RU121978U1 - SYSTEM OF COMMUTATION OF MULTIMEDIA TRAFFIC BASED ON DYNAMIC PRIORITIES - Google Patents

Abstract

A multimedia traffic switching system based on dynamic priorities, containing a memory for traffic packet queues, a switching matrix and a packet switching scheduler, while the packet switching scheduler is connected to the switching matrix and memory for traffic queues, characterized in that it contains a unit for measuring and comparing traffic parameters, a calculation unit and a queue control unit, while the traffic parameters measurement and comparison unit is connected to the memory for queues of traffic packets, as well as to the calculation unit, while the calculation unit is connected to the queue control unit, and the queue control unit is connected to the switching matrix.

Description

The utility model relates to communication technology, namely, packet switched systems, and is intended for use in building heterogeneous traffic switching systems in multiservice communication networks.

One of the problems in the functioning of the switching equipment of multiservice communication networks is the discrimination of low-priority traffic in conditions of high load on the switching system. (Stallings V. Modern computer networks. - St. Petersburg: Peter, 2003. - 783 p.).

One of the ways to increase the volume of served low-priority traffic in conditions of high load on the switching system is to apply service discipline with dynamic priorities. (Kleinrock L. Computing systems with queues. - M.: Mir, 1979. - 600 p.).

A device with a modified circuit and method of transmitting traffic according to the cyclic discipline of service, containing two ports for incoming traffic, serving device and control system (US patent No. 009073968, IPC H04L 12/50 from 03/19/2009). This device eliminates discrimination of low-priority traffic by allocating a certain bandwidth to each source according to priority.

The disadvantage of this analogue is the static distribution of bandwidth between traffic sources. Under certain operating conditions, this drawback does not allow optimal use of the switching system resources for serving low-priority traffic. (Bronstein OI, Dukhovny IM Models of priority service in information-computing systems. - M.: Nauka, 1976. - 220 p.).

There is also a known method of servicing traffic from multiple sources with different service priorities (patent EP No. 1648125, IPC H04L 12/56 of 04/19/2006). This method is based on a combination of two service disciplines: a priority discipline that provides special services according to the level of service and a cyclical service discipline in which the queue selector guarantees a certain throughput according to the weight of the queue. This method also allows you to eliminate discrimination of low-priority traffic by assigning the weight of the queue and cyclic maintenance of each of them according to the assigned weight.

The disadvantage of this analogue is the invariability of the weight of the queue assigned by the administrator. This can lead to inefficient use of the bandwidth resource to serve low-priority traffic in case of changing operating conditions of the serving device. (Bronstein OI, Dukhovny IM Models of priority service in information and computing systems. - M .: Nauka, 1976. - 220 p.).

The closest in technical essence and functions performed analogue (prototype) to the claimed multimedia traffic switching system based on dynamic priorities is the method and device of the packet scheduler (US patent No.2011158091, IPC H04L 12/56 from 06/30/2011). This method includes: buffering the received packet into the corresponding virtual output queue according to the class of service; determination of the weight of each queue in accordance with the number of packets contained in it; servicing each queue on the basis of planning in several stages according to its weight and quality classes. At the first stage, service planning is carried out in strict priority mode, and at the second stage, cyclic servicing of all queues is performed.

With such a combination of the described actions, adaptation of the distribution of resources to changing conditions of the functioning of the service system is achieved by automatically determining the weight of the queue depending on the degree of load with relatively low hardware costs.

However, the prototype device has a drawback: under certain conditions, discrimination of low priority traffic is possible due to the fact that when servicing queues in strict priority mode, the quality of service indicators for the traffic of low priority queues do not meet the requirements in the presence of available bandwidth resources.

The objective of this utility model is to develop a multimedia traffic switching system based on dynamic priorities, which allows to increase the volume of low-priority traffic served by timely changes to the service discipline.

This problem is solved in that the multimedia traffic switching system based on dynamic priorities, comprising a memory for traffic packet queues, a switching matrix and a packet switching scheduler, wherein the packet switching scheduler is connected to the switching matrix and memory for traffic queues, characterized in that it contains a block measuring and comparing traffic parameters, a calculation unit and a queue management unit, wherein the unit for measuring and comparing traffic parameters is connected to a memory for traffic packet queues and as well as calculation unit, wherein the calculation unit is connected to the queue control block and the queue control unit is coupled to a switch matrix.

The listed new set of essential features provides the opportunity to increase the volume of low-priority traffic served, since the additional blocks entered in the scheduler block periodically monitor the parameters of incoming and served traffic according to the service classes, calculate the quality of service traffic and manage the queue depending on the calculation data and quality requirements service. Thus, all streams of multimedia traffic entering the system are provided with special service according to the service classes, discrimination of low-priority traffic flows is not allowed, and the discipline of service is automatically adapted to the changing conditions of the service system functioning, which makes it possible to use service system resources in the most efficient way and thereby ensure greater volume of served low-priority traffic.

The analysis of the prior art made it possible to establish that analogues that are characterized by a combination of features that are identical to all the features of the claimed technical solution are absent, which indicates the suitability of the utility model for the patentability condition of “novelty”.

The claimed utility model is illustrated by drawings, which show:

figure 1 is a General structural diagram of a switching system for multimedia traffic based on dynamic priorities;

figure 2 - diagram of the algorithm of the input blocks of the scheduler.

The dynamic priority multimedia traffic switching system shown in FIG. 1 comprises a memory for traffic packet queues 1, a switching matrix 2, a packet switching scheduler 3. Moreover, a packet switching scheduler 3 is connected to a memory 1 and a switching matrix 2. A packet switching scheduler 3 contains a unit for measuring and comparing traffic parameters 4, a calculation unit 5 and a control unit for queue 6. A unit for measuring and comparing traffic parameters 4 is connected to a memory for queues of traffic packets 1 and to a unit of races 5. etov calculation unit 5 is connected with the queue control unit 6, and a queue management unit 6 connected to the switching matrix 2.

The memory for the queues of traffic packets 1 is designed for online storage of incoming multimedia traffic packets and can be performed on the basis of the Samsung K4D263238F-QC50 chip (Encyclopedia of electronic components. Large integrated circuits // E. Bogatyrev et al. - 2006. - С .224.).

Switching matrix 2 is designed for switching incoming traffic to one of the outputs of the switching system and can be performed on the basis of the Atheros AR8316 chip (Encyclopedia of Electronic Components. Large Integrated Circuits // E. Bogatyrev et al. - 2006. - P.224 )

The packet switching scheduler determines the order of queuing, so that the throughput of the switching system and the efficient use of resources of the switching system are maximized. An additional task of the packet switching scheduler is to eliminate discrimination of low-priority traffic. The unit for measuring and comparing traffic parameters, introduced into the packet switching scheduler, collects data on incoming and served traffic, calculates the statistical characteristics of traffic and compares the quality of service parameters of the service system with the required values. The calculation unit introduced into the packet switching scheduler calculates the main indicators of the quality of the service system based on the data received from the measurement and comparison of the traffic parameters and, in accordance with this, calculates the control actions on the switching matrix, which satisfy the quality of service requirements for the incoming traffic. The queue control unit introduced into the packet switching scheduler, on the basis of the data received from the calculation unit, produces control actions on the switching matrix. The introduced unit for measuring and comparing traffic parameters, the calculation unit and the control unit determine the logic of the switching system and can be implemented in the form of software modules, the logic of which is presented in figure 2. The scheduler can be executed on the basis of the FWIXP420BB (L4520230) microcircuit manufactured by Intel (V. Broydo, O. Ilyina. Computer and system architecture. - SPb .: Peter, 2009. - 720 p.).

The industrial applicability of the utility model is due to the fact that it can be implemented using the modern elemental base, with the achievement of the destination specified in the utility model.

In the well-known work (Kleinrock L. Computing systems with queues. - M .: Mir, 1979. - 600 pp.), A generalized formula is obtained for the average waiting time for servicing packages of the r-th priority for the discipline of s-order servicing:

where ρ _R is the total load of the switching matrix;

ρ _r is the total load of the switching matrix with priority packets not higher than r;

µ _R is the service intensity of the total packet stream;

b _r is a coefficient that determines the priority rise rate for packets of the corresponding classes of service (b ₁ ≥b ₂ ≥ ... ≥b _r ≥0).

As s → 0, we obtain a service system in the order of priorities that do not change in time — relative fixed priorities. For s = 1, we have a system with priorities that linearly change in time - relative dynamic priorities.

In order for all streams of multimedia traffic entering the system to provide special services according to the classes of service, to prevent discrimination of low-priority traffic flows and to automatically adapt the discipline of service to the changing operating conditions of the service system in the switching system, it is necessary to manage the queue, which will perform the following main functions:

- periodic monitoring of key indicators of service quality (delays, packet loss of various priority classes);

- automatic decision-making on changing the discipline of service depending on the load from the incoming traffic and the compliance of the quality of service indicators with the established standards (change in the control parameter s);

- recalculation of the variable (b _{r + 1} / b _r ) ^{1 / s} (relative rate of change of priority) by changing the parameters b _{r + 1} , b _r , s, in order to regulate the average waiting time for the service (delay) to maintain performance indicators within the established standards.

The traffic coming to the input of the switching system can be divided into two parts: traffic that is sensitive to delays (high priority) and traffic that is not sensitive to delays (low priority). Therefore, all traffic can be divided into two priority classes R = 2. Moreover, taking into account the requirements of the quality of service, the packet delay of high-priority traffic t _pp (1) (first priority) should not exceed the required value and the packet delay of low-priority traffic t _pp (2) (second priority) should not exceed .

The average delay of packets of the r-th priority of the switching system with dynamic priorities (DP) consists of the average waiting time for service and the average service time x _r , which allows you to calculate the required waiting time for the service, since the average service time of the packets is calculated:

At the same time, on the basis of the Little formula, the average waiting time for servicing packets of the rth priority is:

Where - average packet queue length of r-th priority; λ _r is the arrival rate of packets of the rth priority.

Based on the "law of conservation of work accumulated in the queue" and the method of diffusion approximation (KI Sychev Multi-criteria design of multi-service communication networks. - St. Petersburg: Publishing House of Polytechnic University, 2008 - 272 s.) The average length of the queue of elastic packets traffic (r = 2) in a switching system of type with dynamic priorities is defined as:

Where - the average queue length for the total priority stream is not higher than r;

, - respectively, the quadratic coefficients of variation for the distribution of the time of receipt and maintenance of packets of the r-th priority;

- the average packet queue length of the r-th priority in the switching system of type with dynamic priorities, defined as .

Average waiting time for service with r-priority packets in a switching system of type with dynamic priorities - is determined by the solution of a system of equations (Kleinrock L. Computational systems with queues. - M .: Mir, 1979. - 600 p.). The solution of the system for the case R = 2:

Based on the results obtained (5) and (6), using the Little formula (3), as well as the formula (4), we obtain the average service waiting time .

The main purpose of the above calculations is to calculate the ratio b ₂ / b ₁ at which the delay value of packets of low priority traffic will meet the requirement . As a result, we get an equation with one unknown - the variable b ₂ / b ₁ , which is calculated as a result of the simplest linear mathematical operations.

After that, it is necessary to determine the delay of high-priority traffic packets at the obtained ratio b ₂ / b ₁ on the basis of a similar sequence of calculations in order to verify the condition .

In case of discrepancy of probability of packet loss of the r-th priority desired value it is necessary to recalculate the required amount of buffer memory for the packet queue taking into account the current operating conditions of the switching system.

To do this, the calculation unit calculates the amount of buffer memory for the packet queue (Kleinrock L. Computing systems with queues. - M.: Mir, 1979. - 600 p.):

Where - requirements for the probability of packet loss.

Expression (7) for calculating the required amount of buffer memory takes into account current parameters of incoming traffic, such as load and variability determined by quadratic coefficients of variation of the distribution of time between packet arrivals and packet service time for individual flows by priority, as well as for total priority traffic not higher , .

The inventive device operates as follows.

Packets of multimedia traffic enter the switching system and exit the system through the corresponding interfaces 1 ′. The data of incoming and served traffic (the duration of time between packets and the duration of the packets) via 1 ″ communication is sent to the unit for measuring and comparing traffic parameters. Characteristics of incoming and served traffic (figure 2) are calculated by the unit for measuring and comparing traffic parameters 4 (figure 1). In case of mismatch of the measured indicators of the quality of service required (Fig. 2), the unit for measuring and comparing traffic parameters 4 transfers the characteristics of the incoming traffic to the calculation unit 5 (communication 4 ′, Fig. 1).

The main task of calculation block 5 is to calculate the main indicators of the quality of the switching system based on the initial data on incoming traffic and the calculation of the control parameters of the service discipline at which the requirements for the quality of service of incoming traffic are met (figure 2).

Calculation block 5 compares the calculated delay of packets of the first priority in the switching system with dynamic priorities delayed (figure 2). When the calculation unit 5 transmits the following parameters to the control unit 6 (communication 5 ′, FIG. 1):

- s = 1 - control parameter indicating the transition from fixed priorities to dynamic;

- variable - determines the relative rate of change of the priority function, i.e. the attitude shows how much faster the high-priority traffic (r = 1) is increasing its priority in relation to the low-priority traffic (r = 2).

In the future, the calculation unit 5 to the control unit of the queue 6 performs periodic (in the case of a service discipline with dynamic priorities) transfer of the calculated control parameters (communication 6 ′, Fig. 1): s and , as well as the calculated value of the required amount of buffer memory taking into account the current operating conditions of the switching system.

The control unit queue 6 based on the obtained control parameters determines the discipline of service (figure 2):

- makes the transition to service discipline with dynamic priorities s = 1;

- determines the rate of increase of priority of packets of continuous b ₁ and elastic traffic b ₂ ;

- changes the amount of buffer memory K _r according to the calculated required value taking into account the current operating conditions of the switching system;

- carries out a return to the discipline of service with fixed priorities (s = 0) in case of compliance with the quality of service indicators of packets of continuous and elastic traffic.

As a result of the operation of the switching system based on queue management, high priority traffic sensitive to delays is provided with special service and at the same time blocking of low priority traffic is not allowed. This is achieved by timely changes to the service discipline according to the calculated control parameters (s, b ₁ , b ₂ , K _r ), which leads to a redistribution of switching system resources from high-priority to low-priority traffic, provided that the quality of service is ensured first. This is the mechanism for automatically adapting resource allocation to changing network conditions, which allows you to serve a larger volume of low-priority traffic. The results of analytical calculations, as well as the simulation performed for various requirements for packet delay and load on the switching system show that the volume of served low-priority traffic increased from 3% to 36% under various modeling conditions.

Claims (1)

A dynamic priority multimedia traffic switching system comprising a memory for traffic packet queues, a switching matrix and a packet switching scheduler, wherein the packet switching scheduler is connected to the switching matrix and memory for traffic queues, characterized in that it comprises a unit for measuring and comparing traffic parameters, a calculation unit and a queue management unit, wherein the unit for measuring and comparing traffic parameters is connected to a memory for the queues of traffic packets, as well as with the calculation unit in, wherein the calculation unit is connected to the queue control block and the queue control unit is coupled to a switch matrix.