KR20110106732A - A buffer space securing method and apparatus for real time data transmission according to data increase/decrease rate - Google Patents

Abstract

The present invention relates to a method for managing a buffer, and more particularly, to a multi-phase adaptive corresponding buffer management method for ensuring real-time delivery of important data in consideration of the state of a buffer according to traffic inflow.

Description

A buffer space securing method and apparatus for real time data transmission according to data increase / decrease rate}

The present invention relates to a buffer management method.

Conventional packet drop methods, such as random early detection (RED), weighted random early detection (W-RED), and tail-drop, lose important data when the buffer overflows. Allowed to be. That is, the conventional buffer management technique performs congestion control by allowing a certain amount of data drop when congestion occurs.

The reason why the loss of important data is allowed in the conventional buffer management method is that when the terminal transmitting the data transmits and receives important data, TCP (transport control) guarantees reliable in-order data delivery. This is to reduce the amount of data transmitted and received by allowing loss because it uses reliable communication method such as protocol.

However, in the case of wireless network environment using multi hopping such as AD HOC network which can be configured by mobile device only without sensor network or infrastructure, packet transmission success rate is significantly lower than that of wired network environment. Reducing the amount of data alone is difficult to ensure real-time data.

In addition, even in a stable network such as a wired network communication network, when network traffic is congested, important transmission data may occur due to congestion, and even if a retransmission protocol is operated, delay time is additionally generated.

That is, the conventional method of discarding data in the buffer in advance, such as random early detection (RED) and weighted random early detection (W-RED), may be performed after the average queue length in the buffer reaches a specific threshold. If the cumulative data is rapidly increased due to network congestion, the packet is discarded. Therefore, the loss of important data due to buffer overflow is likely to occur, and the purpose of packet discarding is for TCP congestion control. Must be partially allocated.

Therefore, in order to guarantee the stable delivery of important data in real time in the network environment as described above, important data to be delivered in real time from each routing node and non-critical data that are not in real time are preempted by buffer management technique. We need a buffer management technique that can be serviced by

The present invention is to solve the above problems, an object of the present invention is to predict the overflow situation according to the degree of change in the queue (queue) in the case of overflow of the buffer due to congestion of the communication network It is to provide a method of securing buffer space to prevent the loss of important data and to ensure data delivery in real time.

In order to achieve the above object, the present invention provides a method and apparatus for determining the possibility of buffer overflow, minimizing the loss of important data through preemptive disposal of non-critical data during network congestion.

The method includes estimating an overflow condition of a buffer according to queue length information of the buffer, and according to a pre-emptive low priority packet drop (PLPPD) algorithm based on the predicted overflow condition of the buffer. Dividing the operation method into a plurality of phases to operate the buffer differently, separating the important data to be processed in the overflow situation from the non-critical data to be discarded, and operating the buffer according to the plurality of aspects. do.

The apparatus includes a receiver for introducing data to be processed, a scheduler that manages the received data by predicting a buffer overflow according to queue length information of a buffer, and a PLPPD based on information received from the scheduler. -emptive low priority) includes an adaptive buffer for processing the data according to an algorithm.

The present invention predicts the overflow situation according to the change of the queue length, and prevents the loss of important data through the buffer space in case of overflow of the buffer due to congestion of the communication network, It is effective to transmit important data in real time.

1 illustrates a pseudo code for a pre-emptive low priority packet drop (PLPPD) algorithm according to the present invention.
2 illustrates state transition diagrams for respective phases based on buffer states according to an embodiment of the present invention.
3 illustrates an example of applying a state transition algorithm to a single network queue structure according to an embodiment of the present invention.
4 is a structural diagram of a data processing system according to an embodiment of the present invention.

Hereinafter, with reference to the accompanying drawings, preferred embodiments of the present invention will be described in detail. In the procedures of each drawing, the reference numerals are added to the respective procedures or the reference numerals are added to the components in the drawings. The same procedures and components are given the same reference numerals even though they are shown in different drawings. Be careful.

1 illustrates a pseudo code for a pre-emptive low priority packet drop (PLPPD) algorithm according to the present invention.

According to Figure 1, at the beginning of the algorithm, the scheduler calculates an "avg value", which is the average length of the buffer queue as individual packets are processed. By calculating the avg, it is possible to check the state of the current buffer and determine the free space of the buffer. The threshold for the average length value (avg value) of the buffer queue per unit time is defined as "Maxth" and "Minth". The Maxth value is the maximum length that the buffer queue can allow, and the Minth value. Is the minimum buffer queue length that the scheduler determines that there is enough free space in the buffer queue.

In addition, apart from the avg value, to calculate the rate of increase or decrease of the space in the buffer queue according to the data accumulated per unit time, the value? Avg is calculated. The reason for measuring the Δavg value is to prevent the possibility of buffer overflow by allowing the scheduler to grasp the amount of change in the remaining amount of the current buffer queue. Threshold value for the change value of the buffer queue length per unit time (Δavg value) is defined as "Pvarth" and "Nvarth", and the Pvarth value indicates that the change amount of the current buffer queue is increasing. The value of Nvarth is a value for judging that the amount of change in the current buffer queue is decreasing.

The algorithm shown in FIG. 1 compares the measured " avg value " and " Δavg value " with each of the thresholds described above to manage the buffer as shown in FIG. 2 below.

2 illustrates state transition diagrams for respective phases based on buffer states according to an embodiment of the present invention.

According to FIG. 2, the above-mentioned "avg value" and "Δavg value" are compared with the "Maxth" and "Minth" values, and the "Pvarth" and "Nvarth" values, respectively, and thus the operation method of the buffer is determined accordingly. Done.

In each step transition condition shown in FIG. 2, when the avg value condition and the? Avg value condition are satisfied at the same time, the transition occurs in preference to the result of the avg value. This is because the avg value is a result of the characteristic of? Avg, and the characteristic of the? Avg value is such that the amount of change over time is larger than that of the avg value.

"Initial state" (S110) is a state in which variables used in the algorithm are initialized before the inflow of data starts.

Once each variable is initialized, by default, the buffer space reserve algorithm transitions to the first phase (S120) state. The first phase (phase 1) is a state before a network congestion state, and as in a general case, a scheduler can normally process all data introduced.

The phase 2-1 phase 2-1 (S130) refers to a state in which a change amount (Δavg value) of the buffer queue length is present between the Pvarth value and the Nvarth value. That is, when it is determined that the inflow of data gradually increases and the risk of buffer overflow gradually increases, the transition from the first phase (S120) to the second phase (Phase 2-1) (S130) is performed. Becomes When the transition to the 2-1 phase (S130) state occurs, the scheduler starts to discard the non-critical data having the lowest rank in the buffer for each packet. Through this, it is possible to secure the buffer space for the important data flowing into the buffer in the future.

The process of distinguishing non-critical data from important data among the imported data may be performed in real time from the moment the data enters the buffer, or may be performed when the risk of buffer overflow is found.

Phase 2-2 phase S-2 (S140) is a change amount of the buffer queue length, although discarding of the lowest data in the buffer is performed in phase 2-1 phase 2-1. (Δavg value) is continuously increasing. That is, in spite of discarding the lowest data in the buffer in phase 2-1 (S130), if the amount of change (Δavg value) in the buffer queue length continues to increase, the second-2-2; Transitioning to phase 2-2 (S140), the scheduler incrementally increases the number of packets to discard from low priority data.

Phase 2-3 (S150) is a case where the change amount (Δavg value) of the buffer queue length falls below the Nvarth value. When transitioning to phase 2-3 (S150), the scheduler determines that network congestion has been eliminated due to the discarding of the non-critical data, so that the scheduler temporarily stops discarding the non-critical data. As a result, when the average length value (avg value) of the buffer queue becomes less than Minth, the transition to the first phase S1 is performed.

Phase 3 (S160) is a case where the overflow of the buffer has already occurred before any action is taken by the scheduler due to the traffic congestion. In this case, packets are discarded from the lowest order by quality of service (QoS) marking level instead of each individual packet unit.

Which state is transferred to is determined based on the measured change amount of the buffer queue length (Δavg value) and the average length value of the buffer queue (avg value). Therefore, when the change amount value (Δavg value) of the buffer queue length (Δavg value) and the average length value (avg value) of the buffer queue (queue) change with the inflow of data, it may transition to another state as shown. The change in the average length value of the buffer queue and the change value of the queue length may be measured at a specified time, or may be measured when there is a change in the incoming data.

That is, when avg <Minth in the first phase, the first phase is maintained as it is, when the Maxth <avg <Minth, the transition to the 2-1 phase, and when avg> Maxth, the transition to the third phase.

In the case of phase 2-1, in the case of Nvarth <? Avg <Pvarth, phase 2-1 is maintained; in the case of avg <Minth, the process returns to the first phase and processes all data normally, and in case of? Avg> Pvarth Transitions to phase 2-2, transitions to phase 3 when avg> Maxth, and transitions to phase 2-3 when Nvarth>? Avg.

In the case of phase 2-2, when Nvarth < DELTA avg < Pvarth, the transition is made to phase 2-1, and when Nvarth >

In the case of phase 2-3, the transition is made to phase 2-1 when Nvarth <Δavg <Pvarth, and transitioned to phase 2-2 when Δavg <Nvarth.

In the third phase, when Maxth < avg < Minth, it transitions to phase 2-1, and when avg < Minth, it transitions to first phase as it is.

 In this embodiment, the transition states are divided into five, but in some cases, it may be possible to divide the transition state into five or less states or more.

3 illustrates an example of applying a state transition algorithm to a single network queue structure according to an embodiment of the present invention.

Referring to FIG. 3, when there is a single network queue, a change amount value (Δavg value) of a buffer queue length due to a constantly changing network state and an average length value of an buffer queue (avg value) are shown. ) Transitions to different states. That is, due to the variable initialized in the Initial state, the buffer basically operates in the first phase (phase 1), but when the average length value (avg value) of the buffer corresponds to Maxth <avg <Minth, the second phase (phase 2) ).

Phase 2 is composed of phases 2-1, 2-2, and 2-3, and the variation value of the buffer length (Δavg value) is measured based on the amount of data flowing into the buffer. ) And the average length of the buffer (avg value) determine the amount of noncritical data to discard. However, as described above, in the present embodiment, the second phase is defined in three states, but it may also be defined as an embodiment above or below.

The third phase (phase 3) is to discard a plurality of packets at once by a predetermined level of quality of service to deliver important data in real time when there is too much incoming data as described above.

4 is a structural diagram of a data processing system according to an embodiment of the present invention.

Referring to FIG. 4, the data processing system according to the present invention may include a receiver 100 for introducing data to be processed, a buffer 200 for processing the data, and a scheduler 300 for managing the buffer.

The scheduler determines an "avg value", which is an average length of a buffer queue as individual packets are processed, and a rate of increase or decrease of space in a buffer queue according to data accumulated per unit time. The buffer determines the possibility of buffer overflow and applies a pre-emptive low priority packet drop (PLPPD) algorithm that minimizes the loss of important data through preemptive disposal of noncritical data during network congestion. Pass data in real time.

In the above description of the preferred embodiments of the present invention by way of example, the scope of the present invention is not limited only to these specific embodiments, the present invention is in various forms within the scope of the spirit and claims of the present invention May be modified, changed, or improved.

100: receiver
200: adaptive buffer
300: scheduler

Claims (12)

In the buffer management method,
Predicting an overflow situation of the buffer according to the queue length information of the buffer;
Dividing the operation into multiple phases for differently operating the buffer according to a pre-emptive low priority packet drop (PLPPD) algorithm based on the predicted overflow condition of the buffer;
Distinguishing important data to be processed and non-critical data to be discarded in the overflow situation; And
And operating a buffer according to the plurality of aspects. The method of claim 1,
The queue length information of the buffer is a result obtained by comparing the variation value (Δavg value) of the buffer queue length and the average length value (avg value) of the buffer queue with a predefined threshold value. Buffer management method, characterized in that. The method of claim 1,
In the plurality of phases according to the PLPPD algorithm, the first phase processes all data, the second phase discards non-critical data in preparation for a buffer overflow, and the buffer overflow occurs due to excessive traffic. And a third aspect to augment non-critical data to be discarded if necessary. The method of claim 1,
The buffer overflow prediction is performed at regular intervals or when there is a change in the amount of data flowing into the buffer. The method of claim 3, wherein
The second aspect is,
If it is determined that the risk of buffer overflow is gradually increasing, phase 2-1 of discarding non-critical data having the lowest priority for each packet, despite discarding non-critical data having the lowest priority, In the second and second aspects of expanding the range of packets discarded gradually when the length of the buffer queue continues to increase, and in the case of the buffer overflow risk disappearing, the second and third stops discarding non-critical data. A buffer management method comprising a phase. The method according to claim 3 or 5,
The third aspect is,
A method for managing a buffer, characterized in that for discarding non-critical data of the lowest priority by each quality of service (QoS) marking level. In a data processing system,
A receiver for introducing data to be processed;
A scheduler for predicting a buffer overflow according to queue length information of a buffer and managing the received data; And
And an adaptive buffer for processing the data according to a pre-emptive low priority (PLPPD) algorithm based on the information received from the scheduler. The method of claim 7, wherein
The queue length information of the buffer is a result obtained by comparing the variation value (Δavg value) of the buffer queue length and the average length value (avg value) of the buffer queue with a predefined threshold value. Adaptive data processing system, characterized in that. The method of claim 7, wherein
The PLPPD algorithm discards the non-critical data in preparation for the first phase processing all the data, the buffer overflow, and the second phase processing the important data, and non-critical data discarding when the buffer overflow occurs due to excessive traffic. Adaptive data processing system, characterized in that the operating method of the adaptive buffer in a third aspect to enlarge the. The method of claim 7, wherein
And the scheduler predicts the buffer overflow at regular intervals or predicts the buffer overflow if there is a change in the amount of data flowing into the buffer. The method of claim 9,
The second aspect is,
If it is determined that the risk of buffer overflow is gradually increasing, phase 2-1 of discarding non-critical data having the lowest priority for each packet, despite discarding non-critical data having the lowest priority, In the second and second aspects of expanding the range of packets discarded gradually when the length of the buffer queue continues to increase, and in the case of the buffer overflow risk disappearing, the second and third stops discarding non-critical data. Adaptive data processing system comprising a phase. The method according to claim 9 or 11,
The third aspect is,
Adaptive data processing system, characterized in that for discarding the non-critical data of the lowest priority by each quality of service (QoS) marking level.

Images