CN101212417A - A Time Granularity-Based Internet Service Quality Assurance Method - Google Patents

Abstract

The invention discloses a method for guaranteeing Internet service quality based on time granularity. By adding a delay control extension header for recording the state information of the data packet in the header of the real-time service data packet, the edge router sets the real-time The relative time limit identifier in the service data packet, the core router assigns the data packet to delay queues of different granularities according to the relative time limit value of the received data packet, and sends the data packets in these queues sequentially; the edge router sends the real-time service test The data packet method measures the congestion level of the network to make a decision on whether to accept a new call. The invention can not only better guarantee the end-to-end delay of the data packet, but also has very low computational complexity in the core router, and has strong expansibility while improving the service quality.

Description

A Time Granularity-Based Internet Service Quality Assurance Method

technical field

The invention relates to a method for guaranteeing service quality of a packet switching network, in particular to a method for guaranteeing Internet service quality based on time granularity, and belongs to the field of data communication.

Background technique

At present, the Internet begins to provide real-time multimedia services such as video telephony, video conferencing, and IPTV. These services require low network delay, but the Internet still lacks an effective service quality assurance solution. The Internet first attempted to adopt the integrated service method to provide service quality assurance for real-time services. It used the Resource Reservation Protocol (ResourceReSerVation Protocol, RSVP, reference: R. Braden, Resource ReSerVation Protocol (RSVP), RFC 2205, September 1997) as the call Signaling, weighted fair queuing (Weighted Fair Queuing, WFQ, reference: Abhay K. Patekh, "A Generalized Processor Sharing ApproachFlow Control in Integrated Services Networks: The Single-Node Case", IEEE/ACM Trans.on Network, Vol 1 , No.3, June 1993.) and other queue scheduling algorithms are used as the control means of the data plane to provide each session flow with a quality of service not lower than the reserved level. Since RSVP adopts a soft-state call system, it needs to update and maintain state information regularly, which limits the number of calls it can hold. The computational complexity of the WFQ queue scheduling algorithm is 0(N), where N is the number of current calls. Although Smoothed Round Robin (SRR, reference: Chuanxiong Guo, SRR: An 0(1) Time Complexity Packet Scheduler for Flows in Multi-Service Packet Networks, IEEE/ACMTransactions on Networking, Volume: 12, Issue: 6 On page (s): 1144-1155, Dec.2004) has a computational complexity of 0(1), but this method still needs to filter each session flow and save the state of each session flow. In order to overcome the scalability problem of integrated services, the method of differentiating services is widely used at present. The idea of differentiated services is to establish a differentiated services implementation domain, divide the data flow into several limited types at the boundary of the implementation domain, and only need to provide corresponding services according to the type of data packets on the core router of differentiated services That's it. This method does not need to maintain the state of each session flow on the DiffServ core router, so the processing efficiency is high and the computational complexity is low. Differentiated Services Although there are some call control methods based on bandwidth brokers (Bandwidth Broker, BB, reference: K. Nichols, A Two-bit Differentiated Services Architecture for the Internet, RFC2638, July 1999), there is no standardized call control protocol. On the data plane, although DiffServ has low computational complexity and provides better service quality than best-effort service, it still cannot provide service quality guarantee for real-time services. Global Earliest Deadline First (GEDF, reference: Matthew Andrews, Lisa Zhang, Minimizing End-to-End Delay in High-Speed Networks with a SimpleCoordinated Schedule, IEEE INFOCOM 1999) and core jitter virtual clock (Core-Jitter VirtualClock , CJVC, reference: Ion Stoica, Hui Zhang, PProviding, Guaranteed Services Without Per Flow Management, ACM SIGCOMM 1999) and other scheduling algorithms encode the delay characteristics of each data packet into the packet header, and the core router no longer keeps each session The state information of each data packet is sent sequentially according to the delay characteristics of each data packet. The computational complexity of this type of algorithm is 0(logN), where N is the number of data packets in the queue.

Contents of the invention

In view of this, the present invention is committed to providing a method for guaranteeing Internet service quality based on time granularity.

This method adopts the basic idea of differentiated services, divides the routers in this domain into edge routers and core routers, pastes a relative delay mark for each real-time service data packet on the edge router, and the core router mark, to control the queuing delay of the data packet, so as to ensure the end-to-end delay of real-time services.

The edge router saves the session information of each real-time service, uses the token controller to control the flow of the real-time service, and sets the relative time limit identifier in the real-time service data packet according to the current remaining token quantity. The relationship between the two is the current remaining token quantity The more, the smaller the value of the relative time limit. If the relative time limit value of the router receiving the data packet is positive, it means that the data packet arrives at the router ahead of time, otherwise, it means that the data packet arrives late. The greater the relative time limit of the data packet, the greater the tolerable queuing delay in the subsequent forwarding process. After a packet passes through a router, its relative timing changes. When the data packet is output from the core router, the relative time limit needs to subtract its queuing delay, plus the forwarding delay constant.

The core router takes the time shorter than the forwarding delay constant as the basic delay unit, and takes the basic delay unit or its integer multiple as the time limit span of the waiting queue. According to the relative time limit of the data packet entering the core router, the route of the data packet can be calculated Time limit: The core router distributes data packets according to the routing time limit of the data packets, and inserts them into the tail of the corresponding queue. When sending, the queue with the lowest routing time limit is sent first, and within each queue, the data packets are sent in the order in which they arrive at the router. The waiting queues established in the core router are divided into two different time granularities. N fine-grained queues use the basic delay unit as their time limit span, and M coarse-grained queues use N times the basic delay unit as their time limit span. N fine-grained queues receive data packets whose relative time limit ranges from 0 to N basic delay units, wherein queue R(n) is used to receive data packets whose routing time limit belongs to [(n-1)×u, n×u); M time-coarse-grained data packet waiting queues T(m) are used to receive ordinary real-time service data packets whose routing time limit is greater than U, and the queue T(m) is used to receive routing time limit belonging to [m×U, (m+1 )×U) packets. When the core router enters the congested state, it discards subsequent test data packets until the congested state is lifted. Every time n times the length of the basic delay unit, the data packets in the first coarse-grained queue are allocated to the fine-grained queue, and the rest of the coarse-grained queues are promoted in turn.

In terms of call control, at the output interface of each router, a certain bandwidth is allocated for real-time services, and a higher scheduling priority is provided for this part of the services. Before sending real-time service data, the edge router at the sending end sends a real-time service test packet to the edge router at the receiving end; when the edge router at the receiving end checks the received real-time service data packet, if it finds a real-time service test failure packet, it rejects the call ; Otherwise, establish the real-time service.

The present invention is achieved through the following technical solutions:

A method for guaranteeing Internet service quality based on time granularity, comprising:

1) Add a delay control extension header for recording packet status information in the header of the real-time service data packet;

2) The edge router forwards the received common real-time service data packet, and sets the relative time limit of the data packet;

3) The core router calculates the routing time limit of the data packet according to the relative time limit of the common real-time service data packet;

4) The core router sets up a data packet waiting queue according to the routing time limit of each common real-time service data packet;

5) The core router sends a common real-time service data packet and updates the relative time limit of the data packet.

The data structure of the delay control extension header is: the type of the next packet header, the length of the extension header, the version of the extension header, the test flag, the service level, and the relative time limit.

In the described method, the method that the edge router forwards the received common real-time service data packet and sets the relative time limit of each data packet is:

1) The edge router allocates a token controller for each real-time service, configures its maximum token depth Δ and token arrival speed α, and the maximum token depth is greater than the length of the largest data packet;

2) When an ordinary real-time service data packet with a length of L arrives at the edge router, if the current remaining token quantity δ is greater than L, then send the data packet, set its relative time limit as β=(Δ-δ)/α, and set The current remaining number of tokens is reduced by L; otherwise, the data packet is sent after the current remaining number of tokens δ is greater than L.

In the described method, the method that the core router sets up a data packet waiting queue according to the routing time limit r of each common real-time service data packet is:

1) Two kinds of time granularities are set in the core router, the fine-grained is the basic delay unit u, and the length of the coarse-grained U is N fine-grained, that is, U=N×u;

2) Divide the busy period into K update periods, the length of each period is U, and record the starting time B(k) of each period;

3) Calculate the routing time limit r of the common real-time service data packet arriving in the k update period, r=t-B(k)+β+σ, wherein t is the current moment, and β is the time when the data packet enters the core router Relative time limit, σ is forwarding delay constant;

4) Establish N time-fine-grained data packet waiting queues R(n) in the core router for receiving ordinary real-time service data packets whose routing time limit belongs to [0, N×u), wherein the queues R(n) are used for Receiving a data packet whose routing time limit belongs to [(n-1)×u, n×u); the routing time limit span of the time fine-grained data packet waiting queue is the basic delay unit u;

5) Set up M time-coarse-grained data packet waiting queues T(m) in the core router, for receiving common real-time service data packets whose routing time limit belongs to [U, (M+1)×U), wherein the queue T( m) for receiving a data packet whose routing time limit belongs to [m×U, (m+1)×U), and the relative time limit span of the time coarse-grained data packet waiting queue is U;

6) Establish a first queue F and a tail queue Z, the first queue F is used to receive packets whose routing time limit is less than zero, and the tail queue Z is used to receive packets whose routing time limit is greater than (M+1)×U;

7) The establishment sequence of the sending queue is queue F, N time-fine-grained data packet waiting queues R(n), M time-coarse-grained data packet waiting queues T(m), and tail queue Z.

The basic delay unit u is smaller than the forwarding delay constant σ.

In the described method, the method that the core router sends the common real-time service data packet is:

1) Send queue F, N time-fine-grained data packets waiting in queue R(n), M time-coarse-grained data packets in waiting queue T(m), and data packets in tail queue Z; restore the state after sending variable initial value;

2) When the update time B(k) arrives, the unsent data packets in the time fine-grained queue R(n) are sequentially added to the tail of the queue F according to the sending order;

3) When the update time B(k) arrives, the data packets in the first coarse-grained queue T(1) need to be allocated to the fine-grained queue, and the rest of the coarse-grained queues are promoted in turn, and the data packets in the queue Z Qualified data packets are added to the queue T(M).

In the described method, the method that the core router updates the relative time limit of the common real-time service data packet is:

1) The core router records the arrival time t_arv of the data packet, and the sending time t_snd of the data packet;

2) When the core router sends the data packet, the relative time limit is updated as β+σ-(t_snd-t_arv), where β is the relative time limit when the data packet enters the core router.

Described real-time service data packet comprises common real-time service data packet, real-time service test data packet and real-time service test failure data packet; In non-congested state, the process of real-time service test data packet and common real-time service data packet in core router is Same, but in the congestion state, the core router changes the real-time service test data packet into the real-time service test failure data packet, and places it at the end of the best-effort service queue; the core router forwards the real-time service test failure data packet by the best-effort service queue.

The establishment method of the real-time service is:

1) On the output interface of each router, allocate a certain bandwidth to reserve for real-time business, and set the sending priority for this part of the business;

2) Before sending real-time service data, the edge router at the sending end sends a real-time service test data packet to the edge router at the receiving end;

3) When the edge router at the receiving end checks the received real-time service data packet, if it finds a real-time service test failure data packet, it rejects the call; otherwise, it establishes the real-time service.

The route establishment method of described real-time service is: add a newly-built flag field in the routing table of router, when the routing entry of router changes, router sets the newly-built flag of this routing entry to be valid, after waiting for newly-built flag to be effective for long time τ, router It is invalid to update the newly-built sign of this route item; When the router outputs the common real-time service data packet, if the newly-created sign of the corresponding route item is valid, then the router sets the data packet as the real-time service test data packet.

The present invention has following technical effect:

1. The calculation complexity of the queue scheduling algorithm proposed in the present invention is low in the core router. In the core router, data packets are grouped into several limited queues according to the relative time limit value, among these queues, they are sent sequentially, and the computational complexity is only 0(1).

2. The calculation complexity of the queue scheduling algorithm proposed in the present invention is also low in the edge router, and it only needs to set the relative time limit of marking according to the token depth in the token controller. This method is simpler than the core dithering virtual clock method, and does not require edge routers to calculate the slack variables of each path.

3. The queue scheduling algorithm proposed by the present invention can control the queuing delay of each data packet in the router, thereby ensuring end-to-end transmission delay.

4. The queue scheduling algorithm proposed by the present invention has higher bandwidth utilization efficiency than the simple FIFO queue scheduling algorithm, and can provide larger guaranteed bandwidth for real-time services under the same network environment.

5. The measurement-based access control technology proposed by the present invention can ensure the end-to-end delay of coarse-grained queue scheduling.

6. The measurement-based access control technology proposed by the present invention does not need to maintain the state information of each session on the core router, and has strong scalability.

7. The present invention can better protect the service quality of real-time services when the network route changes; when the network resources are sufficient, it can provide good service quality for the real-time services whose transmission path changes; when the network resources are insufficient, it can Automatically suspend the real-time business whose transmission path changes, thereby ensuring the service quality of other real-time businesses.

Description of drawings

Fig. 1 is the internal structural diagram of the delay control extension header of data packet among the present invention;

Fig. 2 is a network device connection diagram of the present invention;

Fig. 3 is an internal structural diagram of the edge device of the present invention;

Fig. 4 is the internal queue scheduling structural diagram of the core equipment of the present invention;

Fig. 5 is a call control process diagram of the present invention;

Fig. 6 is a flow chart of the method of the present invention.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments.

The method flow chart of the present invention is shown in FIG. 6 , and this embodiment is based on the IPv6 network environment to realize the service quality assurance system.

protocol extension

In order to meet the requirements of service quality assurance, it is necessary to extend the existing IPv6 standard and add a delay control extension header to the IP packet header. Its value is tentatively set to 102, and the corresponding keyword is DLCT. After the edge router receives the user's real-time service data packet, it adds a delay control extension header to the packet header; when the real-time service data packet reaches the edge router at the peer end, the edge router deletes the delay control extension header and forwards it to the user at the peer end . As shown in Figure 1, the format of the extension header is as follows:

The first octet is the type of the next header;

The second octet is the length of the extension header, indicating the length of the token-level extension header, expressed in the number of octets, and does not include the first 8 octets; its value is 0;

The third octet is the version of the extension header (Message Version), its value is 1, indicating that the current version is 1, and other values are reserved;

The first four bits of the fourth octet are test marks, 0B1000 is a real-time service test data packet, 0B1100 is a real-time service test failure data packet, and 0B0000 is a common real-time service data packet; the last four bits are service level (Service Level) ), whose value is a positive integer, indicating the level of real-time services;

The 5th, 6th, 7th and 8th octets are relative deadlines (Relative Deadline), the value range is an integer, and the unit is 100 microseconds.

Fig. 2 is the structure of IPv6 network environment, the quality of service assurance system is distributed on edge routers and core routers. The system limits the basic delay unit u of each core router to 800 microseconds, the forwarding delay constant σ to 3 milliseconds, and limits the number of end-to-end forwarding hops to less than 16, so that the end-to-end queuing delay is guaranteed to be within 48 milliseconds.

queue scheduling

The service quality assurance method proposed in this patent adopts the idea of differentiating services, and divides the routers in the area into edge routers and core routers. The edge routers save the state information of each data flow, and encode the relative time limit information into the delay control of data packets. In the extension header, the core router does not need to maintain the state information of each session, but only needs to process the data packet according to the relative time limit information in the data packet header.

This method does not consider the line transmission delay and the number of hops in the data transmission path, these issues are the responsibility of the network planning and end systems. The method guarantees the end-to-end transmission delay of the data packet by limiting the queuing delay of the real-time service data packet in the core router to be smaller than the forwarding delay constant σ.

Due to the change of network load, the data packet may reach the next-hop router earlier or later, and this time difference β is called the relative time limit in this paper. A positive relative time limit means early arrival; a negative relative time limit means late arrival. For the data packets arriving earlier, the waiting time in the queue can be longer than the forwarding delay constant, but should be less than the sum of the forwarding delay constant and the relative time limit. The core router can improve the overall delay performance of the network system by allowing the data packets arriving earlier to wait in the queue for a longer time and allowing other data packets with a relatively shorter time limit to be sent first.

Assign a token controller to each real-time service at the edge router, as shown in Figure 3, configure its maximum token depth Δ, token arrival speed α, where the maximum token depth Δ must be greater than the length of the largest data packet.

When an ordinary real-time service data packet with a length of L arrives at the edge router, if the current remaining token quantity δ of the token controller is greater than the length L of the current data packet, the data packet is sent, and the relative time limit β for identifying the data packet is ( Δ-δ)/α, and reduce the current number of remaining tokens by L. Otherwise, wait for the current remaining token quantity of the token controller to be greater than the data packet length, and then send the data packet, identify the relative time limit β of the data packet as (Δ-δ)/α, and reduce the current remaining token quantity L.

When the data packet arrives at the core router, the core router records the arrival time t_arv of the data packet and the sending time t_snd of the data packet, and updates the relative time limit of the data packet when sending to beta+σ-(t_snd-t_arv), where β is the relative time limit when the data packet enters the core router;

As shown in Figure 4, 34 queues are established on the output interface of the core router, respectively F, R(1), R(2), ..., R(16), T(1), T(2) ,..., T(16), Z. When the value of the pointer PR is 1, 2, ..., 16, the current sending queue indicated is R(1), R(2), ..., R(16), and the value is 0, indicating that the pointer Invalid, the initial value of the pointer PR is 0. When the value of the pointer PT is 1, 2, ..., 16, the current sending queue indicated is T(1), T(2), ..., T(16), and the value is 0, indicating that the pointer Invalid, the initial value of the pointer PT is 0.

At the time of initialization, all 34 queues on the output interface of the core router are empty, the start time of the busy period is 0, and the end time of the busy period is E. The busy period is divided into K updates with a length of U (12.8 milliseconds) Time period, the starting time of K time periods is 0, B(1), B(2), ..., B(K-1).

When the ordinary real-time service data packet p arrives in the kth update period of the router, the router time limit r of the data packet is the current time t minus the start time B(k) of the kth update period, plus the data packet entering The relative timing of the core router time, plus the forwarding delay constant σ.

If the router time limit r is less than zero, the data packet is put into the queue F,

If the router time limit r is less than n×800 microseconds and greater than or equal to (n-1)×800 microseconds, then the data packet is put into the queue R(n); if the current pointer PR is zero, the update pointer PR is n ; If the value of the current pointer PR is not zero, and n is smaller than the pointer PR, then the value of PR is changed to n;

If the router time limit r is less than (m+1)×12800 microseconds and greater than or equal to m×12800 microseconds, then the data packet is put into the queue T(m); if the current pointer PT is zero, update the pointer PT to m ; If the value of the current pointer PT is not zero, and m is smaller than the pointer PT, then the value of PT is changed to m;

Otherwise, the packet is placed in queue Z;

When sending, the interface selects data packets from these 34 queues to send, and the selection method is as follows.

If the queue F is not empty, send the head data packet of the queue F;

If the queue F is empty and the pointer PR is not equal to zero, check whether the queue R(PR) is empty. If the queue R (PR) is not empty, then send the head packet of the queue R (PR); otherwise, add one to the pointer PR, and test whether R (PR) is empty again; if it is empty until R (16), then Set the pointer PR to zero;

When the queue F is empty, the pointer PR is zero and the pointer PT is not equal to zero, the queue T(PT) is checked. If the queue T(PT) is not empty, send the data packet of the queue T(PT); if the queue T(PT) is empty, add one to the pointer PT, and then check whether the queue T(PT) is empty. If it is empty until T(16), set the pointer PT to be zero;

When the queue F is empty, the pointer PR is zero and the pointer PT is also zero, check whether the queue Z is empty. If the queue Z is not empty, send the queue head data packet of queue Z; otherwise, the router ends the busy period, and both PT and PR become the initial value zero.

When the update time B(k) arrives, if there are still data packets in the queues R(1), R(2), ..., R(16) that cannot be sent, these data packets are added to the The tail of queue F. Then, the data packets in the queue T(1) are allocated to 16 time fine-grained queues R(1), R(2), . . . , R(16) according to the router time limit. And update the queues T(2), T(3),..., T(16) to queues T(1), T(2),..., T(15), and finally put the queues in the queue Z that belong to Some data packets of T(16) are moved to queue T(16), and both PT and PR are changed to initial value zero.

During the sending process, the limited time D(1), D(2), ..., D(16) of the queues R(1), R(2), ..., R(16) are 800, 1600, ..., 12800 microseconds, if D(i) time arrives, but R(1), R(2), ..., R(i) i queues still have unsent data packets, Then the core router enters the congestion state. After the core router enters the congestion state, when the time D(j) arrives, and the j queues R(1), R(2), ..., R(j) have no data packets waiting to be sent, the core router Enter the non-congested state. In the congestion state, the core router changes the test flag in the header of the real-time service test data packet to 0B1100, thereby changing the data packet into a real-time service test failure data packet and placing it at the end of the best-effort service queue; the core router passes the best-effort service Queues to forward real-time service test failure packets. In the non-congested state, the core router uses the same processing method as the common real-time service data packet to process the real-time service test data packet. The core router adopts the method of best-effort service to process the failed data packets of the real-time service test.

call control

On the output interface of each router, a certain amount of bandwidth is allocated for real-time services, and a higher scheduling priority is provided for this part of the services. The method adopts the RSVP protocol as the call signaling between the terminal and the edge router. As shown in Figure 5, before sending real-time service data, the sender first sends a Path message to the edge router, and establishes relevant state information at the edge router. The edge router at the sending end forwards the Path message to the edge router at the receiving end, and within 1 second thereafter, sends sequentially numbered real-time service test data packets to the edge router at the receiving end according to the resource characteristics required by Path.

The edge router at the receiving end checks the received real-time service data packet, and rejects the call if it finds a real-time service test failure data packet, otherwise, forwards the Path message to the receiving end. The edge router at the receiving end receives the Resv message returned by the terminal, and forwards it to the edge router at the sending end, and forwards it to the sending end. The sending end finally sends a confirmation message ResvConf to the receiving end, so that the call is established.

After the session ends, the terminal sends PathTear or ResvTear to release the resources reserved in the edge router.

route change

Add a new flag field in the routing item of the router. When adding or changing the routing item, set the new flag of the routing item as valid, and start the new flag valid timer. The length of the new flag valid timer is 1 second. After the flag valid timer expires, set the new flag to be invalid. When the router outputs a common real-time service data packet, if the new flag of the corresponding routing item is valid, the router sets the data packet as a real-time service test data packet. During the normal operation of the real-time service, if the real-time service test failure data packet is received, the real-time communication service will be terminated.

Claims (10)

1. A method for guaranteeing Internet service quality based on time granularity, the steps of which are: 1) Add a delay control extension header for recording packet status information in the header of the real-time service data packet; 2) The edge router forwards the received common real-time service data packet, and sets the relative time limit of the data packet; 3) The core router calculates the routing time limit of the data packet according to the relative time limit of the common real-time service data packet; 4) The core router sets up a data packet waiting queue according to the routing time limit of each common real-time service data packet; 5) The core router sends a common real-time service data packet and updates the relative time limit of the data packet. 2. The method according to claim 1, wherein the data structure of the delay control extension header is: the type of the next packet header, the length of the extension header, the version of the extension header, the test flag, the service level, and the relative time limit. 3. method as claimed in claim 2, it is characterized in that the common real-time business data packet that edge router forwards receives and the method that the relative time limit of each data packet is set is: 1) The edge router allocates a token controller for each real-time service, configures its maximum token depth Δ and token arrival speed α, and the maximum token depth is greater than the length of the largest data packet; 2) When an ordinary real-time service data packet with a length of L arrives at the edge router, if the current remaining token quantity δ is greater than L, then send the data packet, set its relative time limit as β=(Δ-δ)/α, and set The current remaining number of tokens is reduced by L; otherwise, the data packet is sent after the current remaining number of tokens δ is greater than L. 4. method as claimed in claim 3, it is characterized in that core router sets up the method for packet waiting queue according to the routing time limit of each common real-time service packet: 1) Two kinds of time granularities are set in the core router, the fine-grained is the basic delay unit u, and the length of the coarse-grained U is N fine-grained, that is, U=N×u; 2) Divide the busy period into K update periods, the length of each period is U, and record the starting time B(k) of each period; 3) Calculate the routing time limit r of the common real-time service data packet arriving in the k update period, r=t-B(k)+β+σ, wherein t is the current moment, and β is the time when the data packet enters the core router Relative time limit, σ is forwarding delay constant; 4) Establish N time-fine-grained data packet waiting queues R(n) in the core router for receiving ordinary real-time service data packets whose routing time limit belongs to [0, N×u), wherein the queues R(n) are used for Receiving a data packet whose routing time limit belongs to [(n-1)×u, n×u); the routing time limit span of the time fine-grained data packet waiting queue is the basic delay unit u; 5) Set up M time-coarse-grained data packet waiting queues T(m) in the core router, for receiving common real-time service data packets whose routing time limit belongs to [U, (M+1)×U), wherein the queue T( m) for receiving a data packet whose routing time limit belongs to [m×U, (m+1)×U), and the relative time limit span of the time coarse-grained data packet waiting queue is U; 6) Establish a first queue F and a tail queue Z, the first queue F is used to receive packets whose routing time limit is less than zero, and the tail queue Z is used to receive packets whose routing time limit is greater than (M+1)×U; 7) The establishment sequence of the sending queue is queue F, N time-fine-grained data packet waiting queues R(n), M time-coarse-grained data packet waiting queues T(m), and tail queue Z. 5. The method according to claim 4, characterized in that said basic delay unit u is smaller than said forwarding delay constant σ. 6. method as claimed in claim 5, it is characterized in that the method that core router sends common real-time service data packet is: 1) Send queue F, N time-fine-grained data packets waiting in queue R(n), M time-coarse-grained data packets in waiting queue T(m), and data packets in tail queue Z; restore the state after sending variable initial value; 2) When the update time B(k) arrives, the unsent data packets in the time fine-grained queue R(n) are sequentially added to the tail of the queue F according to the sending order; 3) When the update time B(k) arrives, the data packets in the first coarse-grained queue T(1) need to be allocated to the fine-grained queue, and the rest of the coarse-grained queues are promoted in turn, and the data packets in the queue Z Qualified data packets are added to the queue T(M). 7. method as claimed in claim 4, it is characterized in that the method that core router updates common real-time service data packet relative time limit is: 1) The core router records the arrival time t_arv of the data packet, and the sending time t_snd of the data packet; 2) When the core router sends the data packet, the relative time limit is updated as β+σ-(t_snd-t_arv), where β is the relative time limit when the data packet enters the core router. 8. method as claimed in claim 6 and 7, it is characterized in that described real-time service data packet comprises common real-time service data packet, real-time service test data packet and real-time service test failure data packet; In non-congested state, real-time service test The data packet is processed in the core router in the same way as the normal real-time service data packet, but in the congested state, the core router changes the real-time service test data packet into a real-time service test failure data packet, and puts it at the end of the best-effort service queue ; The core router forwards the real-time service test failure data packet through the best-effort service queue. 9. The method according to claim 2, characterized in that the method for establishing the real-time service is: 1) On the output interface of each router, allocate a certain bandwidth to reserve for real-time business, and set the sending priority for this part of the business; 2) Before sending real-time service data, the edge router at the sending end sends a real-time service test data packet to the edge router at the receiving end; 3) When the edge router at the receiving end checks the received real-time service data packet, if it finds a real-time service test failure data packet, it rejects the call; otherwise, it establishes the real-time service. 10. The method according to claim 2, wherein the routing establishment method of the real-time service is as follows: a newly-built flag field is added in the routing table of the router, and when the routing item of the router changes, the routing item is set by the router The new flag of the new flag is valid, and after waiting for the new flag to be valid for τ, the router will update the new flag of the routing item to be invalid; when the router outputs ordinary real-time service data packets, if the new flag of the corresponding routing item is valid, the router will set the data The package is a real-time service test data package.

Images