Multimedia content delivery method and system

[0001] The present invention relates to the field of information technology and telecommunications.

[0002] The present invention particularly relates to a method and system for distributing multimedia content combining "push" download and "pull" download using a mechanism called "peer to peer".

Known solutions [0003] It is already known from the scientific publication "On the feasibility of Commercial, Legal P2P Content Distribution" (P. Rodriguez, SM. Tan and C. Gkantsidis, ACM SIGCOMM Computer Communication Review, vol. 36, No. 1, January 2006), the concept of distributing multimedia content via a peer to peer network for commercial purposes. Nevertheless, the cited scientific publication does not reveal the "pull" download stage. [0004] Among the hitherto solutions is also known, from the European patent application EP 1 341 328 (Matsushita Electric Industrial Co.), a data emission system for broadcasting multimedia data and a receiving end device for receiving said multimedia data. The invention described in this European patent application relates to a data receiving device for receiving and transmitting emitted data comprising a series of multimedia data, which device comprises:

- a receiving unit adapted to receive in a sequential manner a series of multimedia data contained in the emitted data;

- a transmitting team adapted to broadcast received data;

- a storage unit adapted to create and store management information in order to collectively manage a series of received multimedia data and assign information corresponding to appropriate multimedia data; and

- a management team adapted to manage the received multimedia data in relation to management information.

[0005] Among the current solutions is also known, from US Patent Application No. US 2004/003090 (Deeds Douglas), a peer to peer information sharing system. This US patent application describes a system and method for sharing data over a network, such as the Internet. The shared data is often audio data, but may correspond to other types of data. Server e.g. An Internet messenger server (serveur de messagerie instantanee) with a network connection receives queries to initiate a data sharing session, which queries contain the addresses of potential contributors. After receiving the request, the messenger server determines whether potential partners are actually connected to the network, and if so, transmits their network address to the unit that sent the request so that it can start transmitting data to each of them. The data is usually played back from a recording made by one of the participants, but the selection is expanded by group members during its transmission, creating a shared listening environment. To avoid potential copyright infringement, the data is restored but not stored for later listening at the receiving terminal level.

[0006] Among the existing solutions, it is also known from the PCT patent application WO 01/10125 (Randall Chung), a video on demand system (VoD or Video on De2 mand), from one server or user to another user. This PCT patent application describes the system and method of data distribution to many end users using a storage system connected via a data transmission network. The end user's equipment for playing audio or video content or for recording includes a storage unit, such as a magnetic disk. Each storage unit can store data, such as digital video content (movies) or digital audio content (songs). The storage unit can also be configured to store parts or fragments of digital video or audio content. The central unit administers, manages and controls the place of storage, playback and recording of digital video or audio content. Data for playing video or audio content, or parts of that data, may reside on one or more physically different storage units, or may reside in a central storage unit.

Presentation of the invention [0007] The present invention aims to eliminate the drawbacks of known solutions by proposing a coding solution particularly adapted to the method of distributing multimedia content which combines the modes "push", "pull" and "peer to peer".

[0008] To this end, the present invention relates, in its most general concept, to a method of distributing multimedia content characterized in that it comprises:

• the first stage consisting of partial download of multimedia content from the content server to the first device of the client in "push" mode, and • the second stage consisting of downloading the missing elements of the mentioned multimedia content in "pull" mode using the "peer to peer" mechanism from the client's second device, while the multimedia content is possible to play with n + p blocks, where nip are natural integers different from zero, wherein n blocks are downloaded during said first stage, while p blocks are downloaded during the second stage, whereas multimedia content can only be used after receiving n + p blocks.

[0009] According to a first variant, the first and second client devices are located under the same DSLAM (Digital Subscriber Line Access Multiplexer) as part of an implementation based on this transmission mode.

[0010] According to a second variant, the client devices are located under adjacent DSLAMs.

[0011] The client devices may be decoders with a hard disk or personal computers.

[0012] According to a specific implementation method, some blocks are systematically introduced into many customer devices, e.g. blocks enabling the beginning of the movie or movie to be displayed in inferior quality so that the user can start watching while waiting for download via the "peer to peer" mechanism.

[0013] The invention also relates to a multimedia content distribution system comprising at least one multimedia content server, communication network, DSLAM and at least two client devices.

[0014] The present invention also relates to a digital device characterized in that it has:

• means for receiving, in a "push" mode, multimedia content in a partial manner from the content server; and • means for receiving by downloading the missing elements of said multimedia content in a "pull" mode via a "peer to peer" mechanism from a second client device;

whereas the multimedia content is possible to play with n + p blocks, where nip are integer non-zero integers, where n blocks are taken in "push" mode, and p blocks are downloaded in "pull" mode using the "peer" mechanism to peer ", but multimedia content can only be used after receiving n + p blocks.

Brief Description of the Drawing [0015] The invention will become better understood in the light of the following description of a method for carrying out the invention, prepared merely as an explanation, with reference to the only figure showing the architecture using the method and system of the invention.

Detailed description of how to implement the invention [0016] First, multimedia content is encoded in N blocks, where N is an integer greater than or equal to 2. Content can be decoded with n + p blocks.

[0017] Multimedia content is e.g. a video document and the system of figure 1 is located on television on demand ("VOD" - "Video on Demand").

[0018] Next, the content is downloaded in "push" mode from the content server, shown in the only figure, via the telecommunications network towards the client devices EC1, EC2, EC3 etc. Each client device receives n blocks. The client device is a hard disk decoder or personal computer and is connected to the telecommunications network via a DSL modem and a DSLAM device. Each client device remembers these n blocks for later use (e.g. on a hard disk). The decoder or computer has the means known as such for decoding and decompressing the content, if this is the case, and if the content is video content, for generating video and audio signals in an appropriate format.

[0019] These n blocks transmitted to ("pushed to") a given client device are defined by the content server. Preferably, these n blocks transmitted to the client device differ from n blocks transmitted to the other client device by means of at least one block. The content server maintains a database indicating which blocks are stored at the level of each client device. This database allows the client device to determine from which peer client device (or from which peer devices) should be asked op missing blocks. Preferably, this database is stored on a device other than the content server but accessible to client devices. [0020] According to a particular variant, the multimedia content is encoded in such a way that any combination of n + p blocks allows the content to be decoded. In this case, the client device will look for p missing blocks on any client device having these complementary blocks.

[0021] In our example, there are three client devices, preferably located under the same DSLAM, n is 3, p is 1 and N is 9. Blocks are coded with some inter-block redundancy. A small selection of p (and especially equal to 1) allows you to minimize the amount of information to be requested in peer to peer mode and thus speed up access to content by the user. This query can take place when the user wants to view the content or when he decides to get the rights to the content.

[0022] Each of the three client devices thus receives three blocks in the "push" mode. To use multimedia content, each client device needs an additional block.

[0023] The client device will thus retrieve the missing block in "pull" mode from another client device using the "peer to peer" mechanism. This device will therefore ask for missing blocks. In this way, the client device will be able to decode content and use it.

[0024] The peer to peer mode avoids the need to download missing blocks from the initial server. This mode also allows you to optimize network resources and avoid overloading the content server.

[0025] These p missing blocks can be distributed on many devices under DSLAM.

[0026] The search request is sent from the client device EC1 to the devices appearing on the list where the devices containing the remaining content are located. This list can be sent earlier by the control server in response to a request from the EC1 client device.

[0027] For example, EC1 sends an inquiry to EC3, EC5 and ECn. EC3 sends one-third of the missing content to EC1, EC5 sends the second of the three missing content to EC1 and ECn sends the last-third to EC1. One can also imagine a mode in which the shares transmitted by EC3, EC5 and ECn to EC1 are not equal, e.g. 55%, 30% and 15%.

[0028] A message can be sent to the control server when some of the content is found valid by the EC1 device.

[0029] It is also conceivable that a priority is set to provide the missing blocks. For example, in SVC (Scalable Video Coding) mode, the base layer will have the highest priority, while the layer called "enhancement layer" will have the lower priority. The content could only be played with some missing blocks, but in lower quality.

[0030] This priority mechanism may be of particular interest for applications having strong time constraints. For example, suppose the EC1 client device can tolerate the absence of the first and second of the three missing content, but needs the last third. In this case, EC1 sends priority information to the content devices that will allow the time limit to be respected.

[0031] A variant of the method consists in dividing the missing content, with which there are strong time constraints, into many parts of equal size or not. These parts will then be transmitted to the client device that requested it.

[0032] The "push" step can be implemented via a digital broadcast network in accordance with one of the DVB standards, eg DVB-T or DVB-H.

[0033] Unicast and multicast mechanisms can be used in the context of this invention. [0034] In a specific embodiment, the following procedure can be followed:

i) The content server sends the entire content only once using the multicast technique.

ii) The control server sends special information to each client device to indicate which part of the content they should remember and which part they can delete.

The advantage of such a mechanism is low bandwidth utilization. The main disadvantage is that client devices must be connected during the push phase. For devices that might not be available during this phase in push mode, you can broadcast content server content to these client devices in unicast transmission.

[0035] During the pull phase, multicast can also be used. In fact, the multicast connection can be used in such a way that the EC1 device can use the EC2 device that already provides the content to the EC3 device.

[0036] The main difference from the unicast transmission approach is that once the control server receives a content query from the client device, the control server initiates two actions:

i) Sending a response to the device that made the request ordering it to "listen" at one specific or many specific multicast addresses / ports.

ii) Sending a control request to one or more devices containing / containing the desired content so that the device or devices start / start broadcasting the content in a multicast connection whose address / port is included in the request. The server can also specify whether the content is broadcast in multicast in carousel mode or only once.

[0037] In the event that the control server detects that the content has already been broadcast (or is being broadcast) by one or more devices, then the control server does not need to send a control request to the broadcast devices. DSLAM, thanks to IGMP mechanisms, will ensure the delivery of the requested content to the device that made such a request.

[0038] In addition, since the content requesting device can "listen" to multiple multicast connections, a second mechanism is needed so that the content requesting device can send a message to the control server when the device finishes downloading the content. At this point, because the control server knows how many devices are listening to the requested content, it may or may not send a control request to the emitting device to order it to stop broadcasting. Of course, this request will only be sent if no device is listening to the multicast connection.

[0039] One variant is to cut shared content into many parts. The larger the parts, the more likely many devices listen to the same parts. The smaller the parts, the greater will be the number of broadcasting devices participating in the same content, which will likely lead to faster download times and shorter broadcasting time by the broadcasting devices.

[0040] In another variant, redundancy coding is introduced so that a larger number of receiving devices can listen to the same multicast connections. For example, to decode content, you only need to receive part of the encoded content. This coding can be systematic or not.

[0041] In another mode of implementation of the invention, the control server may send, in response to the query, a list of all devices that broadcast the requested content. The receiving device chooses how many streams you want to listen to. The receiving device can optimally use the available frequency response. Receiving devices send a special message to the server to indicate which stream they are listening to. This allows the server to determine how many devices are listening to a given emitter.

[0042] Various strategies can be used for the first stage in "push" mode:

a) Multiplexed versus Dedicated Push

b) Full Copies versus Partial Copies Push

c) Prefix Push

d) Global Earliest Deadline First Policy

Multiplexed versus Dedicated Push [0043] The first two strategies for placing content relate to the number of client devices allocated to a given content, e.g. a given movie.

[0044] Assigning a customer to a movie means that parts of the movie or movie are all / remembered / remembered at the client level and that this content can be transmitted to other clients.

[0045] The first strategy, called the multiplex strategy, consists in allocating a large number of clients to a given film. The second strategy, called the dedicated strategy, involves assigning a small number of clients to a given movie. The extreme case of a dedicated strategy is the transmission of content to only one client, while the extreme case of a multiplex strategy is the transmission of content to all clients.

[0046] In a multiplex strategy, the number of potential servers increases, which implies, assuming a wired channel, that the frequency bandwidth allocated to a given content increases linearly with the number of clients allocated to the content.

[0047] The effects of choosing a multiplex versus dedicated strategy are as follows:

- Increasing the number of customers assigned to a given movie increases the rate called supported arrival rate for that movie. This increase also increases the startup latency, but to a lesser extent.

- In the extreme case where the content is transmitted to all clients (full multiplex), it should also be taken into account that each client already has some content and thus needs only to download the remaining part. Thus, in this case the network traffic is reduced.

[0048] It is therefore recommended to increase the multiplexing rate for popular films and to use more dedicated strategies for less popular films.

Full Copies versus Partial Copies Push [0049] The following content placement strategies relate to the amount of data stored at the level of a single customer assigned to a given movie. This is the amount of data that the client device will provide as a server to other clients requesting this content. The device (client-) server can store the film in whole or in part.

[0050] These strategies have the following effects:

- With the same number of assigned customers, increasing the amount of content stored at each client level reduces the parameter called startup latency (start delay) of the movie. However, it should be noted that increasing the number of customers assigned has a greater impact on the parameter called startup latency than increasing the stored quantity.

- With the same number of assigned customers, increasing the amount of content stored at the level of each customer has a very small impact on the indicator called supported arrival rate.

- Reducing the amount of content stored at the level of each customer increases the so-called mark-up overhead generated for this particular movie. More specifically, the client must contact more clients to download content, which increases signaling complexity and increases the number of TCP connections at each client level.

Push prefix:

[0051] This strategy involves push broadcasts of the first seconds or minutes of the content to all clients. This data is not used by the assigned client to transmit as a server to other clients requesting content. This significantly reduces the parameter called startup-latency. This strategy should only be used for very popular movies. In fact, broadcasting a push prefix to all clients when the movie is watched only by a very small number of clients can be considered a waste of storage capacity.

[0052] Depending on the popularity of the film, different push strategies should be used to optimize the use of resources (bandwidth, storage, customers, etc.).

[0053] It is recommended that the following steps be followed:

a) First, determine the number of clients assigned to each movie: choose between multiplex and dedicated push strategy. Multiplexing should be increased for popular films so that high arrival rates are supported for these films. At the same time, to minimize overhead overhead and storage capacity consumption, multiplexing should be reduced for less popular movies.

b) Next, specify the amount of stored content allocated to each movie at the level of each client: it is recommended to increase the stored amount for popular films in order to reduce the total overhead generated by the system. In addition, it is recommended to reduce the amount of content stored for less popular movies, since the impact on the overall overhead will be small and the storage space will be better used by popular movies.

c) Finally, using the rest of the storage space available for the push prefix strategy allows you to adjust the startup latency of each movie. It is recommended to apply this rule only to the most popular movies.

Global Earliest Deadline First Policy [0054] The latter strategy is used in conjunction with the previous ones.

[0055] "First deadline first" or "earliest deadline first" (EDF) is a policy in which a request for the nearest deadline must be handled before all other queries. This is trivial for only one content server that must process all queries because the content server knows the queries and their deadlines. However, in a distributed environment, as in the present invention, the implementation of the EDF (global earliest deadline first policy), in which the query for the earliest deadline should be handled before all other queries in the server device queues, not is a trivial problem. Queries can be spread across many server devices and the system must identify and handle the one that came first.

[0056] Assuming the presence of a central control server that has an index of all content transmitted during the push mode step and its location that receives client queries, a policy called "earliest deadline first" can be implemented at the control server level. This policy will be global because all customer queries will be considered.

[0057] Now suppose that each client device has a list of all content and its locations. To distribute resource consumption, the customer randomly selects a location from which to download the content he wants to use. Assuming the presence of a central control server, which, however, does not have an index of all content and its location, a policy called "global earliest deadline first" (global EDF) can be implemented as follows: each server device must inform the control server of its waiting queue. The central server schedules all queries in a policy function called "earliest deadline first" (EDF) and sends the list to clients.

[0058] In the above description, the invention is given by way of example only. It should be understood that one skilled in the art is able to implement various variants of the invention without departing from the scope of the patent. The example of implementation is especially in the framework of transmission over the telephone network and via the Asymmetric Digital Subscriber Line (ADSL). Other networks and transmission modes may be considered.