CN1298139C - Multi-stage dispatch method for supporting multi-port multi-business - Google Patents

Abstract

The present invention relates to a multi-stage scheduling method supporting multiple ports and much service. The method comprises the steps that (A) different second-stage queues of A, B and C three service grades are respectively configured on all user side ports of a packet ring station for placing data received by the ports in classification; (B) a two-stage scheduling mechanism of service grade scheduling and port weighting scheduling is carried out for the operation of the second-grade queues of the ports; (C) various data obtained by two stages of scheduling enter the first-grade queue of the A, B and C service grades; (D) upper ring scheduling of flexible packet rings is carried out for the first-stage queue. The method of the present invention achieve the purposes that the service of multiple user ports needs to enter a loop circuit of flexible packet rings and all the user ports simultaneously have A, B and C service scheduling, and therefore, the priority scheduling sequence of all the ports of A service scheduling, B service scheduling and C service scheduling is fulfilled, so that the requirements of the A service and the B service for time delay and time delay jitter are guaranteed, and the starvation of all the ports in the same service grade can not occur.

Description

A method for multi-level scheduling supporting multi-port and multi-service

technical field

The invention relates to a multi-level scheduling method supporting multiple ports and multiple services, in particular to a scheduling method in which multiple user ports are supported in RPR (Resilient Packet Ring) equipment, and each user port supports multiple services.

Background technique

Elastic packet ring technology is an emerging technology, the purpose is to build an elastic packet ring with reusable bandwidth, each node has a fair algorithm to ensure bandwidth occupancy, and has ring protection and QOS (Quality of Service, Quality of Service) capabilities. The application is aimed at the backbone ring network of the metropolitan area and the access ring network of the metropolitan area. In December 2000, IEEE specially established the IEEE802.17 standard group to formulate the IEEE802.17 MAC layer standard based on the elastic packet ring technology that can build an elastic packet ring.

The main feature of RPR technology is the classification of services. The services are divided into three types: A, B, and C. Among them, A type is a real-time service, and B type is divided into two parts. The B-CIR (committed information rate) is Committed rate level service, B-EIR (excess information rate) is the service exceeding the committed rate level, and Class C is the best effort service. Define the service class bit of Service Class in the frame structure. In addition, the RPR technology shapes and schedules the services of a single RPR site entering the RPR according to the three categories A, B, and C, and then enters a RPR ring. As shown in Figure 1 and Figure 2, the RPR device generally includes two RPR ring-side ports, and the incoming and outgoing frames from these ports are RPR MAC frames, with A, B, and C service classification identifiers, And a group of user-side ports. Most of the user-side ports are Ethernet ports. The frames entering and outputting from this port are Ethernet frames, and the frame structure does not have A, B, and C service classification identifiers. Multiple RPR devices form an RPR ring through the RPR ring side ports. There is no queue at the user-side port in the existing RPR device. When the data enters a certain user-side port for a certain forwarding operation, it enters the A/B/C queue of a certain RPR loop, and then proceeds to the standard Integer scheduling operations of , enter an RPR loop. There are some problems with this approach: there are often multiple user-side ports, and a certain port may have a large traffic volume, while a certain port has a small traffic volume. According to the existing practice, first-come-first-served, the port with a small traffic volume may be starved as a result. Phenomenon. Another example is that the services entered by a certain user-side port may be Class A services, or Class B and Class C services. It is possible that a large number of Class C services are entered first, and a small amount of Class A services are entered later. According to the existing The practice is first-come-first-served. As a result, a large number of low-priority Class C services are processed first, and high-priority Class A services are processed after Class C services. As a result, the delay and delay jitter of Class A services cannot be solved. solve.

Contents of the invention

The technical problem to be solved by the present invention is to provide a service scheduling method that can better meet the requirements of the elastic packet ring in view of the defect that the elastic packet ring device has multiple user-side ports and the service scheduling between the existing ports is unreasonable.

The technical scheme that the present invention adopts for solving the problems of the technologies described above is:

A multi-level scheduling method supporting multi-port multi-service, comprising the following steps:

Step 1: each user-side port of a single packet ring site is configured with three secondary queues with different service levels, A, B, and C, which are used to classify and place the data received by the port;

Step 2: Run the two-level scheduling mechanism of first service level scheduling and then port weighted scheduling on the second-level queue of the port;

Step 3: All kinds of data dispatched by the two levels enter the first-level queues of A, B, and C service levels;

Step 4: Perform flexible packet ring-on-ring scheduling for services in the first-level queue.

The bandwidth of the minimum class A, B, and C services that each user-side port allows to enter the elastic packet ring can be configured by the user.

The service class scheduling described in step 2 is performed according to the priority order that class A is prior to class B, and class B is prior to class C; the port weighted scheduling described in step 2 is performed according to the weight of each port.

The above-mentioned weights are obtained according to the proportion of service bandwidths of types A, B, and C configured by the user on the user-side port to the same service bandwidth of the site.

Moreover, the port weights of different service levels can be designed to be different values.

Each RPR site has N user-side ports, and the second step includes the following specific steps:

21), judge whether all A service level queues of 1-N ports are empty, if empty continue to circulate;

22), if it is not empty, perform weighted scheduling on all N class A service queues of 1-N ports, and the dispatched class A service is sent to the first class queue of class A service;

23), if all the queues of A service level are dispatched empty, or the first-level queue of A service level is full, then perform B-level service level scheduling;

24), judge whether all B service level queues of 1-N are empty, if empty continue to circulate;

25), if it is not empty, perform weighted scheduling on all N class B service queues of 1-N ports, and the dispatched B class services are sent to the first class queue of B service;

26), if all queues of B service level are dispatched empty, or the first-level queue of B service level is full, then perform C-level service level scheduling;

27), judge whether all the C service level queues of 1-N are empty, if empty continue to circulate;

28), if it is not empty, perform weighted scheduling on all N class C service level queues of 1-N ports, and the dispatched C class service is sent to the first class queue of C class service;

29) If all the queues of service level C are dispatched empty, or the first-level queues of service level C are full, return to the scheduling of service level A; run the above steps cyclically.

The upper-loop scheduling described in step 4 includes shaping and scheduling the services of types A, B, and C in the first-level queue respectively.

Moreover, the shaping and scheduling are performed in the order that class A takes precedence over class B, and class B takes precedence over class C.

While scheduling a certain type of business, the credit value of the corresponding type of business is consumed, and when the credit value is exhausted, the scheduling of this type of business is suspended and another type of business is scheduled instead.

The credit value of the above-mentioned various services is determined by the total bandwidth of A, B, and C services configured on the elastic packet ring site that are allowed to enter the elastic packet ring.

The total bandwidth of A, B, and C types of services allowed by a single RPR station to enter the RPR loop can be configured by the user.

The beneficial effect of the present invention is: because this method configures three secondary queues with different service levels of A, B, and C on each user side port, and designs a method for scheduling the service level of the secondary queue first and then port weighted scheduling Two-level scheduling mechanism, so it can meet the priority scheduling order of class A services entering each port, followed by class B, and last priority of class C, thus ensuring the accuracy of time delay and delay jitter for class A and class B services Requirements, and at the same time, each port will not be starved under the same level of business conditions. The method of the invention effectively solves the scheduling strategy in the case that the services of multiple user ports need to enter the elastic packet ring, and each user port has three types of services of A, B and C at the same time.

Description of drawings

FIG. 1 is a schematic diagram of an RPR device.

Fig. 2 is a schematic diagram of a scheduling framework of an existing device.

Fig. 3 is a schematic diagram of the multi-level scheduling framework of the present invention.

Figure 4 is a flow chart of data reception.

Fig. 5 is a flow chart of two-level scheduling.

Fig. 6 is a flow chart of upper-loop scheduling.

Detailed ways

Below according to accompanying drawing and embodiment the present invention will be described in further detail:

As shown in Figure 3, it is a schematic diagram of multi-level scheduling. Each user-side port is configured with three queues, that is, two-level queues, and each queue corresponds to three service levels of A, B, and C. A second-level scheduler is added between the second-level queue and the first-level queue, which is used to run the two-level scheduling mechanism of the first service level scheduling and then the port weighted scheduling. After the two-level scheduling, the data enters A, B, In the first-level queue of C service level. The user configures the total A-type service bandwidth, B-type service bandwidth, and C-type service bandwidth that the site is allowed to enter the RPR ring, and configures the minimum A-type service bandwidth and B-type service bandwidth that each port is allowed to enter the RPR ring , Class C business bandwidth. The sum of the minimum Class A service bandwidth allowed by each port configured is equal to the configured total Class A service bandwidth of the site, and this correspondence also applies to Class B and Class C services.

This method requires multiple processes to be completed collaboratively, and the specific process is as follows:

The data receiving process is shown in Figure 4. This process is responsible for receiving data, classifying the received data into A, B, and C service levels, and putting the data into the service level queue corresponding to the data service level to which the port belongs. In, that is, in the secondary queue.

The second-level scheduling process is shown in Figure 5, which is responsible for two-level scheduling of the second-level queues. It is set that the RPR site has N user-side ports. First, the first-level scheduling of the service level is performed, that is, in the order of A->B->C. First, determine whether all A service-level queues on ports 1-N are Empty, if empty, continue to cycle through ports 1-N; if not empty, perform second-level scheduling of Class A services, that is, perform weighted scheduling on all N service-level queues of ports 1-N, and the weight is based on It is obtained from the ratio of the class A service bandwidth of the port configured by the user to the class A service bandwidth of the entire site. The scheduled A-type service data is sent to the first-level queue of the A-type service. If all queues of service class A are scheduled to be empty, or the first-level queues of service class A are full, then service class B is scheduled. The scheduling method of service level B is the same as the scheduling method of service level A above. If the queues of service class B of all ports are empty by scheduling or the first-level queues of service class B are full, then service class C scheduling will be performed. The scheduling method of service class C is the same as the scheduling method of service class A above, and the above steps are repeated repeatedly. cycle. It should be noted that the port weights of different service levels of A, B, and C can be designed to be different.

The uplink scheduling process is shown in Figure 6. The normal elastic packet ring uplink scheduling is performed on the services in the first-level queue, including various shaping operations and scheduling processes according to the priority order of A->B->C. When scheduling Class A services, the Class A services in the first-level queue are shaped according to the A0 and A1 credit values, and the corresponding A0 or A1 credit values are consumed, and the shaped data is sent to the queue to be sent and enters the elastic packet ring. When the credit values of A0 and A1 are exhausted, the scheduling of Class A services will be suspended, and Class B services will be dispatched, and Class C services will be dispatched in turn. The process of dispatching Class B and Class C services is similar to that of Class A services. The credit value of different types of services is determined by the total bandwidth of A, B, and C types of services allowed to enter the elastic packet ring at a single RPR site.

Although the business category classification method of the present invention has been illustrated and described, it is obvious that the present invention is not limited thereto. Numerous modifications, alterations, changes, substitutions and equivalents will occur to those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (11)

1, a kind of multi-level scheduling method that supports multi-port multi-service, it is characterized in that: comprise the following steps: Step 1, configure three secondary queues with different service levels, A, B, and C, at each user-side port of a single RPR site, to classify and place the data received by the port; Step 2, running a two-level scheduling mechanism in which service-level scheduling is performed first and then port weighted scheduling is performed on the port secondary queue; Step 3, all kinds of data after the two-level scheduling enter into the first-level queues of A, B, and C service levels; Step 4: Perform flexible packet ring-on-ring scheduling for services in the first-level queue. 2. The method according to claim 1, characterized in that: the bandwidth of the minimum class A, B, and C services allowed to enter the elastic packet ring by each user side port can be configured by the user. 3. The method according to claim 2, characterized in that: the service level scheduling in the second step is carried out according to the priority order that class A takes precedence over class B, and class B takes precedence over class C; said step The port weighted scheduling described in the second is performed according to the weight of each port. 4. The method according to claim 3, wherein the weight value is obtained according to the proportion of service bandwidths of types A, B and C configured on the user-side port to the similar service bandwidth of the site. 5. The method according to claim 4, characterized in that the port weights of different service levels can be designed to be different values. 6. The method according to claim 5, wherein each RPR site has N user-side ports, and said step 2 includes the following specific steps: 21), judge whether all A service level queues of 1-N ports are empty, if empty, continue to perform 21 in a loop; 22), if it is not empty, perform weighted scheduling on all N class A service queues of 1-N ports, and the dispatched class A service is sent to the first class queue of class A service; 23), if all the queues of A service level are dispatched empty, or the first-level queue of A service level is full, then perform B-level service level scheduling; 24), judge whether all B service level queues of 1-N ports are empty, if empty continue to execute 24 in a loop; 25), if it is not empty, perform weighted scheduling on all N class B service queues of 1-N ports, and the dispatched B class services are sent to the first class queue of B service; 26), if all queues of B service level are dispatched empty, or the first-level queue of B service level is full, then perform C-level service level scheduling; 27), judge whether all C service level queues of 1-N ports are empty, if empty continue to execute 27 in a loop; 28), if it is not empty, perform weighted scheduling on all N class C service level queues of 1-N ports, and the dispatched C class service is sent to the first class queue of C class service; 29) If all the queues of service class C are empty by scheduling, or the first-level queues of service class C are full, return to service class A for scheduling. 7. The method according to claim 2, characterized in that the upper ring scheduling in step 4 includes shaping and scheduling the three types of services A, B, and C in the primary queue respectively. 8. The method according to claim 7, characterized in that the shaping and scheduling are performed in the order that class A takes precedence over class B, and class B takes precedence over class C. 9. The method according to claim 8, characterized in that: while scheduling a certain type of business, the credit value of the corresponding type of business is consumed, and when the credit value is exhausted, the scheduling of this type of business is suspended and switched to scheduling Another type of business. 10. The method according to claim 9, characterized in that: the credit value of the various types of services is determined by the total bandwidth of A, B, and C services configured by the elastic packet ring site where they are allowed to enter the elastic packet ring loop of. 11. The method according to claim 10, characterized in that: the total bandwidth of A, B, and C services allowed to enter the RPR loop by the single RPR site can be configured by the user.

Images