Classifications

• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/12—Arrangements for detecting or preventing errors in the information received by using return channel
  • H04L1/16—Arrangements for detecting or preventing errors in the information received by using return channel in which the return channel carries supervisory signals, e.g. repetition request signals
  • H04L1/1607—Details of the supervisory signal
  • H04L1/1642—Formats specially adapted for sequence numbers
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/12—Arrangements for detecting or preventing errors in the information received by using return channel
  • H04L1/16—Arrangements for detecting or preventing errors in the information received by using return channel in which the return channel carries supervisory signals, e.g. repetition request signals
  • H04L1/1607—Details of the supervisory signal
  • H04L1/1685—Details of the supervisory signal the supervisory signal being transmitted in response to a specific request, e.g. to a polling signal
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/12—Arrangements for detecting or preventing errors in the information received by using return channel
  • H04L1/16—Arrangements for detecting or preventing errors in the information received by using return channel in which the return channel carries supervisory signals, e.g. repetition request signals
  • H04L1/18—Automatic repetition systems, e.g. Van Duuren systems
  • H04L1/1829—Arrangements specially adapted for the receiver end
  • H04L1/1835—Buffer management
  • H04L1/1838—Buffer management for semi-reliable protocols, e.g. for less sensitive applications such as streaming video
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L12/00—Data switching networks
  • H04L12/02—Details
  • H04L12/16—Arrangements for providing special services to substations
  • H04L12/18—Arrangements for providing special services to substations for broadcast or conference, e.g. multicast
  • H04L12/1863—Arrangements for providing special services to substations for broadcast or conference, e.g. multicast comprising mechanisms for improved reliability, e.g. status reports
  • H04L12/1868—Measures taken after transmission, e.g. acknowledgments
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L47/00—Traffic control in data switching networks
  • H04L47/10—Flow control; Congestion control
  • H04L47/34—Flow control; Congestion control ensuring sequence integrity, e.g. using sequence numbers

Definitions

• the present invention relates to end-to-end flow control techniques between communicating nodes in a packet communications network, and more particularly to an optimal implementation of an end-to-end reliability mechanism in the delivery. Datas.
• end-to-end reliability is meant here the end-to-end delivery guarantee (also referred to as “end-to-end”) of messages sent from a source node to one or more destination node (s). .
• various errors may occur such as packet corruption, packet disordering, packet loss or duplication.
• various end-to-end reliability mechanisms are provided such as, in particular, adding, by the source node, an end-to-end sequence number to each transmitted message so that the loss of a message can be detected by the destination node by noting a jump in the sequence numbers of the received messages;
• each source-destination pair has a dedicated end-to-end sequence number (one in each sense).
• the end-to-end sequence number of a given pair is incremented by one each time a new message is sent between these two nodes.
• the destination node checks for each source node that the received packets have successive sequence numbers.
• the number of sequence numbers is to be multiplied by the number of virtual networks used. For example, to guarantee the scheduling in a network of 65,536 nodes to 4 virtual networks, it is necessary to store 2 (transmission / reception) x 65,536 (number of nodes) x 4 (number of virtual networks) end-to-end sequence per node. In this case, with a 16-bit sequence number (plus 6 bits for the addition of an error correction code), a memory of 524288 22-bit inputs, that is 11 264 KB, is required. The allocation of such memory space, at each source node and each destination node of the network, is very expensive in resources, and impacts the performance of the network in a proportional manner to its size.
• the destination node informs the source node by returning a "non-acknowledgment" response message specifying the type of error encountered. Subsequently, the source node retransmits the packet (s) concerned.
• the source node stores in a retransmission memory a copy of all the messages it sends. When it receives a positive acknowledgment, the source node may delete the corresponding message from the retransmission memory. On the other hand, when it receives a negative acknowledgment (ie of "non-acknowledgment" type) or if it does not receive a response message for a certain message sent to a destination node, the source node proceeds to its retransmission and if necessary to the retransmission of the following messages to this same destination node.
• a negative acknowledgment ie of "non-acknowledgment" type
• the destination node refuses all the following packets an erroneous packet because the verification of the end-to-end sequence number is then erroneous until the first erroneous message is retransmitted and received without error by the destination node.
• a disadvantage of this retransmission of data lies in the inconvenience of its implementation. Indeed, the source node must retransmit the error message and all subsequent messages to the destination node that detected the error and only this recipient. However, since the messages sent to the different destination nodes are stored in the same retransmission memory (otherwise, a retransmission memory per destination node is practically unrealizable), it is necessary to browse this retransmission memory to identify and retransmit only the required messages to the destination node that detected the error. This certainly affects the performance of the end-to-end reliability mechanism. Furthermore, it is possible that the acknowledgment message is lost even though the corresponding message has arrived at the destination node.
• the source node retransmits this message and requests the retransmission of an acknowledgment message, which causes a duplication of the message on the destination node side.
• Message duplication can also occur in the event of severe congestion of the network slowing down the transmitted message or its acknowledgment so that the source node retransmits this message thinking that it has been lost. As a result, the receiving node will receive the same message twice.
• the maximum possible value of the end-to-end sequence number be much greater than twice the number of messages that a source node can send during a second message.
• duration equal to the sum of the timeout and the worst propagation time between this source node and a destination node.
• the timeout is a timing (RTO for Retransmission Time Out) activated by the source node when sending a data packet and after which the acknowledgment of receipt of this packet is expected.
• the maximum possible value of the end-to-end sequence number must be very much greater than one billion two hundred and fifty million. It must, therefore, be encoded on at least 35 bits. In other words, this amounts to adding in each sent message an end-to-end sequence number of approximately 35 bits. However, adding a sequence number of such a size represents a very expensive overhead which results in a considerable reduction in the useful bandwidth between the source node and the destination node (s).
• a 35-bit sequence number also increases the memory space required for storage at the nodes.
• An object of the present invention is to overcome the aforementioned drawbacks.
• Another object of the present invention is to reduce the complexity of implementing end-to-end flow control techniques, especially in large-scale IP networks.
• Another object of the present invention is to increase reliability in data delivery within IP networks.
• Another object of the present invention is to propose an end-to-end protection method which makes it possible to guarantee the reliability of a large-scale network while limiting:
• the invention proposes, according to a first aspect, a method for managing the end-to-end reliability in the delivery with acknowledgment of data from a source node to a destination node group in a communications network, the destination node group including at least a first destination node, the method comprising the steps of: tagging messages transmitted from the source node to said first node destination by an incremental end-to-end sequence number so that said first destination node expects, following a first message received from said source node, to receive a second message marked with a number of an end-to-end sequence expected from said source node;
• the calculated difference being equal to a predefined threshold, suspending the transmission of messages from the source node to the destination node group;
• the end-to-end sequence number marking a message received by said first destination node and from said source node being different from said expected end-to-end sequence number, deduced by said first destination node the presence of an error in the delivery of data.
• the method presents, according to various realizations, the following characters, if necessary combined: the end-to-end sequence number marking a received message being greater than the expected end-to-end sequence number, the error is a lost message;
• the error is that the received message is a duplicate message
• the predefined threshold is the integer value of half of said predefined maximum value
• the global sequence number is coded on sixteen bits
• the end-to-end sequence number is coded on sixteen bits.
• This method further comprising: when the global sequence number reaches said predefined maximum value, a step of initializing this global sequence number;
• the invention proposes a computer program product implemented on a memory medium, capable of being implemented within a computer processing unit and comprising instructions for the implementation of the method. summarized above.
• Figure 1 illustrates an embodiment of delivery of data from a source node to a plurality of destination nodes
• Figure 2 illustrates a method of storing and accessing data according to one embodiment.
• a source node 10 transmitting messages 1, 2, 3 to destinations, respectively, of a plurality of destination nodes 21, 22, 23 belonging or not to the same virtual network is displayed.
• the source node 10 transmits, to the destination nodes 21, 22, 23, messages 1, 2, 3 via links L1, L2, L3 implementing a transmission protocol with acknowledgment 4-6 (that is, with acknowledgment of receipt or with acknowledgment).
• each message 1-3 is marked, at the source node 10, with an end-to-end sequence number n1-n3 which represents the position of this message in the stream of ordered messages from the source node 10 and to the corresponding destination node 21-23.
• the end-to-end sequence number n1 marking the message 1 represents the order of this message 1 in the set of messages sent by the source node 10 to the destination node 21. It follows that following a message received by a destination node 21, the latter expects the reception of a message, from the source node 10, marked by an expected end-to-end sequence number (n1 + 1) which is successive to the end-to-end sequence number of the received message (n1).
• the end-to-end sequence numbers n1-n3 are coded on 16 bits.
• the limitation of the size of end-to-end sequence numbers n 1 -13 to 16 bits represents a gain of 20 bits per message compared to conventional methods where the size of a sequence number of end-to-end -Bout is 32 bits.
• the sending of any message 1-3 by the source node 10 increments a global sequence number N which represents the position of this message in the global ordered data stream sent by the source node 10 to the destination nodes. -23.
• N represents the position of this message in the global ordered data stream sent by the source node 10 to the destination nodes.
• the number of The overall sequence of the message 2 is its order on the set of messages 1-3 sent by the source node 10 to a destination node group 21-23.
• the global sequence number N is 16-bit coded, being of the same size as the end-to-end sequence number n1-n3.
• the global sequence number N can not take values above two exponents sixteen minus one (2 16 -1).
• the global sequence number increments cyclically, i.e., it is initialized each time it reaches this maximum possible value Nmax (that is, a circular global sequence number or, again, a global sequence number modulo Nmax).
• the global sequence number N is not transmitted with the messages 1-3 on the network 30, and is kept in a retransmission memory 11 associated with the source node 10. It should be noted that more than one destination group of nodes 21-23 can be envisaged, in which case a global sequence number N is defined for each destination node group.
• a predefined size hash table (in this case, 32) makes it possible to obtain on the basis of the destination node 21-23, the end-to-end sequence number n1-n3 of message 1-3.
• the destination node 21 is used here as the key for the hash function to access memory space at the source node 10 which includes the end-to-end sequence number n1 to be assigned to the message 1.
• a memory space is used to store the end-to-end sequence number n1-n3 of each destination node 21-23 (one entry per destination).
• the end-to-end sequence number n1-n3 is obtained by reading the stored end-to-end sequence number in the memory space corresponding to the destination node 21-23. After each read of an end-to-end sequence number n1-n3, an increment of one of the end-to-end sequence number read is performed. That is, the end-to-end sequence number that has just been read, for the transmission of the message 1-3 to the destination node 21-23, is incremented by one. Thus, the next message to be transmitted to this same destination node will have an end-to-end sequence number incremented by one. In this case, the end-to-end sequence number n1 is incremented by one for each message 1 addressed to the destination node 21 which is the key to access this end-to-end sequence number.
• This sequence number n1 is incremented linearly with the number of messages 1 sent by the source node 10 to the destination node 21. This sequence number n1 thus makes it possible to identify the message 1 addressed to the destination node 21 among the set messages addressed to this destination node 21.
• a hash table is defined by destination node group 21-23.
• Doubly linked lists c1-c3, respectively, associated with the end-to-end sequence numbers n1-n3 (and therefore each destination node 21-23) are configured to store messages 1-3 therein, respectively.
• a doubly-linked list includes, for each message in the list, a pointer to the next message, and a pointer to the previous message in the relevant list.
• the doubly chained lists c1-c3, respectively comprising the messages n1-n3, are stored in the retransmission memory 11.
• a copy of the messages 1-3 sent from the source node 10 is inserted according to their destination, respectively, in the chained double lists c1-c3.
• the doubly-linked list c1-c3 avoids, in case of error on an ordered message 1-3, to browse all the inputs of the retransmission memory 11 and, in particular, to retransmit to the destination node 21 -23 affected all messages following the error message.
• the use of doubly-linked lists c1-c3 also makes it possible to reduce the travel time of the retransmission memory 11 in the event that an error is detected requiring the return of several messages to a destination node 21-23, without however, to significantly increase the memory space of the retransmission memory.
• the sending of message on the network 30 is suspended.
• the destination node 21-23 deduces that the received message is a duplicate message.
• the destination node 21-23 deduces that a message is lost.
• the destination node 21-23 can easily distinguish a duplicate message from a lost message by a simple comparison between the expected end-to-end sequence number n1-n3 and the end-to-end sequence number. received.
• the destination node 21-23 is therefore responsible for detecting errors (loss or duplication of messages), and for requesting retransmission of the messages when it deems it necessary.
• the predefined threshold is equal to half of the maximum possible value Nmax of the global sequence number N, or more generally the integer value of half of the maximum possible value Nmax of the global sequence number N.
• the end-to-end flow control uses a window of width Nmax / 2 on the global sequence numbers N and whose upper bound is the global sequence number N of the next message to be sent.
• the consequence is the suspension of the transmission of new message until the reception of the acknowledgment of the oldest message for which no acknowledgment has been received by the source node 10.
• a implicit acknowledgment mechanism is triggered. It consists of browsing the messages in the retransmission memory 11 for the doubly linked list c1-c3 associated via the hash table to the destination node concerned. If a more recent ordered message to the same destination node 21-23 has received an acknowledgment 4-6, it means that the oldest message has also been acknowledged but this acknowledgment has been lost. In this case, it is decided to pay it implicitly and the traffic can resume.
• the retransmission memory 11 is furthermore used to manage the transport layer with a view to retransmitting the messages, thus limiting the increase in the required memory space.
• the retransmission memory 11 in order to increase the number of read and write ports of the retransmission memory 11, is divided into a plurality of memory banks.
• the transmission memory is divided into four banks of memory to multiply the number of read and write ports available by four.
• the distribution of the messages on the different banks of memories can be determined by a hash function dependent on the destination node 21-23 of the message 1-3.
• the retransmission memory 11 can be used both for the application layer (sending of end of message event) and for the transport layer (management of retransmissions) without affecting the performance of a node in terms of the number of messages it can process per second.
• message scheduling is provided for only two out of four virtual networks. This makes it possible to halve the size of the retransmission memory 11 necessary to store the end-to-end sequence numbers.
• the two sequence numbers are stored in the same input and protected by one and the same error correction code. This allows an additional ten percent gain in memory without impacting performance (a single sequence number for one of the two virtual networks is needed at most for each cycle).
• the embodiments described above apply regardless of the type of acknowledgment transmission protocol used between a data source node and one or more destination node (s).
• This transmission protocol can be in point-to-point mode, or in point-to-multipoint mode.
• this transmission protocol with acknowledgment can be TCP (Transmission Control Protocol) or SCTP (Stream Control Protocol).
• node in the above description means any source element and / or destination of data in an IP communications network such as a terminal, a server or a router.
• a message here denotes a packet, or a data frame.
• the embodiments described above allow an orderly and reliable delivery (ie without loss) of data on several links by using a minimum of resources, particularly in terms of memory spaces.
• the various embodiments described above make it possible to optimize, in terms of resources and performance, the flow control in a packet communication network.

Landscapes

• Engineering & Computer Science (AREA)
• Computer Networks & Wireless Communication (AREA)
• Signal Processing (AREA)
• Computer Security & Cryptography (AREA)
• Multimedia (AREA)
• Detection And Prevention Of Errors In Transmission (AREA)
• Communication Control (AREA)
• Data Exchanges In Wide-Area Networks (AREA)

Applications Claiming Priority (2)

Publications (1)

Family

ID=51862393

Family Applications (1)

Country Status (6)

Families Citing this family (3)

Citations (1)

Family Cites Families (13)

• 2014
  • 2014-06-06 FR FR1455147A patent/FR3022094B1/fr active Active
• 2015
  • 2015-05-21 WO PCT/FR2015/051343 patent/WO2015185824A1/fr active Application Filing
  • 2015-05-21 EP EP15732319.7A patent/EP3152876A1/de not_active Withdrawn
  • 2015-05-21 JP JP2016567377A patent/JP6547973B2/ja not_active Expired - Fee Related
  • 2015-05-21 US US15/316,355 patent/US10110350B2/en not_active Expired - Fee Related
  • 2015-05-21 BR BR112016028224A patent/BR112016028224A2/pt not_active IP Right Cessation

Patent Citations (1)

Also Published As

Similar Documents

Legal Events