WO2007124967A1

WO2007124967A1 - Process for the distribution of data blocks broken down into partial blocks in a network that includes multiple nodes as well as relevant nodes and a system

Info

Publication number: WO2007124967A1
Application number: PCT/EP2007/051770
Authority: WO
Inventors: Norbert Oertel; Jürgen PANDEL
Original assignee: Siemens Aktiengesellschaft
Priority date: 2006-04-27
Filing date: 2007-02-23
Publication date: 2007-11-08
Also published as: DE102006019711A1

Abstract

The invention relates to a process for the distribution of data blocks broken down into partial blocks in a network that includes multiple nodes, whereby each partial block distributes to the nodes assigned to it on the distribution tree and the nodes are assigned to the distribution tree in such a way that every internal node is a leaf node of the other distribution trees which add redundancy data to the data block prior to breaking it up into partial blocks and is received at least partially by one of the partial blocks of the internal node of the distribution tree, which completely reconstructs the partial block of the distribution tree based on at least partially received partial blocks and transmits the partial block reconstructed by the internal node to at least one subsequent node in accordance with the distribution tree. The invention also relates to a node and a system that includes a data source and the node.

Description

description

Method for distributing a data block divided into sub-blocks in a multi-node network, and associated node and system

The invention relates to a method for distributing a data block divided into sub-blocks in a multi-node network, as well as an associated node and a system.

For a distribution of data blocks, including, for example, audio and / or visual information, broadcast and / or multicast techniques have been developed, such as Multimedia Broadcast / Multicast Service (MBMS) of the

3GPP Panel (3GPP - Third Generation Partnership Project).

For distributing such data blocks, point-to-point connections are established between a server and one or more terminals (unicast). In an alternative to this, a distribution can be carried out with the aid of a multicast protocol in such a way that the data block (s) is forwarded to a node in a network, which in turn forwards the data blocks to the terminals accessible to it.

Furthermore, an ALM method (ALM - Application Layer Multicast) is known in which some terminals or subscribers receive the data blocks directly from the data source, for example the server, these subscribers passing on the received data blocks to other subscribers. In today's transmission systems, such as DSL (DSL - Digital Subscriber Line), there is a high bandwidth in ei ^¬ ner downlink, but a low bandwidth in an uplink (uplink). Especially when using such asymmetric transmission systems, the ALM method has the disadvantage that the low uplink capacity is heavily utilized by applying the ALM method. To reduce this disadvantage, the split-stream technology has been developed [1], in which the data block is divided into several sub-blocks, which are then sent to different subscribers. These participants forward the sub-blocks to other subscribers, said distribution trees of each ^¬ weiligen sub-blocks are designed so that each internal node of a distribution tree is a leaf node of the other distribution trees. This measure ensures that only one sub-block can not be forwarded in case of failure of a participant. The distribution of Teilblö ^¬ blocks by means of the distribution trees will be each represented by a distribution graph such as is shown in [1]. The participant is called a node.

However, this solution has the disadvantage that the already ge ^¬ rings uplink capacity of a subscriber is further reduced, even if this reduction takes place only temporarily, for example, if the subscriber simultaneously sends emails with large attachments. In this case, the participant can no longer forward the sub-blocks within his distribution tree as intended, so that all subsequent subscribers of the corresponding distribution tree do not or only partially receive the associated sub-block. Supply channels also when transmitting the sub-block on disturbed Übertra ^¬, gung for example, in a wireless Datenübertra ^¬, the sub-block reaches the subscriber or node defective and / or incorrectly or not at all.

Thus, it is an object to provide a method, a node and a system in which a distribution of a data block divided into sub-blocks in a multi-node network is designed such that a reliability in receiving error-free sub-blocks in the network is improved. This object is solved by the independent claims. Further developments can be found in the dependent claims.

The invention relates to a method for distributing a divided into sub-blocks data block in a multiple-node network, each sub-block each distributed on the basis of its associated distribution trees to the nodes, and the nodes are so associated ^¬ trees the distribution that each inner Node of one of the distribution trees is a leaf node of the other distribution trees, the data block before dividing into sub-blocks added redundancy data, at least one of the sub-blocks at ^¬ least partially received by the inner node of the distribution tree, the distribution tree associated sub-block through the inner node as a reconstructed sub-block Basis of the at least partially received partial ^¬ blocks completely reconstructed and the reconstructed part ^¬ block are forwarded by the inner node according to the distribution tree to at least one subsequent node.

By this method it is achieved that the part ^¬ block of an internal node, the following from this node to him node is to be sent, can be completely reconstructed as rekonstruier ^¬ th sub-block, even if that sub-block is not the internal nodes at all, partially complete and / or faulty he ^¬ ranges. This improves reliability in receiving error-free sub-blocks in the network.

The data block can be part of a data stream, e.g. at a

Streaming or a download application. In this case, several data blocks are transmitted one after the other, which represent the data stream. Furthermore, individual symbols of the data block may be redundant in more than one sub-block. In practice, however, it is appropriate that the part describing ^¬ blocks disjoint sets of the data block. Preferably, a method for generating an error correction code is used for generating the redundancy data, in which an error correction of a symbol of the data block is performed on the basis of a freely selectable combination of symbols of the redundancy data and symbols of information data of the data block. A symbol is formed by one or more bits, for example one byte or 64 bits. In this extension, the reconstructed sub-block can be completely reconstructed from any combination of symbols of the information data and / or the redundancy data, whereby a high flexibility in a selection of symbols for complete reconstruction of the reconstructed sub-block is achieved. Furthermore, can be achieved by this high flexibil ^¬ ty, that a high rate it is possible to achieve a complete reconstruction by means of the error correction. This is preferably achieved by a Correction ^¬ rekturcode, by a method of creating Reed-Solomon codes, Fountain codes, LDPC codes (LPDC - Low-Density Parity Check) is formed or Raptor codes.

Preferably, a first number of symbols per sub-block is generated by a minimum value which is determined by the fact that the minimum value of a second number of Symbo ^¬ len of the data block, comprising mationsdaten the redundancy data and information, and a data throughput time period for transmitting of the respective sub-block and this selected minimum value is divided by a third number of the ^inner nodes following nodes. With this development, the first number of symbols is calculated in such a manner are that, firstly, the data throughput of individual links of the network is not overloaded and blocks the other symbols of the data ^¬ not repeatedly transmitted in different sub-blocks.

In a development, a reconstruction of the reconstructed sub-block is started only after receiving at least a fourth number of error-free symbols of the data block, the fourth number of at least one fifth Number of symbols of the information data of the data block corresponds. This achieves on the one hand that the recon ^¬ constructive tion of the reconstructed sub-block is not started before the icons are plentiful. Secondly, the reconstruction is accelerated because it can be performed to that time ^¬ point at which enough, that a fifth number of symbols are present. Thus, this application saves on the use of computing and energy resources in the execution of the method.

The invention further comprises a node, wherein the node is an inner node is a distribution tree, the node following means comprises: a) a first agent wherein said Since ^¬ tenblock redundancy data comprises receiving at least one partially complete sub-block of a data block; b) second means for completely reconstructing the partial block associated with the distribution tree as a reconstructed partial block on the basis of the at least partially received partial block; c) a third means for forwarding the reconstructed sub-block according to the distribution tree to at least one subsequent node.

Thus, the node of the network is able to generate the sub-block to be reconstructed on the basis of the sub-blocks provided with redundancy data. Furthermore, the node can not start generating the reconstructed sub-block until enough symbols have arrived at it, i. the fourth number corresponds to at least the fifth number. For example, a node is represented by a computer in the network.

Finally, the invention also relates to a system comprising a data source and a network with nodes, the data source comprising the following means:

a first data means for generating redundancy data and for inserting the redundancy data into a data block; - A second data means for dividing the data block into at least two sub-blocks;

a third data means for sending the sub-blocks to the nodes of the network on the basis of a distribution tree corresponding to the sub-blocks, and the node comprises the following means:

- at least one partially complete sub-block of the data block, wherein the data ^¬ block comprises a first means for receiving redundancy data; - a second means for reconstructing the complete distribution tree associated with the sub-block reconstructed as sub-block based on the at least one at least partially ^¬ example completely received sub-block;

a third means for forwarding the reconstructed sub-block according to the distribution tree to at least one subsequent node.

The system enables the process to be performed between at least one node and one data source. In this case, the system can be distributed in a distribution service, e.g. be used according to split-stream technology. The data source of the system may calculate the first number of symbols of the subblocks according to the extension described above.

The invention and its developments are explained in detail with reference to the drawing. Show it:

Figure 1 Structure of a network with multiple nodes, in which a data block is divided into three sub-blocks, each sub-block with an associated with him

Distribution tree is forwarded to the nodes;

FIG. 2 shows a flowchart in which defective subblocks in an inner node are completely reconstructed and the reconstructed subblock is forwarded to at least one subsequent node. Elements with the same function and mode of operation are provided in FIGS. 1 and 2 with the same reference numerals.

FIG. 1 shows a network N which comprises at least several nodes N1,..., N9. Such a network is known from [1]. Each node is a computer, wherein the nodes are connected by means of a network, for example according to the LAN standard (Local Area Network). Starting Since ^¬ tenquelle, such as a data server S is a DA tenblock into three sub-blocks Tl, T2, T3 divided from a. The sub ^¬ blocks are sent via a respective assigned distribution tree Bl, B2, B3 to the nodes. The first sub-block Tl is sent via the first distribution tree Bl to the first node Nl, wherein the first node Nl forwards the first sub-block to the second, third and fourth nodes N2, N3, N4. The fourth node N4 then forwards the first sub-block to the fifth, sixth and seventh nodes N5, N6, N7. Finally, the seventh node conveys the first sub-block Tl to the eighth and ninth nodes N8, N9.

The distribution of the first sub-block Tl is carried out on the basis of the first distribution tree Bl, the second sub-block T2 by means of the second distribution tree B2 and the third sub-block T3 by means of the third distribution tree B3. In Figure 1, the first distribution tree Bl with solid arrows, the second distribution tree B2 with dotted arrows and the third distribution tree B3 indicated by dash-dotted arrows. In general, each node belongs to all distribution trees, with one node being assigned as an inner node to a single distribution tree and as a leaf node to the rest of the distribution trees. For example, the fourth node N4 is an inner node of the first distribution tree, whereas the fourth node N4 in the second and third distribution trees B2, B3 corresponds to a leaf node. Furthermore, the arrowhead of the arrows of ^{Verteilungsbäu ¬} me indicates a distribution ^direction of the respective distribution tree associated sub-block. For example, the first one becomes Sub-block Tl passed from the first node Nl to the fourth node N4 within the first distribution tree Bl.

With the aid of FIG. 2, an exemplary embodiment is explained in more detail below, wherein the distribution trees and the network have the same structure as in FIG. According to FIG. 2, the data block comprises information data ID, which comprises, for example, a JPEG-compressed picture (JPEG). In a first means M1, redundant data RD is added to the data block DB. To create the redundancy data, it is in practice advantageous to use an error correction method in the correction of a symbol of the data block a Correction ^¬ rection can be carried out with an arbitrary combination of symbols of the redundancy data and from symbols of the information data. In particular, methods that generate Reed-Solomon codes, fountain codes or adapter codes are suitable for this purpose. Thus, in the Reed-Solomon method, any symbol from any combination of other symbols can be corrected if enough symbols have been received. This will be explained in more detail after description of Figure 2.

In a further method step, the data block, comprising the information and redundancy data, is added to a second data means D2, which forms three sub-blocks T1, T2, T3 from the data block. Furthermore, individual symbols of the data block may be redundant in more than one sub-block. In practice, however, it is appropriate that the part describing ^¬ blocks disjoint sets of the data block. The three sub-blocks are transferred from the data source, taking into account the respective distribution tree, to the nodes of the network.

In the distribution of the sub-blocks in the network, the individual sub-blocks can go spills or due to a faulty wireless transmission UW faulty and / or only partially fully achieve the nodes in routers lost for example due to storage ^¬. That's how it receives fourth node N4, the third sub-block T3 of the sixth node N6 and a relation to the second sub-block T2 partially complete second sub-block T2 'of the fifth node N5. The fourth node does not receive a first sub-block. In the embodiment of Figure 2, the part blocks are ^¬ T2 ', received by a first means Ml T3. The received sub-blocks T2 ', T3 are transferred to a second means M2, which completely reconstructs the sub-block assigned to the distribution tree by the inner node as a reconstructed sub-block TR. In the present Ausführungsbei ^¬ game is the fourth node N4 one of the internal nodes of the first distribution tree Vl, so that the second means to the associated first distribution tree Bl sub-block, that is the first sub-block Tl when the reconstructed sub-block TR = Tl completely reconstructed. The reconstructed sub-block TR is forwarded by means of a third means M3 from the fourth node N4 according to the first distribution tree Bl to at least one subsequent node N7, N8, N9.

One of the node N4, the node N4 is an interior node of a distribution tree Bl, comprises the following means: a) the first means Ml for receiving at least one teilwei ^¬ se full sub-block T2 ', T3 of a data block DB, the data block DB redundancy data RD includes; b) the second means M2 for completely reconstructing the distribution block Bl associated sub-block Tl as a reconstructed sub-block TR based on the at least ei ^¬ nen partially completely received sub-block T2 ', T3; c) the third means M3 for forwarding the reconstructed sub-block TR according to the distribution tree Bl to at least one subsequent node N7.

The second means M2 can additionally be designed such that a reconstruction of the reconstructed sub-block RT is started only after receipt of at least a fourth number A4 of error-free symbols of the data block DB, wherein the fourth number A4 of at least a fifth number A5 of symbols of the information data ID of the data block DB speaks. Is the fifth number A5 = 200, then the recon ^¬ constructive tion of the reconstructed sub-block TR only started after at least A4 = A5 = 200 symbols of the data block, both from the information data as may be derived danzdaten from the redundancy even in the fourth node have been received error-free.

In the following, a procedure for creating the redundancy data RD and the division of the data block into several sub-blocks, for example in a data source S, is shown in more detail. In this case, a data stream R = 320 kbit / s = 40 kbytes / s is to be transmitted to the nodes of the network N. Furthermore, three distribution trees Bl, B2, B3 according to FIG. 1 are used, that is to say k = 3. A data block DB is formed from the symbols which accumulate within a time period T = 5s at the data rates R, that is SBS = R * T = 200 kbytes. In this example, a symbol is made up of 8 bits. To achieve a 100% redundancy, a number of redundancy data symbols of SBR = 200 kbytes is required, then a second number of symbols that A2 of the data block A2 to SBS + = SBR = 400 Kbytes ^¬ he is. The generation of the redundancy data RD by means of an error correction method, for example by means of the Raptor code Ver ^¬ driving, is known to a person skilled in the art, so that it will not be discussed further here.

In the following, a first number Al of symbols per subblock is to be determined. For the respective Verteilungsbäu ^¬ me the available data throughput CNl, CN2, CN3 available:

Bl CNl = 480 kbps = 60 kbps B2 CN2 = 320 kbps = 40 kbps B3 CN3 = 960 kbps = 120 kbps

Each data throughput must be able to transmit in each case a third number A3 of one of the sub-blocks having the first number A1 of symbols, since the third number of subsequent nodes is served by the respective inner node. Thus reduced the maximum bandwidth available per data block per block to a maximum of CN ₁ IAi, with i = 1, ..., k, or per period T to a maximum of CN ₁ -TlA3.

In the present embodiment shown in Figs. 1 and 2, A3 = 3. Generally, the third number A3 is> 1, and it may take a different value for each inner node depending on the structure of the associated distribution tree. Should further be introduced to limit the maximum number of symbols per sub-block, since the jeweili ^¬ gen sub-blocks are to contain disjoint redundancy and information data, then the first number of symbols per sub-block Al to A2 / k limited. Thus, the first number A1 of symbols per subblock results from the following equation:

₁₁ _ (min (Λ2, r-CM)) ₍₁₎

with i = 1, ..., k. According to the above numerical example, the first number Al of symbols per sub-block of one of the associated distribution trees assumes the following values, respectively:

A2 in kbytes CNi * T in kbytes Al in kbytes

Bl 400 300 100

B2 400 200 66

B3 400 600 133

Hereby, the first sub-block Tl comprises 100 kbytes, the second sub-block T2 66 kbytes and the third sub-block T3 133 kbytes, which are transmitted via the respectively associated distribution tree B1, B2, B3.

If only the first and third sub-blocks T1, T3 reach the fifth node N5, a total of 233 symbols are available to the fifth node. Was the error protection method the

Raptor code method is used, can be prepared from 244 emp ^¬ captured symbols both the SBS = 200 symbols of information Onsdaten ID and the SBR = 200 symbols of the redundancy data RD are determined error-free. Furthermore, the non-received first sub-block T 1 can also be completely reconstructed as a reconstructed sub-block TR, which is subsequently sent to the node N 4, N 6, N 8 following the inner account N 5 in accordance with the second distribution tree B 2.

A system SY, comprising the data source S and the network N with nodes N1, ..., N9, can implement and implement the method, the data source having the following means:

- A first data means Dl for generating redundancy data RD and for inserting the redundancy data RD in a data block DB;

- The second data D2 for dividing the data block DB into at least two sub-blocks Tl, T2, T3;

a third data means D3 for sending the sub-blocks T1, T2, T3 to nodes N1, N2, N3 of the network N on the basis of a distribution trees B1, B2, B3 respectively associated with the sub-blocks T1, T2, T3; Furthermore, the node of the system comprises the following means:

- A first means Ml for receiving at least a teilwei ^¬ se complete sub-block T2, T3 of the data block DB, where ^¬ in the data block DB redundancy data RD comprises;

a second means M2 for completely reconstructing the distribution block Bl associated sub-block Tl as a reconstructed sub-block TR based on the at least ei ^¬ nen partially completely received sub-block T2 ', T3;

a third means M3 for forwarding the reconstructed sub-block TR according to the distribution tree Bl to at least one subsequent node N7.

The node, data source, and / or system may be implemented and executed in hardware and / or on software running on a processor. The node can be integrated in a device G, for example a mobile telephone or a computer. The data source can be part of a data server, such as a video or music distribution service. The system can be part of a mobile network, eg according to UMTS standard (UMTS - Universal Mobile Telecommunications System) or GSM standard (GSM - Global System for Mobile Communication), or a data network, such as an IP-based (IP - Internet Protocol) network with at least one DSL, LAN, and / or WLAN -Transmission link (WLAN - Wireless LAN). Furthermore, the nodes and servers using various transmission protocols, such as IP or X.25 protocol can, and various ^¬ Dener transmission techniques, including LAN, WLAN, or ATM (ATM - Asynchronous Transfer Mode), connected to each other to exchange information, such as the sub-blocks be.

The mentioned embodiments can be supplemented by one or more of the following extensions. The method has been explained on the basis of the fourth node N4 and on the basis of the network N according to FIG. In general, each of the nodes that represents at least one inner node of one of the distribution trees may include the first, second and third means Ml, M2, M3. In addition to two subblocks, two, four or more subblocks can be formed from the data block DB. The data block DB can be part of a data stream. The data stream includes, for example, an MPEG4 encoded video stream (MPEG - Motion Picture Expert Group) and / or other multimedia and organizational data, such as a scene description according to MPEG7. The DA is data stream divided into individual data blocks and each Since ^¬ tenblock transmitted separately.

literature

[1] M. Castro, et al. "SplitStream: High-bandwidth content distraction in cooperative environments", rttjL: // j-ro, ect-: r_s. not / irist_b / p3pers / splitεtiear: iptpsOS / paooi. oof

Claims

claims

Method for distributing a data block (DB) divided into sub-blocks (T1, T2, T3) in a network (N) comprising a plurality of nodes (N1, ..., N9), wherein

each sub-block (T1, T2, T3) is in each case distributed to the nodes (N1, ..., N9) on the basis of one of its associated distribution trees (B1, B2, B3),

- the nodes (Nl, ..., N9) are assigned to the distribution trees (Bl, B2, B3) such that each inner node (N4) of one of the distribution trees (Bl) is a leaf node (N4) of the other distribution trees (B2, B3 is), characterized in that a) are added to the data block (DB) in front of a division into sub-blocks (Tl, T2, T3) redundancy data (RD)), b at least one of the sub-blocks (T2 ^1, T3) at least partly ^¬, by the inner node (N4) of the distribution tree

(Bl) is received, c) the sub-block (T1) assigned to the distribution tree (B1) is completely reconstructed by the inner node (N4) as a reconstructed sub-block (TR) on the basis of the at least one partially received sub-block (T2 ¹ , T3), d) the reconstructed sub-block (TR) is forwarded by the inner node (N4) according to the distribution tree (Bl) to at least one subsequent node (N7).

2. Method according to claim 1, wherein a method for the generation of the redundancy data (RD) is provided for the purpose of generating the redundancy data (RD)

Creating an error correction code is used, in which an error correction of a symbol of the data block (DB) based on a freely selectable combination of symbols of the redundancy data (RD) and from symbols of information data (ID) of the data block (DB) is performed.

3. The method according to claim 2, characterized in that the error correction code is formed by a method of creating Reed-Solomon codes, fountain codes or LDPC codes.

4. The method according to any one of the preceding claims, characterized in that a first number (Al) of symbols per sub-block (Tl) is generated by a minimum value, which is determined by the fact that the minimum value of a second number (A2) symbole len of the data block (DB) comprising the redundancy data (RD) and information data (ID), and selected from a data throughput (CNI) times the time (T) for transmitting the second sub-block (Tl) and this selected minimum value by a third number (A3) at the inner nodes (N4) following nodes (N7, N8, N9) is shared.

5. The method according to any one of the preceding claims, characterized in that a reconstruction of the reconstructed sub-block (TR) is started only after receipt of at least a fourth number (A4) on error-free symbols of the data block (DB), wherein the fourth number (A4) at least a fifth number (A5) corresponds to symbols of the information data (ID) of the data block (DB).

6. node (N4), wherein the node (N4) is an internal node of a distribution tree (Bl), in particular for carrying out the method according to one of claims 1 to 5, characterized in that the node comprises the following means: a) a first means (Ml) for receiving at least a part ^¬ as full sub-block (T2 ^1, T3) of a data block (DB), wherein the data block (DB) redundancy data (RD) comprises; b) a second means (M2) for completely reconstructing the partial block (T1) associated with the distribution tree (B1) as a reconstructed partial block (TR) on the basis of the at least partially received partial block (T2 ¹ , T3); c) a third means (M3) for forwarding the reconstructed sub-block (TR) according to the distribution tree (Bl) to at least one subsequent node (N7).

7. System (SY) comprising a data source (S) and a network ^¬ work (N) with nodes (Nl, ..., N9), characterized in that the data source (S) comprises the following means

a first data means (D1) for generating redundancy data (RD) and for inserting the redundancy data (RD) into one

Data block (DB);

- A second data means (D2) for dividing the data block

(DB) into at least two sub-blocks (T1, T2, T3);

a third data means (D3) for sending the sub-blocks (T1, T2, T3) to the nodes (N1, N2, N3) of the network (N) on the basis of a distribution tree belonging to the sub-blocks (T1, T2, T3) ( Bl, B2, B3); the node comprises the following means:

- A first means (Ml) for receiving at least a partially complete sub-block (T2, T3) of the data block

(DB), wherein the data block (DB) comprises redundancy data (RD);

a second means (M2) for completely reconstructing the partial block (T1) associated with the distribution tree (B1) as a reconstructed partial block (TR) on the basis of the at least partially completely received partial block (T2 ¹ , T3);

- A third means (M3) for forwarding the reconstructed sub-block (TR) according to the distribution tree (Bl) to at least one subsequent node (N7).