JP2006510255A - System and method for detecting packet loss in a streaming application using a circular buffer - Google Patents

Abstract

In order to manage packet loss detection in Internet streaming, a circular buffer (ie, a buffer chain forming a ring) is provided. The detection latency is determined by the size of the buffer chain. This can be dynamically adapted to network conditions and application requirements. The present invention can achieve sufficient detection accuracy.

Description

  The present invention relates to packet loss detection in Internet streaming. In particular, the present invention relates to a system and method using a circular buffer implemented by a buffer chain forming a circular for packet loss detection in Internet streaming. More specifically, the present invention dynamically adapts the packet loss detection latency determined by the size of the chain to network conditions and application requirements to achieve sufficient detection accuracy and is easy to implement. About.

  In Internet streaming applications, if important packets such as video I-frames or packets belonging to the base layer of scalable coding are lost, the receiver will retransmit the lost packets to the sender. Can be requested. The receiving side needs to have means for detecting packet loss in a timely manner in order to immediately send a retransmission request. Currently, packet loss detection is performed using a timer or a timing window.

  In the transfer control (TCP) protocol, a timer is used for loss detection. When a packet is transmitted, a timer having a certain timeout value is set on the transmission side of the packet. If the sender expires before receiving a packet acknowledgment (acknowledgment), the sender declares that the packet is lost and retransmits the packet.

  In most streaming applications, loss detection is done using a timing window. The timing window (more precisely the table) has a fixed number of binary entries. Each entry represents the state of the packet (0: lost, 1: received). At a particular point in time, the first entry in this window is associated with a packet identified by a sequence number (eg, a real-time transfer protocol (RTP) packet sequence number). Subsequent window entries are associated with higher sequence number packets in sequence order. Thus, the packet sequence number space can be divided into blocks and each block can be considered to be associated with a window at a particular point in time.

  Currently, two such timing windows associated with two consecutive blocks of packets are used for packet loss detection. Initially, all entries are marked “0”. When a packet is received, the corresponding window entry is marked “1”. When a packet with a sequence number beyond that first window is received, the corresponding second window entry is marked. As soon as a packet with a sequence number exceeding the second window is transmitted, the first window is closed and its entry is checked. Packets associated with entries that remain marked “0” are declared lost. Immediately after the second window, successive new windows are opened and the detection process is resumed.

  The timer method is applicable only to protocols such as TCP that can measure the round trip time of a packet to correctly set the timer timeout value. In streaming applications, most of the time, only one-way media streams are generated. In such a case, the timer method cannot be applied. Instead, a timing window method is used. However, the timing window method has the following limitations.

Window size is fixed-there is no built-in mechanism for window size adaptation, which is desirable when network conditions and application requirements (eg, delay) change.
• Loss detection latency is not uniform for different losses- Loss detection latency is in the range of T to 2T, where T is the average time length of the timing window. If the loss is associated with the first entry in the window, its detection latency is 2T, but if the loss is associated with the last entry in the window, it is T.

  Therefore, a loss detection method that is adaptive and enables uniform loss detection latency is required.

  The system and method of the present invention includes:

A circular chain of buffers that can adjust the chain size according to network conditions and application requirements. Thereby, an adaptive loss detection latency can be realized.
A circular chain of buffers with a fixed chain size. Thereby, a uniform detection latency can be realized when the chain size is fixed.

  Shortening the latency can increase the likelihood of recovering lost packets. If the latency is uniform, the recoverability of any loss can be made equal. Therefore, equalizing the detection latency of any loss is considered preferable for most streaming applications.

  The overhead of implementing the circular buffer of the present invention is low and can be lower than the timing window method.

Referring to FIG. 1a, buffers from 1 to m form a circularly linked array of m buffers B _i 10 (i = 1,..., M). The configuration 12 of each buffer is shown in FIG. As shown in FIG. 1a, m is the length of the circular chain that determines the loss detection latency and can be adapted to network conditions and application requirements. P11 is a pointer that circulates in the chain and points to each buffer in turn. Each buffer B _i 10 in the chain includes two fields F ₁ 13 and F ₂ 14. F ₁ 13 stores a pointer to the next buffer. F ₂ 14 stores the sequence number of a packet that may have been lost.

  In streaming applications, it is assumed that packets are sent in order of packet sequence number. In a lossless and ideal situation, the sequence number of the arriving packet is always greater than the sequence number of the previously arriving packet. When the order of arrival of packets is disturbed or when packets are lost, omissions or gaps are seen in the sequence numbers of received packets. If a drop is observed that spans two or more consecutive numbers, it is likely that this drop represents that one or more packets have been lost. On the other hand, disorder of packet arrival order is common in the Internet. This is because each packet may take a different path in the network, and earlier numbered packets may take longer to arrive than later numbered packets. If an application finds that there is a missing packet arrival sequence, it cannot immediately declare a loss. The application must determine if this omission is merely a disruption of arrival order. The ordered buffer method provides a way to determine whether a loss can be declared.

FIG. 2a shows the code according to the C programming language of a preferred embodiment of the algorithm for achieving the method of the invention. FIG. 2b is a flowchart of the algorithm shown in FIG. 2a. Pointer P12 circulates in the chain of buffers B _i 11. This circulation is driven by receiving a packet at step 20. The sequence number of the received packet is checked against the maximum sequence number s of the packet already received at step 21. If the sequence number of the received packet is less than its maximum sequence number, the received packet is an out-of-order packet.

If the received packet is not an out-of-order packet, step 22 is checked for missing in the sequence and if a missing is found, the following steps are performed.
a. If an unreceived packet is contained in the buffer pointed to by P (if P-> F2 is not zero in step 24), then in step 25, the packet with sequence number P-> F2 is declared lost.
b. Next, regardless of whether P points to an unreceived packet, at step 26 the current maximum sequence number is incremented by 1 and stored in the current buffer, and P is pointed to point to the next buffer in the sequence. Updated (ie, P = P-> F1).
c. In step 27, the number of buffers corresponding to missing is reduced by 1, and steps a to c are repeated until the remaining number of buffers becomes zero.

  Thus, all sequence numbers corresponding to missing are stored in the circular buffer chain, each number occupying one buffer.

If no omission is found, the following steps are performed.
d. If at step 28 the buffer pointed to by P contains an unreceived packet (if P-> F2 is not zero), then at step 29, the packet with sequence number P-> F2 is declared lost and the buffer The sequence number stored in is set to zero.
e. Regardless of whether P-> F2 points to an unreceived packet, at step 30, P is updated to point to the next buffer in the sequence.

Once you ’ve done everything related to the packets in order,
f. In step 31, the current maximum sequence number is set to the sequence number of the received packet.

When an out-of-order packet (a packet with a sequence number lower than the maximum sequence number s of currently received packets) arrives,
g. The number of the received packet is compared with the number stored in the circular buffer, and in step 32 the corresponding record in the buffer is erased (ie set to zero).

Accordingly, the detection latency is determined by the size m of the buffer chain. This pointer, when the pointer again the buffer is not empty (i.e., if F ₂ is not zero) only, because the declaration of the loss takes place.

As shown in FIG. 3, the chain size m that determines the detection latency can be adapted. For example, in a preferred embodiment, initially m = 4. If the observed error rate of the declaration is higher than a predetermined threshold, ie
false_rate> TOLERABLE_RATE
The length may be too short, and it may be necessary to lengthen it by inserting a new buffer and adapting m to the corresponding 36. The longer the network path (ie, the more links that are traversed), the larger m needs to be. This is because the longer the network path that a packet traverses, the more likely it is that the arrival order will be disrupted. The greater the value of m, the less likely it will encounter a packet that has been delivered but out of order in the buffer when the pointer P circulates in the buffer chain B _i 11.

ｆａｌｓｅ＿ｒａｔｅ＝１−ｓｕｃｃｅｓｓ＿ｒａｔｅ
成功率が高すぎる場合、すなわち、
ｓｕｃｃｅｓｓ＿ｒａｔｅ＞ＥＸＰＥＣＴ＿ＲＡＴＥ
である場合は、バッファ・チェーンの長さが長すぎる可能性があり、図３に示すように、バッファを削除することによって短くする必要があろう。 success_rate is initially declared as a predetermined EXPECT_RATE, and then adjusted as follows.

false_rate = 1-success_rate
If the success rate is too high, ie
success_rate> EXPECT_RATE
The buffer chain may be too long and would need to be shortened by deleting the buffer as shown in FIG.

  The present invention can be used in the implementation of a multimedia player that plays media in a storage device connected to a network. Alternatively, it can be used with any type of multimedia receiver that needs to retransmit as an error recovery measure (and therefore must perform packet loss detection). Finally, it can also be used in the implementation of a transfer control protocol in which packet loss detection is performed on the receiving side.

  The method and system of the present invention described and illustrated above is for implementing a circular buffer that can make the detection latency adaptive or fixed. It will be apparent to those skilled in the art that various modifications and variations can be made in the method and system of the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention include modifications and variations that are within the scope of the appended claims and their equivalents.

FIG. 3 shows a preferred embodiment of the circular buffer configuration of the present invention. It is a figure which shows the structure of each buffer of the circular buffer shown to FIG. 1a. FIG. 2 shows a preferred embodiment of an algorithm implementing the circular buffer configuration shown in FIGS. 1a and b. 2b is a flowchart of the algorithm shown in FIG. 2a. FIG. 3 shows a preferred embodiment of an algorithm for adapting the configuration of a circular buffer chain based on network characteristics.

Claims (14)

A system for allowing said receiving side of a plurality of streaming packets transmitted from a given sending side to a receiving side via a network to perform adaptive detection of streaming packet loss,
A circular buffer of size m> 1 entries, each entry having at most one sequence number of a potentially lost streaming packet that has not been received;
Using the circular buffer, a sequence number that has not been received and that may have been lost is detected and stored in the circular buffer, and a sequence number of a lost packet is detected in the circular buffer. Removing from the circular buffer, declaring that the packet is lost, and removing from the circular buffer the sequence number of the potentially lost packet received from the given sender. A packet loss detection module that performs
An adaptation module adapted to adapt the system to network conditions;
A system in which a loss detection latency is determined by the size m of the circular buffer, and the loss declaration may be incorrect.   The system according to claim 1, wherein an initial setting value of m is four. A variable s having an initial value of 1 and adapted to store a maximum sequence number of a streaming packet transmitted from the given sender via the network and received by the receiver;
Starting with the entry in the circular buffer that has an initial position that points to a predetermined position in the circular chain and is next to the entry corresponding to the variable s in order, the m entries in the circular buffer are sequentially And a pointer P adapted to circulate,
For a streaming packet received from the given sender, the packet loss detection module checks the sequence number of the received packet against the variable s;
a. In the received packet sequence, if a missing is found starting from the position pointed to by the pointer P, the sequence number of the packet in which each entry in the circular buffer in the missing may have been lost And the corresponding packet is declared lost, the declared declared total declared_losses is incremented by 1, and each sequence number in the missing is in a sequential entry starting from the position pointed to by the pointer P. Stored in ascending order, P is updated to point to the entry in the circular buffer next to the missing, and s is set equal to the sequence number of the received packet (pkt);
b. If there is no missing in the sequence of received packets, the entry pointed to by the pointer P is examined for the sequence number of the packet that may have been lost and the corresponding packet is declared lost. The total declared loss declared_losses is incremented by 1, the entry is cleared, P is updated to point to the next entry in the circular buffer, and s is the same as the sequence number of the received packet To be set to
c. If an out-of-order packet is found, the circular buffer entry is searched to find an entry containing the same sequence number as the received packet sequence number, and if found, the entry is cleared; The system according to claim 1, wherein if it is not found, the error rate false_declared_losses of the declaration is increased by one. The network condition is at least one of a transmission success rate (success_rate) and a declaration error rate (false_rate), and the success rate is initially set to a predetermined expectation rate (EXPECT_RATE),
The adaptation module adjusts the size m of the circular buffer according to the network conditions;
a. If false_rate> TOLERABLE_RATE (TOLERABLE_RATE is a predetermined threshold), m is increased, an entry is added to the circular buffer,
b.

4. The system of claim 3, wherein m is reduced and entries are removed from the circular buffer. The circular buffer is a circular buffer chain B _i (i = 1,..., M) of multiple (m> 1) buffers, so that each of the multiple buffers is in the chain The system of claim 4, wherein the system includes an entry including a pointer to a next buffer and a value storing a sequence number of an unreceived buffer, and the pointer P points to a buffer in the chain. A system for allowing said receiver of a plurality of streaming packets transmitted from a given sender to a receiver via a network to perform adaptive detection of streaming packet loss,
A circular buffer of size m> 1 entries, each entry having at most one sequence number of a potentially lost streaming packet that has not been received;
Using the circular buffer, a sequence number that has not been received and that may have been lost is detected and stored in the circular buffer, and a sequence number of a lost packet is detected in the circular buffer and the sequence number is detected. Removing from the circular buffer, declaring that the packet is lost, and removing from the circular buffer the sequence number of the potentially lost packet received from the given sender A packet loss detection module that performs
Means for adapting the system to network conditions;
A system in which a loss detection latency is determined by the size m of the circular buffer, and the loss declaration may be incorrect. A variable s having an initial value of 1 and adapted to store a maximum sequence number of a streaming packet transmitted from the given sender via the network and received by the receiver;
Starting with the entry in the circular buffer that has an initial position that points to a predetermined position in the circular chain and is next to the entry corresponding to the variable s in order, the m entries in the circular buffer are sequentially And a pointer P adapted to circulate,
For a streaming packet received from the given sender, the packet loss detection module checks the sequence number of the received packet against the variable s;
a. In the received packet sequence, if a missing is found starting from the position pointed to by the pointer P, the sequence number of the packet in which each entry in the circular buffer in the missing may have been lost And the corresponding packet is declared lost, the declared declared total declared_losses is incremented by 1, and each sequence number in the missing is in a sequential entry starting from the position pointed to by the pointer P. Stored in ascending order, P is updated to point to the entry in the circular buffer next to the missing, and s is set equal to the sequence number of the received packet (pkt);
b. If there is no missing in the sequence of received packets, the entry pointed to by the pointer P is examined for the sequence number of the packet that may have been lost and the corresponding packet is declared lost. The total declared loss declared_losses is incremented by 1, the entry is cleared, P is updated to point to the next entry in the circular buffer, and s is the same as the sequence number of the received packet To be set to
c. If an out-of-order packet is found, the circular buffer entry is searched to find an entry containing the same sequence number as the received packet sequence number, and if found, the entry is cleared; The system according to claim 6, wherein if not found, the error rate false_declared_losses of the declaration is increased by one. A method for adaptively detecting streaming packet loss by the receiving side of a plurality of streaming packets transmitted from a given sending side to a receiving side via a network, comprising:
Providing a circular buffer with a size of m> 1 entries, each entry having at most one sequence number of a potentially lost streaming packet that has not been received;
Receiving a streaming packet having a sequence number from the given sender;
Using the circular buffer and the sequence number of the received packet,
a. Detecting and storing in the circular buffer a sequence number of a packet that has not been received and that may have been lost;
b. Detecting the sequence number of the lost packet in the circular buffer and removing it from the circular buffer, declaring that the packet is lost, so that the loss declaration may be erroneous; and ,
c. Performing any one of the steps of: removing from the circular buffer a sequence number of a possibly lost packet corresponding to the sequence number of the received packet;
Adapting the method to network conditions where loss detection latency is determined by the size m of the circular buffer. Providing a variable s having an initial value of 1;
Setting the provided variable s to s with a higher sequence number in the received streaming packet;
Starting with the entry in the circular buffer having an initial position pointing to a predetermined position in the circular buffer and being next to the entry corresponding to the variable s in order, m of the provided circular buffers Providing a pointer P adapted to cycle through the entries of
For a streaming packet received from the given sender, match the sequence number of the received packet with the variable s;
a. In the received sequence of packets, if a loss is found starting from the position pointed to by the pointer P,
a. 1 Check each entry in the circular buffer in the missing for the sequence number of a packet that may have been lost, declare the corresponding packet lost,
a. 2 If a packet is declared lost, increase the declared declared total lost_losses by 1,
a. 3 store each missing sequence number in ascending order in a sequential entry starting from the position pointed to by the pointer P;
a. 4 Update the pointer P to point to the entry in the circular buffer next to the missing,
a. Setting 5 s to be the same as the sequence number of the received packet;
b. If there is no missing in the received packet sequence,
b. 1 Check the entry pointed to by the pointer P for the sequence number of the packet that may have been lost, declare the corresponding packet lost,
b. 2 If a packet is declared lost, increase the declared declared total lost_losses by 1,
b. 3 Clear the entry pointed to by P,
b. 4 P is updated to point to the next entry in the circular buffer;
b. Setting 5 s to be the same as the sequence number of the received packet;
c. If you find out-of-order packets,
c. 1. Search the circular buffer entry to find an entry containing a sequence number that is the same as the sequence number of the received packet;
c. 2 If found, clear the entry,
c. 3. The method according to claim 8, further comprising: performing one of the steps of increasing the error rate fallyly_declared_losses of the declaration by 1 if not found. The network condition is at least one of a transmission success rate (success_rate) and a declaration error rate (false_rate), and the success rate is initially set to a predetermined expectation rate (EXPECT_RATE),
Said adapting step comprises:
d. If false_rate> TOLERABLE_RATE (TOLERABLE_RATE is a predetermined threshold),
d. Increase 1 m by 1
d. 2 performing the step of adding an entry to the circular buffer;
e. If success_rate> EXPECT_RATE,
e. 1m is reduced by 1,
e. 2. The method of claim 9, wherein the size m of the circular buffer is adjusted according to network conditions by performing any one of the steps of removing entries from the circular buffer.   The method according to claim 9, wherein an initial setting value of m is four. A computer program for use with a processor for adapting detection of streaming packet loss by a receiver of a plurality of streaming packets transmitted from a given sender to a receiver over a network, the computer readable A storage medium; and a computer program mechanism embedded in the computer readable storage medium, the computer program mechanism comprising:
A circular buffer of size m> 1 entries, each entry having at most one sequence number of a potentially lost streaming packet that has not been received;
Using the circular buffer, a sequence number which has not been received and which may have been lost is detected and stored in the circular buffer, and a sequence number of a lost packet is detected in the circular buffer and the circular buffer is detected. Includes instructions for removing from the buffer, declaring that the packet is lost, and removing from the circular buffer the sequence number of the possibly lost packet received from the given sender A packet loss detection routine;
An adaptation routine including instructions for adapting the system to network conditions;
A computer program in which the loss detection latency is determined by the size m of the circular buffer and the loss declaration may be erroneous. A variable s having an initial value of 1 and adapted to store a maximum sequence number of a streaming packet transmitted from the given sender via the network and received by the receiver;
Starting with the entry in the circular buffer that has an initial position that points to a predetermined position in the circular chain and is next to the entry corresponding to the variable s in order, the m entries in the circular buffer are sequentially And a pointer P adapted to circulate,
For a streaming packet received from the given sender, the sequence of the received packet is checked against the variable s according to the instruction of the packet loss detection routine,
a. In the received packet sequence, if a missing is found starting from the position pointed to by the pointer P, the sequence number of the packet in which each entry in the circular buffer in the missing may have been lost And the corresponding packet is declared lost, the declared declared total declared_losses is incremented by 1, and each sequence number in the missing is in a sequential entry starting from the position pointed to by the pointer P. Stored in ascending order, P is updated to point to the entry in the circular buffer next to the missing, and s is set equal to the sequence number of the received packet (pkt);
b. If there is no missing in the sequence of received packets, the entry pointed to by the pointer P is examined for the sequence number of the packet that may have been lost and the corresponding packet is declared lost. The total declared loss declared_losses is incremented by 1, the entry is cleared, P is updated to point to the next entry in the circular buffer, and s is the same as the sequence number of the received packet To be set to
c. If an out-of-order packet is found, the circular buffer entry is searched to find an entry containing the same sequence number as the received packet sequence number, and if found, the entry is cleared; 13. The computer program product according to claim 12, wherein if not found, the error rate of the declaration falsely_declared_losses is increased by 1 and the total declared loss of lost_losses is reduced by one. The network condition is at least one of a transmission success rate (success_rate) and a declaration error rate (false_rate), and the success rate is initially set to a predetermined expectation rate (EXPECT_RATE),
The adaptation routine adjusts the size m of the circular buffer according to the network conditions:
a. If false_rate> TOLERABLE_RATE (TOLERABLE_RATE is a predetermined threshold), m is increased and an entry is added to the circular buffer;
b.

The computer program product of claim 13, wherein m is decremented and entries are removed from the circular buffer.

Images