KR100342523B1 - Method for fair flow control in packet-switched networks - Google Patents

Abstract

A packet switching network includes a plurality of nodes connected to each other, each node having a data queue for connecting a plurality of sources for transmitting and receiving data and for storing data transmitted from the source, Wherein each node is associated with a current queue length and a target value, the method comprising the steps of: determining a local bottleneck virtual node using a data rate at a corresponding source and a minimum data rate to be guaranteed; Estimating the number of lines, allocating a data rate to each of the sources based on a difference between the current queue length and the target value, a differential value of the current queue length, and a number of the estimated local bottleneck virtual circuits , The data rate is transmitted through a feedback signal transmitted to each of the sources It consists of a process of sugar.

Description

[0001] METHOD FOR FAIR FLOW CONTROL IN PACKET-SWITCHED NETWORKS [0002]

The present invention relates to a packet switched network, and more particularly to a method for fair flow control.

The packet-switched network includes an asynchronous transfer mode (ATM) network and the Internet. Equal flow control is a very important factor in information transmission on a network. In particular, fair flow control in an ATM network is associated with an available bit rate (" ABR ") service.

In general, ATM accomplishes network use for information transmission through the following four service methods. That is, a CBR (Constant Bit Rate) scheme that enables a fixed transmission bandwidth service, a Variable Bit Rate (UBR) scheme that allows a variable bandwidth service while ensuring a constant cell loss rate, a UBR (Unspecified Bit) Rate) method, and ABR method. The ABR scheme is a method that changes the transmission bandwidth that the source can send according to the network conditions. That is, it is a way to allow the source to send data according to the available transmission bandwidth of the network.

The ABR service was introduced in ATM networks to support data applications that could not be efficiently supported by bandwidth assurance services such as VBR. Examples of articles or data on this are as follows. S. Sathaye, "ATM Forum Management Specification, Version 4.0", Feb. 1996, F. Bonomi and KW Fendick, "The Rate-Based Flow Control Framework for the Available Bit Rate ATM Service" pp. 25-39, and R. Jain, " Congestion Control and Traffic Management in ATM Networks: Recent Advances and Surveys ", Computer Networks and ISDN Systems 28 (13), (1996) p1723-1738.

While most data applications are highly irregular and have no way of predicting future data traffic demands, they have a well-defined cell loss requirement and may have tolerances for time variations and unpredictable cell delays. More importantly, because of this nature, most data applications can change the data rate depending on network load conditions. Accordingly, the concept of an elastic traffic service that adjusts the input data rate (or transmission flow) in accordance with the available bandwidth in the network has been introduced. A typical example is the ABR service in an ATM network.

In principle, the ABR service does not require complex traffic characteristic modeling and call admission control at each of the sources and switches. With this simplicity, it was expected that the implementation and development of the ABR service would be much easier than the bandwidth guarantee service (CBR or VBR service). However, it has been determined that the design and implementation of ABR-enabled switches is actually much more challenging than originally anticipated. The biggest difficulty is in designing a scalable, stable, and fair ABR flow control algorithm. Especially in the design of an explicit rate (ER) allocation algorithm in asynchronous and decentralized network environments.

The ATM Forum adopts a closed-loop rate-based approach to control the flow of ABR services. Rate-based flow control uses information fed from the network to control the rate at which each source can transmit to the network, as implied by its name. The feedback information is conveyed by a specific control cell called resource management (" RM ") cell. There are three mechanisms for recording the network congestion state on the RM cell: Explicit Forward Congestion Indication (EFCI) marking, RR (relative rate) marking and ER (Explicit Rate) marking. Must be implemented on the switch.

On the other hand, different bottlenecks between the long and varied round-trip delay and the ABR VC (Virtual Circuit) contained in the flow control closed loop make the design of a high performance ER allocation algorithm difficult. It is difficult for ABR queues in the network to be stabilized when the transmission rates of ABR sources are determined according to network state information at different points in time. In particular, using only a binary feedback mechanism (EFCI marking or RR marking, or EFCI marking and RR marking), the ABR cues will have a persistent oscillation in steady state, the amplitude of which is the product of the round trip time delay and the available bandwidth . A detailed description thereof is disclosed in detail in an example of the following papers. E. Hernandez-Valencia et al., "Rate Control Algorithms for the ATM ABR Service", European Transactions on Telecommunications 8 (1) (1997) p7-20 and F. Bonnomi, D. Mitra and JB Serry, (1995) p1267-1283 and KK Ratmarkrishnan and Jain, " A Binary Feedback Sheme for Congestion Avoidance ", IEEE Trans. in Computer Networks with a Connectionless Network Layer, Proc. ACM SIGCOMM'88 (1988) 303-313.

The continuous oscillation of these ABR queues increases the likelihood of cell loss and low link availability by generating periodic buffered < RTI ID = 0.0 > and / or < / RTI > The prerequisite for introducing ER marking, namely ER flow control, is to suppress the continuous oscillation caused by the binary feedback mechanism by ensuring the asymptotic stability of the ABR queue. However, it is still a difficult problem to design the ER allocation algorithm to stabilize the ABR queue asymptotically. This problem results in a feedback control problem with a time delay from a mathematical point of view.

Benmohanmed and Meerkov proposed an ER allocation algorithm that ensures asymptotic stability and allows arbitrary control of closed-loop performance by formulating the rate-based flow control problem as a discrete-time feedback control problem in their first paper. The above-mentioned ER algorithms proposed by Benmohanmed and Meerkov are described in L. Benmohamed and S. M. Meerkov, "Feedback Control of Congestion in Packet Switching Networks: The Case of Single Congested Node", IEEE / ACM Trans. 10 (5) (1999), pp. 693-708, and L. Benmohamed and SM Meerkov, "Feedback Control of Congestion in Packet Switching Networks: The Case of Multiple Congested Nodes", International Journal of Communication Systems, 1997) p227-246.

The ER allocation algorithm proposed by Benmohanmed and Meerkov is shown in Equation 1 below.

Where r [k] is the ER computed by the switch at discrete time k, q [k] is the ABR queue length at discrete time k, and q _T is the target queue length. And Lt; / RTI > is the controller gain, Is the maximum round trip time delay of the ABR VC and I is any constant greater than " 0 ". Despite the rationale described in the above paper, the above algorithm is very complicated to implement and thus has limited practical use. A. Kolarov and G. Ramamurthy, "A Control Theoretic Approach to the Design of Close Loop Rate Based Flow Control for High Speed ATM Networks," Proc. IEEE INFOCOM'97 1 (1997), p293-301, . As can be seen from Equation (1), the ER allocation algorithm By time (j = 0 ~ ), And it requires a large number of floating point multiplications per discrete time slot.

In the paper S. Chong, "Second-Order Rate-Based Flow Control with Dynamic Queue Threshold for High-Speed Wide-Area ATM Networks," preprint 1997, and A. Elwalid, "Analysis of Adaptive Rate-Based Congestion Control for High -Speed Wide-Area Networks, Proc. IEEE ICC'95 (1995) P1948-1953 proposed another ER allocation algorithm based on control theory as a simpler form as shown in Equation 2 below.

In the first paper, we propose the necessary and sufficient condition for the asymptotic stabilization of the closed-loop system when the algorithm of Equation (2) is applied to the case where the round trip time delay of all VCs is the same. And extended to the general case with round-trip time delay.

S. Chong, R. Nagarajan, and YT Wang, "Performance-Based Control of Rate-Controlled Rate-and-Loop Control: First-Order Control Case" , A much simpler form of ER allocation algorithm is proposed. The contents of the ER assignment algorithm of Equation (3) are disclosed in U.S. Patent No. 5,864,538, issued January 26, 1999, entitled " First-order rate-based flow control with adaptive queue threshold for ATM networks " .

r (t) = [-K (q (t) - q _T )] ⁺ , K > 0

Here, [x] ⁺ = max [x, 0] means to select a larger value of "x" and "0".

In the first of two papers related to equation (2), two different stabilization conditions are derived. One of them is a sufficient condition for a general case with a heterogeneous round trip delay and the other is a necessary and sufficient condition for a special case with a homogeneous round trip time delay.

A common drawback of the ER algorithm disclosed in Equations 2 and 3 as compared to Equation 1 is that if the controller gain and queue length threshold are equal to the bandwidth available for ABR traffic and the available bandwidth used by the remote bottleneck VCs The ABR queue length can be undesirably converged to " 0 " because the link at the equilibrium point can not be fully utilized unless it is properly selected according to the instantaneous recognition in fraction. By definition, remote bottleneck VCs in a link are VCs that can not be equally divided on the link because their rate is limited by their PCR (Peak Cell Rate) or bottlenecks on other links in the path. Conversely, if an algorithm like Equation (1) is applied, there is no such an undesirable equilibrium point mentioned above.

It is therefore an object of the present invention to provide a method for guaranteeing maximum link utilization and minimum cell loss regardless of the round-trip time delay of ABR loops.

It is another object of the present invention to provide a method for minimizing the requirement of the ABR queue size by ensuring the asymptotic stability of the ABR queue.

Another object of the present invention is to provide a method for ensuring the fairness of bandwidth utilization among ABR users, and in particular, a method for guaranteeing maximum-minimum fairness, which is an ATM forum standard.

It is another object of the present invention to provide a method for quickly adapting to a change in a network environment such as a change in the number of ABR users and a change in an ABR bandwidth.

It is another object of the present invention to provide a method for carrying out almost all the functions provided in the ATM Forum traffic management standard including functions such as EFCI marking, RR marking, and ER marking.

It is a further object of the present invention to provide a method for achieving high availability, low cell loss, and MAX-MIN fair rate allocation by providing an asymptotically stable operating point.

It is a further object of the present invention to provide a method for increasing the responsiveness and control transient performance for multiple time scales, i.e., changing the cell level ratio of VBR and ABR VC, and changing network load at the cell level arrival and departure of VBR and ABR VC .

It is a further object of the present invention to provide a method for minimizing the number of operations required to calculate an algorithm and reducing the complexity of implementation and scalability in removing VC-specific operations including VC-specific queuing, VC- .

According to another aspect of the present invention, there is provided a packet switching system including a plurality of nodes connected to each other through a packet-switching network, each node connecting a plurality of sources for transmitting and receiving data, Wherein the data queue is associated with a current queue length and a target value, the method comprising the steps of: determining a data rate at a corresponding source and a minimum data rate Estimating the number of local bottleneck virtual circuits by using a difference between the current queue length and a target value, a differential value of the current queue length, and a number of the estimated local bottleneck virtual circuits, And the data rate is transmitted to each of the sources It characterized by constituted by any process that is allocated through the feedback signal.

1 is a network configuration diagram for explaining an ABR service;

2 is a view for explaining each field included in an RM cell;

3 is a diagram illustrating a configuration in which input / output port cards are connected to input / output ports of a switch fabric for ABR service;

Fig. 4 is a specific block configuration diagram of the input / output port card of Fig. 3; Fig.

5 is a block diagram of an ABR service engine block according to an embodiment of the present invention.

Figure 6 shows a network model with nodes of interest;

Figure 7 shows a stable region in terms of U and V according to an embodiment of the present invention;

8 is a diagram showing the relationship between the functions U and V and the asymptotic attenuation rate [alpha].

9 is a diagram for describing queue length thresholds for RR marking and ER marking;

10 is a peer-to-peer model configuration diagram.

11a-11d show experimental results in a peer-to-peer model configuration with only ER marking and no VBR background traffic,

11A shows a source transmission rate a _i (t) of a VC with a PCR of 150 Mbps, FIG. 11B shows a source transmission rate a _i (t) of a VC with a PCR of 25 Mbps, FIG. 11C shows a queue length in the switch SW 1, In switch SW1, the estimate of the number of local bottlenecks VC to be.

12A to 12D are experimental results in a peer-to-peer model configuration with ER marking and VBR background traffic. FIG. 12A shows a source transmission rate a _i (t) of a VC with a PCR of 150 Mbps, a source rate of a VC _i (t), and Figure 12c is a queue length of the switch SW1, Fig. 12d is the number of estimates of the local bottleneck in the VC switch SW1 to be.

13 is a block diagram of a parking lot model;

14A to 14F are experimental results in a configuration of a parking lot model with only ER marking and no VBR background traffic. FIG. 14A shows the source transfer rate a _i (t) of the VC with a PCR of 150 Mbps, a source rate of a VC _i (t), and Figure 14c is a queue length of the switch SW3, Fig. 14d is a queue length of the switch SW4, Figure 14e is an estimate of the number of local bottleneck VC at the switch SW3 14F shows the number of local bottleneck VC estimates in switch SW4 to be.

15A to 15F are experimental results in a configuration of a parking lot model with ER marking and VBR background traffic. FIG. 15A shows the source rate a _i (t) of a VC with a PCR of 150Mbps, FIG. 15B shows a VC with a PCR of 25Mbps of the source transmission rate a _i (t), and Figure 15c is a switch and the queue length in the SW3, Fig. 15d is a queue length of the switch SW4, Figure 15e is an estimate of the number of local bottleneck VC at the switch SW3 15F shows the number of local bottleneck VC estimates in switch SW4 to be.

FIG. 16 is a control flowchart showing an ER allocation control operation performed by the ABR service engine according to an embodiment of the present invention; FIG.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. It is to be noted that the same elements among the drawings are denoted by the same reference numerals whenever possible. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail.

1 is a network diagram for explaining an ABR service. Referring to FIG. 1, a network 2 for an ABR service includes a plurality of nodes 4, 6, 8, and 10 interconnected with each other. A plurality of nodes 4, 6, 8, and 10 are conceptual representations of switches (hereinafter also referred to as " switches "). Each of the plurality of switches 4, 6, 8, and 10 may be connected to a plurality of sources. 1, source A, C (12, 14) are connected to a switch 4, and sources B, D (16, 18) are connected to a switch 8. Each of the sources 12, 14, 16, 18 transmits and receives data to correspondingly connected switches 4, 8. The data transmitted by the source is transmitted to the destination via a path referred to as a VC path including a plurality of nodes. For example, the data transmitted from the source A 12 is transmitted to the destination B 16 via the VC path where the switches 4, 6, and 8 exist.

In the ABR service, a special cell for controlling the RM cell (denoted by RMc in FIG. 1) is placed in order to inform the source of the transmission bandwidth available to the network. In the ABR service, the RM cell is generated at the source 12, 14, 16, 18 or the switch 4, 6, 8, 10, but in the present invention, only the RM cell generated by the source will be described.

The RM cell generated in the source passes through the switches (4, 6, 8) passing through the VC path to the destination, and this cell traveling direction is referred to as forward. At the destination, the RM cell received is sent back to the source, and the direction of the RM cell returned to the source is called the backward direction. In Fig. 1, assuming that the source is 12, 14 and the destination is 16, 18, the forward direction and the reverse direction are indicated. At this time, each of the switches 8, 6, and 4 stores the information on the bandwidth that the switch 8 can allow to the RM cell, and the stored information is transmitted to the sources 12 and 14, The amount of information to be transmitted to the destinations 16 and 18 is appropriately adjusted.

The RM cell contains many kinds of information. Of these, the CCR (Current Cell Rate), MCR (Minimum Cell Rate), ER (Explicit Rate), NI (No Increase) and CI (Congestion Indication) fields are important. FIG. 2 shows fields included in the RM cell. Referring to FIG. 2, a source-receive (S-R) field is a field for storing source and destination related information, and a CCR field is a field for storing a transmission bandwidth value of a source when a source first generates an RM cell. The MCR field is a value stored when a source generates an RM cell, which means the minimum data rate of each VC. The ER field is used in the reverse direction when the RM cell generated by the source goes through all the forward paths and then returns to the source. Each switch in the reverse direction writes its available bandwidth information to the RM cell, where the ER field in the RM cell stores the information. When the RM cell arrives at the switch, it stores the smaller value of the available bandwidth stored in the ER field and the available bandwidth calculated in the switch as a new value of the ER field. That is, the source receives the ER value calculated in the switch having the smallest usable bandwidth among the paths of the VCs. For this purpose, the source generating the RM cell initially uses the peak cell rate (PCR), which is the maximum bandwidth that can be transmitted to the ER field. Also, NI and CI are fields used for relative rate control. The NI field is a field that informs the source that the switches on the path of the VC do not increase the transmission bandwidth of the source any more depending on the degree of traffic congestion. Is a field for informing the source that the transmission bandwidth of the source is reduced.

For ABR service, each switch needs to write the amount of allowable transmission bandwidth to the ER field in the RM cell when the reverse RM cell passes. To do this, each switch needs to calculate the available bandwidth and a switch algorithm to calculate it. In the end, what we want to achieve through the switch algorithm is the value of the transmission bandwidth that the switch can tolerate for each VC. The value of this transmission bandwidth is known to the source, and by referring to this value, the source adjusts the transmission rate so that the normal ABR service is performed.

As shown in FIG. 3, each of the switches 4, 6, 8, and 10 of FIG. 1 includes an input / output port card 22 for each of the input / output ports of the switch fabric 20 for ABR service. The input / output port card 22 includes an input / output buffer management unit 30, an ABR service engine 32, and an output interface 34, as shown in FIG. The input / output buffer management unit 30 is connected to the switch fabric 20 and manages input / output queuing. The input / output buffer management unit 30 includes an ABR queue 36. The output interface 34 performs the user network interface function of the ATM layer and the ABR service engine 32 performs the batch processing of the ABR algorithm and related functions according to the embodiment of the present invention while the RM cell passes through it.

An example of the specific block configuration of the ABR service engine 32 of FIG. 4 is shown in FIG. 5, the ABR service engine 32 includes an EFCI marking unit 40, a Q estimating unit 42, an ER calculating unit 44, a queue average value calculating unit 46, an inverse RM cell recording unit 48 ).

In FIG. 5, the EFCI (Explicit Forward Congestion Indication) marking unit 40 marks an EFCI bit notifying EFCI congestion in a forward data cell applied when EFCI congestion occurs. The queue average value calculator 46 of FIG. 5 receives the queue lead signal and the queue write signal. The queued signal is a signal applied from the outside when the cell is exited from the queue, and the queued signal is an externally applied signal when a new cell is input to the queue. The queue average value calculator 46 increases the value of the queue instant variable when the queue light signal is applied and decreases the value of the queue instant variable when the queue read signal is applied, Calculate the average value of the queue values during the calculation period. The queue average value calculator 46 calculates an average value of the calculated queue values To the ER calculation unit 44. [ The queue average calculation unit 46 also determines EFCI congestion using the queue read signal and the queue write signal, and outputs a signal EFCI_CG indicating EFCI congestion to the EFCI marking unit 40 in which EFCI congestion has occurred. Also, according to the embodiment of the present invention, congested based on a low threshold q _LT and a high threshold q _HT preset in the ABR queue, and signals CG (congested) and VCG (very_congested) . The congestion and very congestion signals CG, VCG will be used for relative rate control according to embodiments of the present invention.

In FIG. 5, the Q estimation block 42 is a block for estimating the number of local bottleneck VCs. The estimator 42 reads the CCR and MCR values in the RM cell and compares the calculated ER value r (t) periodically to calculate an estimate of the number of local bottleneck VCs . Estimate The interval W for calculating the discrete time ER will be described in detail in the section " (5) Discrete time ER algorithm and Q estimation ". The estimate Is provided to the ER calculation unit 44. [

5, the ER calculator 44 periodically calculates and updates the ER value and outputs the most updated ER value r (t) to the reverse RM cell register 48 when the reverse RM cell arrives.

In FIG. 5, the backward RM cell write unit 48 records the ER value r (t) periodically calculated by the ER calculation unit 44 in the ER field in the RM cell. More specifically, the backward RM cell write unit 48 obtains a value obtained by adding the MCR value in the RM cell to the ER value in the current RM cell and the ER value r (t) provided by the ER calculation unit 44 (T) + mi is recorded in the ER field in the RM cell only when r (t) + mi is smaller than r (t) + mi. This is to perform explicit transmission rate control. The reverse RM cell recorder 48 also records the bits of the binary logic state in the NI field and the CI field in the RM cell according to the congestion related to the relative rate control provided by the queue average value calculator 46 and the very congestion signals CG and VCG do. This is to perform relative rate control.

The control operation for ER allocation performed by the ABR service engine 32 of FIG. 4 will be briefly described with reference to the control flowchart of FIG. 13 and the specific block configuration of FIG.

The ABR service engine 32 estimates the Q value by the estimating unit 42 in step 100 and proceeds to step 102. [ In step 102, the ER calculation unit 44 updates the periodic ER calculation and the calculated ER value r (t). Thereafter, when the RM cell arrives at the backward RM cell recording unit 46 (step 104 of FIG. 7), the ABR service engine 32 proceeds to step 106 of FIG. In step 106, the most recently updated ER value r (t) is read by the reverse RM cell recorder 46 and the MCR value mi is read from the received RM cell. Then, in step 108, an ER allocation value ri (t) = r (t) + mi of VCi is calculated by the backward RM cell recording unit 46. Then, in step 110, the calculated ri (t) is recorded in the ER area of the reverse RM cell.

The switch algorithm proposed in accordance with the embodiment of the present invention aims to make the queue state of the ABR queue 36 a convergence value and to fix the queue length value in the steady state to the target value q _T. The fact that the length of the queue is kept constant includes the fact that the input traffic amount entering the ABR queue 36 becomes equal to the output traffic amount. For example, if the input traffic volume is significantly larger than the output traffic volume, the ABR queue 36 may be required at a time when the target queue length value q _T is reached. In this case, unexpected cell loss is caused. Therefore, the embodiment of the present invention implements a switch algorithm that takes into consideration changes in the buffer amount and the queue length state (buffer state) of the ABR queue 36.

In the embodiment of the present invention, a continuous time ER allocation algorithm as shown in Equation (4) is proposed to have the form of a controller which can easily and simultaneously remove an undesired equilibrium point in the ABR queue 36. [

Here, r (t) is the calculated ER value, Is the derivative of r (t). Q is the set of locally bottle-necked VCs in the link, and Q | is the cardinality of Q. q (t) is the length of the ABR queue 36 at time t, and q_T is the length (i.e., target value) of the target ABR queue 36. A and B are gains of the controller, and in the embodiment of the present invention, the values are set differently according to the length of the ABR queue 36, and this setting has asymptotic stability.

Comparing the ER allocation algorithm according to the embodiment of the present invention shown as equation (4) with the algorithm shown in equation (2), the algorithm shown in equation (4) according to the embodiment of the present invention is represented by the damping term Instead of r (t) . ≪ / RTI > Is a differential value of q (t), which will be described in detail with reference to equations (21) and (23) to be described later.

This change in the embodiment of the present invention removes the undesirable equilibrium in the closed loop system resulting in the ABR queue 36 being able to use available bandwidth and remotely bottlenecked VCs Converges to the target value q _T at any time, irrespective of the available bandwidth. This means that full utilization of available bandwidth in steady state is guaranteed all the time.

Another notable feature of embodiments of the present invention is that the controller gains A, B are normalized by the number of local bottleneck VCs | Q |. This normalization makes the asymptotic decay rate of the closed loop system irrelevant to the number of local bottleneck VCs | Q |. The estimation of the number of local bottlenecked VCs | Q | has become an important topic in rate-based ABR flow control research. For example, M. K. Wong and F. Bonomi, "A Novel Explicit Rate Congestion Control Algorithm", IT Proc. &Quot; An Efficient Rate Allocation Algorithm for ATM Networks Providing Max-Min Fairness ", Technical Report UCSC-CRL-95-29 (1998) p 2432-2439 and L. Kalampoukas, A. Varma and KK Ramakrishnan, IEEE GLOBECOM'98 4 , Computer Engineering Dept., University of California, Santa Cruz, June 1995 and A. Charny, K. Ramakrishnam and A. Lauck, "Time Scale Analysis and Scalability Issue for Explicit Rate Allocation in ATM Networks", IEEE / ACM Trans. on Networking 4 (1996) p569-581, R. Jain et al., "ERICA Switch Algorithm: A Complete Description", ATM Forum / 96-1172 (1996).

The difficulty that lies in the dynamic nature of the estimation process is the dynamic nature of the ER allocation processor, that is, until the closed-loop system reaches a steady state, the ER update will affect the update | Q | So that the nonconforming design of the | Q | estimation algorithm can make the closed-loop system unstable. A stable, stable and measurable estimation algorithm is presented in connection with the ER algorithm according to the embodiment of the present invention.

According to the embodiment of the present invention, the control theory ER algorithm expressed by Equation (4) ensures so-called MAX-MIN fair rate allocation because the reason is as follows. All VCs based on the calculation of Equation (4) in the switch always assign the same ER (hereinafter referred to as " common ER ") to the shared link. Therefore, the RM cell transmits the minimum allocation ER through the path to the source, and the source transmits the data to the minimum ER. The ABR queue 36 in the switch will have a delay and its value will be attenuated if the VC is bottleneck elsewhere and the source transmits data at a lower rate than the common ER assigned by the switch. If the ABR queue 36 falls below the target queue length q _T , the switch increases the common ER to be the target queue length q _T according to the algorithm shown in equation (4). So the local bottleneck VCs will use the same bandwidth that is not used by the remote bottleneck VCs in the end.

In fact, an algorithm like Equation 4 should be replaced with a discrete time concept if applied to a situation where the network is operating. The algorithm of Equation (4) proposed according to an embodiment of the present invention is expressed by a discrete time algorithm as shown in Equation (5).

Where T is the duration of the discrete time slot. The algorithm of equation (5) is a special case of the algorithm of equation (1). That is, if I = 1, , And = 0, (5). &Quot; (5) " As opposed to the algorithm of equation (1), the transformed algorithm no longer allows any control of the closed-loop dynamic properties. However, the inventors of the present invention argue that the arbitration of control capability is very expensive in terms of implementation cost and is not required for ABR flow control design, as discussed above. To support this assertion, embodiments of the present invention will illustrate that the transformed algorithm can accept control over closed-loop performance.

In the control theory ER algorithm (Equation 4 and Equation 5) according to an embodiment of the present invention, queue length control is a primary interest and fair rate allocation is a secondary result of queue length control. Other types of ER algorithms to be described below. In equations (14) to (18), maximum-minimum fairness is of primary interest and queue length control is secondary. The other type of ER algorithm follows the available bandwidth for ABR traffic and requires each switch to share fairly the remote bottleneck VCs in the VC bottleneck link and the number of local bottleneck VCs. Based on this information, the switch updates the per-VC ER assignments in a way that asymptotically achieves the MAX-MIN fair rate allocation and the target link utilization.

(1) ER rate flow control

Hereinafter, it will be described that the ER allocation algorithm of the present invention is mathematically modeled and the ER allocation algorithm of the embodiment of the present invention guarantees the MCR required at each source and satisfies the maximum-minimum fairness through the modeling.

Figure 6 is a diagram showing a network model with nodes of interest. 5, a plurality of VCs each include a plurality of corresponding sources. The VCi 54 corresponds to the source i 50 and the VCj 56 corresponds to the source j 52. [ 36 is an ABR queue, q _T indicated on the ABR queue 36 is a target value of the ABR queue 36, and is connected to the ABR queue 36 Is the available bandwidth of the shared link of the plurality of VCs. Also And Each means a forward path delay and a backward path delay in the VCi 54, Wow Is equal to the round trip time delay in VCi 54 . And Each means forward path delay and reverse path delay at VCj 56, And Lt; RTI ID = 0.0 > VCj < / RTI > (56) .

First, we assume the following for the analysis of the network model.

Assumption 1) Traffic is considered as determined flow, and network queuing processor and flow control mechanism are considered to be continuous in time. This assumption makes the model a closed-loop system by differential equations.

Assumption 2) VCi's round-trip time delay Includes forward path delays, including propagation, queuing, transmission, and processing time And reverse path delay . Here, it is assumed that the round trip time delay is constant.

Assumption 3) The sources are persistent. Here, the term "persistent" means that the source can sufficiently transmit data at the data rate assigned to the network.

Assumption 4) There is no arrival or departure of VCs until the system reaches a normal state.

Assumption 5) Available bandwidth in the link Is constant until the system reaches a steady state. The buffer size of the ABR queue 36 in the link is assumed to be infinite.

The transmission rate when the source i 50 transmits data at the source time t is denoted by a _i (t), and the ER value allocated to the VCi (54) as the interested node performs the ER allocation calculation at the node time t r _i (t). Also, the latest minimum value of the ER value assigned to the VCi 54 by the node is denoted by b _i (t), the peak rate constrained by the VCi 54 is denoted by p _i (t) It should be noted. P _i (t) is a PCR (Peak Cell rate) in VCi (54) as a standard term in the ATM ABR.

Ignoring the linear increase and the exponential decrease based on the binary feedback, the operating characteristic of the source is modeled as Equation (6).

Here, N is a set of all VCs including nodes to which the route is interested. The model as in Equation (6) shows that the source i (54) The ER value assigned by the node of interest before, And the minimum ER value on the route other than the node of interest, that is, bi (t), and the PCR of the VC, that is, pi.

Meanwhile, the operation characteristic of the ABR queue 36 per class of interest at the node is given by Equation (7).

The ER allocation algorithm according to the embodiment of the present invention includes a queue length q (t), a derivative value of the queue length , An estimate of the number of local bottleneck VCs Which is performed by each switch based on the current network state including the current network state. These switch algorithms are given by the following equations (8) and (9).

In Equation (9), the controller gains A and B are A, B > 0. And m _i in Equation (8) is the minimum data rate at the node for which the source i (50) requires a guarantee during the period in which VCi is alive, and MCR (Minimum Cell Rate) is the standard term in the ATM ABR. here , Call admission control is guaranteed, as expressed in Equation (10).

r (t) is the ER allocation value per VC That is, a common part, that is, a common ER value, which is most needed for algorithm computation. The VC-specific computation required in accordance with an embodiment of the present invention is merely to add m _i , which is the guaranteed MCR value at source i (50) to the common ER value r (t). This simple operation allows embodiments of the present invention to reduce the complexity of calculations with an increase in the number of VCs and to have scalability. Each RM cell of the current ABR service supporting VCi 54 carries the MCR value m _i at the round trip.

In the embodiment of the present invention, the common ER value r (t) is updated by " background calculation ". &Quot; Background computation " means that the node periodically calculates the common ER value regardless of whether the RM cell is received. The advantage of performing the background computation as described above is that it can calculate the most recently updated common ER value r (t) in advance when the RM cell arrives at the corresponding node.

Node sewn to update a common ER value r (t) obtained in accordance with Equation (9) through the background calculated in MCR field of the RM cell in VCi (54) through reading the m _i, updates the read m _i and the last the common ER value r (t) by substituting the equation (8) to obtain the ER value r _i (t) to be assigned to VCi (54) records the value r _i (t) to the ER field of the RM cell.

Another notable feature of the algorithm according to embodiments of the present invention is that the estimate of the number of local VCs In the normalization of the controller gains A and B. The normalization is not absolutely necessary and is a recommendation. The normalization allows the Peru system's asymptotic decay rate to be independent of the number of local VCs. The normalization of the controller gains A and B will be described in detail in "(4) Main muscle and asymptotic damping rate" below.

The remote bottleneck VC and the local bottleneck VC are defined in a steady state for a predetermined network load. In a link, remote bottleneck VCs are defined as VCs that can not be shared fairly on the link because of the limited data rate due to PCR or because of bottlenecks on other links in the path. In a link, a local bottleneck VC is defined as VCs that can be fairly shared on the link.

In order to mathematically define local bottleneck VCs and remote bottleneck VCs, , , , The set Q of all local VCs is given by Equation (11). Here, a _is, r _is, b _is means a _i (t) _, r _i (t), and b _i (t) in the steady state.

Then, the set N-Q of all the remote VCs is given by Equation (12).

(5) Discrete-Time ER Algorithm and | Q | Estimation " In the estimation algorithm, It is not efficient to converge quickly to | Q |, but during the transient period Will have a greater robustness than | Q |. This stiffness characteristic is necessary for the stability of the ABR flow control.

(2) steady state and fairness

The steady state characteristics of the ABR closed loop dynamic characteristic will be described below when the ER algorithm according to the embodiment of the present invention is applied. That is, the analysis result according to the embodiment of the present invention will be explained. The closed-loop dynamic characteristics are such that the derivative values of the system variables are all " 0 & , , Is assumed to have an equilibrium point. Based on this assumption, the following equations (13), (14) and (15) are obtained from Equations (6), (8) and (9).

,

r _is = r _s + m _i , ∀ _i ∈ N,

q _s = q _T

here, to be. Other notations have been defined in the past. Since q _s = q _T > 0, the buffer equation of Equation (7) has the same meaning as Equation (16).

The following equations (17) can be obtained by combining the equations (13), (14), and (16) with the definitions of the equations (11) and (12).

Equation (17) means Equation (18).

As described above, the embodiment of the present invention has the following characteristics in a steady state and results as shown in Equation (19) below.

(I) the queue length is equal to the target queue length (q _s = qT), (ii) the available bandwidth in the link is used as full as possible ), (iii) the individual MCRs are guaranteed on each link and the bandwidth minus the sum of the MCRs There is a steady-state only solution (equilibrium point), which is equally shared in terms of max-min fairness. In other words,

The above-mentioned characteristic means that when the ER algorithm according to the embodiment of the present invention is applied, the ABR closed loop system has a unique and vibration-free operating point, and in the vibrationless operating point, Not only MCR guarantees but also maximum (MIN-MIN) fairness is realized. The queue length of the ABR queue 36 is equal to the target value q_T. That is, the equilibrium point is completely independent of the available bandwidth for ABR traffic and some of the available bandwidth used by remote bottleneck VCs, including PCR-constrained VCs.

What is surprising is that the ER algorithm of the embodiment of the present invention is simple while producing such a good equilibrium point. The ER algorithm according to an embodiment of the present invention is based on q (t) and It requires information only, and does not require any other measurement or monitoring results. Estimate Is used to normalize the controller gains A and B, but only to promote the convergence rate, it has nothing to do with the equilibrium point. The Benmohamed's algorithm disclosed in Equation (1) also yields a unique system equilibrium point having the same properties as the algorithm of the embodiment of the present invention, but there is a high degree of implementation complexity. In contrast, if the algorithms disclosed in equations (2) and (3) are applied, the closed loop system has an equilibrium point that varies based on available bandwidth and some of the available bandwidth used by the remote bottleneck VCs, The queue length can be converged to " 0 ". For example, S. Chong, R. Nagarijan and YT Wang, "Design-Stable ABR Flow Control with Rate-Feedback and Open-Loop Control: First-Order Control Case," Performance Evaluation 34 (4) , S. Chong, " Second-Order Rate-Based Flow Control with Dynamic Queue Threshold for High-Speed Wide-Area ATM Networks ", preprint 1997 and paper A. Elwalid, " Analysis of Adaptive Rate-Based Congestion Control for High- Wide-Area Networks, Proc. IEEE ICC'95 (1995) P1948-1953.

(3) Asymptotic stability

The stability of the ER algorithm according to the embodiment of the present invention will be described below. In general, the stability characteristics of the equilibrium point given in Equation 19 can be examined in the case of multiple nodes. However, because of the complexity of the coupling dynamic characteristics between the nodes, the analysis may be too complex, and therefore, an embodiment of the present invention does not attempt to solve the overall stability problem in the combined multi-node configuration. In the embodiment of the present invention, only the simulation is investigated in the section " (6) Simulation results " below. On the other hand, in L. Benmohamed and SM Meerkov, "Feedback Control of Congestion in Packet Switching Networks: The Case of Multiple Congested Nodes", International Journal of Communication Systems, 10 (5) (1997) p227-246, The node-to-node dynamic characteristics are separated, and the local stability conditions derived from the adjacent points of the equilibrium point are also available for the FCFS service scheme. Depending on these results, it is assumed that an adjacent point having the characteristic R exists also in the case of the present invention. Consider a subset of R that contains the equilibrium point. Here, the following are satisfied.

a) That is, the dynamic characteristics of the remaining nodes are in a steady state.

b) And , That is, local bottleneck VCs And the remote bottleneck VCs As shown in FIG.

c) Saturated nonlinear values in equations (7) and (8) are not activated. That is, q (t) and r (t) are positive values.

d) - The estimation process is in a steady state. In other words, Is a constant.

At the adjacent points of these equilibrium points, the dynamic characteristic equations (6, 7) and (9) can be simplified to the following equations (20), (21) and (22).

By combining the expression (20) and the expression (21), the following expression (23) can be obtained.

Error function e (t) = q (t ) - q T as by defining and combining equation (22), the derivative and the derivative of equation (8) in the equation (23), to be obtained for the closed-loop equation as equation (24) have.

Equation 24 is a second-order delayed differential equation. The characteristic equation of the closed loop equation is given by the following equation (25).

Equation 25 has an infinite number of roots. For the asymptotic stability of the closed-loop equation of equation (24), all roots of the characteristic equation of equation (25) must have negative real parts. This is described in R. Bellman and KL Cooke, "Differential-Difference Equations" (Academic Press, New York, 1963) and G. Stepan, "Retarded Dynamical Systems: Stability and Characteristic Functions" have.

The equation of D (s) = 0 can rely on Pontryagin's criterion, assuming discrete delays of rational ratios to find the necessary and sufficient condition with stable roots. See, for example, R. Bellman and K. L. Cooke, "Differential-Difference Equations" (Academic Press, New York, 1963), S. J. Bhatt and C.S. Hsu, Stability Criteria for Second-Order Dynamical Systems with Time Lag, (1996) p113-118. In the general case involving continuous delays or discrete delays of the ratios, the stepan's Criterion provides a way to derive the necessary and sufficient condition, as described in G. Stepan, "Retarded Dynamical Systems: Stability and Characteristic Functions" (Longman Scientific Technical, 1989). Constructing an explicit form is very complex in the case of multiple connections of various round trip time delays.

Instead, embodiments of the present invention derive the necessary and sufficient condition for asymptotic stability when all round trip time delays are the same. With τ_i = τ, ∀_i, the closed-loop equation of equation (24) is represented by the following equation (26).

The equation of Equation (26) can be normalized such that the time delay? t = τξ, and the equation of Equation 26 is expressed by the new variable ξ, the following Equation 27 is obtained.

In the equation of the above equation (27) , to be. The characteristic equation of Equation (27) is as shown in Equation (28).

We can use the Fontian criterion as we have already posted to find the necessary and sufficient condition that all roots of H (z) = 0 have negative real parts.

The above-mentioned theory leads to the following conclusion. As described above, , ,

The closed loop equation of Equation (26) has a gradual stability only when the inequality condition as shown in Equation (29) is satisfied.

In the two inequalities, < RTI ID = 0.0 > Is the only solution of U = ω sin ω. Also, the gains A and B of the controller can be determined by the above inequalities. 7 shows a stable region in terms of U and V according to an embodiment of the present invention.

(4) Main muscle and asymptotic damping rate

In the following, the determination of the rate at which a stable closed loop system approaches a steady state according to an embodiment of the present invention will be described. The solution of the normalized closed-loop equation of Equation (27) can be expressed by the series of the following Equation (30).

Where p _n (?) Is the appropriate polynomial and z _n ,? _N are the roots of the corresponding characteristic equations in equation (28). As the root with the maximum real part Let's consider the main roots shown as. , The following equation (31) is derived from the equation (30).

Where C is a constant dependent on the initial condition of equation (27) The Represents asymptotically approaching one. From Equation (30), the following Equation (32) is derived.

From here, Represents the Euclidean norm, and c is a constant dependent on the initial condition of equation (27). The circular variable t (= ) Can be expressed by the following equation (33).

here, It should be noted that the circular system is the asymptotic decay rate going to the equilibrium point. The reciprocal of the asymptotic decay rate Is the time constant of the circular closed loop system, and the time constant A small perturbation around the equilibrium point The time taken to decrease by the factor of. Where alpha is the asymptotic decay rate, Is the time constant of the normalized system.

Fig. 8 is a graph showing the relationship between the functions U and V and the asymptotic attenuation rate [alpha]. Referring to FIG. 8, it is a concave function with a maximum value of about 0.3 near the point (U, V) = (0.6, 0.1). The contour line at? = 0 corresponds to the boundary of the stable region shown in Fig. Once determined for a given (U, V) pair, β can be easily determined by substituting z = -α + jβ in the characteristic equation of equation (28) and calculating the real and imaginary parts.

(5) discrete-time ER algorithm and -calculation

So far, the continuous-time model of the ABR flow control problem has been described. However, since the feedback information is actually relayed through the RM cells, the sampled form rather than the continuous time is rather useful. Furthermore, since the RM cell itself must compete for the link bandwidth along the round-trip time path, the feedback information is not periodic and the time intervals of the RM cells are time varying.

According to an embodiment of the present invention, the ER algorithm in Equation (8) and Equation (9) is implemented as discrete time possibly in the switch as shown in Equation (34). The common ER is periodically updated at a period of T by Equation (34).

, A, B > 0

In Equation (48) Quot; means the low-pass filtered queue length, i.e., the queue average value. In an embodiment of the present invention By using a period-averaging filter to obtain .

Note that the discrete-time ER update equation of equation (34) corresponds to equation (9) when T → ∞. In contrast to the periodic computation of the common ER values r [k], r [k + 1], the ER value per VC allocation is performed aperiodically at the moment when corresponding RM cells arrive in the forward or reverse direction. That is, when the RM cell of VC _i arrives at time t, the switch computes Equation (35).

r _i (t) = r (t) + m _i

Then, the result according to Equation (35) is recorded in the ER field of the RM cell. In the equation of Equation (35), r (t) denotes the latest value r [k + 1] of the common ER value r [k], which is periodically calculated in the embodiment of the present invention, Updated. m _i is available from the MCR table per VC that is held in or from the RM cell that is received and is updated at the start and the arrival of the call. If the embodiment of the present invention determines to take the value of m _i from the received RM cell, the per MCR table and its access is no longer needed. Then, in the algorithm according to the embodiment of the present invention, only the necessary per-VC operation can be performed with only the addition as in Equation (35).

On the other hand, many theoretical systems have been proposed to estimate or track local bottleneck VCs. For example, M. K. Wong and F. Bonomi, "A NovelExplicit Rate Congestion Control Algorithm", IT Proc. &Quot; An Efficient Allocation Algorithm for ATM Networks Providing Max-Min Fairness ", Technical Report, UCSC-CRL-95-29, Computer ", IEEE LABELCOM'98 4 (1998) p2432-2439 and L. Kalampoukas, A. Varma and KK Ramakrishnan ACK Transmission on Networking 4 (" Time < / RTI > Scale Analysis and Scalability Issue for ATM Networks ", IEEE Trans. 1996) p569-581 and R. Jain et al., "ERICA Switch Algorithm: A Complete Description", ATM Forum / 96-1172 (1996). ON "sources sharing a link in algorithms such as the number proposed in C. F., Su, G. de Veciana and J. Warlrand," Explicit Rate Flow Control for ABR Services in ATM Networks, preprint, The basic idea of estimating the number is attractive because it does not require computation per VC.

In an embodiment of the present invention, the algorithm is modified to estimate the number of local bottleneck VCs without performing computation per VC.

Assume that at the switch time t ^j , the j ^th RM cell arrives at the switch. According to the ABR specification [1], if the j ^th RM cell is an RM cell of VC _i , the j ^th RM cell carries the a _i (t ^j - τ _i ^f ) value of the CCR field and the m_i value of the MCR field . The switch monitors the arrival of the RM cell synchronously over fixed time intervals of W seconds. for the l ^th interval, the number of local bottleneck VC can be approximated by the following equation 36.

In Equation 36, 1 {·} denotes an indicator function, CCR (t ^j ) denotes a value in the CCR field, and MCR (t ^j ) denotes a value in the MCR field. r (t ^j) represents the latest value of the common ER value at time t ^j. In addition, the NRM is negotiated at the time of setting up the VC connection, as a cycle in the cell unit for forwarding the forward RM cell. ^ j th moment RM cells arrive, if the current cell rate minus the MCR is equal to or more recent values of the common value of the ER in the switch, j ^th VC RM cell belongs is calculated as a local bottleneck VC. δ is a limit value to prevent underestimation of the number of local bottleneck VCs particularly close to the steady state. When the system approaches a steady state, the current cell rate of the local bottleneck VC converges to the sum of the MCR and the common ER value, and if there is no threshold value δ, it is erroneously calculated as a remote bottleneck VC due to a small perturbation at the current cell rate . However, by having the limit value delta, it is possible to effectively avoid the underestimation described above. As a result of the simulation in the embodiment of the present invention,? = 0.9 is preferable. In addition, the expectation number of RM cell arrival of VC in W seconds, It should be noted that the sum of these values over the W second time interval gives an accurate estimate of the number of local bottleneck VCs since the value of the indicator function is normalized by the value of the indicator function. An iterative estimate based on this estimate for each time interval is calculated as: < EMI ID = 37.0 >

In Equation 37, [lambda] is an average factor, int [a] represents the lowest positive number of a or more, and the saturation function sat _a [b] is defined by the following equation (39).

For all t . The actual number of local bottlenecked VCs may be " 0 " for any networking load, but in the embodiment of the present invention, to avoid dividing by a value close to " 0 & .

Now the problem is how to choose the interval W and the mean factor λ. As the number of VCs sharing one link increases or as the available bandwidth decreases for a given interval W and the arrival wait time of the RM's VCs increases, the switch will allocate a smaller number of VC's RM cells So that eventually | Q | The estimate of _l starts to fluctuate greatly. To solve this problem, it is also possible to adjust W according to the number of VCs sharing the links and changes in the available bandwidth. But this is not easy to implement. Instead, the periodic averaging operation in equation (37) It is possible to select a large? At a value close to 1, while expecting to eliminate the variability of? In the embodiment of the present invention, through the simulation, the number of VCs sharing a wide range of one link, Was 0.98 regardless of the selection of the interval W. [0060]

The inventors of the present invention have found that, - to further improve the estimation algorithm. It is assumed that controller gains A and B are selected as shown in Equation (40).

The (U, V) pair exists in the stable region shown in FIG. - the estimates are sufficiently accurate that for all times The selection of the controller gains A, B is made. However, , The actual (or effective) values U 'and V' of U and V that control the normalized closed-loop equation, ie, Equation 27, are given by Equation 41 by substituting Equation 40 in Equation 26:

Consider the stable region shown in FIG. if (U ', V') exist on the straight line connecting the point (U, V) and the origin (0, 0), and thus exist in the stable region. Contrary, (U ', V') moves upward along a straight line including the origin (0, 0) and the point (U, V), resulting in a deviation from the stable region. Can be allowed to change only the asymptotic decay rate without affecting the stability of the system, Underestimation of the system should be avoided as it makes the system unstable. The above is the reason that the threshold value delta is necessary and introduced in the above equation (36). However, δ does not solve all situations, especially in situations where the new ABR VC participates in a small initial cell rate (the initial cell rate in the ATM ABR terminology). If the new VC participates in a small initial cell rate, it is very likely to be computed as a remote bottleneck VC during the transient period. This is because, if the limit? Is not sufficiently small, it is very likely that CCR (t ^j ) - MCR (t ^j ) in (36) is smaller than? R (t ^j ). In this situation, if the new VC is actually a remote bottleneck VC, this is not a calculation mistake, but if the new VC is actually a local bottleneck VC An initial underestimation of the system can be made to dissipate the system. In order to solve this problem, in the embodiment of the present invention - The estimation algorithm is improved as follows:

As soon as a new VC arrives at time t, it deliberately updates the VC as Equation 42 and Equation 43 as if it were a local bottleneck VC.

By doing so, The initial underestimation of From Equations (42) and (43) . This is a very desirable characteristic for system stability as described above. In the section " (6) Results of simulation " to be described later, it is assumed that, through simulations, - Verify the performance of the estimation.

Next, in the embodiment of the present invention, it is considered that RR (Relative Rate) marking is applied in relation to ER marking. While ER marking plays an important role in achieving MAX-MIN fairness as well as MCR assurance, RR marking plays an auxiliary role in suppressing transient overshoot of possible queue lengths (i.e., Minimizing losses).

Figure 9 shows a state in which queue length thresholds for such RR and ER marking are set. In FIG. 9, q _T is the target queue length, q _LT is the low threshold of the ABR queue, and q _HT is the high threshold of the ABR queue 36. Where q _T is related to ER marking, and q _LT and q _HT are related to RR marking. The threshold q _HT is a threshold for setting CI bits in the CI field of FIG. 2, and Q _LT is a threshold for setting NI bits in the NI field of FIG. If the queue length of the ABR queue 36 is greater than the threshold q _HT , the switch sets the CI bit of the CI (Congestion Indication) field in the reverse RM cell to "1" to indicate an excess of the queue length. If the queue length of the ABR queue 36 is between the thresholds q _LT and q _HT , the NI bit of the NI (No Increase) field in the reverse RM cell is set to "1" to indicate that the source should not increase the transmission bandwidth any more Set. In the embodiment of the present invention, if the target queue length q _T is set to be much smaller than the low threshold value q _LT , if the network load does not change suddenly, the queue length may stay near q _T by control of ER marking , The RR marking is hardly activated. If the queue length exceeds q _{HT due} to a sudden change in load, in embodiments of the present invention, the linear increase and exponential decay modes are initiated and the linear increase and decrease until the ER marking can again control the queue length. Enables exponential decay mode.

In the "(6) simulation results" section, the RR marking effectively limits the worst queue length in the transient period through simulations, and at the same time the ER marking performs the actual control again to return the queue length behavior behavior, To an asymptotic stable mode.

(6) Simulation results

In the following, simulation results are presented to confirm the above-described analyzes and to test the superior performance of the algorithm according to the embodiment of the present invention. The simulation model used was developed on the NIST ATM simulator platform.

We considered two different network topologies: peer-to-peer configuration and parking lot configuration, which are fair standards. Table 1 summarizes the recommended values for the design parameters of the algorithm proposed by the present invention and uses these variables in the following simulation studies.

ER assignment algorithm AB q _T T Estimation algorithm W δ λ 250 cells 0.9 0.98

( = max { , }, = One cell transmission time)

First we consider a peer-to-peer configuration. FIG. 10 shows a diagram of a peer-to-peer model including a VBR source. In Fig. 10, s1, ..., s20, and vbr1 are the sources for transmitting data, and d1, ..., d20, and dvbr1 are the destinations for receiving the transmitted data. Among the sources, vbr1 is the VBR source, and dvbr1 among the destinations is the VBR destination.

Referring to the peer-to-peer configuration of FIG. 10, the 20 ABR VCs have the same path, and each of the links has the same capacity set at 600 Mbps. In order to implement the WAN environment, the distance between each of the sources s1, ..., s20 and the first switch SW1 in Fig. 10 is set to 1,000 Km. Assuming that the signal propagation speed is 2.5 × 10 ⁵ Km / sec and the queuing time is ignored, Is approximately 10 msec. The VC models used in the simulation configuration of FIG. 10 are summarized in Table 2 below, and all sources are persistent. In the following Table 2, the units in PCR, MCR, ICR, arrival time, and departure time are Mbps.

Source PCR MCR ICR Arrival Time Departure Time Aperture ratio 0 to 1 1 to 2 2 to 3 3 to s1-s4 150 0 10 0 ∞ 36.1 33.3 30 32.5 s5-s9 150 10 10 0 ∞ 46.1 43.3 40 42.5 s10 150 0 10 2 ∞ 30 32.5 s11-s14 25 0 25 0 ∞ 25 25 25 25 s15-s19 25 10 25 0 ∞ 25 25 25 25 s20 25 0 25 1 3 25 25

It should be noted that the PCR values, MCR values, ICR values, and the arrival and departure times of the VCs are changed to examine the impacts of PCR-restricted sources, differences in MCR and ICR, call activity on network performance. For purposes of comparison, we compute the theoretical fairness ratios satisfying the MCR guarantee as well as the MAX-MIN fairness, and include the results in Table 2. Referring to Table 2, it can be seen that the fairness ratio of each VC is changing at the corresponding time according to the arrivals and departures of other VCs. The sources s1-s10 are bottlenecked in the switch SW1, and the sources s11-s20 are bottlenecked by their PCR restriction. For example, for the maximum-minimum fairness, ABR sources s1-s4 are set to 36.1 Mbps in 0 to 1 second, 33.3 Mbps in 1 to 2 seconds, 30 Mbps in 2 to 3 seconds, The data must be transmitted at 32.5 Mbps. In the present invention, 32 data cells are generated between two adjacent omni-directional RM cells. That is, NRM = 32.

11A to 11D show experimental results in a peer-to-peer model configuration with only ER marking and no VBR background traffic, and FIGS. 12A to 12D show experimental results in a peer to peer model configuration with ER marking and VBR background traffic Experimental results are shown.

11A shows the source rate a _i (t) of a VC with a PCR of 150 Mbps, and Fig. 11B shows a source rate a _i (t) of a VC with a PCR of 25 Mbps. When the actual source rate shown in FIG. 11A and FIG. 2, which is the same as the theoretical equilibrium rate. The transmission rates of the sources s1-s4 and s10 coincide with the common ER r (t) calculated at the switch SW1 because their MCR is 0 Mbps and the transmission rates of the sources s5-s9 are 1 Mbps larger than the common ER since their MCR is 10 Mbps . And the sources s11-s20 are independent of their MCR value and are limited by their PCR. The initial transient characteristics are due to the initial condition that the ER value r (0) = 0 in both switches SW1 and SW2. That is, it takes time for the common ER value to jump to the operating point, but this is a phenomenon that hardly occurs during normal operation. FIG. 11C shows the queue length at the bottleneck node SW1. A join of source s20 in one second and source s10 in two seconds results in a surge of queue length and a leave of source 20 in three seconds causes a sudden drop in queue length, ≪ / RTI > However, according to the embodiment of the present invention, the proposed algorithm is quickly re-stabilized at the queue length q _T = 250 cells. 11D, in the switch SW1, the estimated number of local bottlenecks | Q | _avg (t). By computing an estimate of the smallest positive integer equal to or greater than the estimate, we can obtain the integer estimate | Q | The following "[x]" means "x =" and "y)" means "y>". As can be seen from FIG. 11D, the integer estimates are 18 at [0, 1) seconds, 9 at [1,2, ) Is 18 in the second. Note that, except for the initial time interval [0, 1) second, the integer estimate is 9, [2, ) Seconds to 10 | Q | And is completely consistent with the truth value of. The difference in the time interval [0, 1) seconds is calculated by multiplying the estimate | Q | _avg (t) is intentionally increased. In a given simulation scenario, we reach 18 VCs at 0 s, This is because _avg (t) initially has 18 degrees.

For VBR background traffic, the results do not change vertically as shown in Figures 12A-12D. That is, the macroscopic (time-averaged) operating characteristics are almost the same as those with non-VBR traffic disturbance. VBR background traffic has a peak rate, a length of on and off periods of 10 Mbps, 20 Seconds and 20 Off < / RTI > 12A and 12B, the transmission rates of the sources (excluding the PCR restriction source) almost completely follow the ON / OFF pattern of the VBR background traffic. The on / off behavior characteristic of VBR background traffic causes repetitive surge and drop of ABR queue length. However, the ER allocation algorithm according to the embodiment of the present invention allows the queue length to quickly recover to the target value as shown in FIG. 12C.

Next, considering the parking rack configuration shown in FIG. 13, a case of a plurality of bottleneck nodes and VCs having different round trip delays will be described below. 16 ABR VCs with different source positions are included and the performance of the links is set equal to 600 Mbps except that the link between switch SW3 and SW4 and the link between SW4 and SW5 are 300 Mbps. The VC models used in this simulation configuration can be summarized in Table 3. Here all of the sources are assumed to be persistent.

Source PCR MCR ICR Arrival Time Departure Time Fair rate Bottle Location s1, s5, s9 150 0 10 0 ∞ 21.67 SW3 s13 150 0 10 0 ∞ 120 SW4 s2, s6, s10 150 10 10 0 ∞ 31.67 SW3 s14 25 0 25 0 ∞ 130 SW4 s3, s7, s11 25 10 25 0 ∞ 21.67 SW3 s15 25 0 25 0 ∞ 25 PCR s4, s8, s12, s16 25 10 25 0 ∞ 25 PCR

For purposes of comparison, the present invention calculates MCR guaranteed maximum-minimum fairness theoretical fairness rates for a given scenario, and the results are included in Table 3. To further clarify this scenario, we included the theoretical bottleneck locations for each VC in Table 3. The bottleneck position is at a position where each individual fair share is judged. 14A to 14F show simulation results when there is no VBR background traffic and only ER marking is present. 14A to 14F are experimental results in a configuration of a parking lot model with only ER marking and no VBR background traffic. FIG. 14A shows the source transfer rate a _i (t) of the VC with a PCR of 150 Mbps, the source rate a _i (t) of the VC, Figure 14c is a queue length of the switch SW3, Fig. 14d is a queue length of the switch SW4, Figure 14e is an estimate of the number of local bottleneck VC at the switch SW3 | Q | gt; _avg (t), < / RTI > FIG. 14F shows the number estimate Q | _avg (t). 14A and 14B, the actual source transfer rates in steady state are completely consistent with the theoretical fairness ratios given in Table 3, regardless of their round trip time delays and bottleneck locations. The initial transient characteristic is due to the initial condition of r (0) = 0 at all switches, which is a phenomenon that can hardly occur during normal operation. In the given scenario there are two congested nodes SW3 and SW4. The expected queue length at these congested nodes SW3 and SW4 converges to the target value 250 cells as shown in Figs. 14C and 14D. 14E and 14F show the number estimate Q of the local bottleneck VC in each of the switches SW3 and SW4. _avg (t). As a result of our observations, the estimates for SW3 and SW4 in steady state remain near 9 and 2, respectively. This is consistent with the data in Table 3.

Finally, when the RR marking is applied together with the ER marking, the results according to the embodiment of the present invention will be described below. We set the queue length thresholds for RR marking to q _T = 250 cells, q _LT = 750 cells, q _HT = 1000 cells. This choice of queue length thresholds is made by predicting that the ER marking plays a key role in ensuring asymptotic normal maximum-minimum flow control, while the RR marking is expected to play an additional role in limiting overshooting of the queue length will be. In the embodiment of the present invention, the same parking lot configuration scenario having VBR background traffic is used. The VBR background traffic has a peak rate, a length of on and off periods of 10 Mbps, 20 Seconds and 20 Off < / RTI > The results are shown in Figs. 15A to 15F. 15A to 15F are experimental results in a configuration of a parking lot model with ER marking and VBR background traffic. FIG. 15A shows the source rate a _i (t) of a VC with a PCR of 150Mbps, FIG. 15B shows a VC with a PCR of 25Mbps the source of the transmission rate a _i (t), Fig. 15c is a queue length of the switch SW3, Fig. 15d is a queue length of the switch SW4, Figure 15e is an estimate of the number of local bottleneck VC at the switch SW3 | Q | _avg (t), and FIG. 15F shows the number estimate Q of the local bottleneck VC at switch SW4. _avg (t). Compared with the case of FIGS. 14A to 14F, the overshoot overshoot of the ABR queue length in SW3 actually dropped from 6000 cells to 3200 cells. Obviously this drop is due to the activation of additional RR marking. However, these gains are paid for temporary vibrations of the queue length and the source transmission rate, as can be seen in Figs. 15A, 15B and 15C. We have observed that the ER allocation algorithm according to the embodiment of the present invention can be effectively controlled and that the queue length behavior behavior is recovered from the oscillatory mode to the asymptotic mode.

While the present invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments. Therefore, the scope of the present invention should not be limited by the described embodiments but should be determined by the equivalents of the claims and the claims.

As described above, the present invention minimizes the requirement of the ABR queue size by ensuring maximum link utilization and minimum cell loss regardless of the round-trip time delay of the ABR loop and ensuring the asymptotic stability of the ABR queue. Also, the present invention guarantees the fairness of the bandwidth utilization among ABR users, in particular, ensures the maximum-minimum fairness as an ATM forum standard, and adapts quickly to changes in the network environment such as the number of ABR users and the ABR bandwidth. It also performs almost all of the functions provided in the ATM Forum traffic management standard, including functions such as EFCI marking, RR marking, and ER marking, and provides an asymptotic stabilization operating point for high availability, low cell loss, (MAX-MIN fair rate) assignment is achieved. The present invention also improves responsiveness and control transient performance for multiple time scales, i.e., cell level rate changes of VBR and ABR VCs, and network load changes at cell level arrival and departure of VBR and ABR VCs, Minimizing the number of operations needed to reduce the complexity of the implementation and eliminating per-VC behavior, including VC-specific queuing, VC-specific computation, and VC-specific table accesses.

Claims (1)

Wherein the packet switching network includes a plurality of nodes connected to each other, each node having a data queue connecting a plurality of sources for transmitting and receiving data and storing data transmitted from the source, And a method for fair flow control in the packet switched network, the method being related to a target value, Estimating the number of local bottleneck virtual circuits using the data rate of the corresponding source and the minimum data rate to be guaranteed; Allocates a data rate to each of the sources based on a difference between the current queue length and a target value, a differential value of the current queue length, and a number of the estimated local bottleneck virtual circuits, And a feedback signal transmitted through the feedback signal.