RU2329607C2 - Method and device to control queue buffer - Google Patents

Abstract

FIELD: information technology.

SUBSTANCE: queue buffer control system, where the queue length parameter QL is compared to the threshold value Lth to initiate the overload notification procedure, contains the procedure of making a decision whether or not to report the overload. Also, the Lth threshold automatic adaptation procedure S5 is provided, based on the estimated communication channel capacity (LC) value. The S5 automatic threshold adaptation procedure takes effect in one of at least two adaptation modes, where the first adaptation mode relates to the delay minimisation in the queue and adapts the Lth threshold value on the basis of n·LC, where n≥1, and the second mode relates to use maximisation and adapts the Lth threshold value based on m·LC, where m≥1 and m>n.

EFFECT: actualisation of two different operation modes, one for reduction of the queue delay, the other - for use maximisation.

36 cl, 4 dwg

Description

Description

Technical field

The present invention relates to a method and apparatus for managing a queue buffer, said queue buffer placing data blocks in one or more queues.

State of the art

In communication systems based on data blocks, that is, in which the information to be transmitted is divided into many blocks, and individual blocks are sent over a communication network, it is known to provide queue buffers in communication channels in a network so that data blocks transmitted through such a communication channel can to be buffered. The buffer may be a transmit or output buffer (i.e., a buffer for data blocks to be sent on the outgoing communication channel), or a receive or input buffer (i.e., a buffer for data blocks that are sent on the outgoing communication channel).

Such blocks for transporting data can have many names, such as protocol data blocks, frames, packets, segments, combinations, etc., depending on the specific context, the particular protocol used and some other conventions. In the context of this document, all such data blocks will be referred to collectively as data blocks.

The procedures for placing data blocks in a queue, moving them into a queue, and removing data blocks from a queue are referred to as manipulating a queue or managing a queue.

A variety of queue management principles are known, including those in which a queue length parameter (e.g., absolute queue length or average length) is monitored and compared with a threshold length value to perform an overload notification procedure if the threshold value is exceeded. Such an congestion notification procedure may consist of discarding a data block and / or marking data blocks (for example, setting the congestion notification flag in data blocks). Various examples of such control principles, such as dropping on filling or random early detection (RED), are described in the introductory part of EP-A-1 249 972, so there is no need to repeat them here.

EP-A-1 249 972 proposes a scheme in which a threshold length value to be compared with a queue length parameter is automatically updated based on one or more characteristics of a communication channel through which data blocks from the queue are to be sent. By adapting the threshold to one or more characteristics of the communication channel, a very flexible form of active queue management is obtained that provides improved throughput and reduced latency, especially over communication channels that have time-varying characteristics, such as wireless communication channels. In particular, EP-A-1 249 972 proposes to evaluate a communication channel capacity based on a data rate in a communication channel and adapting a pore value based on an estimated communication channel capacity.

Object of the invention

An object of the present invention is to provide an improved method and apparatus for managing a queue buffer.

SUMMARY OF THE INVENTION

This problem is solved by the method and device, as described in the independent clauses in this application. Preferred embodiments are described in the dependent claims.

In accordance with the present invention, in a system in which managing a queue buffer or administering a queue buffer comprises determining a length parameter value related to the queue length, comparing this value with a threshold, and performing an overload notification procedure if the threshold is exceeded, and which includes an adaptation procedure , said automatic threshold adaptation procedure operates in one of at least the first and second adaptation modes, where the first adaptation mode is associated with minimizing the delay in queue, and adapts the threshold value (Lth) based on n · LC, where LC represents the estimated value of the capacity of the communication channel, and n = 1, and the second adaptation mode is associated with maximizing utilization and adapts the mentioned threshold value (Lth) based on m · LC , where m> 1 and m> n.

According to the present invention, the queue buffer control method and the queue buffer controller are configured such that at least two different modes of operation are possible, one of which is provided to reduce queue delay, and the other is provided to increase utilization. In mode to increase utilization, the threshold value is set higher than in mode to reduce queue delay. The reason for this is that the higher the threshold, the longer the average queue length becomes, but the sender should be able to place more load (for example, increase the transmission window or transmission speed) in the network transmitting the data blocks, since congestion notifications are transmitted less often due to a higher threshold. This improves utilization, since it means that in general there will always be data blocks to send over the communication channel, but it also increases the queue delay. On the other hand, if it is desirable to reduce the queue delay, the threshold at which the congestion notification procedure can be set should be lower, thereby leading to the fact that senders of data blocks reduce their network load more often, while reducing the average queue delay. However, load reduction also poses a high risk of reduced use, since the load can be reduced to such an extent that there are periods in which there are no data blocks for transfer, i.e. in which the communication channel is underutilized. Therefore, the present invention provides a queue management method and system that provides a flexible balancing of the trade-off between delay and use, by providing appropriate modes for each situation.

The installation of the first and second mode can be performed in any desired way. For example, the installation can be carried out manually by the operator, for example, the operator sets the parameter so that the control procedure is used to identify which mode should be used. According to another example, setting a mode may be performed automatically. Preferably, the automatic installation takes into account the number of events indicating the loss of data blocks that occur outside the transmission device in which the controllable buffer is provided. In particular, in a preferred embodiment, the automatic mode setting procedure sets the threshold adaptation to the second mode (higher threshold) if the number of data loss indication events exceeds a predetermined value. Exceeding the preset value indicates that data block losses are occurring, and these data block losses will cause the senders of the data blocks to reduce their load, so that additional execution of congestion notifications (for example, skipping a data block and / or marking a data block) could would lead to unnecessary further load reduction, which would ultimately lead to underutilization of the communication channel.

In other words, the described embodiment assumes that the flow control performed on the sender and / or receiver side is such that the sender will reduce its load on the network transmitting data blocks (i.e., send fewer data blocks) if it is determined that Lost data blocks. The purpose of the load notifications issued from the managed queue buffer is, therefore, to indicate to the sender the need to reduce the load on the network. As a result of this, this embodiment has the advantage of allowing the regulation of congestion notifications depending on potential data loss outside the data block transmission device in which the buffer is provided. This avoids congestion notifications (omissions of data blocks and / or data block markings) in cases where loss of data blocks would occur in such a way that congestion notifications could lead to the sender reducing its load to a greater extent than this is necessary, which, in turn, could lead to underutilization of the communication channel, in which the queue buffer is provided. Underutilization means a state in which the communication channel is down, that is, it is not transmitting any data blocks. It is highly desirable to exclude the underutilization of the communication channel, since an idle communication channel means a waste of resources.

According to another preferred embodiment of the invention, the queue buffer control method and the queue buffer controller are configured such that events that indicate potential data loss in a stream queued in the queue buffer are also taken into account for dynamically adapting the length threshold value to preset adaptation mode. In other words, after the preset mode is set (for example, manually or automatically, and automatic installation can be performed based on the measured number of loss indication events), which means that the threshold value is adapted in a basic way based on the ratios n · LC or m · LC where LC represents the capacity of the communication channel and m> n, dynamic subsequent tuning may be performed depending on the measured number of loss indication events.

It should be noted that loss indication events can be any event that has the ability to indicate potential data loss. For example, such an event may be a skip of a data block from a sequence of data blocks in the queue (which means potential data loss upstream of the queue buffer) or loss indication information contained in the acknowledgment data blocks sent from the recipient of the queued stream , to the sender of the stream queued. Such loss indication information may be expressed in explicit form, for example, an explicit notification from the receiver of the stream to the sender of the stream that a particular data block of the sequence is missing, or in an implicit form, for example, sending repeated acknowledgments for the last data block received correctly in the sequence .

The ability to automatically adapt the threshold for implementation in the first mode or second mode means that the specified queue in the buffer is controlled in accordance with one of the modes. It is possible that the buffer contains many queues, each of which is associated with one of the available modes (of course, there can be more than two available modes of operation, for example, the third mode, in which the length threshold is adapted based on q · LC, where q> m). In other words, each queue has its own corresponding threshold length value for comparison with the corresponding measured length parameter, and each length threshold is individually adapted. In such a situation, an embodiment is preferred in which the incoming data blocks to be queued are divided into categories associated with the modes, and then placed in a queue managed in accordance with the specified mode. For example, a buffer controller can parse a data block for specific information, such as a protocol identifier or port identifier in the header, and assign delay sensitive data blocks (for example, data blocks transporting segments from a Telnet application) to a queue controlled by in the mode to reduce the delay, and assign data blocks that are sensitive to bandwidth (for example, data blocks transporting segments from the ftp application), in turn, trolled mode to maximize utilization.

Brief Description of the Drawings

Further aspects and details of the present invention are explained in the following detailed description and preferred embodiments with reference to the drawings, in which the following is presented:

FIG. 1 is a schematic diagram of a block diagram of a buffer and a buffer controller in accordance with the invention;

FIG. 2 is a flowchart of an embodiment of a method in accordance with the invention;

FIG. 3 is a flowchart of another embodiment of a method in accordance with the invention;

FIG. 4 is a flowchart of a basic embodiment of a method in accordance with the invention.

Detailed Description of Embodiments

Although some of the embodiments described below may refer to specific protocols, such as TCP / IP, the present invention can be applied to any system transporting data blocks using a queue management scheme where congestion notification is performed depending on events when the queue length parameter reaches the threshold length value. The present invention is not limited to any particular length threshold and congestion notification scheme, and therefore, for example, applies to any known RED scheme, to schemes that discard data blocks when the queue is full, for example, dropping the end segment, random dropping or dropping the front segment, and to any known scheme that performs overload notification explicitly instead of or in addition to any dropping of data blocks.

It is preferable to apply the method and apparatus of the present invention in connection with the active queue management disclosed in EP-A-1249972. All disclosure of this document and its corresponding US application are incorporated herein by reference.

In FIG. 1 is a schematic representation of a queue buffer controller 10 providing an implementation of the present invention for managing data blocks in a queue buffer 20. Reference numeral 3 denotes a communication network through which the data blocks 30 enter the queue buffer 20 for placing in the queue 21 before being transmitted via the communication channel 40. Reference numeral 50 relates to a data block transmission device, in which a buffer 20 and a controller 10 are provided. The data block transmission device may be, for example, a router or a server connected to a communication network 3.

The data blocks 30 queued 21 may belong to one or more streams. A stream is generally identified by a source and destination address, a service access point identifier (SAP), and a protocol identifier. The definition and concept of a stream is well known in the art, for example, from TCP / IP, in which case the source and destination addresses are called IP addresses, and the SAP identifier is the port address, so no further explanation is required here.

The following describes specific elements for implementing the principle of the present invention in the controller 10 of the queue buffer. It should be noted that the queue buffer controller generally contains more elements than these elements shown, in particular, well-known elements for processing received data blocks and buffer control, which are not described in detail for simplicity. In particular, the controller 10 may have additional elements to implement the system described in EP-A-1 249 972.

Reference numeral 101 relates to a queue length determination unit for determining a length parameter value associated with a queue length 21. In addition, a comparison unit 102 is provided for comparing a specific length value with a length threshold value Lth provided by the threshold adaptation unit 104, which is configured to the ability to automatically adapt threshold values by estimating the LC value of the communication channel capacity based on the data rate DR of the communication channel 40. The comparison unit 102 is connected to the congestion notification unit 103, which performs the congestion notification procedure if the determined length value is greater than the threshold length value. As explained in more detail below, the length parameter to be determined may be selected in any suitable or desirable way, for example, be the absolute length of the QL queue or the average length of the QLav queue, and the congestion notification procedure may equally be selected as suitable or desirable, for example, as a procedure for discarding data blocks and / or a procedure for marking data blocks in explicit form.

The congestion notification unit 103 preferably comprises a decision unit 1031 for deciding whether or not to perform congestion notification with respect to one or more data blocks in queue 21. In other words, the congestion notification unit 103 is configured such that optionally performs congestion notification if the queue length parameter exceeds a threshold length value. This is, in principle, known from the prior art, for example from the RED system or other systems described in EP-A-1 249 972, where the length threshold value is a first or lower threshold, and a second or higher threshold is also provided, wherein if the queue length parameter exceeds the first threshold but does not exceed the second threshold, then a probabilistic decision is made to perform congestion notification with respect to one or more data blocks.

In accordance with the present invention, the threshold adaptation unit 104 operates in one of at least two adaptation modes. The first adaptation mode is associated with minimizing the queue delay and adapts the threshold value Lth of the length based on the relation n · LC, where n ≥ 1. For example, n may be 1, and the first adaptation mode sets Lth = LC or Lth = LC + Δ1, where Δ1 is a positive coefficient smaller than LC, for example 0 <Δ1 ≤ LC / 10. The second adaptation mode is associated with maximizing utilization and adapts the threshold value of length Lth based on the relation m · LC, where m> 1 and m> n. For example, m can be equal to 3, and the second adaptation mode sets Lth = 3 · LC - Δ2, where Δ2 is a positive coefficient smaller than LC, for example 0 <Δ2 ≤ LC / 10. If a function of dynamic adaptation is provided with respect to external events (such as loss indication events) after setting the mode, then the above settings are initial and then can be changed in accordance with the appearance of such events. If the dynamic adaptation function is not provided, then the value of Lth can be stored as it was established, but it should be noted that the value of LC is generally not static, that is, the LC will change in time, which will lead to the possibility of changing the corresponding value of Lth.

The coefficients m and n can be arbitrary positive numbers, but are preferably positive natural numbers.

The installation of the first mode or the second mode can be carried out in any suitable or desirable way. For example, this can be done manually by the operator, using the appropriate mechanism, as shown by switch 106 in FIG. 1, which can be installed in the first mode M1 or in the second mode M2. Element 106 may be a valid switch, but is preferably a software element by which a user can set a threshold adaptation unit to a desired mode.

The value of the capacity of the communication channel can be understood as the minimum amount of data that the sender of the stream in question must send, so that the bandwidth that the communication channel 40 allocates to this stream is fully used. Full use means that the fraction of the bandwidth allocated to this stream is always fully utilized. In other words, if we consider a simplified example of a communication channel 40, which serves only a single stream, this means that the sender of this stream sends such a volume of data that the communication channel 40 is constantly busy, that is, it constantly transmits data blocks, without any waiting time between them. In other words, the capacity of the communication channel is determined by the amount of data that the sender sends, so that any additional data sent will not lead to an increase in throughput, since additional data is queued.

Therefore, the capacity of the communication channel can also be understood as the product of the data rate provided by the communication channel 40 for the stream under consideration by the round-trip time (RTT) associated with the stream for the case of an unloaded network. An unloaded network means that there is no queue delay. As a result, the RTT value for an unloaded network is equal to the difference between the actual RTT of the stream and all queue delays for said stream. The capacity of the communication channel is also sometimes defined as the capacity of the trunk for a hypothetical trunk between the end points of the flow, and this trunk has a "width" DR and a "length" equal to the RTT value for an unloaded network.

Therefore, the estimation of the communication channel capacity value will generally consist in determining the time value indicating the RTT value of the unloaded network and multiplying this value by the data transfer rate DR provided by the communication channel 40 for the stream in question. The determination of this RTT value of an unloaded network may be carried out in any suitable or desirable way. One example of this is the calculation of the sum of the constant value RTTwc and the value of RTT provided by the communication channel 40. The RTTwc value is the worst case estimate of the full RTT value of an idle network, excluding the contribution made by the communication channel 40, and may have a value of 150 to 300 ms, more preferably 150 to 250 ms. Using this principle has the advantage that it is not necessary to obtain flow specific information.

An alternative way to estimate the RTT value of an unloaded network for a given stream is to calculate the queue delay in buffer 20, for example, by storing the average time that the buffered data block 30 is in queue 21, and calculating the difference between the actual RTT of the stream and this queue delay . The actual RTT value for the stream may, for example, be entered into the data blocks of the stream by the sender and read by the controller 10.

It should be noted that the estimate of the LC capacity of the communication channel, in General, will not be identical to the current instantaneous capacity of the communication channel. The process of estimating the capacity of the communication channel is preferably such that the estimated value exceeds the actual capacity of the communication channel, that is, the estimate is conservative. This can be implemented in any suitable or desirable way, for example, using the aforementioned worst case estimates for an RTT unloaded network and / or by adding predefined positive coefficients to one or more parameters used in estimating the capacity of the communication channel. In other words, you can add a predefined coefficient to the RTT value of an unloaded network and / or to a RTT value of a communication channel and / or a DR value, and you can add a predefined positive coefficient ε to the calculation result, that is, replace the calculated LC value with LC + ε: LC ← LC + ε.

The selection of parameters n and m for the first and second mode is preferably carried out in accordance with the reaction that the sender of the data blocks will exhibit upon receipt of the congestion notification. If the sender is of a type that reduces its network load in accordance with the coefficient k, for example, divides its transmission window by k, if window-based flow control is used, then n is preferably chosen to be equal to k-1, and m is preferred selected as k ² -1. The reason for this is explained below.

As noted above, the estimated LC link capacity is such that as soon as the sent data blocks become larger than this value, a queue occurs. As a result, if the threshold value Lth of the length is reached, then the amount of data sent is equal to LC + Lth. As a result, if congestion notifications are triggered, then the window-based sender will have a transmission window equal to LC + Lth.

As a condition for setting Lth to a mode that serves to minimize queue delay, it is desirable to obtain the correct balance between keeping the average queue length small and keeping the communication channel busy, that is, it is desirable to set the queue length as small as possible without causing a state of underuse. This leads to the fact that after the notification of congestion is established, that is, when the transmission window of size Lth + LC is divided by k, the resulting window size will be equal to LC, since this means that the load is sufficient to keep the communication channel busy:

,

,

which leads to Lth = LC · (k-1).

As a condition for setting the Lth value to the mode, which serves to maximize utilization, it is taken into account that in addition to the characteristics of congestion notifications to reduce the load that the sender places on the network, data loss indication events occur that also lead to the sender reducing its load. Then the following assumption of the worst case is made: after the first reduction of the transmission window by a factor of k, a data loss event occurs (which is not an intentional discarding of data in the context of an overload notification), causing a further decrease corresponding to the coefficient k. In such a scenario, which assumes data loss outside the transmission device in which the managed buffer is located, this leads to the condition:

,

,

which, in turn, leads to Lth = LC · (k ² -1).

In the case of the TCP / IP protocol, k = 2, that is, the transmitter reduces the transmission window when receiving an overload notification or determining an event indicating the loss of a data block. Because of this, if a managed buffer is used to queue TCP segments, then n is preferably 1, and m is preferably 3.

In accordance with a preferred embodiment of the present invention, a loss indication event detector 105 is further provided. The detector 105 is configured to detect an event outside said data block transmission device 50, said event indicating a potential loss of a data block in a stream placed in queue 21.

In the example shown in FIG. 1, a loss indication event detector 105 is connected to both a threshold adaptation unit 104 and a decision block 1031, so that both of them can take into account detected loss indication events or signals output from them. However, it is also within the scope of the invention that only the threshold adaptation unit 104 takes into account the results of the loss indication event detection unit 105, or when only the decision block 1031 takes into account the results of the loss indication event detection unit 105.

Other features of the procedures carried out by the above elements are explained below in connection with examples of the method according to the invention, which may be embodied in the buffer controller of FIG. one.

It should be noted that the above elements 101-105 may be provided as hardware, software, or any suitable combination of hardware and software. Preferably, the controller 10 is a programmable data processor, and elements 101-105 are software elements, that is, pieces of software code.

In FIG. 4 is a flowchart of a basic embodiment of a method according to the present invention, which method may be performed using the controller 10 shown in FIG. 1. At the first step S1, the value of the length parameter associated with the queue length is determined 21. This parameter associated with the queue length can be associated with the queue length in any desired or suitable way, for example, be a real or instant value QL of the queue length or parameter derived from the actual or instantaneous value of the queue length, such as the average QLav value.

In the example of FIG. 4, the parameter associated with the queue length is the actual QL of the queue length. If it is desired to use the average value QLav of the queue length, this average value can be determined in accordance with any known suitable averaging algorithm, and such an algorithm may typically consist in updating the old value by calculating the sum of the old average value multiplied by the first weight coefficient, and instantaneous value of the length of the queue multiplied by the second weight coefficient. For example, the Qlav value can be calculated as follows:

Qlav (new) = Qlav (old) x (1-1 / 2 ^wf ) + (QL x 1/2 ^wf ),

where QL represents the instantaneous value of the length of the queue and wf is the exponential weight coefficient, adjustable between 0 and 1.

According to FIG. 4, in step S2, the queue length parameter QL is compared with the queue length threshold value Lth. If the threshold value Lth of the queue length is exceeded, the congestion notification procedure S3 is executed, otherwise the congestion notification procedure S3 is skipped.

As noted above, the congestion notification procedure may be selected to be carried out in any suitable or desirable manner. For example, it may include discarding / marking one or more predefined or randomly selected data blocks from queue 21, or discarding / marking one or more newly arriving data blocks before placing them in queue 21.

The congestion notification procedure may also include a decision-making procedure on whether or not to perform congestion notification with respect to one or more data blocks from queue 21. As noted above, such a decision-making procedure may, for example, depend on a probability function, such as this is known in the prior art.

In the example shown in FIG. 4, the control procedure then continues to step S5, in which the automatic threshold adaptation procedure is performed. As defined above, the automatic threshold adaptation procedure S5 is performed in one of at least the first and second adaptation modes, the first adaptation mode being associated with minimizing the queue delay and adapting the threshold value Lth based on the relation n · LC, where n = 1, and the second adaptation mode is associated with maximizing utilization and adapting the threshold value Lth based on the relation m · LC, where m> 1 and m> n.

The mode in which step S5 is implemented can be manually set by the operator or can be set automatically by a subroutine designed to perform this task. An example of such a routine is described below.

In FIG. 2 shows an embodiment of a method that comprises all the steps of the embodiment shown in FIG. 4, so that a repeated description thereof is not required, with the procedure for detecting the loss indication event S4 being added after steps S2, S3. This loss indication event detection procedure S4 can detect events that indicate a potential loss of a data block in one or more streams placed in a queue 21 where this potential loss of a data block occurs outside of the data block transmitting device 15. For example, the loss indication event detection procedure S4 may comprise monitoring the identifiers of the data blocks of the queued stream, where skipping the data block from the sequence indicates a potential loss of the data block. Alternatively or additionally, the procedure for detecting a loss indication event may comprise monitoring the loss indication information in acknowledgment data blocks sent from the receiver of the stream to the sender of the stream, this loss indication information may be presented in explicit or implicit form.

In accordance with the example of FIG. 2, the automatic threshold adaptation procedure S5 takes into account the results provided by the loss indication event detection procedure S4 in order to automatically set the first or second adaptation mode. For example, the loss indication event detection procedure S4 may output a reference value indicating the number of such loss indication events (which will be explained in more detail below), and if this reference value exceeds a predetermined value, the adaptation procedure is set to work in the second mode (optimization of use) otherwise, it is set to the first mode (delay optimization).

In addition to using the result of the loss indication event detection procedure to automatically set the adaptation mode, it is additionally or alternatively possible to use the result of the loss indication event detection procedure for dynamically adapting the threshold value Lth in the first or second mode. In other words, after this mode is set (for example, manually or automatically, and automatic installation can be performed based on the measured number of loss indication events), which means that the threshold value is, in principle, adapted based on the ratio nxLC or mxLC, further dynamic adjustment may be performed depending on the measured number of loss indication events.

For example, if the loss indication event detection procedure indicates a number of potential data block losses outside the transmission device 50, then the threshold Lth used to start the congestion notification procedure may be increased relative to the value that was set in the predetermined mode. For example, it can be increased relative to the estimated LC value of the communication channel capacity, which could be the initial choice for the Lth value in the first mode. The increase will be only a fraction of the LC, for example, up to one tenth of the LC, as some type of “fine tuning” in a given mode.

Increasing the threshold value Lth makes it less likely that the queue length parameter QL will reach the threshold value Lth in step S2, so it is less likely that congestion notification will be performed. Therefore, it is possible to avoid the fact that congestion notifications will be executed, although the sender of the stream has already reduced the load on the network due to the loss of data blocks that occurred regardless of the congestion notification procedure. Losses of data blocks outside the transmission unit 50 together with congestion notifications from the transmission unit 50 can cause the sender of the stream to reduce its network load too much, which, in turn, could lead to the underutilization of communication channel 40.

Attention is drawn to the fact that the particular order of steps shown in FIG. 2 and 4 are for example purposes only. In particular, steps S1, S2, and S3, which together form the decision procedure for starting the congestion notification procedure, are independent of the adaptation procedure implemented by step S5. Therefore, steps S1-S3 can be ordered independently of step S5. Similarly, step S4 may be implemented independently of steps S1-S3 and S5, for example, as a procedure that is executed in parallel with others and / or is selectively called as a subroutine.

With respect to the examples shown in FIG. 2 and 4, it should be noted that all the steps shown will, in principle, be contained in a larger-scale way of managing the queue buffer 20, and this larger-scale method has more steps and procedures, but these additional steps and procedures are not shown, since they do not concern of the present invention. The methods of FIG. 2 and 4, like all embodiments of the methods according to the invention, can be implemented as software, where steps S1-S3 can be implemented, for example, in one program chain, while step S5 can be implemented in another independent program chain . Step S4 may be implemented in yet another independent program chain.

In the example shown in FIG. 2, the adaptation procedure for Lth in step S5 may take into account the result of the procedure for detecting the loss indication event S4 in any desired or suitable manner. In principle, the threshold adaptation procedure should be implemented in such a way that the threshold value Lth of the length increases with the occurrence of potential data loss outside the data block transmission device 50 to make the start of the congestion notification procedure S3 less likely, as explained above.

According to a preferred embodiment of the invention, the method of detecting a loss event is performed in such a way that a counting procedure is provided for counting the number of events indicating loss of data blocks occurring outside the transmission device 50 in the queued stream under consideration, and a procedure is provided for obtaining a characteristic count value from the counted numbers. The automatic threshold adaptation procedure then comprises the step of setting the adaptation mode and / or fine-tuning the threshold value Lth depending on this characteristic reference value.

The process of obtaining the characteristic value of the reference can be selected by any suitable or desirable method. For example, it may contain determining the number of events indicating losses incurred outside the data block transmission device 50 in the queued stream in each of p corresponding predefined time intervals, where p is a natural number, and then selecting the maximum of these numbers as characteristic reference values. If p is 1, then this simply means a count of the number of loss indication events within a predetermined interval and the output of a count value for each consecutive interval. In order to make the characteristic value of the reference less susceptible to short-term fluctuations, it is preferable to choose p greater than 1, for example p = 5. In this case, the procedure holds the counted number of loss indication events in the five most recent intervals and derives the maximum from the loss indication events of the most recent five intervals as the characteristic value of the reference.

The predetermined interval or intervals during which the number of loss indication events is counted can be selected in any suitable or desirable manner, for example, as having a predetermined fixed length. Advantageously, they are determined dynamically, as the time between two consecutive decisions to execute congestion notification for one or more data units by means of the decision procedure contained in the congestion notification procedure. In other words, when the decision procedure in the congestion notification procedure decides to perform congestion notification, a new interval is triggered, and therefore a new count of loss indication events begins. This count of loss indication events continues until the next congestion notification.

Another possibility of obtaining the characteristic value of the reference is to determine the average number of events indicating the losses that occur outside the transmission device 50. Such averaging can be carried out by controlling the number of events indicating losses in successive predefined intervals, as in the example described above, and then averaging the numbers thus calculated over a number of intervals. Such an averaging operation is preferably performed as current averaging, for example:

AV (new) = AV (old) x (1-q) + Num x q,

where AV represents the average value, q is the weighting factor, defined as 0 <q <1, and Num represents the number of loss indication events in the most recent completed interval.

As already mentioned, in the general case, the threshold length adaptation procedure S5 will increase the threshold length value Lth if the characteristic reference value indicates an increase in loss indication events.

Regarding the count of loss indication events, it should be noted that the selection of what is counted as a loss indication event can be made in any suitable or desirable manner. For example, if the procedure for detecting a loss indication event includes monitoring sequence identifiers in the stream under consideration, then each detection of a missing data block in the sequence (“skip” in the sequence) can be counted as a loss indication event. This process can also be modified in that the recognition of such a gap is not directly counted as a loss indication event, but the procedure first waits for a predetermined wait period, and the loss indication event is counted only if the missing data block does not appear during the wait period . The introduction of such a waiting period makes the procedure more resistant to reordering of data blocks in the transmission channel in network 3.

As another example, if the procedure for detecting a loss indication event comprises monitoring loss indication information in acknowledgment data blocks, then, for example, the occurrence of loss indication information may be counted as a loss indication event. For example, in the case of TCP / IP, a loss indication event may be the occurrence of duplicate acknowledgment. As such, each occurrence of a duplicate acknowledgment may be counted as a loss indication event. However, when monitoring the acknowledgment data blocks for loss indication information, it is preferable that the buffer controller operate in the same manner as the sender of the data blocks with regard to identifying data block loss events. In other words, the buffer controller must apply the same criteria to determine the data block loss event as the sender of the data block for the purpose of flow control. For example, in the case of the TCP / IP protocol, the sender of the data block will determine that an event has occurred when the data block was lost after receiving a predetermined number (for example, 3) of duplicate acknowledgments. In this case, it is preferable that the buffer controller also counts duplicate acknowledgments and determines that a data block loss event has occurred after counting a predetermined number (e.g. 3) duplicate acknowledgments.

If the congestion notification performed by the buffer controller corresponds to gaps in the data blocks, then the controller should preferably be able to register these gaps in conjunction with the sequence number of the skipped data block in order to be able to identify acknowledgment messages that refer to the specified skipped data block , and do not count such loss indication information as an event that indicates data loss outside the transmission device 50 chi.

Next, another embodiment of the present invention is described with reference to FIG. 3. Steps S1 and S2 are identical to those already described with reference to FIG. 2 and 4, so there is no need to repeat this description. However, in the example of FIG. 3, the congestion notification procedure S3 consists of steps S31, S32 and S33, which are different from the example shown in FIG. 2 or 4. Step S31 represents a decision procedure S32 on whether to perform congestion notification with respect to one or more data blocks. For example, this could be the above example of checking a probability function. If the result of this decision procedure is affirmative, then step S31 is performed, that is, a procedure for determining whether an event has occurred indicating a potential loss of a data block outside the data transmission device 50 for the stream in question, and if such an event has occurred, is the decision to notify overload in step S31 is canceled. In other words, in this case, congestion notification is not performed.

In the specific example of FIG. 3, step S32 is to determine whether a loss indicating event has occurred within the predetermined guard time interval GT after it is detected that the length parameter QL has exceeded the threshold length value Lth.

In other words, after detecting that the QL parameter has exceeded the value of Lth, it is monitored whether the LIE event has occurred indicating loss during GT time or not. If a loss indication event occurs, then the decision about the congestion notification made in step S31 is canceled or blocked, that is, it is not executed, since it is assumed that the sender of the stream will reduce its network load based on the specified loss of data blocks outside the transmission device 50, so that the implementation of the congestion notification could lead to an excessive reduction in the data blocks sent by the sender of the stream, thereby possibly leading to the underutilization of the communication channel 40.

On the other hand, if no loss indication events are detected in the procedure in step S32 during the guard time interval GT, then an overload notification decided in step S31 is performed in step S33.

In the above example of FIG. 3, a decision procedure, which consists of steps S31, S32, and S33, takes into account the result of detecting a loss indication event. In the example of FIG. 2, it was the automatic threshold adaptation procedure S5 that took into account the result of detecting a loss indication event. It should be noted that these two possibilities of using the result of the detection of a loss indication event can also be combined, i.e. both the automatic threshold adaptation procedure and the decision procedure take into account the result of detecting a loss indication event.

The following describes another embodiment. The ability to automatically adapt the threshold threshold S5 or the ability of the threshold adaptation unit 1094 to operate in the first mode or in the second mode means that the specified queue in the managed buffer is controlled in accordance with one of the modes. It is possible that the buffer supports many queues, each of which is associated with one or more of the available modes. Of course, there can be more than two modes, for example, the third mode, in which the length threshold is adapted based on sxLC, where s> m. In other words, each queue has its own corresponding threshold length value for comparison with the corresponding measured length parameter, and each length threshold is individually adapted. In such a situation, an embodiment is proposed in which incoming data blocks to be queued are selected by categories associated with the modes, and then placed in a queue managed in accordance with the selected mode. For example, a buffer controller can parse a data block for specific information, such as a protocol identifier or an identifier in the header, and assign delay sensitive data blocks (for example, data blocks transporting segments from a Telnet application) to a queue managed in mode to reduce latency, and assign data units sensitive to bandwidth (for example, data blocks transporting segments from an ftp application) to a queue managed in mode to maximize olzovaniya.

Although the present invention has been described with specific examples, they should not be interpreted as limiting, since the scope of the invention is defined only by the claims. The reference numbers indicated in the claims serve only to simplify the perception of the claims and should not have any restrictive effect.

Claims (36)

1. A method for controlling a queue buffer (20) in a data block transmission device (50), said queue buffer being used to place data blocks (30) in a queue (21) and connected to a communication channel (40), comprising determining (S1) the value (QL; QLav) of the length parameter associated with the queue length, comparing (S2) this value (QL; QLav) with a threshold value (Lth) and performing (S3) the congestion notification procedure for one or more data blocks from said queue if said value (QL; QLav) is greater than said threshold value (Lth) length, and an automatic threshold adaptation procedure (S5) that comprises estimating a communication channel capacity (LC) value based on a communication channel data rate (40) of a communication channel (40) and adapting a threshold value (Lth) based on an estimated communication channel capacity (LC), characterized in that the automatic threshold adaptation procedure (S5) operates in one of at least the first and second adaptation modes, where the first adaptation mode is associated with minimizing the delay in the queue and adapts the threshold value (Lth) based on n · LC, where LC represents the estimated value the capacitance of the communication channel, and n≥1, and the second adaptation mode is associated with maximizing utilization and adapts the said threshold value (Lth) based on m · LC, where m> 1 and m> n. 2. The method according to claim 1, in which the queue buffer is configured to receive data blocks from the sender, which performs window-based flow control and divides its transmission window by k, k> 1, upon receipt of an overload notification or upon detection of a block loss data, with n = k-1 and m = k ² -1. 3. The method according to claim 2, in which k = 2, n = 1 and m = 3. 4. The method according to any one of claims 1 to 3, in which the setting of the first mode or second mode is performed manually by the operator. 5. The method according to any one of claims 1 to 3, comprising a procedure for automatically setting a mode for automatically setting a first mode or a second mode. 6. The method according to claim 5, comprising a procedure for detecting a loss indication event for detecting potential loss of data blocks outside said data block transmission device (50) in a stream queued in said stream buffer. 7. The method of claim 6, wherein the procedure for detecting a loss indication event comprises monitoring sequence identifiers of data blocks in said queued stream. 8. The method of claim 6, wherein the procedure for detecting a loss indication event comprises monitoring loss indication information in acknowledgment data blocks sent from the receiver of the queued stream to the sender of said queued stream. 9. The method of claim 6, wherein the procedure for detecting a loss indication event comprises a counting procedure for counting the number of events indicating loss of data blocks occurring outside said data block transmission device in a queued stream, and the procedure for obtaining the characteristic value of the reference from the counted numbers. 10. The method according to claim 9, in which the aforementioned procedure for obtaining the characteristic value of the reference determines the number of events indicating losses incurred outside the device for transmitting data blocks in a queued stream in each of p corresponding predetermined intervals, and p is a natural number, and selects the maximum value from the mentioned numbers as the characteristic value of the reference. 11. The method of claim 10, wherein said predetermined intervals are defined as the time between two successive decisions about performing congestion notification for a data block in said queued stream. 12. The method according to claim 9, in which the aforementioned procedure for obtaining the characteristic value of the reference determines the average number of events indicating the loss occurring outside the said device for transmitting data blocks in the queued stream, as the said characteristic value of the reference. 13. The method according to any one of claims 6-12, wherein said automatic mode setting procedure takes into account the result of said loss indication event detection procedure. 14. The method of claim 13, wherein said automatic threshold adaptation procedure takes into account the result of said loss indication event detection procedure for dynamically adapting said threshold value (Lth) in said first or second adaptation mode. 15. The method according to any one of claims 6-12, 14, wherein said congestion notification procedure comprises a decision step for deciding whether to perform the congestion notification procedure with respect to one or more data blocks, the decision making step depends on the outcome of the procedure for detecting a loss indication event. 16. The method of claim 15, comprising monitoring whether an event has occurred that indicates a potential loss of a data block in the queued stream during a predetermined time interval (GT) after it is detected that said value of said parameter (QL, QLav) of the length related to the length of said queue is larger than the threshold value (Lth) of the length, and at the said decision stage, congestion notification is not performed if an event indicating a potential loss of the data block oshlo during said predetermined period of time (GT), and otherwise, said congestion notification is made. 17. The method according to claim 1 or 16, wherein said queue buffer is configured to store at least a first queue and a second queue, said automatic threshold adaptation procedure adapting the first threshold value (Lth_1) associated with the first queue to according to said first adaptation mode, and adapts the second threshold value (Lth_2) associated with the second queue in accordance with said second adaptation mode. 18. The method of claim 17, comprising a discrimination and placement procedure for distinguishing data blocks to be queued based on their contents and for placing data blocks in said first or second stage, depending on the result of the discrimination. 19. The controller (10) of the queue buffer for controlling the buffer (20) of the queue in the data block transmission device (50), said queue buffer (20) being used to place the data blocks (30) in the queue (21) and connected to the channel (40 ) communication containing a queue length determining unit (101) for determining a parameter value (QL; QLav) of a length associated with the length of said queue (21), a comparison unit (102) for comparing said value with a threshold value (Lth), an overload notification unit (103) for performing an overload notification procedure if said value is greater than said length threshold value, and a threshold adaptation unit (104) for automatically adapting said length threshold value (Lth) by estimating a communication channel capacitance value (LC) based on a data rate (DR) of said communication channel (40) and adapting said length threshold value (Lth) based estimated value (LC) of the capacity of the communication channel, characterized in that said threshold adaptation unit (104) operates in one of at least the first and second adaptation mode, where the first adaptation mode is associated with minimizing queue delay and adapts the threshold value (Lth) based on n · LC, where LC represents the estimated value communication channel capacities and n≥1, and the second adaptation mode is associated with maximizing utilization and adapts the said threshold value (Lth) based on m · LC, where m> 1 and m> n. 20. The queue buffer controller according to claim 19, wherein the queue buffer is adapted to receive data blocks from the sender, which performs window-based flow control and divides its transmission window by k, k> 1, upon receipt of an overload notification or upon detection loss of a data block, wherein said threshold adaptation block (104) is configured such that n = k − 1 and m = k ² −1. 21. The queue buffer controller according to claim 20, in which k = 2, n = 1 and m = 3. 22. The queue buffer controller according to any one of claims 19-21, comprising an installation mechanism for manually setting the first mode or second mode by the operator. 23. The queue buffer controller according to any one of claims 19-21, comprising a mechanism for automatically setting a mode for automatically setting a first mode or a second mode. 24. The queue buffer controller according to claim 23, comprising a loss indication event detection unit (105) for detecting potential loss of data blocks outside said data block transmitting device (50) in a stream queued in said stream buffer. 25. The queue buffer controller according to claim 24, wherein said loss indication event detection unit (105) comprises a control unit for monitoring sequence identifiers of data blocks in said queued stream. 26. The queue buffer controller according to claim 24, wherein said loss indication event detection unit (105) comprises a control unit for monitoring loss indication information in acknowledgment data blocks sent from the receiver of the queued stream to the sender of said queued stream . 27. The queue buffer controller according to claim 24, comprising a counter for counting the number of events indicating loss of data blocks occurring outside said data block transmission device in said queued stream, and a counted number processor to obtain a characteristic count value from the counted numbers. 28. The queue buffer controller according to claim 27, wherein said counted number processor is configured to determine the number of loss indication events occurring outside said data block transmission device in a queued stream in each of p of corresponding predetermined intervals, wherein p is natural number, and selecting the maximum value from the mentioned numbers as the characteristic value of the reference. 29. The queue buffer controller according to claim 28, wherein said predetermined intervals are defined as the time between two successive decisions about performing congestion notification for a data block in said queued stream. 30. The queue buffer controller according to claim 27, wherein said counted number processor is configured to determine an average number of loss indication events occurring outside said data block transmission device in a queued stream as said reference characteristic value. 31. The queue buffer controller according to any one of paragraphs.24-30, wherein said automatic mode setting mechanism is configured to take into account the output of said loss indication event detection unit. 32. The queue buffer controller according to claim 31, wherein said threshold adaptation unit (104) is adapted to take into account the output of a loss indication event detection procedure for dynamically adapting said threshold value (Lth) in said first or second adaptation mode. 33. The queue buffer controller according to any one of claims 24-30, 32, wherein said congestion notification unit (103) comprises a decision unit (1031) for deciding whether to perform the congestion notification procedure with respect to one or more data blocks, and the decision block (1031) is configured to take into account the output of the block (105) detection events of the indication of loss. 34. The queue buffer controller according to claim 33, wherein said loss indication event detection unit (105) is configured to monitor whether an event has occurred that indicates a potential loss of a data block in the queued stream over a predetermined time interval ( GT) after it has been found that said value of said length parameter (QL, QLav) related to the length of said queue is larger than said length threshold value (Lth), said decision block (1031) being configured to Stu not perform congestion notification if an event indicating a potential data unit loss occurred during said predetermined period of time (GT), and to otherwise perform said congestion notification. 35. The queue buffer controller according to claim 19 or 34, wherein said queue buffer is configured to store at least a first queue and a second queue, said threshold adaptation unit being adapted to adapt a first threshold value (Lth_1) associated with a first queue, in accordance with said first adaptation mode, and an adaptation of a second threshold value (Lth_2) associated with the second queue, in accordance with said second adaptation mode. 36. The queue buffer controller according to claim 35, comprising a discriminating unit and premises for distinguishing data blocks to be queued based on their contents and for placing data blocks in said first or second stage depending on the result of the distinguishing.

Images