FR2778804A1 - METHOD FOR ACCESSING A NETWORK OF THE INTERNET TYPE VIA A TELECOMMUNICATION SATELLITE AND ARCHITECTURE FOR THE IMPLEMENTATION OF SUCH A METHOD - Google Patents

Abstract

The invention relates to a method for accessing an Internet network (RI) by a user terminal (SIT) via a spatial segment comprising a two-way data transmission link (/ 1 f, / 2f, / 1r, / 2r ) by telecommunications satellite (Sat) and a service provider (ISP '), coupled to the terrestrial segment of the network (RI). The “TCP” transport layer of the terminals (SIT) is distributed in two half-modules (31 -32, 40-42) arranged at the ends of the space segment. Transmissions in this segment are carried out according to an internal proprietary protocol. The half-modules (31-32, 40-42) include specific interfaces (32, 40). Use is also made of a negative type acknowledgment mechanism associated with a "decoy" mechanism which consists in the service provider (SIT) sending the acknowledgment back without being transmitted to the terminal (ISP '), via the space segment.

Description

METHOD FOR ACCESSING A NETWORK OF THE INTERNET TYPE VIA A

  TELECOMMUNICATION AND ARCHITECTURE SATELLITE

  FOR THE IMPLEMENTATION OF SUCH A PROCESS

  The present invention relates to a method of accessing a network of the Internet type via a transmission satellite.

of data.

  The invention also relates to a system for implementing this method. The last few years have seen a very significant development of communications via satellite

  telecommunication. Advances in technology have enabled significant miniaturization of communication terminals, and in particular the creation of portable terminals.

  On the other hand, there has also been an equally significant and rapid expansion of the Internet network to which millions of computers15 of all types and of all powers are currently connected. A user connected to this network can therefore have access to resources

  very important disseminated on all the planet, in particular in databases of very diverse natures.

  In the context of the invention, the term "Internet" must be understood in a broad sense: it includes the "Internet" network proper but also so-called "intranet" or "extranet" networks, generally all networks on which transmissions are made under

  particular protocols which will be recalled below, for example the "TCP" protocol (for "Transfer Control Protocol").

  Generally, access to the Internet is via a service provider or "ISP" (for Internet Service Provider), according to the Anglo-Saxon terminology commonly used.

  entities, such as certain large companies or universities, the coupling with the network is direct and is obtained via specialized servers which connect5 internal networks, for example of the intranet type, with the Internet network.

  The need was therefore very quickly felt to combine these two areas, that is to say to make possible

  access to the Internet via telecommunications satellites.

  FIG. 1, appended to the present description, illustrates an example of general architecture of a system of

  transmission allowing a terminal SIT of an end user U to have access to a remote host computer system RH, via the Internet network RI, on the one hand, and via a satellite link, on the other hand.

  The SIT terminal includes a computer system and conventional transceiver circuits, under the

  general reference 20. The transmitter-receiver part of the assembly 20 is coupled to a transmit-receive antenna 21 pointed at a telecommunication satellite.

  Sat munication. The transmitted and / or received data use a bidirectional transmission link 12, earth - satellite. Similarly, the ISP service provider is equipped with a computer system and conventional transceiver circuits, under the general reference 10. The transceiver part of the assembly 10 is coupled to a transceiver antenna 11 pointed at a Sat telecommunications satellite. The assembly 10 communicates with the Sat satellite via a bidirectional transmission link 11, earth - satellite, via the antenna 11. The coupling to the Internet network RI is symbolized by a bidirectional transmission link 12. Finally, the remote haste system RH comprises a computer system 30. The coupling to the Internet network RI is symbolized by a transmission link

bidirectional 61.

  The major difficulty caused by such data transmission, SIT to RH, is essentially linked to the ground - satellite - earth links part, that is to say the space segment: links 11 and 12. Indeed, the usual protocols implemented for transmissions within an Internet type network, and in particular the above-mentioned "TCP" protocol, are not optimized for access10 via a telecommunications satellite link. This protocol makes these transmissions uncompetitive compared

  broadband terrestrial transmissions under development, which will be available in the near future.

  Although the "TCP" protocol has been developed to satisfactorily meet the needs that arise on a wide variety of networks, it does not already operate

  not all the possibilities of some Internet-type subnets.

  In particular, the "TCP" protocol offers very poor performance on networks which are characterized by a long propagation time and a

  high bandwidth, networks often called "large capacity networks" or "Long Fat Networks" ("LFNs"), according to English terminology.

  The major problem results from the limits of the data rate imposed by the dimensions of what is called "variable window" or "sliding window" according to English terminology, and also from the ineffective processing of lost segments on high capacity links. The "TCP" options, which are standardized features to improve the performance of the original protocol, are

  often insufficient to completely resolve these problems.

  On long propagation time networks, the "TCP" protocol has significant limitations during the so-called "three-way connection establishment" procedure. This procedure consists in the exchange of three "TCP" packets which are both necessary and sufficient to synchronize the opening of the connection. For sending each packet, the sending host must wait for the reception of the previous packet returned by the other end of the link. In the

  high capacity networks or "LFNs" mentioned above, this leads to a very long connection opening time.

  The so-called "slow start" mechanism, for its part, although very useful for operations on the Internet, can very seriously limit the speed when it is applied to transmission channels with high propagation time and high bandwidth. This mechanism consists of sending packets at a low speed at the start of the connection and testing if there is congestion on the network. If the connection is not congested, then the transmission rate is continuously increased, until it reaches a steady state. The network condition testing process depends on the propagation time on the links. It follows

  that, for a network of the aforementioned "LFNs" type not congested, a very long time is necessary to reach the steady state. Again, the consequence is that the wide bandwidth of the network is underutilized, since small amounts of packets are transmitted for too long.

  These problems arise very acutely with regard to access to an Internet network via a telecommunications satellite transmission link. In particular, it is necessary to find a solution to the problem of high propagation times. Indeed, if we consider a geostationary satellite, the propagation time of an electromagnetic wave between a ground station and a satellite, or vice versa, is of the order of 125 ms. A full outward and return trip therefore requires a time interval of 0.5 s. It should be noted that these high delays did not

  were considered during the original design of transmission protocols on the Internet.

  Referring again to FIG. 1, it can easily be seen that a connection between a user terminal SIT and a remote host RH includes a satellite earth-satellite-earth link, 12 and 11, characterized by a relatively bandwidth broad and, as indicated above, a significant transmission delay. It also includes a series of standardized terrestrial links, which normally have shorter transmission times, but which are more sensitive to packet loss or alteration. Terrestrial links are often characterized by a relatively low speed and are also subject to loss of sequences. In fact, according to the "TCP" protocol, the data packets are arranged in sequences transmitted by different routes (routing) and they arrive at their destination in a random order. The complete message must then be re-ordered

  to be restored to its original form.

  The result is that the user - host connection results in long transmission times, low

  data rate, and a transmission path with data loss and sequencing.

  In summary of what has just been recalled, it can be seen that the connection between an end user and a remote host successively borrows two types of networks,

  with very different transmission characteristics, even if a common communication protocol is implemented. We also note that we do not benefit from the advantages specific to each of these networks, but that we cumulate above all their deficiencies.

  The invention aims to overcome the drawbacks presented by the known art for accessing an Internet type network using at least one telecommunications satellite link, while maintaining full end-to-end compatibility of the characteristics of the protocol.

  "TCP", which is fundamental for an Internet architecture.

  To do this, the invention provides a method for decoupling the network part comprising the telecommunications satellite links between a

  end user and a service provider (space segment), on the one hand, of the conventional Internet terrestrial network part, between this same service provider and a remote host, on the other hand.

  In general, the architecture of communication networks is described by various layers. By way of example, the "OSI" standard ("Open System Interconnection") defined by the "ISO" comprises seven layers which range from the so-called low layers (for example the so-called "physical" layer which relates to the physical transmission medium ) to the so-called high layers (for example the so-called application layer), passing through intermediate layers, in particular the so-called "transport" layer. A given layer20 offers its services to the layer immediately above it and requires the layer that immediately

  other services, via appropriate interfaces. They communicate using primitives. They can also communicate with layers of the same level. In some architectures, one or the other of these layers may be nonexistent.

  In an Internet environment, there are five layers, and more precisely, going from the upper layer to the lower layer, they include: the application layer ("http", "ftp", "e- mail ", etc.), the transport layer (" TCP "), the layer

  network addressing ("IP"), the data link layer ("PPP", "Slip", etc.) and the physical layer.

  In the context of the process of the invention, the above-mentioned decoupling essentially concerns the so-called "transport" layer. The "TCP / IP" module used by the subscriber on his terminal, usually consisting of two overlapping layers of software, is replaced by a module specific to the invention. A module of this type is also included in the installation of the service provider, on its end in communication with the satellite. The same is true for the "IP" network addressing layer. In other words, the "TCP / IP" module of a terminal according to the prior art is divided into two parts, arranged on either side of the satellite communication link. On the other hand, the upper layer called "application" is kept in

each user terminal.

  This architecture makes it possible to maintain full compatibility with the Internet environment, while abolishing the traditional limits encountered when implementing the "TCP" protocol on networks with

big capacity.

  In a preferred embodiment, the method according to the invention uses an acknowledgment procedure of the so-called "negative" type, instead of a conventional acquisition procedure, that is to say of "positive" type, for links between specific modules, ie on links between the terminal of the end user and the service provider, via the telecommunication satellite. This procedure consists of reporting only the receipt of missing data, to

  within a predetermined time interval, rather than signaling each correct receipt, which leads to a waste of time.

  Still according to a preferred embodiment, use is made of a mechanism that can be called "decoy", or "spoofing" according to English terminology. This technique overcomes the flow limit due to the small size of the aforementioned variable windows. This technique is preferably used with the negative acknowledgment technique used on the earth - satellite - earth section. The service provider, more specifically the specific half-module, when it receives data from the remote host, sends it an "artificial" acknowledgment of receipt, without waiting for an acknowledgment of receipt which should be provided to it, as it is the case in a

  standard architecture, through the end user's terminal, which greatly accelerates the exchange process.

  Finally, still according to a preferred embodiment of the method of the invention, the module specific to the invention is transparent vis-à-vis the other protocols used on networks of the Internet type, and in particular of a protocol known as acronym "UDP", for "User Data Protocol". Indeed, this particular protocol does not suffer from the same limitations as the "TCP" protocol. It is, in particular, much less sensitive to propagation times

  high presented by telecommunication satellite links.

  The subject of the invention is therefore a method of transmitting data between at least one user terminal and a first computer system known as a remote host, under Internet type protocol, via a network comprising a terrestrial segment and at at least one space segment consisting of a bidirectional link by telecommunication satellite, to which said terminal is connected, said land and space segments being interconnected by a second computer system called service provider, characterized in that, the transmissions being associated with 30 a stack of protocol layers, each layer communicating via interfaces with the layers immediately above and below it, and said stack comprising at least one upper layer of software applications, a first intermediate layer, immediately lower, called the transport layer, of "TCP" type, and a second interm layer diary, called "IP" type network addressing, the method comprises: - the distribution of said "TCP" layer of each user terminal in two half-modules so that they each comprise a layer of "TCP" transport and a specific interface layer, - the installation of one of said half-modules at the ends of said spatial segment, a first half-module being installed in said IT system providing a service and a second half-module being installed in the user terminal, - the transfer of the "IP" network addressing layer relating to the user terminal in the IT system of the service provider, - the transmission of data over said bidirectional link using an internal transport protocol based on a specific data transmission mode, - bidirectional conversion between said Internet type protocol and said internal protocol by said specific layer of the first and s econd half-modules, and the creation of an interface between said transport layer of the second half-module and a layer

  software applications installed in the user terminal25.

  The invention also relates to a computer architecture for implementing this method.

  The invention will be better understood and other characteristics and advantages will appear on reading

  the following description, made with reference to the appended figures, among which:

  - Figure 1 schematically illustrates the transmission chain between an end user terminal and a remote host system, via a telecommunications satellite and a terrestrial network for data transmission under Internet protocol, according to known art; - Figure 2 illustrates this same chain by highlighting the software layers involved in the transmission of data; - Figures 3a and 3b schematically illustrate an example of a data transmission chain implementing the method according to the invention and the corresponding system architecture; - Figure 4 schematically illustrates a variant of this architecture; - and Figure 5 schematically illustrates a system architecture for the transmission of data under

  two different internet protocols.

  Before describing the method of accessing an Internet type network according to the invention, via a satellite

  telecommunications, we will first of all specify the architecture of a system according to the prior art, from an interaction point of view of the main software layers, with reference to FIG. 2.

  For this FIG. 2, and the following figures, the elements or circuits identical, or at least similar, to those of FIG. 1 bear the same

  references, and they will only be re-described as necessary.

  In this figure 2, to fix the ideas, only two end users, Ua and Ub, and their terminals SITa and SITb have been represented. The components of the system 30 associated with these users carry references supplemented by the letters a and b, respectively, but

  are, a priori, functionally identical.

  We will describe the architecture of only one of the two terminals, the user terminal Ua. This includes, in addition to the hardware and circuits described above, software modules referenced 22a which are broken down according to the following software layers: - a lower layer 224a, called a driver or satellite network manager; - a first intermediate "IP" network addressing layer 223a; - a second intermediate "transport" layer, split into two layers of the same level: a layer for the "TCP" protocol 221a and a layer for the "UDP" protocol 222a;

  - and an upper layer of applications 220a.

  The SITb terminal has the same architecture, and there is no need to describe its components, which are referenced identically, with the exception of the final letter b.

  Internet protocols, and in particular the two aforementioned protocols, are well known per se and standardized.

  We will first consider the only "TCP" protocol, which presents the limitations which were recalled when it is implemented in a broadband network and long transmission times, as is the case for a link by telecommunications satellite. The "TCP" protocol currently obeys the "IPV4" standard (version 4 of this protocol). As mentioned, the number of subscribers connected to the Internet is growing very rapidly. However, an address conforming to the "IPV4" protocol comprises only four bytes, ie 232 theoretical addresses, but in reality much less due to the hierarchical structure of the Internet architecture organized in domains. Projections for the near future (years 2005-2011) raise fears of a shortage of addresses. Also, in 1995, recommendations were made to adopt a new protocol ("IPV6) and were published (working groups" Internet Engineering Task Force "and" IPng ", for" Internet Protocol new generation "). This

  new version of the "TCP" protocol allows, in particular, a much larger number of available addresses, but5 also brings other improvements.

  A user terminal, for example the SITa terminal, therefore communicates with what will be called hereinafter the "satellite subnetwork", which comprises the bidirectional links 11a and 12a (llb and 12b for SITb), as well as the circuits internal to Sat satellites and, as described below, the pilot 131 included in the system

  IT 13 of the ISP service provider.

  The main function assigned to the IT system of the ISP service provider is the routing of data to an addressed subscriber (in the direction host system - subscriber) or, on the contrary to an addressed host system (in the direction subscriber - host system). It can naturally, in addition to this basic function, fulfill other functions, such as temporary storage of data (function20 "mailbox, etc.), but it will in any case include at least the following software layers: - two lower layers made up of pilots: a first pilot, 131, on the "satellite subnetwork" side, forming the pendant of pilot 224a (or 224b) of terminal 22a25 (or 22b), and a second pilot, 132, on the network side Internet terrestrial RI;

  - And an "IP" network addressing layer 130.

  More specifically, the service provider communicates, via the pilot 132, with the Internet network RI via a main route 12, commonly called

  "backbone", according to Anglo-Saxon terminology.

  Likewise, the remote host RH can be connected to the Internet network RI by a main artery, a conventional telephone link of a switched network, a transmission link of the "ISDN" type ("Digital Integrated Network"

  Service ") or via another service provider (not shown). The connection is symbolized by a bidirectional bus 61.

  A priori, the transmissions are carried out according to a "client-server" mode, the various terminals being the

  "clients" of the "server" constituted by the remote RH host.

  These reminders being made, we will now describe, with reference to FIG. 3a, an exemplary embodiment of a

  architecture for implementing the method according to the invention.

  In this FIG. 3a, only the case of the "TCP" protocol has been considered. Application of the method of the invention to

  another Internet protocol, in particular the "UDP" protocol, will be described later in connection with FIG. 5.

  According to an important characteristic of the invention, a decoupling is carried out between the "satellite subnetwork" (space segment) and the rest of the Internet network, that is to say the conventional terrestrial part of this network. To do this, the lower and intermediate software layers previously residing in the end user terminals (of which only one, SIT ', has been represented in FIG. 3a), have been "broken" into two parts. A first part 31-32, is located in the user terminal SIT ', and second

  part 40-41, is located in the IT system of the ISP 'service provider.

  The lower layers, 40 and 32 respectively, play a role analogous to the pilots 131 and 224 (a or b) of the known art. They are however specific, because they form the interface between physical supports (links 11 and 12), with the same characteristics as those of the known art (FIGS. 1 or 2) and half-modules constituting the specific transport layers to the invention, 41 and 31. The lower layers manage the internal exchanges in the subnetwork aun quaumirzou puadmoD S Tallaassed alaD ZI aledTuTrad a-T Ve, 1 oa t IToTD5oT Ilnpo ael GIua JuoUiSTDogd snld 'IH azqsaaaeq qauzaeuI nTas al einpomu a alua aesodsTp 's (. Xl aqe.) T1aassvd aun, p Funm sa, dSI a GoTn-s ap T2slesad np anbTq4uxoguT 0o aUis ss aT' enbT: sT9aoxeD aiane aun uolas SalnGTIdDe seina - qtinaQn STA uoTeDTunMmoD ap saATTUmTad sea1mu saI quesTrTn qe senbTsselD.aDi. selnpou sep enb aDTAtas amGUI al juVaggo SGZToU saqroq11 sap MwwoD 4UGssTfe UOTJUGAUTiT ap apagoad çz el uoIs sanbTgTDgds salnpou-TUIGp sel 'euinsez us qTIlaqs nwespnnpnn npnpnnnn ell (qúZZ no úEat "aI" aqonoD: z; aznr;) nuuoD qm, I uo Gs ZIS IEUTUxua un suep GquLIduT ITei TFnb. ,, aI "anoD ie qsa oz Za nespa ap e5Essasp2, p eqDnoD -el 'enLTqd saadie-TD G91ITTTqp aq; VA Tnb ellalassed 'rrp quaealdoad auaequI GasGa nees9r o DA-eS aIIaTssed aun lt, a "I, neassa ap eSessip-ep a-DnoD aun Ted en Vr-suoD 1sa UautquIt AUAUUAAUUAUAUUAUUAUAUUAUUAUUAUAUUAT 9qgD np qa 'Oou suoTeDTIdde, p aqDnoD el Ied gniTisuoD qsG qUl2GuI quawa UoITAuGI', LIS TeUTUULI a9-4D na enBT; TDeds qUGmUUUOXTAUe a laulaquI quaauuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuu coat package offer 4uen-4TsuoD QI dSI GDTAxas ap aTVqepsad np anbTqluIojuT GUIsIsS a- suep qxed aD: rne, p '0lo', Z1IS TeuTTLq Gl suep qIed eun, p TE 'senbTEoT s; aqDnoD nO sTln nap ser (ZT qe TT SUOSTeTI) -TIIajes nfeasa-snos np saqg9wTaxa xnmp sl ans ,,% aaqD e. anoAqla as 4aS xnap ua adnOD qsa nuuoD XeI uoles (q no e) z [IeuTula un suep aquesezd (q no?) ITZZ dDJ.L / nbTun aLDnoD e 'SeUle s; aalne, p us senbTTDeds sGa uuo pp uoT uoles' aaTIlLes bl "IP" network address layer 54, relating to

  ISP service provider, and a pilot 53, playing a similar role, if not identical, to the pilot 132 of the known art (FIG. 2).

  The "IP" layers 42 of module 4 and 54 of gateway 5 communicate with each other via a bus

  bidirectional 420, forming a physical support, and any conventional pilots (not shown) located on either side of this bus 420.

  Due to the provisions which have just been recalled, compatibility with the Internet environment

remains whole, this from start to finish.

  Indeed, the virtual architecture of the system according to the invention can be schematized by that represented in FIG. 3b. The characteristics of the terrestrial Internet network, i.e. the path from the host

  remote RH, via link 61, the Internet network RI proper and the main artery 12 remain unchanged.

  These components have been shown in solid lines in FIG. 3b. The same applies to the applications layer 30 of the terminal SIT '.

  On the other hand, the communication path between these two series of components can be represented simply by a virtual pilot 224 ′ seen by the terrestrial network and virtual intermediate layers of “IP” network addressing and “TCP” protocol. There is therefore a good conservation of end-to-end compatibility, of course in both directions of communication. The protocol conversions are carried out by the two half-modules specific to the invention (layers 31 and 41), located at the two ends of the space transmission chain. Between these two modules, 31 and 41, and via the specific pilots, 32 and 40, the data and control signal transmissions are carried out in a specific mode,

  based on round trip connections.

  The link following the direction ISP'-SIT '(or go) uses an internal transport protocol, of the so-called "proprietary" type, advantageously based on a standard widely used for transmission transmissions of digital video signals known by the acronym "DVB- S "(for" Digital Video Broadcasting "type S). This standard is described, for example, in the following document: "Digital Broadcasting 10 Systems for Television, Sound and data Services: Frame Structure, Channel Coding and Modulation for 11/12 GHz

  Satellite Services ", ETS 300 421," European Telecommunications Standards Institute ".

  For the SIT'-ISP 'link, a technique quite similar to that which has just been mentioned can also

be retained with profit.

  The physical layer for such links is characterized by the virtual absence of errors with regard to data transfers. It is easy to understand20 that a mechanism based on the accumulation of acknowledgments and the use of variable windows, mechanism

  implemented by the "TCP" protocol, becomes ineffective and useless, if this type of connection is implemented.

  In addition, problems related to packet sequencing losses are nonexistent on satellite links. Heavy re-sequencing mechanisms

can therefore be omitted.

  In summary, the "transport" part of the protocol internal to the space segment, that is to say in practice the communication section between the modules 31 and 41, is based on a specific data transmission standard which, quite at the same time, guarantees reliable data transfers, overcomes most of the characteristics specific to the "TCP" protocol, since they are complex and, in any case, useless, and maintains compatibility

  of the Internet environment, as has been shown more particularly with reference to FIG. 3b.

  According to another characteristic of the invention, in a preferred embodiment, use is made of a particular mechanism for acknowledging receipt of data on the spatial segment. This mechanism can be called "negative acknowledgment". The normal mechanism, which can be called "positive acknowledgment", requires that, for each correctly received data packet, an acknowledgment of

  reception. This mechanism causes significant time losses, because statistically, on a link with a low error rate, which is the case for the space segment, most packets are delivered correctly.

  However, according to the negative acknowledgment procedure, only an incorrect reception is reported. More precisely, when the recipient, whoever it is, detects a dead time in the sequence of the 20 packets received exceeding a predetermined time interval, an alert packet, or "NAK" signal, is sent so as to signal that there is a missing package. This mechanism requires the presence of a buffer memory or an equivalent device, at the level of the transmitter, so as to temporarily store the data which have just been transmitted, until a predetermined time-out period occurs.

  has elapsed ("time out"). Beyond this period, if the sender has not received a negative acknowledgment or "NAK" signal, it is assumed that the packet in memory has been correctly received by the recipient. The corresponding data can therefore be deleted.

  According to another characteristic, and still in a preferred embodiment, the method of the invention uses a mechanism called "spoofing" or "decoy", implemented in conjunction with the negative acknowledgment mechanism. This mechanism makes it possible to avoid the problems of speed limitation due to the small window dimensions on a link with a large capacity, or "high latency" 15 according to the English terminology. By default, most standard window sizes are 8,096 bytes

  microcomputer operating systems ("OS"). The maximum size currently allowed by the "TCP" protocol is 64 KB (16 bytes in the header of a "TCP" packet), 10 which allows a maximum bit rate of 1024 KB / s, if we admit a 0.5 s round trip time for the space segment.

  Manufacturers offer software that allows windows whose dimensions, by default, are included in the range 8 KB to 24 KB, which allows a bit rate going from 15,128 kbits / s to 384 kbits / s. Bodies in charge of defining Internet standards ("the Internet Engineering Task Force") have also proposed modifications to the "TCP" protocol ("RFC 1323 - TCP extensions for High Performance") which could theoretically extend the size of windows up to '' to a maximum of 1.07 GB, which would allow a flow, also theoretical, of 17.12 Gbits / s. However, the increase in throughput, under the "TCP" protocol, is accompanied by increased risks of packet loss. The method according to the invention, in its aforementioned variant, aims to reduce the transmission times "end to end

  end ", known as the Anglo-Saxon parameter" Round Trip Time "or" RTT ".

  The characteristics of the above mechanism will be

  explained with reference to FIG. 4. The corresponding procedure can be divided into four main phases.

  We will first describe the characteristics of a system architecture allowing the implementation of a

such mechanism.

  The software layers 30 of the remote haste system RH, referenced 32, comprise at least one "TCP" layer 321 and one application layer 320. The computer system of the service provider, referenced ISP ', comprises in addition layers 40 and 41 , two buffer memories, 44 and 45. The gateway 5 has not been shown in FIG. 4. To better highlight the mechanism, the link 11 in FIG. 3a has been split into two distinct links: an uplink / if and a downlink 11r, associated with the buffers 45 and 44, respectively. A distinction was also made between "haste - service provider" and reverse links: 134 and

143, respectively.

  Similarly, the end user terminal, referenced SIT ', includes two buffer memories, 33 and 34, associated with uplinks to the Sat satellite, 12r and

descending, 12f, respectively.

  The data sent by the remote host RH to the terminal SIT 'on the link 134 are the subject of an acknowledgment of receipt as soon as they reach the installation of the service provider ISP' and that they "penetrate" into the specific module 4. This arrangement therefore makes it possible to subtract from the global propagation delay, that is to say the above-mentioned parameter "RTT", the outward and return propagation time in the space segment. As indicated, this propagation time constitutes the main fraction of the

  "RTT", due to the propagation time back and forth of electromagnetic waves to travel the distance "earth -

  satellite - earth ", that is to say approximately 0.5 s. The standard variable window 30 is maintained at the level of the ISP 'service provider so as to maintain normal interactions with the

rest of the RI Internet network.

  The data in the direction of the user terminal SIT ′ is then transferred by the specific protocol, using the buffer memory 45, so as to guarantee high reliability. Since the negative acknowledgment mechanism is used, the ISP service provider 'does not have to wait for any acknowledgment signal from the final recipient to transmit the data to the SIT terminal', which is a gain sensitive time. These data are transmitted to the memory

  buffer 34 of terminal SIT ', via links 1lf and 12f-

  In the other direction, the data sent by the SIT terminal 'for any recipient on the network

  Internet RI, for example distant haste RH, are subject by application 3 to specific layers (half-module 31-

  32), then via the protocol specific to the ISP 'service provider (half module 40-42), via the buffer memories

33 and 44.

  Finally, the data is submitted in a conventional manner to the "IP" module resident in the system.

  IT service provider ISP ', to be delivered to the remote host RH by the conventional terrestrial Internet network RI, by the link 143. This, in turn, transmits an acknowledgment of receipt, by the link 134 to the ISP service provider '.

  The SIT 'user terminal should not wait for this acknowledgment to transmit data

  additional. The reliability of the data exchange between the two entities SIT 'and RH is guaranteed by the negative acknowledgment mechanism described above.

  In accordance with this mechanism, the data exchanged is temporarily stored in the buffer memories, 33 and 34 for the SIT terminal ', and 44 and 45 for the ISP service provider30. If, at the end of a predetermined time interval, one or other of these entities has not received an alert packet or "NAK", the data stored in the buffer concerned, corresponding to a previously transmitted packet can be deleted. Otherwise, they are re-issued, a transmission error

  being detected in the manner which has been described previously and causes the emission of the signal "NAK".

  The method according to the invention guarantees absolute transparency with respect to the other protocols used in transmissions on networks of the Internet type, in particular for the protocol known by the acronym "UDP" (for "User Datagram Protocol"). Indeed, these protocols do not have the same sensitivity to the effects

  Undesirable broadband and long delay links.10 It is therefore unnecessary to have recourse to the provisions which have just been described for the "TCP" protocol.

  FIG. 5 illustrates an alternative architecture in accordance with the invention which allows multiplexing and

  demultiplexing of data traffic, between two protocols.

  On the user terminal side, referenced SIT ", the software module 3" is subdivided into two sub-modules, Ml and M2. The

  first module, Ml, comprises the application layers 30 and the specific layers 31-32 of FIGS. 3a or 4.

  The second submodule, M2, includes, in common with the first submodule M1, the upper application layers 30, a "UDP" protocol layer and a

"IP" network addressing layer.

  In the system of the service provider SIT ", we find the module 4 unchanged, this module comprising the specific layers 40-41 and the network addressing layer" IP "42. The latter communicates, as before, with the layer d "IP" network addressing, referenced 54 "of a

  5 "gateway, via bus 42 and any pilots (not shown).

  The module 4 and the sub-module M1 communicate with each other via the space segment, referenced S $. We represented

  the "earth - satellite - earth" links by a single bidirectional link S $ a.

  The 5 "gateway includes a pilot connected to the Internet RI network via the main route 12.

  The network addressing layer "IP" 54 "of the gateway 5" is connected to the submodule M2 also by the spatial segment Sl, and more precisely by a bidirectional link Slb. There are therefore two parallel data transmission channels, both using the space segment, between the ISP service provider "and the SIT user terminal" .10 However, the data transmitted is processed differently, depending on the protocol used at the level of the

transport layer.

  Since the data packets include, in the protocol fields, information indicating which transport protocol is used, "TCP" or "UDP" in the example described, the data streams can be

  demultiplexed (direction RH to SIT ") or multiplexed (direction SIT" to RH) by the network addressing layer "IP" 54 "in module 5" (gateway) of the ISP service provider ".

  In other words, the data under the "UDP" transport protocol are transmitted directly from the ISP service provider "to the SIT terminal", or vice versa, by the Slb link, as they would have been if the modules specific to the the invention was not implemented.25 They do not, moreover, pass through these modules. For this type of transmission, under the "UDP" transport protocol,

  the architecture of the system is, in itself, quite similar to an architecture in accordance with known art, for example that described with reference to FIG. 2.

  On the other hand, with regard to the data transmitted under the "TCP" protocol, the path followed is quite

  similar to that described with reference to FIGS. 3a or 4, once the demultiplexing process has taken place in the layer 54 ".

  In the SIT terminal ", there is no demultiplexing, nor indeed multiplexing, strictly speaking. The data transmitted by the service provider ISP" arrive in two distinct channels towards the application layer 30. In in the opposite direction, it is the application layer 30 which directly transmits the data to be transmitted to one or the other channel: in module M1 for the transport protocol "TCP" or submodule M2 for the

"UDP" transport protocol.

  Multiplexing is systematic towards the Internet network RI, via the main artery 12, that is to say in

  exit from the network addressing layer "IP" 54 ", since all the data, whatever the transport protocol used, take the same channel on this network.

  In summary, from a logical point of view, part of the logical layers of the user terminal SIT "are

  implanted in the ISP service provider system ". Naturally the" IP "network addressing layer relating to the SIT terminal" I, but physically located in the ISP system "is characterized by the same addresses as that located in the SIT terminal ".

  On reading the above, it is easy to see that the invention achieves the goals it has set for itself.

  It brings many advantages, including the following: First, it allows better control of the data flow, as well as more efficient management. The decoupling between the long-delay space network segment and the rest of the Internet network, that is to say the conventional terrestrial network, in fact allows a very significant improvement in the control of data flows and in the mechanisms for managing dimensions. This is due to the fact that the reduction of end-to-end delays, achieved by the management mechanisms specific to the invention makes these

  more efficient mechanisms and faster response.

  This provision meets one of the most delicate needs posed by an environment as changeable and inhomogeneous as an Internet type network can be.

  The invention also allows a very significant reduction in additional transmissions on telecommunications satellite links. Indeed, for each "TCP" data segment, a minimum of 40 bytes is added to the data that can be described as "useful". Out of these 40 bytes, 20 bytes constitute the "IP" header necessary for addressing and 20 bytes constitute the "TCP" header. As noted, the new version of the protocol, "IPv6", allows more address space

  important, which requires an even longer header.15 Since the data transfers on the space segment are based on a "proprietary" type protocol, the

  additional "TCP" and "IP" headers are no longer required, which increases the efficiency of data packets.

  Traffic is reduced in the space segment. Indeed, traffic control on satellite links are limited to the exchanges normally existing between the "TCP" layers and the application layers. The "TCP" data segments carrying important "IP" and "TCP" headers, for data management commands and connection management, are thus avoided on the links.

  by telecommunication satellite.

  The invention also allows a limited connection opening time, on both sides. Indeed, according to a very important characteristic of the invention, the "TCP" interface has been moved and is now located directly on the Internet itself (terrestrial segment). This

  This arrangement eliminates the problems relating to the so-called "slow start" mechanism.

  The invention also makes it possible to reduce the memory resources required on the user side. This is due to the implementation of the so-called "negative" acknowledgment mechanism. The transmission buffer requirements are very significantly reduced compared to a conventional "TCP-to-TCP" system according to the known art. Indeed, in the case of a loss of a data packet, a "NAK" signal,

  for data transmitted on the return link, is returned to the service provider within the limits of a predetermined fixed time interval.

  In a system according to known art, the "TCP" module stores the transmitted packets until an acknowledgment of receipt is received.

  reception says "positive" or "ACK". The time elapsed until reception is extremely variable. This is due to the changing environment characterizing the Internet15. All transmitted packets must therefore be stored until receipt of acknowledgments of receipt concerning them.

  For these reasons, while a conventional system requires sufficient memory resources to accommodate the most extreme situations, that is to say an "ACK" signal received after a very long delay, the invention for its part only requires one

  minimum buffer, since the required size is always set at a small value, a direct function of the above-mentioned fixed time interval.

  As has been shown, the process of the invention can be made entirely "transparent" to the

  protocols other than the "TCP" transport protocol, in particular for the "UDP" protocol.

  Finally, the invention retains end compatibility

  in the end with the Internet environment.

  It should be clear, however, that the invention is not limited to only examples of embodiments explicitly

  described, in particular in relation to FIGS. 3a to 5.

  In particular, the so-called "proprietary" transport protocol for data transmissions on links in the space segment (earth satellite - earth) is not limited to a single protocol based on the only well known data transmission standard " DVB-S "above, even if the latter is particularly suitable for such links. Other data transmission standards can be implemented without departing from the scope of the invention. Indeed, due in particular to the division of the "TCP" modules of the user terminals into two parts "straddling" the ends of the space segment, the latter, from a logical point of view, remains "invisible" from the

rest of the Internet.

  Numerical values have been specified only to fix ideas. They are essentially linked to the current state of standardization (version) of the

"TCP" transport.

  Finally, in an alternative embodiment not described, it is possible to design a network having several spatial sections, for example for connecting two user terminals through the Internet, each

  being connected to a service provider via separate telecommunications satellites.

Claims (15)

  1. Method for transmitting data between at least one user terminal (SIT) and a first computer system called remote host (RH), under Internet-type protocol, via a network comprising a terrestrial segment (RI) ) and at least one space segment consisting of a bidirectional link (11, 12) by telecommunication satellite (Sat), to which said terminal (SIT) is connected, said land (RI) and space segments being interconnected by a second system IT, said service provider (ISP), characterized in that, the transmissions being associated with a stack of protocol layers, each layer communicating by interfaces with the layers which are immediately above and below it, and said stack comprising at least an upper layer of software applications (30), a first intermediate layer, immediately lower, called the transport layer, of the "TCP" type, and one two ith intermediate layer, called "IP" type network addressing, the method comprises: - the distribution of said "TCP" layer of each user terminal in two half-modules so that they each include a "TCP" transport layer 25 (31, 41) and a specific interface layer (32, 40), - the location of said half-modules at the ends of said spatial segment, a first half-module being installed in said computer system service provider (ISP) and a second half-module being installed in the user terminal (SIT), - the transfer of the "IP" network addressing layer relating to the user terminal (SIT) in the system IT service provider (ISP), - data transmission on said bidirectional link (11, 12) using an internal transport protocol based on a specific data transmission mode, - bidirectional conversion between said internet type protocol and the said internal protocol by said specific layer (40, 32) of the first and second half-modules, - and the creation of an interface between said transport layer (31) of the second half-module and a layer   software applications (30) installed in the user terminal.   2. Method according to claim 1, characterized in that said specific data transmission mode on said bidirectional link (11, 12) by telecommunication satellite (Sat) is the standard for broadcasting   digital video signals of the "DVB-S" type.   3. Method according to claim 1, characterized in that, said user terminals (SIT ') and said remote host (RH') being able to be transmitter and / or receiver of data packets, it implements a procedure of acknowledgment of receipt comprising the following steps: - for each transmission of a data packet, by a transmitter to a recipient, the storage by the transmitter of a copy of said data packet; the transmission by said recipient of an alert packet signaling a packet of transmitted data missing on detection of a dead time in the reception of a sequence of data packets exceeding a predetermined timeout period; - the conditional retransmission of said data packet stored on reception by the transmitter of said alert packet; - or the deletion of this data packet after   the expiration of said delay period.   4. Method according to claim 3, characterized in that the transmission of data packets from said remote host (RH) to one of said user terminals (SIT ') comprises the following steps: - transmission by the host distant (RH) from a data packet; - the generation of an acknowledgment by said service provider (SIT ') upon receipt of said data packet, by said first half-module, and the transmission of this acknowledgment to said remote host; - And the transmission of said data packet to the user terminal (SIT ') by implementation of said   internal protocol, via the first and second half   modules. 5. Method according to claim 3, characterized in that the transmission of data packets from one of said user terminals (SIT ') to said remote host (RH) comprises the following steps: - the submission of a packet of data by said software application layer (30) from the data terminal to the first half-module and the transmission of said data packet to said service provider (ISP '), by implementing said internal protocol on said bidirectional link (l1f , 12f, 11r, 12r) of the space segment, via the second half-module; - the submission of said data packet by an "IP" network addressing layer (42) of the service provider (ISP ') to said terrestrial segment of the Internet network (RI), for transmission to the remote host (RH) under said Internet-type protocol; - and the generation of an acknowledgment by the remote host (RH), on receipt of the data packet to   destination of the service provider (ISP ').   6. Method according to any one of the claims   above, characterized in that said network (RI) carrying data streams under said transport protocol "TCP" and at least one second transport protocol, the transmission of data streams from said remote host (RH) to one of said user terminals (SIT ") comprises the following steps: - demultiplexing said data streams transmitted by the remote host (RH) by an" IP "network addressing layer (54") of said service provider ( SIT "); - the transmission of said demultiplexed data streams, under the" TCP "transport protocol, to the user terminal (SIT"), via a first bidirectional transmission link (Sla) of the space segment implementing said protocol internal, via the first and second half-modules; - and the direct transmission of said demultiplexed data streams, under said second transport protocol, to the user terminal (SIT "), via a second bidirectional transmission link (Slb) of the space segment implementing the second protocol and specific layers (M2) of this protocol, located in the user terminal (SIT ") and comprising an interface with said layer software applications (30).   7. Method according to claim 6, characterized in that the transmission of the data streams from one of said user terminals (SIT ") to said remote host (RH) comprises the following steps: - the selective submission of data by said software application layer (30) to the second half-module (Ml) when the data must be transmitted under the "TCP" T transport protocol or to specific layers (M2) to said second transport protocol when this data must be transmitted under this protocol; - the transmission of said data under the "TCP" protocol to said service provider (ISP "), via said first link (Sla) of the space segment and the first and second half-modules; - direct transmission of said data under the second protocol to said service provider, via said second link (Slb) of the space segment; - multiplexing of data under the two protocols by a network addressing layer "IP" (54 ") of the service provider (ISP"); - and the submission of the multiplexed data, by said network addressing layer "IP" (54 "), to said terrestrial segment of the Internet network (RI), for a transmission to the remote host (RH).   8. Method according to claims 6 or 7,   characterized in that said second protocol for transport is the "UDP" protocol.   9. IT architecture for implementation   of the method according to any one of the claims   , characterized in that said user terminals (SIT) comprise a software application layer (30), in that these user terminals (SIT) and the computer system of said service provider (ISP) each comprise a half-module in which are implanted a "TCP" transport layer (31, 41) and a specific interface layer (32, 40) with said bidirectional transmission link (11, 12) by telecommunication satellite (Sat) and ensuring bidirectional conversion between said "TCP" transport protocol and said internal protocol, and in that the service provider's computer system (ISP) further comprises a relative "IP" network addressing layer (42) to the network addresses of the terminal user (SIT).   10. Computer architecture according to claim 9, characterized in that said bidirectional communication satellite link comprises a forward channel (/ lf, 12f) and a return channel (11r, 12r) between said user terminals (SIT ') and the computer system of said service provider (ISP '), and in that said system (ISP') and the terminals (SIT ') each comprise two buffer memories (44-45, 33-34), associated respectively with the outgoing channels ( / if, l2f) and return (11r, 12r).   11. IT architecture according to claims   tions 9 or 10, characterized in that the system of said service provider (ISP) further comprises a gateway (5), disposed between said first half-module and said terrestrial segment of the network (RI) and in that this gateway ( 5) comprises at least one "IP" network addressing layer (54) relating to the service provider (ISP) and an interface layer (53) called the pilot network management.   12. IT architecture according to the claims   tions 11, characterized in that, said network (RI) carrying data streams under said transport protocol "TCP" and at least one second transport protocol, said user terminals (SIT ") comprise a stack of layers ( M2) comprising at least one "IP" network addressing layer (36) and a transport layer (35) specific to said second protocol communicating with said software application layer (30), in that said spatial segment comprises first (SLa) and second (SLb) bidirectional data transmission links, the first link (SLa) ensuring transmission under said internal protocol and the second link (SLb) under said second protocol, and in that the computer system of said service provider (ISP ") comprises means for demultiplexing (54") data streams received from said remote host (RH), carried under one or the other of said two protocols, so as to transmit select ectively, towards said user terminals (SIT "), the streams given under the" TCP "transport protocol via said first half-module, said first link (SLa) and said second half-module, and the data streams under the second transport protocol via said second link   (SLb) and said layer stack (M2).   13. Computer architecture according to claim 12, characterized in that the computer system of said service provider (SIT ') comprises means for multiplexing (54 ") data streams received from said user terminals (SIT"), conveyed under said transport protocol "TCP" or said second transport protocol, in a data stream transmitted to said remote host (RH) via said gateway (5) and said segment terrestrial (RI).   14. IT architecture according to claims   tions 12 or 13, characterized in that said stack (M2) comprises at least one "UDP" transport layer (35)   and an "IP" network addressing layer (36).   15. Computer architecture according to any one   claims 12 to 14, characterized in that said   multiplexing and multiplexing means are constituted by said network addressing layer "IP" (54 ") of said gateway (5).