BE1023707B1 - MULTI-PATH TCP IN HYBRID ACCESS NETWORKS - Google Patents

Abstract

A method of exchanging data over a TCP connection (234) between a client node (100) and a networking node (103). The TCP connection includes a primary MPTCP sub-stream (122) between a customer premises equipment, an HCPE (101), and a hybrid access gateway, a HAG (102). By the HCPE, we convert (410, 413) of the first TCP segments (401, 406) to the first MPTCP segments (402, 405) of the primary MPTCP sub-stream and vice versa and destination addresses of the first TCP segments are used in information destination address information of the first MPTCP segments, and source addresses of the first MPTCP segment are used as source address information of the first TCP segments. By the HAG, we convert (411, 412) of the second TCP segments (403, 404) into second MPTCP segments (402, 405) of the primary sub-stream and vice versa

Description

FIELD OF THE INVENTION [01] The present invention generally relates to the field of network connectivity provided to clients by a hybrid access network. In such a hybrid access network, a client can access a server in an external network such as the Internet through hybrid subscriber premises equipment or HCPE that connects to a hybrid access gateway or HAG on more than one access network.

[02] More particularly, the invention relates to the provision of multi-path transmission control protocol (TCP) capabilities between such HCPE and such HAG within a single-path TCP connection between the client and the server.

BACKGROUND OF THE INVENTION [03] Subscriber premises equipment connects clients or local area networks to an access network of an Internet service provider or I SP. The SP then provides Internet access to clients by connecting the access network with the Internet via a gateway on the provider's core network.

[04] With the higher bandwidths provided by the latest wireless technologies such as for example LTE, hybrid subscriber premises or HCPE facilities have been introduced providing network access on both wired and wireless technologies. In order to further amplify the bandwidth for the user, ways of using multiple access networks simultaneously have been introduced. One of them is the use of multipath protocol TCP (MPTCP) which is specified by the I ETF in RFC 6824. The protocol makes it possible to establish a TCP connection between a client and a server node . The connection then comprises several sub-streams, sub-streams MPTCP, on which the data is transferred.

[05] However, a disadvantage of the protocol is that both the endpoints, i.e. the client and the server, must support MPTCP in order to make full use of it. Therefore, the protocol does not support the establishment of different sub-streams between intermediate nodes of a TCP connection, such as between the HCPE and the gateway on the different access networks.

[06] In EP2882148, a solution is provided to address the above disadvantage and, thus, to provide multipath capabilities between a HCPE and a gateway. Both the HCPE and the gateway serve as an intermediate proxy between a client connecting to a server. When the client connects to the server, the HCPE will convert the client's TCP synchronization segment into an MPTCP synchronization segment and address it to the gateway with the server address included in an optional field of the segment. The gateway in turn converts the MPTCP segment back to a TCP segment and the address to the server. In this way, a three-way dialogue is performed and a TCP connection between the client and the server is established.

[07] A disadvantage of the above solution is that the server sees the gateway as the source of the connection and not the client or the HCPE. The use of HAG and HCPE is thus not transparent to the server. Another disadvantage is that the HCPE must provide the server address as an optional field. This increases the size of critical time synchronization segments and is further not optimal because the optional fields are limited in size. Another disadvantage is that the traffic inside the access network and the central network of the SP is addressed to the gateway and not to the server. This is problematic for legal reasons due to the fact that the SPs are forced to keep track of network activity and thus servers that a subscriber and so a client is trying to reach.

In order to do this, the SP will have to link the intercepted traffic to information from the gateway. This is complex and therefore expensive.

[08] It is an object of the present invention to overcome the above disadvantages and to provide a way to provide an end-to-end transparent TCP connection between a client and a server while using multi-path capabilities between the HCPE and a bridge. SUMMARY OF THE INVENTION [09] This objective is achieved, in a first aspect, by a method for exchanging data over a TCP connection between a client node and a networking node. The TCP connection includes a primary MPTCP sub-stream on a first access network between a hybrid subscriber premises equipment, a HCPE serving as the first proxy node and a hybrid access gateway, an HAG serving as a second node. representative. The method comprises the following steps: the conversion, by the HCPE, of first TCP segments into first MPTCP segments of the primary MPTCP substream and vice versa; and - using, by the HCPE, destination address information of the first TCP segments as destination address information of the first MPTCP segments; and - using, by the HCPE, source address information of the first MPTCP segments as source address information of the first TCP segments; and - converting, by the HAG, second TCP segments into second MPTCP segments of the primary sub-stream and vice versa while preserving source and destination address information.

[10] The client is thus connected to the networking node, for example a server, via the HCPE and the HAG. HCPE provides client access to more than one access network, for example a digital subscriber line (DSL) and an LTE network. The HAG in turn serves as a gateway between the access networks and an external network such as the Internet in which the networking node is located. In other words, the HAG is between the external network and the HCPE. The TCP connection comprises a first monotrajet part between the client and the HCPE, then a multi-path part between the HCPE and the HAG and then a single-path part between the HAG and the server. The multi-path portion includes the primary MPTCP sub-stream, i.e., the sub-stream that was first configured upon establishment of the MPTCP connection. The multi-path portion may include one or more additional or auxiliary substreams between the HCPE and the HAG established on other access networks. The TCP connection can be initiated from the client node or from the networking node.

[11] TCP segments sent from the client to the networking node are intercepted by the HCPE which converts them into MPTCP segments. If the HCPE decides to send a segment to the primary substream, it does not change the destination address in the segment. Depending on whether the HCPE applies Network Address Translation (NAT) or not, it can change the source address information in the segment. Subsequently, the HAG will intercept the segment as all traffic to HCPE will pass through the HAG. The HAG then converts the MPTCP segment back to a TCP segment without changing the source or destination address information. In this way, the HAG and the MPTCP are transparent to the networking node and the networking node believes that it is receiving single-path TCP segments from the client or HCPE. In the other direction, segments sent by the networking node to the client are intercepted by the HAG and converted to MPTCP segments. Again, if the HAG decides to send the segments to the primary MPTCP substream, it does not change the source or destination address of the segments. At the HCPE level, the segment is converted to a TCP segment while preserving the source address and passed to the client. Depending on the NAT function at the HCPE level, the destination address can be translated to the client.

[12] In other words, both the client and the networking node experience a single path TCP connection while a multipath is supported for the connection between the HCPE and the HAG. This is an advantage that the use of multipath is transparent to both the client and the networking node. This also implies that network logging for legal purposes can be done at any location before or after the HAG. Transparency is also advantageous for traffic engineering for IP sources and destinations, for class of service (CoS) marking, for a traffic report by, for example, the traffic flow information export protocol. I nternet protocol (I PF IX) or the sampled flow protocol (sflow) and for security applications such as, for example, intrusion detection system (I DS) and Internet provider security (I DS) tags ( I PS). In addition, it is an advantage that the server address can be provided in the destination field of the MPTCP segments and does not need to be provided in an optional field, thus improving protocol compatibility and reducing the size. segments.

[13] Since the HCPE and the HAG maintain an MPTCP connection state, the TCP connection may further include an auxiliary MPTCP substream on a second access network between the HCPE and the HAG. The method then further comprises: at the level of the HCPE, the conversion of third TCP segments into third MPTCP segments of the auxiliary MPTCP sub-stream and vice versa; and at the level of the HAG, the conversion of fourth TCP segments into fourth MPTCP segments of the auxiliary sub-stream and vice versa.

[14] As provided by the MPTCP protocol, the HCPE or the HAG can then decide on which sub-stream to send the MPTCP segment based on current network performance or network policies. Similar to the primary substream, both HCPE and HAG convert between TCP and MPTCP states and vice versa.

[15] According to one embodiment, the conversion further comprises: by the HCPE, the use of destination address information of the HAG for the third MPTCP segments; and - by the HAG, using destination address information of the networking node for the fourth segments sent to the networking node.

[16] In order to route the segments differently for the auxiliary sub-stream, the HCPE now sends an MPTCP segment with the data of the TCP segment to the network address of the HAG in the second access network. In this way, the segments are routed on the auxiliary sub-stream to the HAG. At the HAG, the destination address is again replaced by that of the networking node.

[17] Due to the fact that all segments of the sub-sub-stream need to be addressed to the HAG directly, the network operator may use policies that prevent the client or HCPE from using the second network. access to access the networking device directly. In this way, the use of the second access network can be controlled in an easy and simple manner.

[18] According to one embodiment, the method further comprises, in order to establish the auxiliary MPTCP substream: - sending, by the HAG to establish the auxiliary MPTCP substream, a segment comprising an address of the HAG in the second access network on the primary MPTCP substream.

[19] The HAG thus announces its address to establish the auxiliary MPTCP substream to HCPE. This has the advantage that the HAG can, at any time, determine when the HCPE, and thus the client, can use the auxiliary MPTCP substream.

[20] The HCPE may initiate the auxiliary sub-stream by sending a segment including a request to establish the second MPTCP sub-stream to the HAG. Alternatively, the HAG may initiate the auxiliary sub-stream by sending a segment including the request to establish the second MPTCP sub-stream to the HCPE. In this case, it is not necessary to send the HAG address in the second access network on the primary MPTCP substream if the HAG knows the alternate address of the HCPE.

[21] Alternatively, both the primary and auxiliary MPTCP substreams can be established by: - by the HAG, the reception of a synchronization segment from the HCPE indicative of a request to establish the MPTCP substreams primary and auxiliary; and - subsequently, by the HCPE, receiving a synchronization and acknowledgment segment from the HAG; and - the establishment by HCPE of the primary and auxiliary MPTCP substreams; and - the receipt by HAG of an acknowledgment segment from HCPE; and - subsequently, by the HAG, the establishment of the primary and auxiliary MPTCP substreams.

[22] The primary and auxiliary MPTCP substreams are thus established on the HCPE side after the exchange of two messages, i.e., the initial synchronization segment and then the receipt of the corresponding acknowledgment from the HAG. In other words, the HCPE establishes the primary and secondary TCP substreams based on the synchronization segment and the received acknowledgment. There is no other need for additional dialogue messages.

[23] On the HAG side, the sub-flows are established after the exchange of the three messages, i.e., the initial synchronization segment, the acknowledgment with the synchronization segment from the HAG and then the acquittal from HCPE.

[24] It is thus an advantage that the two sub-streams can be established by a single three-way dialogue. The number of segments exchanged and the duration of establishment are thus independent of the number of auxiliary substreams. In addition, the same amount of TCP segments and MPTCP segments are exchanged. Both the HCPE and the HAG thus perform a one-to-one conversion of the segments during the establishment of the TCP connection.

[25] In addition, the exchanged messages can be made backward compatible with segments used in the current MPTCP protocol, that is, respectively the SYN segment with the MP_CAPABLE option, the SYN + ACK segment with the MP_CAPABLE option and the ACK segment containing the MP_CAPABLE option. If the message can be the same, the interpretation given to it by the client and the server is different and, therefore, the sub-flows are established only by a single three-way dialogue.

[26] In another embodiment, the HAG comprises a plurality of gateway servers each acting as a gateway between the first access network and the networking node; and the method further comprises: - by the HAG, load balancing of the primary MPTCP substream on one of the gateway servers.

[27] Since the HAG is a throttling to the external network, the primary MPTCP substreams are load balancing on the different servers, ie, all segments of a certain TCP connection are affected to the same gateway server. Since the auxiliary sub-stream is assigned to a separate address, the load balancing of the sub-sub stream can then be accomplished by providing the HCPE with the network address of the gateway server to which the primary sub-stream is assigned. .

[28] According to a second aspect, the invention relates to a method for exchanging data over a TCP connection between a client node and a networking node; the TCP connection comprising a primary MPTCP sub-stream on a first access network between a subscriber premises equipment, a HCPE serving as the first proxy node and a hybrid access gateway, an HAG serving as the second proxy node ; the method comprising: by the HCPE: converting first TCP segments into first MPTCP segments of the primary MPTCP sub-stream and vice versa; and o using destination address information of the first TCP segments as destination address information of the first MPTCP segments; and o using source address information of the first MPTCP segment as source address information of the first TCP segments.

[29] According to a third aspect, the invention relates to a method for exchanging data over a TCP connection between a client node and a networking node; the TCP connection comprising a primary MPTCP sub-stream on a first access network between a subscriber premises equipment, a HCPE serving as the first proxy node and a hybrid access gateway, an HAG serving as the second proxy node ; the method comprising: by the HAG: converting second TCP segments into second MPTCP segments of the primary sub-stream and vice versa while preserving the source and destination address information.

[30] According to a fourth aspect, the invention relates to a hybrid subscriber premises equipment, HCPE, configured to perform the steps performed by the HCPE according to the first and second aspects.

[31] According to a fifth aspect, the invention relates to a hybrid access gateway configured to perform the steps performed by the HAG according to the third aspect.

According to a sixth aspect, the invention relates to a system comprising the HCPE according to the second aspect and the HAG according to the third aspect.

Brief Description of the Drawings [33] Figure 1 illustrates a hybrid subscriber premises equipment, a hybrid access gateway and a server used to establish a TCP connection according to one embodiment of the invention; and [34] Figure 2 illustrates segments exchanged to establish a multipath TCP connection according to one embodiment of the invention; and [35] Figure 3 illustrates exchanged segments for establishing an auxiliary sub-flow according to one embodiment of the invention; and [36] Figure 4 illustrates segments exchanged on both the primary and secondary sub-streams of a multipath TCP connection according to one embodiment of the invention; and [37] Figure 5 illustrates a hybrid access gateway according to one embodiment of the invention.

[38] Figure 6 illustrates a suitable computer system as a further embodiment of a hybrid subscriber premises equipment and / or a hybrid access gateway according to an embodiment of the invention.

Detailed Description of the Embodiments [39] Figure 1 illustrates a system for exchanging data over a TCP connection between a client 100 and a networking node 103 according to one embodiment of the invention. The TCP connection can be initiated by the client 100 in which case the node 103 acts as a server or vice versa. For the remainder of this description, the networking node 103 will be designated as the server. The system further comprises hybrid subscriber premises equipment 101, further designated as HCPE. The HCPE 101 serves as a gateway for the client 100 and any other networking device within the local network 113. The HCP 101 provides the client with access to the access networks 110, 111 of an Internet service provider ( I SP). The system further comprises a hybrid access gateway 102 allowing communication between the access networks 110, 111 and an external network 112 such as, for example, the Internet. Server 103 resides in this external network 112. Both HCPE 101 and HAG 102 are denoted as "hybrid" because they are able to communicate with each other on more than one access network. . An access network can be a wired access network such as for example an ADSL, ADSL2, VDSL, VDSL2, fiber or cable access network. In such a case, the HCPE will have a wired communication interface. An access network may also be a wireless access network such as for example an LTE access network, Wi-Fi, satellite or any other wireless access network. In such a case, the HCPR will also include a wireless interface.

[40] As will be described in the embodiments below, the client 100 can establish a TCP connection with the server 103, i.e., a connection in which both the client 100 and the server 103 maintain TCP status information to maintain a reliable connection. When sending data over the TCP connection to the server, the client sends the TCP segments to HCPE 101. The HCPE retains both a TCP state and a multipath TCP connection state, MPTCP. MPTCP is an extension of the TCP protocol. A version of the protocol is published by the I ETF in RFC 6824. The HCPE then converts the TCP segments into MPTCP segments and sends them on one of the access networks 110, 111 to the HAG 102. this, the HCPE maintains a primary MPTCP sub-stream 122 on the first access network 110 with the HAG 102 and an auxiliary MPTCP sub-stream 124 on the second access network 111 with the HAG 102. In this way, the TCP connection can benefit from the aggregated bandwidth of access networks 110 and 111. Also, HAG 102 retains both a TCP connection state and an MPTCP connection state. When the HAG receives the MPTCP segments from the HCPE, it converts them back into TCP segments and transfers them to the server 103. The TCP segments from the server are sent in a similar manner to the client 100.

[41] For addressing, the TCP port numbering scheme can be combined with the internet protocol addressing scheme, IP, which is commonly referred to as TCP / I P. The IP protocol may for example be I Pv4 or the most recent I Pv6.

[42] Figure 2 illustrates the establishment of the TCP connection 234 between the client 100 and the server 103 according to one embodiment of the invention. In the example, the client initiates the connection, but the establishment can also be performed by the node 103 in which case the client 100 acts as a server node. The establishment is performed by the exchange of segments TCP 201 to 209. In each segment of Figures 2 to 4, the segment type is underlined followed by the source address denoted by "SA" and the destination address denoted by "DA". For the addresses in the segments, "CL" indicates the address of the client 100, "SRV" indicates the address of the server 103, "HCPE" indicates an address of the HCPE 101 and "HAG" indicates an address of the HAG 102.

[43] In a first step, the client transmits a TCP synchronization segment 201 or in addition short a segment SYN on its networking interface 220 to the server 103 by adding the network address assigned to the networking interface 227 of the server in the destination address field of the synchronization segment 201. Since the HCPE serves as a gateway for the client 100, the segment will be received at the networking interface 221 of the HCPE 101.

[44] The HCPE includes two external network interfaces 222 and 223 connecting the HCPE respectively to the first and second access networks 110 and 111. Then, in step 210, the HCPE 101 converts the received SYN segment 201 into a segment SYN 202 which indicates multipath capabilities, i.e., a SYN segment containing the MP_CAPABLE option. Depending on whether the HCPE implements a network address translation, NAT, the HCPE can replace the source address of the segment 201 with the address of the HCPE, i.e. the address assigned to the interface 222. A NAT can for example be made when using the IPv4 while it is not normally useful when using the I Pv6. In both cases, the HCPE 101 does not change the destination address during the conversion step 210.

[45] In a new step, HCPE 101 then transmits segment 202 on its first networking interface 222 and then on primary sub-stream 122 to the server. As the HAG acts as a gateway for the first access network to the network in which the server is located, the segment 202 will be routed to the network interface 224 of the HAG and thus received by the HAG 102. Next, in step 211, the HAG converts the MPTCP segment 202 back to a TCP 203 segment by removing the MP_CAPABLE option. During conversion, the HAG does not change the source or destination address of the segment. The HAG then transfers the converted segment.

[46] When the server 103 receives the SYN segment, everything happens as if the client 100 or the HCPE 101 attempted to establish a single-path TCP connection with the server. If there is no NAT applied, both the HCPE 101 and the HAG 102 remain completely transparent to the server. In return, the server 103 responds to the client 100 or HCPE 101 by a synchronization and acknowledgment segment 204, a SYN + ACK.

[47] This segment 204 is again intercepted by the HAG 102 serving as a gateway to the HCPE 101 and the client 100. In the step 212, the HAG performs a similar conversion of the segment 204 as in step 211, but adds now the multi-path capability option, commonly referred to as MP_CAPABLE. Again, both the source and destination addresses remain unchanged.

[48] Since HCPE serves as a gateway to the client for segments from the HAG, the converted segment 205 will be received at the HCPE 101. In step 213, the HCPE removes the MP_CAPABLE option and thus converts the segment MPTCP 205 into a TCP segment 206. Depending on whether a NAT is performed, the source address may be changed to that of the client 100.

[49] As for segments SYN 201-203, the client now acknowledges the TCP connection through segment 207. This segment is again converted 214 by the HCPE into an ACK + MP_CAPABLE 208 segment and converted 215 back to an ACK segment. When the ACK segment 209 is received by the server 103, the TCP connection 234 is considered established at the level of both the client 100 and the server 103 by the three-way dialogue. Between the HCPE and the HAG, the connection will be treated as the primary substream 122 of an MPTCP connection.

[50] In conjunction with the establishment of the TCP connection 234 or at any time thereafter, a second sub-flow may be established between the HCPE 101 and the HAG 102.

[51] Figure 3 illustrates the steps performed by HCPE and HAG to establish the second sub-stream 124 according to one embodiment of the invention. The steps performed conform to the MPTCP protocol to establish an auxiliary sub-stream. After establishing the first sub-stream 122, the HA sends an MPTCP segment 301 containing an ADD_ADDRESS option to the CPE in which the HAG interface network address 225 on the second access network 111 is announced. Subsequently, the HCPE and the HAG exchange other segments such as the segment SYN + MP_JO IN 302 and the segment SYN + ACK + MP_JO IN 303 in order to establish the second sub-stream MPTCP 124. For the exchange of MPTCP segments, the HCPE will use the second network interface 223 and thus the second access network.

[52] In order to exchange data on both sub-streams 122, 124, both HCPE 101 and HAG 102 keep track of the MPTCP state, i.e. data sequence, DSN, MPTCP, as defined by the MPTCP protocol.

[53] According to another embodiment, the second sub-stream 124 can also be implicitly established during the establishment of the first sub-stream. In this way, additional network traffic and delay caused by the exchange of segments 301-303 can be avoided. This implicit establishment is possible because the HC 101 is known to HAG 102 since they are both part of the same subscriber network, i.e. the same ISP. This other embodiment will now be described with reference to FIG.

[54] When the HAG has received the first SYN + MP_CAPABLE segment 202, the HAG will also obtain the addressing information from the second networking interface of HCP 223. The manner in which this address information can be obtained is explained further in this section. -after.

[55] The reception of the segment 202 is interpreted by the HAG 102 as a request to configure both the first MPTCP sub-stream 122 and the auxiliary MPTCP sub-stream 124. This is thus not only an indication that the HCPE 101 supports multipath TCP. This request can be directly included in the first segment 202. In a variant, the request can also be made indirectly, for example, by a predefined configuration in the server according to which segments 202 should always be considered as such a request. In this way, no further additional information must be present in segment 202. HCPE 101 may also modify or configure such a configuration via an out-of-band connection or channel between HCPE 101 and HAG. 102.

[56] When the request is included in a segment 202, this can be done in several ways. One way is to define a new TCP option that includes such a request. Another possibility is to place the request inside the payload of segment SYN 202. Yet another possibility is to place the request inside an option in the network packet, that is, say, as I P option. This is particularly advantageous in IPv6 where there is no strict limit on the length of such an option and so on the length of the query.

[57] When the HCPE receives the segment SYN + ACK + MP_CAPABLE 205, it interprets it as a confirmation that the HAG 102 is ready for communication on the two sub-streams 122 and 124.

[58] Subsequently, HCPE 101 confirms the establishment of the first and second substreams with the ACK + MP_CAPABLE 208 segment to the HAG. Upon receipt, HAG 102 is confirmed that HCPE 101 has established both sub-streams 122, 124 and also establishes first sub-stream 122 and sub-stream 124.

[59] Obtaining the address information of HCPE 101 for the second sub-stream by HAG 102 can be accomplished in a number of ways as described below.

[60] In a first way of carrying out the addressing information, these are included in the SYN + MP_CAPABLE segment 202 and the HAG 102 then extracts this information from the segment 202. According to a first example, a new TCP option is defined which includes this addressing information. In a second example, the addressing information is embedded as payload data in the SYN segment 202. In a third example, the addressing information is embedded as an option in the network packet, where that is, as an I P option. This is particularly advantageous in I Pv6 where there is no strict limit on the length of such an option and so on the length of the information of addressing. The incorporation of the addressing information may further be combined with the incorporation of the request as indicated above. Preferably, the addressing information is provided such that backward compatibility with the MPTCP protocol is guaranteed.

[61] A second way of obtaining the addressing information is through the use of an out-of-band communication mechanism or channel, i.e., by communication between the HCPE 101 and the HAG 102 outside MPTCP substreams. For example, a separate connection may exist between the HCPE 101 and the HAG 102 to exchange new information regarding the subflows. Such a connection can be used by management applications running on both the HCPE 101 and the HAG 102 which manage the multi-path sub-flow establishment.

[62] In a third way, the addressing information is independently deduced by the HAG 102 by a predefined logical relationship, i.e., according to a predefined rule indicating how the addressing information can be derived. . Some examples of such rules are: - there is a mathematical relationship between the network address and / or a port of the HCPE 101 for both sub-streams. For example, the network address and / or the port for the sub-sub stream can be obtained by incrementing the network address and / or a port used for the first sub-stream. - The addressing information is identical to that used for previous substreams established by the HCPE. - The HAG has access to a database that lists all addresses of all HCPEs.

[63] The HAG 102 and the HCPE 101 may also support a connection establishment from the external network, i.e., when the server 103 initiates the establishment of a TCP connection 234 with the client 100.

[64] When TCP connection 234 is established according to one of the above embodiments, data segments may be exchanged between client 100 and server 103. Figure 4 illustrates how segments 401 through 408 are converted by HCPE 101 and HAG 102 to use both sub-streams 122, 124 while maintaining transparency for both client 100 and server 103.

[65] When a TCP segment 401 is sent from client 100 to server 103 along primary sub-stream 122, segment 401 will be intercepted at the HCPE, and forwarded as MPTCP segment 402 to the first sub-stream. or primary sub-stream to the server 103. During this conversion step 410, the destination address information is preserved. The source address information may be changed to that of the interface 222 when NAT is applied by the HCPE. The MPTCP segment 402 will in turn be intercepted by the HAG and converted back to TCP data segment 403. During the conversion step 411, the source and destination address information is preserved by the HAG. Finally, segment 403 will arrive at server 103 as a single-path TCP segment from client 100 or HCPE 101.

[66] The conversion in the opposite direction from the server 103 to the client 100 on the first MPTCP sub-stream 122 is performed in a similar manner. A TCP segment 404 from the server 103 will be intercepted by the HAG 102 and converted to a MPTCP segment 405. During the conversion step 412, the source and destination address information is preserved. Segment 405 will in turn be received or intercepted by HCPE. At the HCPE, the segment 405 is converted to a TCP 406 segment. During this conversion step 413, the source address information is preserved. The destination address information may be changed depending on whether NAT is applied by the HCPE. The client 100 then receives the single-path TCP segment 406 transparently, i.e., as if coming from the server 103.

[67] When the TCP segment 401 is sent from the client 100 to the server 103 along the auxiliary sub-stream 124, the segment 401 will again be intercepted at the level of the HCPE 101 and transmitted as the MPTCP segment 407 on the second subset. auxiliary flow or sub-stream to the server 103. During this conversion step 410, the destination address information is replaced by the address information of the HAG 102, i.e., the assigned address. at the network interface 225 of the HAG. During this same step 410, the source address information is changed to that of the interface 223. The segment 407 is then transferred to the second access network 111 to the HAG. As the MPTCP segment 407 is addressed to the HAG, the segment 407 will be routed to the network interface of HAG 225. As the auxiliary sub-stream 124 is linked to the primary sub-stream 122, the HAG will identify the segment 407 as belonging At the TCP connection 234. Therefore, the HAG converts the segment, during the conversion step 411, into a TCP 403 segment and replaces the destination address with that of the server. If a NAT is not applied at the HCPE, the source address is replaced by that of the client 100, otherwise by the address assigned to the HCPE interface 222, i.e. interface used for the primary substream 122.

[68] The conversion in the opposite direction from the server 103 to the client 100 on the auxiliary MPTCP substream 124 is performed in a similar manner. A TCP segment 404 from the server 103 will be intercepted by the HAG 102 and converted to an MPTCP segment 408, i.e., an MPTCP segment for the ancillary MPTCP sub-stream 124. During the conversion step 412, the source address of the server is replaced by the source address of the HAG, i.e., the address assigned to the network interface 224 of the HAG. The destination address is replaced by that of the HCPE 101, i.e., the network address of the HCPE network interface 223 on the second access network 111.

[69] The segment will then be routed along the second access network 111 to the HCPE 101. At the level of the HCPE 101, the segment 408 at the interface 223. Since the HCPE keeps a state MPTCP, it deduces that the segment 408 belongs to the TCP connection 234. Therefore, subsequently, during the conversion step 413, the segment 408 is converted to segment 406. In step 413, the source address of the HAG is replaced by the server address and the destination address is replaced by the client's address. Finally, segment 406 arrives at client level 100.

[70] Figure 5 illustrates the arrangement of the HAG according to one embodiment of the invention. The HAG 102 includes a server 501 including proxy server logic 505 for performing the conversion steps as explained with respect to the previous embodiments. The segments received on the primary MPTCP sub-stream 122 and the auxiliary MPTCP sub-stream 124 are converted by the proxy server logic 505 into TCP segments and vice versa.

[71] Also other data packets 510 may be received at the networking interface 224 such as for example packets according to other networking protocols including UDP or I CMP. These packets will not be converted by the proxy server logic, but are directly transferred to the external network on the networking interface 226. In this way, the MPTCP functionality in the HCPE and the HAG 102 also remains completely transparent for others. protocols.

[72] HAG 102 may further include a plurality of HAG servers 501 through 503 to balance the load of the data traffic. In this regard, the HAG 102 also includes load balancers 520 and 521. The load balancer 520 balances packets from different HCPEs on one of the HAG servers 501-503. The load balancer 521 balances packets from the outside network on one of the HAG servers 501-503. Several policies can be implemented in the load balancing device such as for example: load balancing MPTCP flow, that is to say, by newly established TCP connection between a client and a server. - HCPE load balancing, that is, each HCPE is assigned a single HAG server for all its connections. This can be done by checking the source I P address of the packets received on the primary interface 224. - Source prefix load balancing. In this way, a cluster of HCPEs is assigned a single HAG server.

The above load balancing mechanisms ensure that the load balancing devices on both the HCPE side and the external network side 520 and 521 consistently route packets of a given stream to the same proxy. This also ensures that the load balancing decision is not changed for a current stream, i.e., for an ongoing TCP connection between a client and a server.

[73] During operation, that is, when the multipath TCP works on both the primary and auxiliary substreams, a situation may arise whereby the networking address assigned to the interface network 222 for the primary MPTCP sub-stream 122 becomes unavailable, for example by an interface reset, while the existing auxiliary sub-stream 124 is still in place. In such a case, the HCPE 101 may have a REMOVE_ADDR option present in a segment to the HAG 102 on the still active slave substream 124, i.e. it shows that the networking address used is no longer valid. This then triggers a cleaning of the MPTCP state at the HAG 102 for any sub-stream attached to this address. This closure of connections that lose their primary substream is necessary because the address could become allocated to another HCPE during the life of the MPTCP session. The common address could be used simultaneously by two different users, which could lead to errors on the network.

[74] Therefore, the HAG 102 may terminate the TCP connection 234 if the network address used for the primary sub-stream is lost and subsequently release all associated resources. This can be done by the following steps: - the HAG receives a segment from the HCPE 101 with the REMOVE_ADDR option indicative of the fact that the networking address used for the primary substream is lost. - With this, the HAG 102 resets the TCP connection 234 to the server. HAG 102 sends a segment with TCP option MP_FASTCLOSE to HCPE 101. Upon receipt, HCPE 101 resets TCP connection 234 to client 100.

[75] Fig. 6 shows a suitable computer system 600 as a further embodiment of HCPE 101 or HAG 102. The computer system 600 can generally be formed as a suitable universal computer and includes a bus 610, a processor 602, a local memory 604, one or more optional output interfaces 616, a communication interface 612, a storage element interface 606, and one or more storage elements 608. The bus 610 may include a or more than one conductor that provides communication between the components of the computer system 600. The processor 602 may include any type of conventional processor or microprocessor that interprets and executes programming instructions. A local memory 604 may include a random access memory (RAM) or other type of dynamic storage device that stores information and instructions for execution by the processor 602 and / or a read-only memory (ROM) or other type of device. static storage device which stores information and static instructions for use by the processor 602. The storage element interface 606 may include a storage interface such as for example a SATA interface or an SCS I interface for connecting the bus 610 to one or more storage elements 608, such as one or more local disks, for example SATA disk drives, and to control the reading and writing of data on and / or from these storage elements 608. Although the storage elements 608 above are described as a local disk, generally any other suitable computer readable medium such as a semi-cond driver nctors or flash memory cards could be used. The system 600 described above can also run as a virtual machine over the physical hardware. The steps performed on the HCPE and HAG devices according to the above embodiments may be partially or completely implemented as programming instructions to be executed on the processor 602. The communication interface 612 may furthermore correspond at the networking interfaces of HCPE or HAG 221,222, 223, 224, 225, 266.

[76] Although the present invention has been illustrated with reference to specific embodiments, it will be apparent to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention The invention can be implemented with various changes and modifications without departing from the scope thereof. The present embodiments should therefore be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and any changes that are in the meaning and scope. The equivalence of the claims is therefore intended to be encompassed here. In other words, it is intended to cover any and all modifications, variations or any and all equivalents that fall within the scope of the underlying underlying principles and whose essential attributes are claimed in this application patent. It will further be understood by the reader of this patent application that the words "comprising" or "understanding" do not exclude other elements or steps, that the word "a" or "an" does not exclude plurality, and that a single element, such as a computer system, a processor or other integrated unit can perform the functions of several means claimed in the claims. Any reference number in the claims shall not be construed as limiting the respective claims concerned. The terms "first", "second", "third", "a", "b", "c", and the like when used in the description or in the claims are introduced to distinguish elements or steps similar and are not necessarily descriptive of a sequential or chronological order. Similarly, the terms "above", "below", "on", "under", and the like are introduced for purposes of description and not necessarily to denote relative positions. It is to be understood that the terms so used are interchangeable under appropriate circumstances and that the embodiments of the invention are capable of operating according to the present invention in other sequences, or in orientations different from those described (s) or illustrated above.

Figure 1: 100 Customer (CL)

101 HCPE

102 H AG 100 Server (SRV)

Figure 2: 220 Customer (CL)

101 HCPE

102 HAG 103 Server (SRV) 122 Primary MPTCP Substream

Figure 3: 220 Customer (CL)

101 HCPE

102 HAG 103 Server (SRV) 124 Ancillary MPTCP Sub-Flow 122 Primary MPTCP Sub-Flow

Figure 4: 220 Customer (CL)

101 HCPE

102 HAG 103 Server (SRV) 124 Ancillary MPTCP Sub-Flow 122 Primary MPTCP Sub-Flow

Figure 5: 520 Load balancing device 521 Load balancing device 501 HAG server 1 502 HAG server 2 503 HAG server N 505 Authorized representative

Claims (12)

A method for exchanging data over a TCP connection (234) between a client node (100) and a networking node (103); wherein the TCP connection comprises a primary MPTCP substream (122) on a first access network (110) between a subscriber premises equipment, a HCPE (101) serving as the first proxy node and a gateway of hybrid access, an HAG (102) serving as the second proxy node; the method comprising: by the HCPE: converting (210, 213, 214) first TCP segments (201, 206, 207) into first MPTCP segments (202, 205, 208) of the primary MPTCP sub-stream and vice versa; and o using destination address information of the first TCP segments (201, 207) as the destination address information of the first MPTCP segments (202, 208); and o using source address information of the first MPTCP segment as source address information of the first TCP segments; and - by the HAG: the conversion (211, 212, 215, 217) of second TCP segments into second MPTCP segments of the primary sub-stream and vice versa while preserving the source and destination address information, - the connection TCP (234) comprising an auxiliary MPTCP substream on a second access network between the HCPE (101) and the HAG (102); the method further comprising: o at the HCPE, converting third TCP segments into third MPTCP segments of the auxiliary MPTCP substream and vice versa; and o at the HAG, the conversion of fourth TCP segments into fourth MPTCP segments of the auxiliary sub-stream and vice versa. The method of claim 1, further comprising: - by the HCPE, the use of destination address information of the HAG for the third MPTCP segments; and - for the HAG, using destination address information of the networking node for the fourth segments sent to the networking node. 3 - Process according to claims 1 or 2, further comprising, for establishing the auxiliary MPTCP substream: - sending, by the HAG to establish the sub-MPTCP auxiliary flow, a segment comprising an address of the HAG in the second access network on the primary MPTCP substream. 4 - Process according to any one of claims 1 to 3, further comprising: - sending, by the HCPE, a segment comprising a request to establish the second substream MPTCP to the HAG. 5 - Process according to any one of claims 1 to 3, further comprising: - sending, by the HAG, a segment comprising a request to establish the second sub-flow MPTCP to HCPE. 6 - Process according to claim 1 or 2, comprising, for the establishment of the primary MPTCP sub-stream and auxiliary MPTCP sub-stream: - by the HAG, the reception of a synchronization segment from the HCPE indicative of a request to establish the primary and auxiliary MPTCP substreams; and subsequently, by the HCPE, receiving a synchronization and acknowledgment segment from the HAG; and - the establishment by HCPE of the primary and auxiliary MPTCP substreams; and - the receipt by the HAG of an acknowledgment segment from the HCPE; and - subsequently, by the HAG, the establishment of the primary and auxiliary MPTCP substreams. The method of one of claims 1 to 6, wherein the HAG comprises a plurality of gateway servers, each acting as a gateway between the first access network and the networking node; and wherein the method further comprises: - by the HAG, the load balancing of the primary MPTCP substream on one of the gateway servers. A method for exchanging data over a TCP connection (234) between a client node (100) and a networking node (103); wherein the TCP connection comprises a primary MPTCP substream (122) on a first access network (110) between a subscriber premises equipment, a HCPE (101) serving as the first proxy node and a gateway of hybrid access, an HAG (102) serving as the second proxy node; the method comprising: by the HCPE: converting (210, 213, 214) first TCP segments (201, 206, 207) into first MPTCP segments (202, 205, 208) of the primary MPTCP sub-stream and vice versa; and o using destination address information of the first TCP segments (201, 207) as the destination address information of the first MPTCP segments (202, 208); and o using source address information of the first MPTCP segment as source address information of the first TCP segments, - the TCP connection (234) comprising an auxiliary MPTCP sub-stream on a second network access between HCPE (101) and HAG (102); the method further comprising: o at the HCPE, converting third TCP segments into third MPTCP segments of the auxiliary MPTCP substream and vice versa; and o at the HAG, the conversion of fourth TCP segments into fourth MPTCP segments of the auxiliary sub-stream and vice versa. A method of exchanging data over a TCP connection (234) between a client node (100) and a networking node (103); wherein the TCP connection comprises a primary MPTCP substream (122) on a first access network (110) between a subscriber premises equipment, a HCPE (101) serving as the first proxy node and a gateway of hybrid access, an HAG (102) serving as the second proxy node; the method comprising: - by the HAG: o converting (211, 212, 215, 217) second TCP segments into second MPTCP segments of the primary substream and vice versa while preserving the source and destination address information, the TCP connection (234) comprising an auxiliary MPTCP substream on a second access network between the HCPE (101) and the HAG (102); the method further comprising: o at the HCPE, converting third TCP segments into third MPTCP segments of the auxiliary MPTCP substream and vice versa; and o at the HAG, the conversion of fourth TCP segments into fourth MPTCP segments of the auxiliary sub-stream and vice versa. Hybrid subscriber premises equipment, HCPE (101), configured to perform the steps performed by the HCPE according to any one of claims 1 to 8. 11 - Hybrid access gateway (102) configured to perform the steps performed by the HAG (102) according to any one of claims 7 and 9. A system comprising HCPE (101) according to claim 10 and HAG (102) according to claim 11.