JP2001060956A - Transport layer multi-link communication method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make communication connected with Internet or the like efficient and improve its speed, by satirically supplying a route at every external connection in a network layer, binding each kind of communication executed at every external connection in both end transport layers and supplying them as one kind of communication to a high order layer. SOLUTION: IP addresses and port numbers are given at the both ends of respective cub-flows (A) 210 and (X) 220 in the same way as TCP. A candidate address and the candidate port number coming them designate an address and a port at the Source side of a total flow 200. Besides, a sequence number concerning a transfer packet is given as the sequence number of TCP. In addition to it, a total sequence number, for example, as the arrangement order of data in the case of binding the sub-flows (A) 210 and (X) 220 is given as the option of TCP. At the side of a Destination, received data is re-constituted in order of the total sequence number.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

TECHNICAL FIELD The present invention relates to a host connected via the Internet or a private network,
More specifically, the present invention relates to a transport layer multilink communication method suitable for improving communication efficiency in a multihome connection environment having a plurality of external connections to the Internet or the like.

[0002]

2. Description of the Related Art In communication in the Internet environment,
Currently, the TCP / IP protocol group is used for communication between devices such as a host and a router. In the TCP / IP protocol group, each protocol is hierarchically configured, and as shown in FIG. 2, the configuration is “link layer”, “network layer (IP)”, “transport layer (TCP or UDP)”. It consists of an "application layer".
The "link layer" controls physical aspects such as a physical interface and a driver. The “network layer” controls the movement of packets from the end to the end (End-to-End) on the Internet, and is based on IP (Internet Protocol) and IP.
CMP (Internet Control Message Protocol) and the like are used. The “transport layer” controls the flow and multiplexing of data between two hosts, and provides reliable communication TCP (Trnsmission Control Protocol) and UDP (UserData) which provides simple datagram transfer.
gram Protocol). The “application layer” performs processing dependent on each application.

On the other hand, in connection between organizations in the Internet environment, there is a single connection form in which a single external organization is connected to perform all Internet communication through the single external organization, and a direct connection with a plurality of external organizations. In addition, there are used two connection forms of a multi-home connection form in which the external connection is appropriately used to communicate with the Internet.

In the Internet, the "network layer"
The routing of packet delivery is controlled by the IP, which is the protocol described above. In IP, a route is selectively used, and usually, a route of a packet destined for a certain destination is operated so as not to change (stable) as much as possible. In the delivery of packets destined for a multihomed network in a multihomed environment, the multihomed network broadcasts routing information addressed to its own network to each of a plurality of external connections, and the external network sends By determining the delivery route based on the route information,
A plurality of routes have been selected. In the delivery of packets from a multihomed network, a route is determined in the own network based on route information obtained from each of a plurality of external connections, so that selection from a plurality of routes can be performed. Is being done.

[0005] In the data link layer, a multi-link technique is used as a technique for increasing the speed by using a plurality of low-speed links. This is known as PPP (Point-to-Ponit).
It is implemented as an extension of the Protocol, and is mainly used to bundle a plurality of homogeneous data links.

[0006]

In the current framework of "route control in the network layer", when a multihomed network communicates with a specific partner of an external network, a route is established in the network layer. However, since this cannot cope with dynamically (temporally) fluctuating traffic conditions in the network, the route is not always the best route for the transport layer communication. In other words, the routing control at the network layer
Path control according to path weight, priority, path length, etc., does not represent a state in which actual congestion at the moment is considered, and the selected path is not necessarily the fastest. Absent.

Further, from the viewpoint of a multi-home network, since a plurality of routes are selectively used, a plurality of external connections cannot be effectively used up. Therefore, even if there is room in a band other than the path used for the current communication, the path cannot be used. Further, the amount of routing information increases as compared with the number of multihomed networks, which places a burden on routing.

An object of the present invention is to enable a dynamic communication path selection according to an actual data transfer situation and efficient use of a plurality of external connections in a multihomed connection environment.
It is an object of the present invention to improve the efficiency and speed of communication between a host, a router, and the like connected via the Internet or a private network.

[0009]

According to the present invention, in the network layer, a plurality of routes for each external connection are statically provided, and each communication (each link) performed for each external connection route is transported at both ends. Bundle in layers (this is called multilink),
It is provided as one communication to the upper layer.

[0010] Flow control according to the communication performance of the path is performed by communication (subflow) for each path, and distribution of the load according to the state of each subflow in the total flow obtained by bundling the subflows, and subflow as needed. Process deletion / addition of.

According to the present invention, in a multihomed connection environment, communication can be performed using an optimum path from among a plurality of paths according to the communication efficiency at the time of communication in the transport layer. In addition, efficient communication can be performed using a plurality of routes.

[0012]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a conceptual diagram of a protocol stack in the transport layer multilink communication of the present invention. Here, the configurations of the ring layer, network layer, transport layer, and application layer are basically the same as those in FIG. 2, but the transport layer is divided into TCP, UDP, and multilink. A set of TCP and UDP of the link layer, the network layer, and the transport layer virtually exists for each path of a plurality of external connections (three sets in FIG. 1).
Each communication (link) performed for each route is bundled (multilink) in a multilink layer of a transport layer, and provided as one communication to an upper application layer. Hereinafter, an embodiment in which the TCP is extended to realize the transport layer multilink communication of the present invention will be described in detail.

Each host utilizing the transport layer multilink communication has a plurality of IP addresses physically or logically. Each of the plurality of IP addresses of one host corresponds to each external connection.
The route used for packet transfer according to the source IP address is the corresponding external connection. That is, a plurality of routes are realized by a combination of a plurality of source IP addresses (source addresses) and destination IP addresses (destination addresses).

FIG. 3 shows a conceptual diagram of the transport layer multilink communication when there are two external connections as an example.
In FIG. 3, reference numeral 10 denotes a host on the transmission side (source side);
0 is the host on the receiving side (destination side).
The host 10 is connected to external networks A and X through routers 101 and 102, respectively. In response to each of these external connections, the host 10 sets “AB.
s1. s2 "and" XYt1.t2 ". The IP address of the host 20 is “D”. The packet 110 is transferred from the host 10 to the router 101 and the host 20 via the route of the network A by the combination of the source address (source address) “ABs1.s2” and the destination address (destination address) “D”. Is done. The packet 120
Is transmitted from the host 10 to the router 1 by the combination of the source address “XYt1.t2” and the destination address “D”.
02, transferred to the host 20 via the network X route. Note that a TCP / IP packet also includes a source port number (source port number) and a destination port number (destination port number), but these are omitted in FIG. 3 for simplicity.

Here, communication for each path passing through each external connection is called a subflow, and end-to-end communication that bundles the subflows is called a total flow. The application uses this total flow in the same way as conventional communication.

FIG. 4 is a conceptual diagram showing the relationship between the subflow and the total flow. In FIG. 4, 200 is the total flow, and 210 and 220 are the sub-flows A and X. Each of the subflows 210 and 220 performs flow control and congestion control in accordance with the communication performance of the path in the same manner as ordinary TCP. Processing such as allocation to transfer data and deletion / addition of subflows is performed.

Next, the processing in TCP will be specifically described. Both ends (source, destination) of each subflow are provided with an IP address and a port number similar to those of normal TCP. Here, one IP address among the IP addresses assigned to a plurality of subflows for each end of communication is called a representative address. Similarly, any one of the port numbers is called a representative port number. The representative address and the representative port number are used to designate an address / port on the local side (source side) of the total flow. As a TCP sequence number, a sequence number (subflow sequence number) related to a transfer packet for each subflow is assigned, and each subflow is processed as one connection of the conventional TCP. In addition to this, a total sequence number is given as an option of TCP as an arrangement order of data when bundling subflows.
The total sequence number indicates the order of data before being divided into each subflow, and the receiving side reconstructs the data received in each subflow in the order of the total sequence number.

FIG. 5 shows a specific example of a sub / total sequence number assigned to a transfer packet of each subflow. This means that data having a configuration as shown in FIG.
0 to the host 20 indicate that data 1, data 2, and data 5 are transferred on the path of the subflow (A) 210, and data 3 and data 4 are transferred on the path of the subflow (X) 220, respectively. In this case, the host 10 on the source side attaches 1, 2, and 3 as TCP sequence numbers (subflow sequence numbers) to data 1, data 2, and data 5, respectively, and similarly assigns TCP sequence numbers 1 and 2 to data 3 , Data 4 respectively. This is the same as the conventional one. In addition to this, host 1
In the case of 0, total sequence numbers 1 to 5 are added to data 1 to data 5 using a TCP option. FIG.
Although not shown, each packet is provided with a pair of source / destination IP address and port number, whereby data 1, data 2 and data 5 are transferred on the path of subflow (A) 210 and data 3 And data 4 are transferred on the path of the subflow (X) 220. In the destination host 20, each subflow 210, 2
The data received via 20 is bundled as a total flow 200, and the data 1 to 5 shown in FIG. 6 are reconfigured according to the total sequence number assigned to each data.

In FIG. 5, a set of an IP address and a port number enclosed by a broken line "ABs1.s2, P1"
Indicates a representative address / representative port number. Of course,
Other “XYt1.t2, P2” may be used as the representative address / representative port number.

The connection of each subflow is the same as the TCP connection procedure. Procedures for starting up and bundling a plurality of subflows and releasing unnecessary subflows include SETUP, ATTACH, DETACH, and SHU.
There are five types, TDDOWN and RECOVERY.

SETUP is a process for setting up a multilink connection.
There are ential setup and concurrent setup that connects multiple subflows in parallel and bundles them later.

In Sequential Setup, first, MP (Multi-link E) is applied to a subflow passing through the first route.
(xtension Protocol) _REQUEST is set up and the TCP is set up by 3-way handshake. If the communication partner supports the transport layer multilink, an ACK with an option of MP_GLANTED is returned at the time of 3-way handshake. The MP_GRANTED message is provided with an identifier (flow identifier) for identifying the total flow (one for each direction of connection).
D is given). If the connection to the first route fails, a similar setup is performed by selecting a subflow passing through another route. In the case of Sequential Setup, the local address and port of this first connection are the representative address and representative port. When the setting of the subflow is completed, communication is started using the subflow (each packet is given a total sequence number in addition to a normal sequence number). As for subflows of other routes, subflows are added by ATTACH processing described later. FIG. 7 is a conceptual diagram of Sequential Setup.

On the other hand, in the Concurrent Setup, a connection of an arbitrary number of subflows larger than one of a plurality of paths is set up simultaneously. In this case, first, the local address and port number are acquired for each path set up at the same time. Then, when setting up each subflow, a flow identifier is added to the option of MP_REQUEST. This flow identifier is used to identify that a plurality of TCP connections (a plurality of subflows) are one multilink connection. If the person receiving the connection supports the transport layer multilink, the other party's address and port of the connection request received first are set as the representative address and representative port of the other party (connection start side) of the connection, Add MP_GRANTED option with its representative address, representative port and flow identifier
-Returns the way handshake ACK. For the second and subsequent subflows, MP_GRA indicating the representative address / representative port of the total flow and the flow identification ID
Return NTED with ACK. FIG. 8 shows Concurrent Set.
The conceptual diagram of up is shown.

In the ATTACH, a subflow is added (see FIG. 7). At the time of ATTACH, a subflow is connected to a partner corresponding to multilink by attaching an MP_ATTACH option and a flow identifier.
On the other side, if the total flow represented by the representative address and the representative port is still open, MP_ATTAC
Return HACK with HED. When the setup of the sub-flow is completed, a part of the data of the total flow is started (hereinafter, each packet is given a total sequence number in addition to a normal sequence number). This ATTACH
Subflow addition by the above-mentioned Sequential Setup
Other than that, it may be generated when the transfer amount of the total flow suddenly increases, when the state of other subflows largely fluctuates, or when a periodic trigger for a certain period of time occurs.

DETACH deletes a subflow.
Deletion of subflows by DETACH is the same as termination of normal TCP connection for each subflow. However, the subflow used as the representative port number is D
When performing ETACH, the port number cannot be reused until the total flow is completed. Deletion of a subflow occurs when the total performance does not change even when a subflow is added, or when the transfer is smaller than other flows.

SHUTDOWN is a process for completing communication in the total flow. In the case of SHUTDOWN,
Each subflow is terminated in the same manner as normal TCP, and if the subflow is already terminated but the port number is used as the representative port number, the port number is released.

RECOVERY is a process performed when a subflow is suddenly disconnected. In the steady state, the retransmission process is performed in each subflow. Therefore, in the total flow, it is only necessary to arrange the data in the order of the total sequence number. The data after the point where it is not confirmed is retransmitted as a total flow.

Finally, a layer lower than the network layer required to realize the above processing will be described.

In order to realize the transport layer multilink communication of the present invention as described above, it is necessary to configure a plurality of network layers for each external connection. The solution is to multiplex physical data links.
Means such as “multiplexing of logical data link” and “multiplexing of IP network” can be considered. In the "multiplexing of physical data links", as shown in FIG. 9, one physical data link is prepared for each external connection, and a plurality of end hosts are connected to each data link. For example, one host 10 has a plurality of Ethernet (registered trademark) interfaces, each of which is connected to a different physical network. As shown in FIG. 10, "multiplexing of logical data links" refers to a virtual LAN (virtual LAN, VLA) prepared by the data link layer.
N) Using a function, a single physical interface is divided into a plurality of virtual logical data links, and each logical data link is constructed as a network connected to each external connection. Specifically, IEEE802.
Ethernet VLA found in 1Q
A VC (Virtual Circuit) of N or ATM corresponds to this. In the “multiplexing of IP networks”, as shown in FIG. 11, a plurality of IP addresses are assigned to each host for each external connection, and an intermediate router uses a set of a source IP address and a destination IP address. It controls the route. Specific examples of such a router function include froute and SuMiRe.

As described above, the present invention has been specifically described based on the embodiments. However, the present invention is not limited to the embodiments, and it is needless to say that various modifications can be made without departing from the gist of the present invention. is there. For example, in the embodiment, communication between end hosts is assumed. However, when a router physically or logically has a plurality of addresses, multilink communication can be similarly performed between routers. Further, in the embodiment, the example in which the TCP is extended to realize the multilink connection is described. However, the multilink connection can be similarly implemented by extending the UDP.

[0031]

As described above, according to the present invention, in a communication from a network having a plurality of external connections to an external network, a plurality of external connection paths are bundled in the transport layer to provide a plurality of external connection paths. Efficient use of external connections and dynamic communication path selection according to the actual data transfer situation become possible. Thereby, communication on the Internet or the like can be performed more efficiently and at higher speed.

[Brief description of the drawings]

FIG. 1 is a conceptual diagram of a protocol stack for transport layer multilink communication according to the present invention.

FIG. 2 is a diagram showing a protocol stack in a normal TCP / IP suite.

FIG. 3 is a diagram showing a relationship between an address and packet transfer of transport layer multilink communication according to the present invention.

FIG. 4 is a diagram illustrating a relationship between a subflow and a total flow of the transport layer multilink communication according to the present invention.

FIG. 5 is a diagram illustrating a relationship between a subflow sequence number and a total flow sequence number of transport layer multilink communication according to the present invention.

FIG. 6 is a diagram showing a data structure for explaining a relationship between a subflow sequence number and a total sequence number in FIG. 5;

FIG. 7 is a conceptual diagram of Sequential Setup in the transport layer multilink communication of the present invention.

FIG. 8 is a conceptual diagram of Concurrent Setup in the transport layer multilink communication of the present invention.

FIG. 9 is a diagram showing a multiplex configuration of a physical data link for realizing the transport layer multilink communication of the present invention.

FIG. 10 is a diagram showing a multiplex configuration of a logical data link for realizing the transport layer multilink communication of the present invention.

FIG. 11 is a diagram showing a multiplex configuration of an IP network for realizing the transport layer multilink communication of the present invention.

[Explanation of symbols]

 10, 20 host 30 network 101, 102 router 110, 120 packet 200 total flow 210, 220 subflow

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Toshihiro Takada 2-3-1 Otemachi, Chiyoda-ku, Tokyo Inside Nippon Telegraph and Telephone Corporation (72) Inventor Kensuke Fukuda 2-3-1 Otemachi, Chiyoda-ku, Tokyo No. 1 Nippon Telegraph and Telephone Corporation F term (reference) 5K030 GA03 HC01 HC13 HD03 LB06 LE03 MB13 5K033 AA01 AA03 BA05 CC01 DA06 DB18 5K034 AA01 AA07 DD03 KK28 9A001 CC02 CC06 JJ12 JJ25 KK56

Claims (8)

[Claims] In communication in a multi-home connection environment having a plurality of external connections to a network, communication for each external connection path is bundled in a transport layer and provided as one communication to an upper layer. A transport layer multilink communication method, comprising: 2. The transport layer multilink communication method according to claim 1, wherein a communication for each external connection path is defined as a subflow, a communication obtained by bundling the subflows is defined as a total flow, and each subflow has a communication performance of the path. A transport layer multi-link communication method characterized by performing flow control according to the sub flow and allocating loads according to the state of each sub flow and deleting / adding sub flows in the total flow. 3. The transport layer multilink communication method according to claim 1, wherein the transfer packet includes, in addition to a communication sequence number for each route, a total order indicating a bundled communication for each route. A transport layer multilink communication method, characterized by providing a sequence number. 4. The transport layer multilink communication method according to claim 1, wherein when starting up a multilink connection, the number of connections is increased one by one. Method. 5. The transport layer multi-link communication method according to claim 1, wherein when starting up a multi-link connection, a plurality of connections are established in parallel and bundled later. Multilink communication method. 6. The transport layer multilink communication method according to claim 1, wherein a physical data link is prepared for each external connection, and an end host or a router is connected to each data link. Transport layer multilink communication method. 7. The transport layer multi-link communication method according to claim 1, wherein a single physical interface is divided into a plurality of virtual logical data links, and each logical data link is connected to each external connection. A transport layer multilink communication method, which is constructed as a connected network. 8. The transport layer multilink communication method according to claim 1, wherein a host or a router connected via a network has a plurality of IP addresses physically or logically, and each IP address is A transport layer multilink communication method characterized by being associated with an external connection.