CN111416830A

CN111416830A - Self-adaptive P2P streaming media data scheduling algorithm

Info

Publication number: CN111416830A
Application number: CN202010232269.8A
Authority: CN
Inventors: 刘晓光; 赵子毅; 张晴晴
Original assignee: Beijing Isurecloud Technology Co ltd
Current assignee: Beijing Isurecloud Technology Co ltd
Priority date: 2020-03-27
Filing date: 2020-03-27
Publication date: 2020-07-14

Abstract

The invention discloses a self-adaptive P2P streaming media data scheduling algorithm, in a streaming media playing system, the streaming media data is divided into a plurality of TF frames with the same length, then the TF frames are placed into a buffer area of a sending source, the length of each frame is TF _ len bytes, and the TF _ sequence _ number is sequentially stored in the buffer area according to the serial number of a transmission frame. The availability of TF frames in the node Buffer is recorded with a Buffer Map. The buffer map BM additionally records the highest sequence number maxseq-minseq +1Bits existing in the current byte, each byte bit correspondingly represents the availability of the TF frame, and if available, it is set to 1, otherwise it is set to 0. Compared with the traditional P2P streaming media scheduling algorithm, the method overcomes the defects of the traditional strategy, and can automatically adjust the scheduling strategy according to the heterogeneity of the network and the dynamic property of the nodes. The experimental result shows that the least-priority scheduling algorithm can effectively shorten the starting delay of the system and improve the video quality of the terminal user.

Description

Self-adaptive P2P streaming media data scheduling algorithm

Technical Field

The invention relates to the technical field of P2P streaming media data, in particular to a self-adaptive P2P streaming media data scheduling algorithm.

Background

In recent years, streaming media applications based on P2P technology have become a research focus. The early P2P scheme does not adapt well to node dynamics and network heterogeneity with a tree structure. PRO (Peer to Peer Receiver-driver Overlay), DONet (Data-driver Overlay Network) and the like improve the stability and robustness of the system through an Overlay Network of a Gossip protocol structure.

In a system based on the DONet, media streams are divided into a series of small data blocks and exchanged among different neighbor nodes, the system performance mainly depends on the construction of an overlay network and a data scheduling algorithm, and a plurality of Gossip-based algorithms have been proposed to construct the overlay network so as to improve the randomness and reliability of the nodes. The invention provides a novel data scheduling algorithm, namely an adaptive P2P streaming media data scheduling algorithm, by analyzing the problems of the existing scheduling algorithm.

There are some related studies on the P2P data scheduling policy. Such as Random strategy (Random), least first (RF) and Round Robin (Round Robin) have been proposed and employed in many streaming media systems.

The random policy is a very simple P2P data scheduling policy according to which, for each data block that is missing, each node randomly selects a node from the partner nodes that hold the data block and then requests the selected partner node. This strategy can be used in a DONet-based system, but its performance is not stable, especially in heterogeneous network environments.

A round robin robust strategy is used in a layered streaming media system. According to this strategy, all requested data blocks are proportionally distributed to one partner. If only one partner holds the data block, then obtaining the data block from the partner; otherwise, obtaining from the partner node with the largest possible bandwidth. This strategy achieves good load balancing.

The least-priority strategy is a heuristic algorithm with quick time response, and the main idea of the strategy comprises the following steps: firstly, requesting data blocks with fewer potential providers; if there are multiple potential providers for a block of data, the partner with the most remaining bandwidth and sufficient time available will be selected.

The existing algorithm has the following problems:

in CoolStreaming/DONET, the scheduling strategy is generalized to a variation of the parallel machine scheduling problem. Therefore, in a streaming media system, an approximately optimal schedule can be achieved. However, it does not perform well in streaming applications where real-time requirements are high. In a streaming media playback system, there are few potential providers of a data block for the following reasons: 1) the data block has a smaller sequence number and it has slid out of the buffer sliding window in some buddies; 2) it is a very new data block, has a very large sequence number, and is not widely spread, so only the data source node and the partner directly connected to the data source will have this data block. For the first case, it is appropriate to request data blocks with fewer potential providers first using a least-priority policy. However, for the second case, where bandwidth is insufficient, requesting fewer provider blocks first may result in blocks with smaller sequence numbers not arriving until the playout deadline. In addition, the least-first policy requires that the precise bandwidth of each partner node be known in advance, which is difficult to measure in a real-time environment.

Disclosure of Invention

The invention provides an adaptive P2P streaming media data scheduling algorithm,

the invention analyzes the problems of the current popular scheduling algorithm by comparing with the traditional P2P streaming media scheduling algorithm, and provides a self-adaptive P2P streaming media data scheduling algorithm-the least-priority scheduling algorithm. The algorithm overcomes the defects of the traditional strategy, and can automatically adjust the scheduling strategy according to the heterogeneity of the network and the dynamic property of the nodes. The experimental result shows that the least-priority scheduling algorithm can effectively shorten the starting delay of the system and improve the video quality of the terminal user.

Drawings

FIG. 1 is a schematic diagram of an algorithm according to an embodiment of the present invention.

FIG. 2 is a graphical illustration of an algorithm and RF strategy according to an embodiment of the present invention.

Detailed Description

In order to facilitate an understanding of the invention, the invention is described in more detail below with reference to the accompanying drawings and specific examples. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.

One embodiment of the present invention is an adaptive P2P streaming media data scheduling algorithm: in a streaming media playback system, streaming media data is divided into a plurality of TF (Transfer frame) frames having the same length, and then placed in a buffer area of a transmission source, each frame having a length of TF _ len bytes, and sequentially stored in the buffer area according to a transmission frame sequence number TF _ sequence _ number (sequence _ number: sequence number). The availability of TF frames in the node Buffer is recorded with a Buffer Map (BM for Buffer Map). The buffer map BM further records the highest sequence number maxseq-minseq +1Bits (where maxseq is the maximum sequence number and minseq is the minimum sequence number) existing in the current byte, and each byte bit correspondingly indicates the availability of the TF frame, and if the availability is set to 1, the byte bit is set to 0 otherwise.

To obtain a minimized start-up delay and a smoother play-out experience, the scheduling algorithm should first fetch the TF frame with the smaller transport frame Sequence number TF _ Sequence _ number. At the same time, the algorithm must be able to adapt to highly dynamic network conditions and be able to achieve load balancing. To meet these requirements, an adaptive data scheduling algorithm (least-first scheduling algorithm) is proposed.

In the least-first scheduling algorithm, it is assumed that TF frames with smaller transport frame sequence number TF _ sequence _ number have higher priority. The specific algorithm specifically comprises the following steps:

step 1: requesting a TF frame if it has a smaller transport frame Sequence number TF Sequence number and only one provider;

step 2: otherwise, it will be inserted into the buffered data buffered _ data set;

and step 3: in each cycle of scheduling, the TF frame in the buffer data buffer _ data is considered first and then the others are considered;

and 4, step 4: if a TF frame corresponds to multiple potential providers, the provider with the minimum request number req _ num is selected;

and 5: if there are still two or more providers with the same request number req _ num, then the node that best matches the current node IP address will be selected.

In the above description, the request number req _ num is a variable defined in advance and is initialized to 0. for each node, when there is another partner node requesting a TF frame from it, the request number req _ num is incremented by 1; when it successfully sends out a TF frame, the request number req _ num is decremented by 1. With this strategy, the scheduling algorithm will be adjusted based on the actual transmission status of the streaming media data, and each node can always get data quickly from the partner that has the ability to provide data according to the least limited scheduling algorithm. In other words, a partner node with sufficient resources (e.g., network bandwidth) can always provide more data to other nodes, and the computational complexity of the algorithm is O (M × N), where M is the maximum value of the transmit frame sequence number TF _ sequence _ number in the sliding window of the buffer and N is the number of partner nodes for this node.

A partner node may not provide data at any time to receive and its access bandwidth may change over time, thus requiring a retry mechanism to re-request TF frames that have been requested but have not been introduced for a long time. However, too many re-requests may cause network congestion and result in more TF frames not being reachable by the playout deadline. To avoid this, we define a maximum request number max _ req of a variable, which refers to the maximum number of TF frames that the node can request from a partner at the same time, and this variable can be adjusted according to the actual size of the TF frames, if the size of the TF frames is too large, the maximum request number max _ req should be reduced to avoid network congestion due to redundant re-requests; otherwise, if the TF frame size is too small, the maximum number of requests max _ req may be increased to fully utilize the network bandwidth.

On the basis of the above, the present invention provides another example of analysis: according to the above description, in each node, the least-priority scheduling algorithm selects an appropriate data provider according to an actual data transmission state, and thus is particularly effective in a heterogeneous network. Fig. 1 is a schematic diagram of an example of data scheduling.

As shown in fig. 1, the intermediate node has three partner nodes, and the numbers on the arrows represent the available bandwidth from the partner node to this node. The number in the rectangle next to each partner node represents the TF frame situation it holds. We assume that: the best match of the IP address of partner 2 with the central node is followed by partner 1 and finally partner 3.

To better illustrate the problem, we set the maximum number of requests max _ req to 2 in this example. According to the least-first scheduling algorithm strategy, the intermediate node simultaneously requests a maximum request number max _ req of TF frames from each partner node, and at the moment, the node can re-request one TF frame from other partners whether the partner successfully sends the previously requested TF frame or not. Partner nodes with more bandwidth left take less time to send a TF frame and therefore will request more TF frames from these partner nodes. During this period, the node has requested 6, 9, 3 TF frames from

partners

1,2,3, respectively, and this ratio is just in line with the available bandwidth, which also indicates that the least-first scheduling algorithm policy is an adaptive scheduling policy.

The performance evaluation of the invention, in order to better analyze the performance of the least-first scheduling algorithm, also realizes the RF scheduling strategy in the streaming media playing system for fair comparison. As mentioned earlier, our goal is to minimize play-out start-up delay, and therefore the time it takes to buffer the same number of TF frames is the primary indicator to evaluate the performance of each algorithm. And, the results of each experiment were averaged over 10 identical experiments. In the experiment, the overlay network has 50 nodes, each node has 4-6 partners. The scheduling algorithm is executed periodically, and we set this circumference to 1s (seconds). The media coding rate used in the experiment was 500Kbps, the size of each TF frame was about 100K, and the value of the maximum request number max _ req was set to 5. As shown in fig. 2, fig. 2 presents a graph of a least-first scheduling algorithm and an RF strategy. As can be seen from fig. 2, in order to buffer 60s of play data, the least priority scheduling algorithm takes only 10s, while RF takes 20s, and in most cases, buffering data for the same play time, the least priority scheduling algorithm takes less time than RF. It follows that the least-first scheduling algorithm is significantly better than the RF algorithm strategy. The RF curve fluctuates at the beginning because it requests some new database. The above experimental results show that the least-first scheduling algorithm can shorten the play start delay and make the player fluent with each bit under limited conditions, especially under the condition that the bandwidth of each partner node is insufficient, but the available bandwidth is enough for playing the full-quality media. The least-first scheduling algorithm can guarantee better quality of service qos (quality of service), it is fully distributed, and it can achieve better performance than the conventional policy without knowing the partner node bandwidth and without any load balancing mechanism.

The technical features mentioned above are combined with each other to form various embodiments which are not listed above, and all of them are regarded as the scope of the present invention described in the specification; also, modifications and variations may be suggested to those skilled in the art in light of the above teachings, and it is intended to cover all such modifications and variations as fall within the true spirit and scope of the invention as defined by the appended claims.

Claims

1. An adaptive P2P stream media data scheduling algorithm, characterized in that, in the stream media playing system, the stream media data is divided into multiple TF frames with the same length, and put into the buffer of the sending source, each frame is TF _ len bytes in length, and stored in the buffer in sequence according to the transmission frame sequence number TF _ sequence _ number; the availability of the TF frame in the node buffer area is recorded by a buffer map BM; the buffer map BM also records the highest sequence number maxseq-minseq +1Bits existing in the current byte, where: maxseq is a maximum sequence number, minseq is a minimum sequence number, each byte bit correspondingly represents the availability of the TF frame, if the availability is set to be 1, otherwise, the byte bit is set to be 0; firstly, obtaining a TF frame with a smaller transmission frame serial number TF _ Sequence _ number; setting the TF frame with the smaller transport frame sequence number TF sequence number to have the higher priority specifically comprises the following steps:

step 1: requesting a TF frame if it is determined that it has a smaller transport frame Sequence number TF _ Sequence _ number and only one provider;

and step 3: in each cycle of scheduling, firstly considering a TF frame in buffer data buffer _ data;

2. The algorithm of claim 1, wherein the request number req _ num is a predefined variable and is initialized to 0, and for each node, the request number req _ num is incremented by 1 when there are other partner nodes requesting a TF frame from it; when it successfully sends out a TF frame, the request number req _ num is decremented by 1.

3. The algorithm of claim 2, characterized in that each node defines a maximum number of requests max _ req for a variable, which refers to the maximum number of TF frames a node requests from a partner at the same time; the variable is adjusted according to the actual size of the TF frame, and if the size of the TF frame is overlarge, the maximum request number max _ req is reduced; otherwise, if the TF frame size is too small, the maximum request number max _ req is increased.