximum Transmission Unit (PMTU) for each radio path is fixed. Accordingly, there is a loss of data occupied by IP-in-IP encapsulation/decapsulation, that is, an overhead of an external IP header. In addition, the PMTU is configured differently for each radio path. For example, LTE PMTU=1460 bytes/IPv4, WiFi PMTU=1500 bytes, 5G PMTU=1420 bytes/IPv6, etc. may be configured. Therefore, in order to transmit the same data through a plurality of radio paths, a transmission data unit size should be set based on a minimum value among PMTUs of each radio path, for example, 5G PMTU=1420 bytes/IPv6.

However, if IP-in-IP overhead is added, the actual data transfer rate is inevitably lower. Because the header overhead required for encapsulation/decapsulation is additionally required, a larger amount of transmission is used even if the same data size is transmitted. Of course, a PMTU larger than the determined minimum PMTU value may be used, but in this case, fragmentation occurs at the IP level, which may increase the transmission delay time.

In addition, the IP-in-IP encapsulation/de-encapsulation method uses User Datagram Protocol (UDP) as a transport. In UDP transmission, there is a possibility that data may disappear in the middle because there is no procedure to check whether the data from the sending end has arrived normally at the receiving end. Since this method is a best-effort method that can be lost due to congestion or other factors in intermediate nodes, it is inevitably disadvantageous compared to the case of using another transport (TCP, etc.) in terms of transmission reliability.

Another method of redundant transmission is a method of using Transmission Control Protocol (TCP) as a transport. The TCP method shows higher reliability than the UDP method because it includes a procedure to check whether the data sent from the transport layer is normally delivered.

However, in this method, due to the characteristics of TCP, when one wireless path has a higher delay time than the other wireless path during overlapping transmission for each wireless path, the same data must be transmitted similarly to a path that may have a relatively low transmission speed. Therefore, if the transmission rate converges toward the slow latency side, the entire TCP transmission itself may be slowed in the end-to-end side.

As described above, overlapping transmission and merged transmission have their respective limitations in view of only the purpose of reducing transmission delay time. Accordingly, an embodiment of the present invention provides ultra-precision data communication in which redundant transmission and merged transmission are selectively applied according to service requirements.

Ultra-precision data communication is a method required by 5G killer applications such as smart cities and smart factories. It provides precise control of network performance such as reliability/survivability).

In this case, ultra-precision data communication according to an embodiment of the present invention enables adaptive transmission to a changing network state by selectively applying redundant transmission or merged transmission in units of packet flows as well as for service purposes.

In addition, according to an embodiment of the present invention, when duplicate transmission is selected, duplicate packets are transferred through a plurality of radio paths, and a packet received from a radio path having the shortest transmission delay time among each radio path is prioritized. . Accordingly, while minimizing the overall packet delivery delay, it is possible to evenly distribute the entire packet through an available path, thereby minimizing service fragmentation.

In addition, according to an embodiment of the present invention, during overlapping transmission, some packet sequences are repeatedly transmitted through a plurality of wireless paths having different transmission rates, and other packet sequences are transmitted through a single wireless path having a relatively high transmission rate (or a large transmission capacity). It uses a scheduling method that transmits without overlapping. In this way, it is possible to alleviate the downlink convergence problem that occurs when simply redundant transmission through a plurality of paths, that is, the loss problem in which the overall transmission rate is down-converged to a wireless path having a relatively slow transmission speed. In addition, you can benefit not only in latency and reliability, but also in transmission speed.

In addition, an embodiment of the present invention provides ultra-precision data communication in which a combination of at least two network interfaces is selected from among a plurality of network interfaces according to service requirements and network conditions, and merged transmission or redundant transmission is selectively applied through them. Therefore, it is possible to provide reliable communication by minimizing the packet error rate that may occur due to packet drop, queue/buffer overflow, etc. that may occur in an intermediate node having a best-effort delivery system. have.

Now, an ultra-precision data communication method and apparatus according to an embodiment of the present invention will be described with reference to the drawings.

1 is a block diagram of an ultra-precision data communication system according to an embodiment of the present invention.

Referring to FIG. 1 , the ultra-precision data communication system includes a terminal 100 and a Multi-Path Aggregation & Duplication (MPAD) gateway 200 connected to the terminal 100 through a plurality of networks. Here, the ultra-precision data communication is simultaneously transmitted through multiple paths, but since redundant transmission and merged transmission are selectively applied, it may be referred to as MPAD.

The terminal 100 may have multiple communication interfaces, and may be connected to the MPAD gateway 200 at one time through the multiple communication interfaces. The terminal 100 may use three types of network interfaces: an LTE interface 101 , a 5G interface 103 , and a WiFi interface 105 . Of course, it is not limited to these three types of network interfaces, and it does not necessarily have to be three. The terminal 100 is defined as a terminal capable of multi-path transmission using at least two network interfaces.

The terminal 100 includes an MPAD agent 107, which performs packet scheduling for selecting a transmission mode among redundant transmission and merged transmission, and selection of an optimal network interface combination. The MPAD agent 107 may be implemented as a terminal internal logic. In this case, the MPAD agent 107 additionally performs authentication, status management, and the like, which are basically necessary for multi-path transmission. Inside the terminal, the MPAD agent 107 and various applications communicate through sockets.

The MPAD gateway 200 is located at the contact point of multiple networks, for example, it may be located at the contact point of the LTE network 301 , the 5G network 303 , and the WiFi network 305 . The MPAD gateway 200 is connected to the core network 400 and the WiFi network 305 to which the LTE network 301 and the 5G network 303 are connected. And it is connected to the content server 600 through the Internet (500).

Here, it is assumed that the content server 600 does not support multiple communication interfaces. It is assumed that the content server 600 is a general server that communicates with Transmission Control Protocol (TCP) through a single path, at this time, the MPAD agent 107 and the MPAD gateway 200 may explicitly designate a packet flow through a proxy protocol. This structure can be applied to a centralized network deployment structure such as a 5G central center.

The MPAD gateway 200 receives all packets destined for the Internet 500 through the core network 400 and the WiFi network 305 . In this case, the destination address of the received packet is the MPAD gateway 200 rather than the content server 600 .

When the MPAD agent 107 implements the proxy function, the MPAD agent 107 changes the destination address of the packet to the MPAD gateway 200 . The MPAD gateway 200 may operate as a proxy server that receives all multi-path packets and delivers them to the content server 600 . The MPAD gateway 200 is implemented as a proxy server, and connects a session through a signaling procedure with the terminal 100 .

The terminal 100 and the MPAD gateway 200 transmit and receive data in a redundant transmission or a merge transmission method.

When using multi-path communication, the terminal 100 may transmit/receive data to and from the content server 600 through the MPAD gateway 200 . The terminal 100 may transmit/receive data to and from the content server 600 without going through the MPAD gateway 200 when single-path communication is used. Here, it is assumed that the content server 600 is a general server that performs TCP communication through a single path.

The MPAD agent 107 connects to the MPAD gateway 200 and performs a connection procedure when a service application is driven and a packet is generated. The MPAD agent 107 and the MPAD gateway 200 may exchange signaling information according to the Socket Secure (SOCKS) protocol defined in RFC1928 and RFC1929. Since a proxy connection method conforming to the SOCKS protocol defined in RFC1928 and RFC1929 is known, a detailed description thereof will be omitted.

When a session is created, the MPAD agent 107 is connected to the MPAD gateway 200 through a plurality of subflows. At this time, if the 5G network is the default network, the primary subflow is connected to the 5G network and the other radio paths are connected to the secondary subflow.

In an embodiment of the present invention, there may be four transmission schemes as follows.

1) Uplink merge transmission: MPAD agent 107 sets the transmission mode to merge transmission and requests a session connection

2) Uplink redundant transmission: MPAD agent 107 designates the transmission mode as redundant transmission and requests a session connection

3) Downlink merge transmission: MPAD agent 107 requests a session connection to port 10000 of MPAD gateway 200

4) Downlink redundant transmission: MPAD agent 107 requests a session connection to port 20000 of MPAD gateway 200

As such, whether merge/duplicate transmission is determined according to the up/down transmission direction for each session, which is an implementation specific matter, may take a method of explicitly selecting a scheduler by modifying the socket program.

For example, the MPAD agent 107 transmits the same data to multiple networks instead of the conventional round trip time (RTT)-based optimal network selection method for scheduling of the Multi-path Transmission Control Protocol (MPTCP), that is, merged transmission. A redundant transmission (redundant) mode may be used.

The socket option for packet flow can be set to redundant transmission mode regardless of the system-wide packet scheduler setting.

Packet scheduling means deciding which path to send each packet out of among multiple paths that can actually be transmitted. You can transmit or merge transmission by selecting an optimal path.

In this case, the scheduling subject is the MPAD agent 107 in the case of uplink transmission, and the MPAD gateway 200 in the case of downlink transmission.

2 is a block diagram of an ultra-precision data communication system according to another embodiment of the present invention.

At this time, FIG. 2 has a structure substantially similar to that of FIG. 1, but shows a different embodiment in the arrangement structure of the MPAD gateway 200'. Accordingly, descriptions of the same contents as those of FIG. 1 will be omitted, and only exemplary configurations different from those of FIG. 1 will be described.

The ultra-precision data communication system according to the embodiment of FIG. 2 corresponds to a non-proxy-based on-path arrangement structure. That is, the MPAD gateway 200 ′ is disposed in-line on the routing path immediately behind the core network 400 and operates without using a proxy protocol. In this case, the MPAD gateway 200 ′ monitors packets transmitted and received by the core network 400 and/or the WiFi network 305 to and from the Internet 500 in real time, and a packet including a preset IP address and/or port is displayed. If it is found, it intercepts it, connects a session with the terminal 100 ′ instead of the content server 600 , and processes the packet in a multi-path communication method.

This on-path deployment structure is used for a decentralized network deployment structure such as 5G edge centers.

Detailed operations of the terminal 100 ′ and the MPAD gateway 200 ′ described in FIGS. 1 and 2 will be described.

3 is a block diagram of a terminal according to an embodiment of the present invention, FIG. 4 is an exemplary diagram illustrating redundant transmission according to an embodiment of the present invention, and FIG. 5 is a delay for each radio path according to an embodiment of the present invention. It is a diagram explaining time.

First, referring to FIG. 3 , the terminal 100 includes an LTE interface 101 , a 5G interface 103 , a WiFi interface 105 , an MPAD agent 107 , 107 ′, an application 109 , and a socket communication unit 111 . , a duplicate transmission processing unit 113 , a merge transmission processing unit 115 , and a single transmission processing unit 117 .

The MPAD agent 107 means a configuration with a proxy function corresponding to the embodiment of FIG. 1 , and the MPAD agent 107 ′ means a configuration without a proxy function corresponding to the embodiment of FIG. 2 .

When the MPAD agent 107 operates in the proxy mode, the destination address of the OUTER IP header of the packet is set to the IP address of the MPAD gateway 200 operating as a proxy server.

In addition, since the operation of the terminal 100 is common to all the embodiments of FIGS. 1 and 2 , the common operation will be described below.

The terminal 100 may transmit packets through the LTE interface 101 , the 5G interface 103 , and the WiFi interface 105 . In the case of a general transmission method, the terminal 100 transmits and receives a service packet through one network interface selected by the highest routing priority among the LTE interface 101 , the 5G interface 103 , and the WiFi interface 105 . When the delay time of the selected uppermost routing path deteriorates, the terminal 100 must change to the next or next lane routing path, but in the case of TCP, when the source address of the packet flow is changed by changing the routing, Independent of the session being processed, a new session should be added and used. Accordingly, service interruption may occur. In order to compensate for this, a multi-path aggregation transport protocol such as MPTCP may be used to segregate/aggregate packets based on the available transmission amount of each network interface 101 , 103 , 105 depending on the situation. In this case, in the case of a service that needs to transmit data within a specific time, for example, a mission-critical or hard real-time guarantee, the service may be vulnerable or unavailable depending on delay time fluctuations. Accordingly, the terminal 100 can take advantage of packet delivery delay time by transmitting duplicate data without dividing the packet to be transmitted through multiple paths based on the available network interfaces 101 , 103 , and 105 .

In an embodiment of the present invention, the MPAD agents 107 and 107' of the terminal 100 may transmit a packet by selecting one transmission mode from among redundant transmission, merged transmission, and single transmission based on service requirements.

The MPAD agents 107 and 107' may select a transmission mode on a per application basis and/or per packet flow basis. Although only one application 109 is illustrated in FIG. 3 , a plurality of applications 109 may exist. In this case, the MPAD agents 107 and 107 ′ can select different transmission modes for each application 109 according to the service purpose. have. That is, application A may select only merge transmission and application B may select only redundant transmission. For example, the application 109 that provides a service sensitive to delay time may only perform redundant transmission.

In addition, the MPAD agents 107 and 107' can select a transmission mode in one application 109 in units of packet flows. When looking at the entire data from the point of view of one service application, single transmission, merged transmission, and redundant transmission are mixed.

The MPAD agents 107 and 107' may drive the corresponding sockets 111A, 111B, and 111C according to the transmission mode. However, based on the uplink, all sockets 111A, 111B, and 111C are created even in the case of the same application 109, and the sockets 111A, 111B, and 111C of the transmission mode selected for each packet flow are selectively driven. can do.

In order for the application 109 of the terminal 100 to transmit and receive packets, socket communication is connected with the MPAD agents 107 and 107' in charge of communication inside the terminal.

The socket communication unit 111 creates sockets 111A, 111B, and 111C for communication between the application 109 and the MPAD agents 107 and 107'. The sockets 111A, 111B, and 111C are generated by the programming code of the application 109 and are different for each transmission mode. The socket communication unit 111 creates a redundant transmission socket 111A, a merge transmission socket 111B, and a single transmission socket 111C.

When the socket connection of the application 109 is detected, the MPAD agents 107 and 107' determine the transmission mode based on the purpose of the application 109 or the destination information (IP address/port) of the packet flow, and the transmission mode The sockets 111A, 111B, and 111C corresponding to are connected to the application 109 . Then, the packet flow received through the connected sockets 111A, 111B, and 111C is scheduled and output to one of the duplicate transmission processing unit 113 , the merge transmission processing unit 115 , and the single transmission processing unit 117 .

In this case, the MPAD agents 107 and 107' control session creation with the MPAD gateway 200, and separately create sessions corresponding to the sockets 111A, 111B, and 111C. That is, the duplicate transmission session, the merged transmission session, and the single transmission session are divided and generated.

When a socket connection is requested from the application 109, the MPAD agents 107 and 107' determine the transmission mode based on the destination IP address and destination port of the packet flow, and connect the socket corresponding to the transmission mode. For example, when multiple packet flows are generated in one application 109, MPAD agents 107 and 107' select a redundant transmission mode for packet flows requiring low latency (eg, control messages, etc.), For packet flows (eg, media download, etc.) requiring As such, even within one application 109, it can be divided into a redundant transmission mode and a merged transmission mode.

The MPAD agents 107 and 107' can explicitly select the redundant transmission mode, the merged transmission mode, or the single transmission mode by their own rules when there is a packet flow generated by the application 109 .

The MPAD agents 107 and 107' store a matching table in which the transmission mode is matched to the application ID or destination information indicating the service characteristics of the packet. The service characteristics of the packet may include an application ID from which the packet is generated or destination information (IP address and/or port) of the packet. This matching table is stored in advance and may be updated in real time by an external server (not shown).

The MPAD agents 107 and 107' determine the transmission mode mapped to the destination information (IP address/port) and/or the application ID of the packet requested by the application 109 when the socket connection request of the application 109 is detected. can

When a socket is created, the MPAD agents 107 and 107' open a corresponding port and perform packet scheduling. A socket is defined by a protocol, an IP address, and a port number.

The application 109 is connected to the redundant transmission socket 111A, and the MPAD agents 107 and 107' open port #1 for multi-path redundant transmission. The MPAD agents 107 and 107' output the packet flow transmitted through the redundant transmission socket 111A to the redundant transmission processing unit 113 .

The redundant transmission processing unit 113 performs redundant transmission through a plurality of available paths for a packet flow transmitted through the redundant transmission socket 111A. In this case, the redundant transmission processing unit 113 performs redundant transmission through at least two network interfaces selected by the MPAD agents 107 and 107'.

The redundant transmission processing unit 113 includes an MPTCP or MPUDP layer 113A, a TCP/UDP layer 113B, an IP layer 113C, and an L2/L1 layer 113D.

The TCP/UDP layer 113B includes a number of subflow layers, and each subflow is mapped with a network interface 101 , 103 , 105 . The MPTCP or MPUDP layer 113A may duplicate and distribute packets to be transmitted to each subflow layer 113B, and may preferentially process packets received first from each subflow layer 113B and discard the rest.

Each subflow layer 113B operates the same as the existing TCP layer and is treated as a separate session.

The MPTCP or MPUDP layer 113A divides the duplicate packet of each subflow into TCP segments or UDP segments. The IP layer 113C corresponding to each subflow generates a packet in which an IP header is appended to TCP segments or UDP segments. The packet is delivered to the MPAD gateway 200 through the network layer 113D of L2/L1. According to one embodiment, the MPAD agents 107 and 107' may perform packet scheduling so that all packet flows are transmitted equally and redundantly to each network interface 101, 103, 105.

According to another embodiment, the MPAD agent (107, 107') transmits a portion of the packet redundantly to the plurality of network interfaces (101, 103, 105) and another portion of the packet to the plurality of network interfaces (101, 103, 105) Packets can be scheduled to be transmitted without overlapping to the network interfaces 101, 103, and 105 having the fastest speed or the least delay time among them. That is, the MPAD agents 107 and 107' transmit packets to multiple network interfaces 101, 103, and 105 in duplicate, while some packets can be simultaneously transmitted only to the network interfaces 101, 103, and 105 with a large transmission capacity. . Such a redundant transmission method can eliminate loss caused by down-converging to the performance of a network interface having the slowest delay in the case of simply overlapping transmission. When TCP is used as a transport layer, if one network interface among the plurality of network interfaces 101, 103, and 105 has a larger delay time than the other network interface, the same packet must be sent to a route with a low delay time, so the transmission speed is reduced. Latency converges to the slower side, which can slow down the entire TCP transmission itself. As such, in order to compensate for the disadvantage of downward convergence to the speed of a relatively slow network interface, some of the packets are transmitted redundantly, and scheduling is performed to send more non-duplicated packets to a relatively fast network interface. You can benefit not only from reliability but also for transmission speed.

In addition, in the case of redundantly transmitted packets, only the packets received with the lowest delay are taken and packets transmitted slowly are discarded to ensure low-latency transmission speed, and some packets are transmitted only through the path with the minimum delay time, thereby broadening the bandwidth. .

The redundant transmission processing unit 113 duplicates a portion of the packet flow to at least two network interfaces 101, 103, and 105, and the remainder except for a portion of the packet flow is the most among the plurality of network interfaces 101, 103, and 105. It is possible to transmit without overlapping only to at least one of the network interfaces 101, 103, and 105 having a high transmission speed or a minimum transmission delay. In this case, it may be transmitted without overlapping only to one network interface 101 , 103 , and 105 , or it may be divided and transmitted to at least two network interfaces 101 , 103 , 105 without overlapping.

Referring to FIG. 4 , the MPAD agents 107 and 107 ′ when the original packet buffer 10 allocated by the redundant transmission socket ( 111A in FIG. 3 ) is filled with a packet flow, a plurality of networks 301 , 303 , 305 ) decides which packet to send to each. Here, packets generated by the application ( 109 in FIG. 3 ) are stored in the original packet buffer 10 .

At this time, the MPAD agents 107 and 107' store the packets stored in the original packet buffer 10 in the divided packet buffers 11, 13 and 15 of the subflow. At this time, the MPAD agents 107 and 107' may split the original packet flow. That is, the MPAD agent 107, 107' distributes some packets 1, 2, 3 equally to the divided packet buffers 11 and 13 to be transmitted to the 5G network 301 and the LTE network 303, respectively, The remaining packets 4 and 5 are distributed only to the divided packet buffer 15 to be transmitted to the WiFi network 305 . That is, the MPAD agents 107 and 107' duplicate some of the original packets (1, 2, 3) to the 5G network 301 and the LTE network 303, and the rest (4, 5) to the WiFi network 305 ) can be scheduled for a single transmission.

The MPTCP or MPUDP layer 113A divides the packet added to the divided packet buffers 11, 13, and 15 of each subflow into TCP segments or UDP segments. The IP layer corresponding to each subflow generates a packet with an IP header attached to TCP segments or UDP segments. The packet is delivered to the MPAD gateway 200 through the L2/L1 network layer.

Here, when performing redundant transmission based on TCP, the MPTCP layer 113A performs redundant transmission according to its own flow control procedure.

On the other hand, in the case of performing redundant transmission based on UDP, if the MPUDP layer 113A directly duplicates a packet, it may be determined to be the same packet and discarded. Therefore, IP encapsulation is performed so that it can be recognized as overlapping but different packets.

Also, MPTCP and MPUDP may be merged to operate.

In this way, when the socket option is applied in the redundant transmission mode, the MPAD agents 107 and 107' transmit the uplink data transmitted from the application 109 to the MTU (Maximum Transmission Unit) mapped for each network interface 101, 103, and 105. ) by size, the amount to be transmitted is defined for the same or unequal conditions, and this is specified at every packet scheduling time through the available network interfaces 101, 103, and 105.

In the case of using TCP as the transport layer, unlike the redundant transmission method using UDP as a transport, there is no transmission overhead such as IP-in-IP and possible packet drop when using UDP. High-reliability transmission is possible. In addition, since TCP is used as a transport, it is possible to transmit with higher reliability compared to the UDP transport method, which is transmitted using the best-effort method, and since it is TCP, there is an advantage that 100% reliable transmission is possible.

Compared to the UDP transport method, even if the MTU size determined by the PMTU is different for each path, the advantage that TCP can take is that there is no overhead such as IP-in-IP, so the amount of transmission data (payload size) that can be transmitted at one time is large. Therefore, there is an advantage in terms of transmission speed.

The MPAD agent (107, 107') may preferentially process a packet delivered from the network interface (101, 103, 105) with the minimum delivery delay among duplicate packets delivered through the plurality of network interfaces (101, 103, 105). have. Through this, it is possible to minimize packet delivery delay and at the same time enable even distribution of the entire packet through an available path, thereby minimizing service fluctuations.

The MPAD agents 107 and 107' are connected to the application 109 through the merge transmission socket 111B, and open port #2 for multi-path merge transmission. The MPAD agents 107 and 107' output the packet flow transmitted through the merge transmission socket 111B to the merge transmission processing unit 115 .

The merge transmission processing unit 115 performs merge transmission through a plurality of available paths for a packet flow transmitted through the merge transmission socket 111B. In this case, the merge transmission processing unit 115 performs the merge transmission through at least two network interfaces selected by the MPAD agents 107 and 107'.

The merge transmission processing unit 115 includes an MPTCP layer 115A, a TCP layer 115B, an IP layer 115C, and an L2/L1 layer 115D.

The TCP layer 115B includes a number of subflow layers, each subflow mapped with a network interface 101 , 103 , 105 . Each subflow operates as an independent TCP session, so the MPTCP layer 115A provides independent TCP connection control and congestion control functions for each subflow. The MPTCP layer 115A performs connection control for a plurality of subflows, distribution of traffic between subflows, retransmission processing, and packet alignment functions.

The TCP layer 115B divides the packet of each subflow into TCP segments. The IP layer 115C corresponding to each subflow generates a packet with an IP header attached to TCP segments. The packet is delivered to the MPAD gateway 200 through the network layer 115D of L2/L1.

The MPAD agents 107 and 107' are connected to the application 109 through a single-path transmission socket 111C, and the MPAD agents 107, 107' open port #3 for single-path transmission. The MPAD agents 107 and 107' output the packet flow transmitted through the single-path transmission socket 111C to the single transmission processing unit 117 .

The single transmission processing unit 117 includes a TCP layer 117A, an IP layer 117B, and an L2/L1 layer 117C. The single transmission processing unit 117 may be connected to a plurality of network interfaces 101 , 103 , and 105 , but may be connected only to a default network interface (eg, 105 ) rather than selecting an optimal interface combination.

The TCP layer 117A divides the packet into TCP segments. The IP layer 117B generates a packet with an IP header appended to the TCP segments. The packet is delivered to the content server 600 via the Internet 500 via the L2/L1 network layer 117C.

The MPAD agents 107 and 107' can schedule packets in two ways in case of redundant transmission.

According to an embodiment, the MPAD agents 107 and 107' may always transmit the same packet to the plurality of network interfaces 101, 103, and 105 when a packet flow is transmitted from the application 109. This method is a method of transmitting according to the service requirements of the application 109 . Since all packets are transmitted equally to the plurality of transmission networks 301, 303, and 305, delay time can be reduced and transmission reliability can be improved.

According to another embodiment, when a packet flow is transmitted from the application 109, the MPAD agent 107, 107' transmits it to an available network interface among a plurality of network interfaces 101, 103, and 105, but transmission is no longer possible. For the network interface, the corresponding packet may be discarded. That is, the MPAD agents 107 and 107' may ignore the duplicate transmission request according to the network situation even though there is a request from the application 109 . The MPAD agents 107 and 107' may determine the optimal scheduling by judging the network condition even if there is a duplicate transmission request from the application 109. Therefore, this method can always select the optimal network while satisfying the transmission speed and delay time.

In proportion to the transmission amount of the slow path, if the overlapping transmission through multiple paths is divided into a part of the entire packet to be transmitted (hereinafter, packet group A), packet group A is the same for all data paths regardless of the path at the sender. is transmitted Packet group A transmitted in this way is all delivered through multiple paths prepared to the receiving side. Of these, only the data received with the lowest delay is taken and the relatively slow transmitted data is discarded to ensure the lowest delay transmission rate in terms of transmission delay. can do.

On the other hand, at the time of transmitting packet group A, the MPAD agents 107 and 107' are distinguished from packet group A, which is duplicately transmitted through that path, if there is a path through which more packets can be transmitted among the multiple paths it can transmit. It can be scheduled to transmit an additional packet (hereinafter, packet group B). Unlike packet group B, which is scheduled in this way, is transmitted through all paths regardless of path by packet group A, it is transmitted through only a part of the path. This is possible by the TCP inherent or sequence-based congestion control algorithm in subflow units of MPTCP, and since the transmittable amount, cwnd (contention window size), is managed differently for each path, packet groups A and B are delivered differently for each path.

In addition, the MPAD agents 107 and 107' can extend the applicable range of the transmission mode and the policy for handling it in various ways, such as for each specific user, for each specific location, for each specific address band, and the like.

Meanwhile, depending on the type of service, it may be more important for the packet reception period to be constant than to reduce the absolute delay time. However, jitter in the network is not an adjustable value, and it is difficult to predict its change. To solve this problem, the MPAD agents 107 and 107' reduce the difference in the reception period through the buffer, thereby making the packet reception period constant.

If the delay time difference between the different network interfaces 101, 103, and 105 is large during the merge transmission process, packets are received out-of-order. things can happen In addition, since the upper level understands it as a single interface transmission, it is misjudged as a situation in which the jitter of the network is very large. In order to prevent this problem, the MPAD agents 107 and 107' can store the packet in the reception buffer before forwarding the packet to the upper layer and adjust the transmission speed. Then, re-ordering in the buffer is also possible and the problem due to jitter is reduced because packets can be delivered at a constant rate.

The MPAD agent (107, 107') slows the speed of outputting packets to the upper layer as the delay time difference between the network interfaces (101, 103, 105) is larger, and the delay time difference between the network interfaces (101, 103, 105) is small. By increasing the speed at which packets are output to the higher layer, the transmission efficiency can be increased while precisely adjusting the reception period.

In case of packet loss, as in merged transmission, the performance difference between interfaces should be considered. However, if the delay time of a specific network interface 101, 103, 105 is short and there is no loss, packets received through the other network interfaces 101, 103, 105 are discarded and the network interface 101 with the shortest delay time , 103, and 105), it is necessary to accept only the packets received, so it is possible to adjust the speed of packet transfer from the buffer to the upper layer based on the network interfaces 101, 103, and 105 with the best performance. In this way, redundant transmission In the case of determining the packet forwarding speed to the upper layer, it is determined by combining the minimum delay time and the difference between the network interfaces.

Although not shown in the figure, in FIG. 3, a receive buffer is arranged between the TCP/UDP layer 113B and the MPTCP/MPUDP layer 113A and between the TCP layer 115B and the MPTCP layer 115A for each subflow, and the receive buffers are It is possible to adjust the difference in the reception period through The MPAD agents 107 and 107' measure the delay time difference between the network interfaces 101, 103, and 105 in the merge transmission mode, and as the delay time difference increases, the packet is output from the TCP layer 115B to the MPTCP layer 115A. The speed of outputting packets from the TCP layer 115B to the MPTCP layer 115A is relatively increased as the delay time difference is smaller.

The MPAD agents 107 and 107' operate in the same manner as in the merged transmission mode when packet loss occurs in the redundant transmission mode. On the other hand, if there is no packet loss, the speed at which packets are output from the TCP layer 115B to the MPTCP layer 115A is adjusted based on the network interfaces 101, 103, and 105 with the best performance.

In addition, an optimal combination of network interfaces is selected from among the plurality of network interfaces 101, 103, and 105 to transmit packets. Accordingly, it is possible to minimize a packet error rate that may occur due to a packet drop, a queue/buffer overflow, etc. that may occur on an intermediate node having a best-effort delivery system, thereby providing highly reliable communication. An optimal combination of network interfaces can be adaptively transmitted according to changing network conditions.

Referring to FIG. 5 , the MPAD agents 107 and 107 ′ have delay times for each network interface 101 , 103 , 105 between the terminals 100 and 100 ′ and the MPAD gateway 200 and the individual network interfaces 101 . , 103, 105) is the difference between the transmission time of the packet transmitted through and the arrival time of the ACK. The link capacity is defined as the maximum transmittable amount of each network interface 101 , 103 , 105 within a range in which packet loss does not occur. This latency and link capacity are continually used to compute the optimal combination of network interfaces.

The optimal combination of network interfaces may vary depending on whether the purpose of redundant transmission is to reduce latency and increase reliability, or whether to increase bandwidth is the main goal of merged transmission, and may also vary depending on the congestion control algorithm used.

In the case of redundant transmission, an interface having a short absolute delay time is mainly selected, and in the case of merge transmission, an interface combination having a large link capacity and a small delay time difference may be selected.

In addition, the optimal network interface combination is also changed according to the congestion control algorithm. Among the widely used algorithms, the bottleneck bandwidth and round-trip propagation time (BBR) algorithm is relatively more effective than the difference in delay time. The interface combination can be selected by giving weight to the capacity). In addition, in the case of the CUBIC algorithm, the interface combination can be selected by giving a relatively greater weight to the difference in delay time between network interfaces.

An optimal network interface combination is determined through Equation 1 below.

Here, d _i is the propagation delay of the network interface (i). C _i is the link capacity of the network interface (i). α and β are weights. n is the number of network interfaces. d _max is the maximum allowable propagation delay of the service. C _min is the minimum capacity required by the service. δ is a scalable coefficient.

Equation 1 is for all network interface combinations, that is, {5G}, {LTE}, {WiFi}, {5G, LTE}, {5G, WiFi}, {WiFi, LTE}, {5G, LTE, WiFi} It is an expression that calculates the value of the objective function. The MPAD agent 107, 107' may derive the optimal interface combination for which the value of the objective function is the minimum.

The objective function may determine which of delay time and link capacity to be more important according to a weight factor.

The objective function is designed so that it is possible to select whether to value the minimum delay among the delay times or the delay time difference between network interfaces. MPAD agents 107 and 107' may select an optimal combination by adjusting weights according to overlap/merge transmission and congestion control schemes.

은 네트워크 인터페이스들의 지연 시간을 계산하는 수식이다. 수학식 1의

은 네트워크 인터페이스들의 평균 링크 용량을 계산하는 수식이다. 따라서, α를 증가시켜 네트워크 인터페이스들의 지연 시간에 비중을 두거나 α를 감소시켜 네트워크 인터페이스들의 링크 용량에 비중을 둘 수 있다.of Equation 1

is a formula for calculating the delay time of network interfaces. of Equation 1

is an equation for calculating the average link capacity of network interfaces. Accordingly, it is possible to increase α to give weight to the delay time of the network interfaces, or decrease α to give weight to the link capacity of the network interfaces.

은 네트워크 인터페이스들의 평균 지연 시간을 산출하는 수식이다. 수학식 1의

은 각 네트워크 인터페이스의 지연 시간 편차를 산출하는 수식이다. 따라서, β를 증가시켜 네트워크 인터페이스들의 평균 지연 시간에 비중을 두거나 β를 감소시켜 각 네트워크 인터페이스의 지연 시간 편차에 비중을 둘 수 있다.of Equation 1

is an equation for calculating the average delay time of network interfaces. of Equation 1

is a formula for calculating the delay time deviation of each network interface. Accordingly, increasing β may give weight to the average delay time of the network interfaces, or decreasing β may give weight to the delay time deviation of each network interface.

MPAD agents 107, 107' may increase α to weight the interface delay or decrease α to weight the interface link capacity. MPAD agents 107, 107' may increase β to weight the average delay time or decrease β to weight the delay time variance. That is, as β increases, a link combination with a small average delay time is selected, and as β decreases, a link with a small delay time deviation is mainly selected.

For example, in a service with maximum latency (d _max )=30 ms, minimum link capacity (C _min )=10 Mbps, 5G latency (d _5g )=10 ms, LTE latency (d _lte )= 20 ms, WiFi latency _{Assume that} (d wifi ) = 50 ms, 5G link capacity (C _5g ) = 8 Mbps, LTE link capacity (C _lte ) = 5 Mbps, WiFi link capacity (C _wifi ) = 100 Mbps. At this time, the Wifi link capacity (capacity) is 100 Mbps, but since the WiFi delay time (d _wifi ) is larger than the service maximum delay time (d _max ), the corresponding link, that is, the Wi-Fi link is not selected.

In addition, in the case of 5G or LTE, although both delay times are satisfied, single link transmission cannot be selected because it has a capacity smaller than the _{service minimum link capacity (C min ).} Therefore, in the end, a {5G, LTE} combination is selected, through which redundant transmission and/or merged transmission is performed. If the WiFi delay time (d _wifi ) becomes 20 ms and all delay conditions are satisfied, {WiFi}, {5G, LTE}, {5G, WiFi}, {LTE, WiFi for an arbitrary weight factor }, {5G, LTE, WiFi} is calculated by Equation 1 and a combination with the smallest value is selected. In this way, the weight is determined by the importance of each of the average delay, the delay deviation, and the link capacity according to the characteristics of the service.

The MPAD agents 107 and 107' measure the delay time of each network interface and the link capacitor (or the transmission capacity of the link) at regular intervals, and calculate Equation 1 using these.

At this time, the MPAD agents 107 and 107' determine a condition to increase the relative weight among network interface delay and link capacity and a condition to increase the relative weight among delay time deviation and average delay time based on the destination information of the packet. The MPAD agents 107 and 107' receive information from the application 109 on the network interface delay, link capacity, and the condition with high weight among delay time deviation and average delay time, and adjust α and β based on the information.

In this case, the MPAD agents 107 and 107' may select an initial interface set by using statistical information-based prior information when starting merge transmission or duplicate transmission. To this end, the MPAD agents 107 and 107' may organize and retain the average delay time and max throughput information up to the MPAD gateway 200 for each network interface in the past communication in units of a predetermined time. The past communication refers to the entire communication (or a predetermined period in the past) that was performed before a specific point in time when a packet to be transmitted/received by the application 109 occurred. In addition, the subjects of the communication are the MPAD agents 107 and 107' and the MPAD gateway 200 . For example, when the terminal 100 wants to transmit a new packet, the average delay and throughput of each 5G, LTE, and Wi-Fi path between the terminal 100 and the MPAD gateway 200 is monitored for the last 3 days. use the result.

If such information is managed separately, it is difficult to predict the performance of each network interface at the start of transmission. The optimal combination of network interfaces can be configured more efficiently.

However, in the past max throughput information, when the amount of transmission/reception packets is smaller than the link capacity, a difference occurs from the actual link max capacity. Therefore, it should not be interpreted as an indicator of the maximum transmittable performance of that network interface.

If it is necessary to select an initial network interface without using prior information, the MPAD agents 107 and 107' randomly select one network interface, create a connection, and then add other network interfaces in sequence. have. In this case, the initial access delay time is relatively fast.

In addition, the MPAD agents 107 and 107' can select and apply a packet transmission method after creating a connection for the entire network interface. In this case, a wide bandwidth can be secured from the beginning.

The MPAD agents 107 and 107' continuously identify and update an optimal interface set even after determining the initial interface. MPAD agents 107 and 107' determine the link characteristics (link capacity and delay time) of individual interfaces, which can be identified from packet loss and RTT values for the network interface currently being used for transmission, and A network interface that is not busy can be actively identified by sending some data.

6 is a block diagram of a gateway according to an embodiment of the present invention.

Referring to FIG. 6 , the MPAD gateways 200 and 200 ′ receive packets from the terminal 100 and forward them to the content server 600 , and transmit the packets received from the content server 600 to the terminal 100 . . It is assumed that the content server 600 does not support multiple communication interfaces. MPAD gateways 200 and 200' are network devices for multi-path transmission. Gateways 200 and 200' are LTE interface 201, 5G interface 203, WiFi interface 205, Ethernet interface 207, MPAD server engine 209, duplicate transmission processing unit 211, merge transmission processing unit ( 213), and a TCP transmission processing unit 215 .

The redundant transmission processing unit 211 includes an MPTCP/MPUDP layer 211A, a TCP/UDP layer 211B, an IP layer 211C, and an L2/L1 layer 211D. The duplicate transmission processing unit 211 processes the duplicate packet received from the terminal 100 and then transmits it to the content server 600 , duplicates the packet received from the content server 600 , and transmits the duplicate packet to the terminal 100 . have. A specific operation is the same as that of the terminal 100 described with reference to FIGS. 3 to 5 .

The merge transmission processing unit 213 includes an MPTCP layer 213A, a TCP layer 213B, an IP layer 213C, and an L2/L1 layer 213D. The merge transmission processing unit 213 merges the packets received through the plurality of communication interfaces and transmits them to the content server 600 , and divides the packets received from the content server 600 to the terminal 100 through a plurality of communication interfaces. send to A specific operation is the same as that of the terminal 100 described with reference to FIGS. 3 to 5 .

The TCP transmission processing unit 215 includes a TCP layer 215A, an IP layer 215B, and an L2/L1 layer 215C. The TCP transmission processing unit 215 corresponds to a single-path transmission operation by performing communication with the content server 600 , and is the same as the operation of the terminal 100 described with reference to FIGS. 3 to 5 . However, the TCP transmission processing unit 215 is different from FIG. 3 in that it is connected to the content server 600 through the Ethernet interface 207 .

The MPAD server engine 209 performs a server operation corresponding to the MPAD agent 109 . At this time, the MPAD server engine 209, like the MPAD agent 109, also selects a transmission mode, selects an optimal network interface combination, and performs packet scheduling and reception processing for each transmission mode. However, since the MPAD server engine 209 does not know the network state experienced by the terminal 100 when selecting the transmission mode and the optimal network interface combination, it follows the request of the terminal 100 or responds to the request of the content server 600 . can operate accordingly.

The MPAD server engine 209 may select a transmission mode/optimal network interface combination by applying a policy based on a request to duplicate or merge a specific packet flow from the MPAD agent 107 . Alternatively, the MPAD server engine 209 may select a transport mode/optimal network interface combination according to the policy of the MPAD gateways 200 and 200' itself. Alternatively, when the MPAD server engine 209 receives a packet flow from the content server 600, it receives an on-demand request for duplicate transmission or merged transmission from the content server 600, and selects a transmission mode based thereon. . At this time, it is possible to receive an on-demand request for the optimal combination of network interfaces as well as the transmission mode. In order for the MPAD server engine 209 to receive an on-demand request, the content server 600 and the MPAD gateways 200 and 200' may establish a separate interworking structure. The interworking structure may use a mobile backend infrastructure, or a dedicated structure may be newly built.

7 is a flowchart illustrating an ultra-precision data communication method according to an embodiment of the present invention.

Referring to FIG. 7 , the computing device executing the operations of the MPAD agents 109 and 109 ′ and/or the MPAD server engines 209 and 209 ′ described in FIGS. 1 to 6 is redundantly transmitted in units of applications or packet flows. , and selects one transmission mode from among merge transmission and single path (S101).

When redundant transmission or merged transmission is selected (S103), the computing device applies different weights to the average delay time, delay time deviation, and link capacity according to the type of transmission mode and/or the type of congestion control algorithm to apply different weights to a plurality of network interfaces A combination of at least two network interfaces is selected from among (S105). In this case, step S105 uses Equation 1 described above.

The computing device performs redundant transmission or merged transmission through at least two selected network interfaces (S107).

On the other hand, when single path transmission is selected, the computing device transmits the packet through a single path through the default network interface (S109).

8 is a diagram illustrating redundant transmission through multiple paths according to an embodiment of the present invention, and illustrates a case in which a delay time of a 5G interface is minimal.

Referring to FIG. 8 , it is assumed that the terminals 100 and 100 ′ can transmit through three types of network interfaces: 5G, LTE, and WiFi.

In the case of a general transmission method, the terminals 100 and 100' transmit and receive packets through one network interface according to the highest routing priority among available network interfaces. However, if the delay time of the uppermost routing network interface is deteriorated, it is necessary to change to a suboptimal or suboptimal network interface.

However, in the case of the TCP protocol, when the source address of a service is changed due to a change in the routing path, a new session is added independently of the existing session, so that the service is interrupted. To compensate for this, using a multi-path aggregation transport protocol such as MPTCP, data can be segregated/aggregated based on the available transmission amount of each network interface depending on the situation. However, in the case of a service that needs to transmit packets within a specific time, for example, mission-critical or hard real-time, it may be vulnerable or unavailable depending on delay time fluctuations. have.

Accordingly, the delay-sensitive service can minimize the packet delivery delay time by transmitting the packet redundantly.

That is, the MPAD gateway (200, 200') stores the received (incoming) data (eg, packets No. 1 to No. 6) in a buffer, and transmits it back to the terminals 100 and 100'. In the packet scheduling step, The same redundant data (packets 1 to 6) is transmitted through the 5G, LTE, and WiFi routes to which the terminal is connected.

However, depending on the type of the network interface and the network condition, the arrival time of the packet is different for each network interface. It is assumed that the gray dotted line is the point at which it can be delivered through the terminal-side network interface and delivered to the application (socket). At this time, the 5G interface has a low latency, so packets 1, 2, and 3 are already possible, and the LTE interface is only capable of packet 1, whereas the WiFi interface has not yet received a single packet. In this case, packets 1, 2, and 3 of 5G with the lowest propagation delay are delivered to the application (socket), and packet 1 received through the LTE interface is discarded. That is, application services are possible through the 5G interface with the lowest transmission delay.

9 is a diagram for explaining redundant transmission through multiple paths according to another embodiment of the present invention, and describes a case in which WiFi transmission delay time is minimum.

In this case, it is assumed that a gray dotted line is a point in time that is received through the terminal-side network interface and can be delivered to an application (socket).

Referring to FIG. 9 , the situation changes in FIG. 8 to increase the transmission delay of the 5G interface and relatively minimize the transmission delay of the WiFi interface.

On the terminal 100, 100' side, based on the packets received from a plurality of available network interfaces, packets #1, #2, #3, and #4 of the WiFi interface that arrive first are first taken and delivered to the application (socket), and later Packet 1 of the incoming 5G interface and packet 1 of the LTE interface are discarded and the WiFi interface with the lowest transmission delay is taken. Accordingly, the overall transmission delay of the packet can be minimized.

10 is a diagram for explaining redundant transmission through a plurality of paths according to another embodiment of the present invention, and describes a transmission method pursuing high reliability.

In this case, it is assumed that a gray dotted line is a point in time that is received through the terminal-side network interface and can be delivered to an application (socket).

Referring to FIG. 10 , since the duplicate packets transmitted to the terminal through the 5G, LTE, and WiFi interfaces are based on IP communication by each network characteristic or intermediate nodes, the characteristics of the IP protocol are the best effort characteristics of incoming data. Drops may occur when the amount increases.

In addition, a situation in which smooth transmission is impossible due to a packet error or retransmission due to noise or interference may occur even on wireless access.

Therefore, in the example of 5G, LTE, and WiFi interfaces, when the duplicate packets delivered respectively reach the terminals 100 and 100', the dropped packets for each network interface are ignored and the original packet is returned to the original packet based on the normally arrived packet(s). Combinations are also possible.

Therefore, when the multi-path overlapping transmission according to the embodiment of the present invention is performed, it is possible to process in a substitutable structure when a packet error/drop occurs as well as a gain in delay time.

The embodiment of the present invention described above is not implemented only through the apparatus and method, and may be implemented through a program for realizing a function corresponding to the configuration of the embodiment of the present invention or a recording medium in which the program is recorded.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improved forms of the present invention are also provided by those skilled in the art using the basic concept of the present invention as defined in the following claims. is within the scope of the right.

Claims (12)

As an ultra-precision data communication method of a terminal connected to a multi-network through a plurality of network interfaces,
When a transport packet is generated, determining a transport mode matching the application ID or destination information indicating the service characteristics of the transport packet;
When the determined transmission mode is redundant transmission, combining the plurality of network interfaces to select a network interface combination having the minimum average delay time of the network interfaces included in each combination;
When the determined transmission mode is merge transmission, combining the plurality of network interfaces to select a network interface combination in which the delay time deviation of the network interfaces included in each combination is minimum; and
Transmitting the transport packets redundantly or merged transmission using the selected network interface combination
Including, ultra-precision data communication method. In claim 1,
The step of selecting a network interface combination having the minimum average delay time comprises:
measuring each delay time and link capacity for the plurality of network interfaces; and
Selecting a network interface combination in which an objective function using each of the delay time and link capacity is minimized,
The objective function is
The average delay time of the network interfaces included in each combination is calculated to have a higher weight than the delay time deviation, and the total delay time that is the sum of the average delay time and the delay time deviation is the average of the network interfaces included in each combination. An ultra-precise data communication method, which is calculated to have a weight higher than the link capacity. In claim 1,
The step of selecting a network interface combination having the minimum delay time deviation comprises:
measuring each delay time and link capacity for the plurality of network interfaces; and
Selecting a network interface combination in which an objective function using each of the delay time and link capacity is minimized,
The objective function is
The delay time deviation of the network interfaces included in each combination is calculated to have a weight higher than the average delay time, and the average of the network interfaces included in each combination is higher than the total delay time that is the sum of the average delay time and the delay time deviation. An ultra-precise data communication method, wherein link capacity is calculated to have a high weight. In claim 1,
The duplicate transmission is
A high-precision data communication method, wherein a portion of the transport packet is repeatedly transmitted through all network interfaces included in the selected combination, and another portion of the transport packet is transmitted without overlapping through some network interfaces included in the selected combination. In claim 4,
Some of the network interfaces are
A high-precision data communication method in which a network interface having a maximum network performance based on a transmission speed or a delay time is selected from among a plurality of network interfaces included in the selected combination. In claim 1,
After the step of transmitting,
receiving duplicate transport packets using the selected network interface combination; and
Discarding duplicate transport packets received from the remaining network interfaces except for duplicate transport packets received from a network interface having a minimum delay time among network interfaces included in the selected network interface combination;
Further comprising a, ultra-precise data communication method. In claim 1,
After the step of transmitting,
storing transport packets received using the selected network interface combination in a receive buffer; and
Adjusting the output speed of the transmission packet stored in the reception buffer to the application layer according to the transmission mode
Further comprising a, ultra-precise data communication method. In claim 7,
The adjusting step is
In the case of merge transmission, the larger the delay time difference between the interfaces in the selected network interface combination, the slower the output speed, and the smaller the delay time difference, the higher the output speed is, the ultra-precise data communication method. In claim 7,
The adjusting step is
In the case of redundant transmission, when packet loss occurs, the larger the delay time deviation between the interfaces in the selected network interface combination, the slower the output speed, and the smaller the delay time deviation, the higher the output speed.
If there is no packet loss, the output speed is adjusted based on the packet reception speed of a network interface having the largest link capacity among interfaces in the selected network interface combination. As an ultra-precision data communication method of a terminal connected to a multi-network through a plurality of network interfaces,
Determining a redundant transmission mode matching the application ID or destination information indicating the service characteristics of the transport packet;
combining the plurality of network interfaces to determine a network interface combination having a minimum average delay time; and
Transmitting the transport packet in duplicate using the determined network interface combination;
The transmitting step is
A portion of the transport packet is repeatedly transmitted through all network interfaces included in the selected combination, and another portion of the transport packet is transmitted without overlapping through some network interfaces included in the selected combination. In claim 10,
The transmitting step is
Check the available link capacity of the network interfaces included in the determined network interface combination, and exclude network interfaces whose available link capacity does not satisfy the minimum link capacity according to the service characteristics of the transport packet from the redundant transmission path. communication method. In claim 10,
The determining step is
measuring latency and link capacity for each of the plurality of network interfaces;
combining the plurality of network interfaces and calculating a delay time deviation, average delay time, and average link capacity of the network interfaces for each combination using the delay time and the link capacity, and
A network interface combination in which the objective function in which the average delay time is given a weight higher than the deviation of the delay time and the total delay time that is the sum of the average delay time and the deviation of the delay time is weighted higher than the average link capacity is the minimum step to determine
Including, ultra-precision data communication method.

Images