WO2020142867A1 - Traffic shaping method and related device - Google Patents

Abstract

Embodiments of the present application provide a traffic shaping method and a related device. The method comprises: a device obtains a scheduling cycle of a scheduler and a bucket filling cycle of a token bucket, wherein the number of clock cycles included in the bucket filling cycle of the token bucket is equal to the number of clock cycles included in the scheduling cycle of the scheduler; and the device schedules the X output ports of the device according to the scheduling cycle and fills the token bucket according to the bucket filling cycle, wherein the number of clock cycles occupied by a first port in the scheduling cycle is positively related to the proportion of the bandwidth of the first port in the total output bandwidth of the device, the first port being any one of the output ports of the device, and X is a positive integer greater than or equal to 1. By using the embodiments of the present application, implementation costs can be significantly reduced.

Description

Flow shaping method and related equipment Technical field

The present invention relates to the field of communication technology, and in particular, to a traffic shaping method and related equipment.

Background technique

In the data transmission scenario, due to the limited sending capacity/bandwidth of the output port, the scheduler needs to ensure that the average bandwidth of the scheduled data cannot exceed the load bandwidth of the output port. It mainly consists of two parts. On the one hand, the scheduler is used to separate the data to be sent according to specific rules (such as calendar algorithm, weighted round robin (WRR), weighted fair queueing, weighted fair queueing, WFQ) etc.) dispatched to the output port. Taking calendar as an example, the output port scheduling specifications generally use time division multiplexing scheduling. The basic principle is that the user configures the calendar calendar table in advance, where the scheduling result of each scheduling is directly pre-configured, and the data structure is directly polled according to the working clock cycle during working hours For each data in the data, if other conditions are met (such as the traffic shaping conditions below, etc.), the scheduling result is directly obtained, otherwise the scheduling result is invalid this time, and the next data continues to be polled in the next clock cycle.

The other is to use traffic shaping to control the data output bandwidth of each port queue. The industry has included a variety of methods to complete traffic shaping. Typically, the token bucket mechanism is used to restrict data transmission. The basic principle of token communication is shown in Figures 1A and 1B. When the data stream needs to be output, it will first take out the number of tokens corresponding to the data size from the token bucket according to the size of the data to transmit data. That is to say, for data to be transmitted, there must be enough tokens in the token bucket. If the number of tokens is not enough, the data will be discarded or cached. This can limit the data flow to be less than or equal to the speed of token generation, thereby achieving the purpose of limiting traffic. According to the token bucket mechanism, the average bandwidth of data transmission is equal to the bucket filling speed, and the bucket depth of the token bucket is the maximum transmission burst size. At present, the more frequently used token bucket filling scheme is the regular bucket filling.

The basic principle of timed bucket filling is that the filling time of each token bucket is fixed, and different bandwidths can be achieved by configuring different numbers of bucket filling tokens. To implement this algorithm, at least the following data structures need to be introduced: the number of tokens in the current token bucket, the filling value per unit time, used to determine the time count of the filling time, and the maximum bucket depth of the token bucket. Considering that there are multiple ports in the device, it is logically that multiple ports set their respective token bucket control bandwidths. A typical design is to share a token bucket circuit, store the data structure of different ports in different addresses of memory (such as static random-access memory (SRAM)), and fill the bucket with time-division multiplexing , Bucket deduction operation (that is, the operation of deducting tokens from the token bucket). At this time, the bucket filling period needs to be greater than the number of ports, and each token bucket is filled at most once in each bucket filling period, so as to ensure that bucket filling requests of different ports do not conflict in time. The typical value of the bucket filling period is: the bucket filling period is equal to min{port number r, 2^q}, where q takes a value that satisfies 2^q>r.

In the mechanism of timed bucket filling, the logical bucket corresponding to each output port needs to be filled once in the bucket filling period. Therefore, in order to meet the average bandwidth of data transmission, the bucket needs to be filled more every bucket filling period. Token. Too many one-time filling tokens will cause the bucket depth to become larger, resulting in a higher implementation cost. How to reasonably perform bucket filling to reduce the implementation cost is a technical problem being studied by those skilled in the art.

Summary of the invention

The embodiments of the present invention disclose a traffic shaping method and related equipment, which can significantly reduce the implementation cost.

In a first aspect, an embodiment of the present application provides a traffic shaping method. The method includes:

The device obtains the scheduling cycle of the scheduler and the bucket filling cycle of the token bucket, where the number of clock cycles included in the token bucket filling cycle is equal to the number of clock cycles included in the scheduler's scheduling cycle;

The device schedules the X output ports of the device according to the scheduling period and fills the token bucket according to the bucket filling period, wherein the first port occupies the scheduling period The number of clock cycles is positively related to the ratio of the bandwidth of the first port to the total output bandwidth of the device. The first port is any one of the output ports of the device, and X is greater than or A positive integer equal to 1.

By executing the above method, the scheduling period of the scheduler and the bucket filling period of the token bucket are set to be the same, so that the speed of scheduling data matches the speed of the bucket filling token, so the token bucket does not need to be configured with a large bucket depth On the one hand, the implementation cost of configuring a larger bucket depth is reduced. Specifically, since the scheduling period and the token bucket filling period are set to be the same, the bucket deduction and bucket filling occur at the same time every clock cycle, so the bucket deduction and filling Buckets can be combined into one read and write, instead of deducting and filling buckets, which significantly reduces read and write complexity, that is, reduces the cache performance requirements of the memory used; on the other hand, reduces traffic Burst improves the transmission performance of the device.

With reference to the first aspect, in a first possible implementation manner of the first aspect, the device schedules the X output ports of the device according to the scheduling cycle and the token according to the bucket filling cycle Bucket filling, including:

Configure a token parameter for the first port, the token parameter includes a first token parameter, the first token parameter includes a remaining token amount K, a first token amount n1, and a mark bit parameter, the The marker bit parameter is used to mark one clock cycle within the h clock cycles; the initial value of the remaining token amount is equal to the total number of tokens m;

When it is in turn to schedule the first port when scheduling the first port, and determine whether the current clock cycle is the clock cycle marked by the flag bit in the token parameter;

If not, fill the token bucket with the first token amount n1 and update the remaining token amount K in the token parameter, where the updated remaining token amount is equal to the remaining order before the update Card amount minus the first token amount n1;

If yes, fill the token bucket with the current token amount K, and initialize the remaining token amount K in the token parameters to the total token amount m.

In this implementation, only the remaining token amount K, the first token amount n1, and the marking parameters need to be configured for the first port, and there is no need to separately configure each corresponding clock cycle of the first port, which reduces The complexity of the configuration, and this process is relatively simple to implement, which can save hardware and software resources.

With reference to the first possible implementation manner of the first aspect, in a second possible implementation manner of the first aspect, the token parameter further includes a second token parameter, where the second token parameter includes The remaining token amount K, the first token amount n1 and the flag bit parameters, wherein the initial value of the remaining token amount K in the second token and the remaining token amount K in the first token The initial value is different, the first token amount n1 in the second token is different from the first token amount n1 in the first token; the first token parameter is used in the first port The bandwidth of the token bucket is used as the basis for filling the token bucket with the token bandwidth unchanged, and the second token parameter is used as a basis for filling the token bucket with the token bucket after the bandwidth of the first port is changed.

In this implementation, multiple sets of parameters are configured when configuring the token parameters, because the port bandwidth may change. If the configuration is changed and the bucket is filled according to the configuration before the change, there may be problems. If it is temporarily changed after the change The configuration will also cause some delay, that is, the scheduling and bucket filling operations cannot be performed immediately after the port bandwidth changes. This optional solution predicts in advance what kind of changes may occur on the port, and then pre-configures a set of spare token parameters according to this, namely the second token parameter, the parameter type in the second token parameter and the first order The parameter types of the card parameters are the same, that is, the specific values may be different. The parameters in the second token parameters can meet the scheduling and bucket filling requirements after the port bandwidth changes.

With reference to the first aspect, or any one of the foregoing possible implementation manners of the first aspect, in a third possible implementation manner of the first aspect, the device outputs X output ports of the device according to the scheduling period Scheduling and bucket filling the token bucket according to the bucket filling cycle include:

Mapping a value representing the amount of tokens for each clock cycle within the scheduling cycle; wherein the sum of the values corresponding to the h clock cycles is equal to the total number of tokens m;

Scheduling the first port in the h clock cycles in the scheduling cycle, and filling the token bucket according to the value corresponding to each clock cycle in the h clock cycles .

In this implementation, a value is mapped in advance for each clock cycle in the scheduling cycle, and each subsequent scheduling and bucket filling to each clock cycle will fill the bucket according to the value corresponding to the clock cycle, because each clock cycle The values are independent of each other and do not affect each other, so even in which clock cycle an accident occurs will not affect the bucket filling in other subsequent clock cycles.

In a second aspect, an embodiment of the present application provides a traffic shaping device. The device includes a processing unit, a scheduler, and X output ports. X is a positive integer greater than or equal to 1, where:

The processing unit is configured to obtain a scheduling period of the scheduler and a bucket filling period of the token bucket, where the number of clock cycles included in the token bucket filling period and the clock period included in the scheduler scheduling period Equal number

A scheduler, configured to schedule the X output ports according to the scheduling period and fill the token bucket according to the bucket filling period, wherein the first port occupies the scheduling period The number of clock cycles is positively related to the proportion of the bandwidth of the first port in the total output bandwidth of the device, and the first port is any one of the X output ports.

By running the above unit, the scheduling period of the scheduler and the bucket filling period of the token bucket are set to be the same, so that the speed of the scheduling data matches the speed of the bucket filling token, so the token bucket does not need to be configured with a large bucket depth On the one hand, the implementation cost of configuring a larger bucket depth is reduced. Specifically, since the scheduling period and the token bucket filling period are set to be the same, the bucket deduction and bucket filling occur at the same time every clock cycle, so the bucket deduction and filling Buckets can be combined into one read and write, instead of deducting and filling buckets, which significantly reduces read and write complexity, that is, reduces the cache performance requirements of the memory used; on the other hand, reduces traffic Burst improves the transmission performance of the device.

With reference to the second aspect, in a first possible implementation manner of the second aspect, the scheduler is configured to schedule the X output ports according to the scheduling period and the order according to the bucket filling period Filling the bucket with a brand bucket, including:

Configure a token parameter for the first port, the token parameter includes a first token parameter, the first token parameter includes a remaining token amount K, a first token amount n1, and a mark bit parameter, the The marker bit parameter is used to mark one clock cycle within the h clock cycles; the initial value of the remaining token amount is equal to the total number of tokens m;

When it is in turn to schedule the first port when scheduling the first port, and determine whether the current clock cycle is the clock cycle marked by the flag bit in the token parameter;

If not, fill the token bucket with the first token amount n1 and update the remaining token amount K in the token parameter, where the updated remaining token amount is equal to the remaining order before the update Card amount minus the first token amount n1;

If yes, fill the token bucket with the current token amount K, and initialize the remaining token amount K in the token parameters to the total token amount m.

In this implementation, only the remaining token amount K, the first token amount n1, and the marking parameters need to be configured for the first port, and there is no need to separately configure each corresponding clock cycle of the first port, which reduces The complexity of the configuration, and this process is relatively simple to implement, which can save hardware and software resources.

With reference to the first possible implementation manner of the second aspect, in a second possible implementation manner of the second aspect, the token parameter further includes a second token parameter, where the second token parameter includes a remaining order The card amount K, the first token amount n1 and the mark bit parameters, wherein the initial value of the remaining token amount K in the second token is different from the initial value of the remaining token amount K in the first token, The first token amount n1 in the second token is different from the first token amount n1 in the first token; the first token parameter is used for the bandwidth of the first port unchanged Is used as a basis for token bucket to fill bucket tokens, and the second token parameter is used as a basis for token bucket to fill bucket tokens after the bandwidth of the first port is changed.

In this implementation, multiple sets of parameters are configured when configuring the token parameters, because the port bandwidth may change. If the configuration is changed and the bucket is filled according to the configuration before the change, there may be problems. If it is temporarily changed after the change The configuration will also cause some delay, that is, the scheduling and bucket filling operations cannot be performed immediately after the port bandwidth changes. This optional solution predicts in advance what kind of changes may occur on the port, and then pre-configures a set of spare token parameters according to this, namely the second token parameter, the parameter type in the second token parameter and the first order The parameter types of the card parameters are the same, that is, the specific values may be different. The parameters in the second token parameters can meet the scheduling and bucket filling requirements after the port bandwidth changes.

With reference to the second aspect, or any one of the foregoing possible implementation manners of the second aspect, in a third possible implementation manner of the second aspect, the scheduler is configured to check the X number according to the scheduling period The scheduling of the output port and the bucket filling of the token bucket according to the bucket filling period include:

Mapping a value representing the amount of tokens for each clock cycle within the scheduling cycle; wherein the sum of the values corresponding to the h clock cycles is equal to the total number of tokens m;

Scheduling the first port in the h clock cycles in the scheduling cycle, and filling the token bucket according to the value corresponding to each clock cycle in the h clock cycles .

In this implementation, a value is mapped in advance for each clock cycle in the scheduling cycle, and each subsequent scheduling and bucket filling to each clock cycle will fill the bucket according to the value corresponding to the clock cycle, because each clock cycle The values are independent of each other and do not affect each other, so even in which clock cycle an accident occurs will not affect the bucket filling in other subsequent clock cycles.

With reference to the first aspect, or any of the foregoing possible implementations of the first aspect, or the second aspect, or any of the foregoing possible implementations of the second aspect, in an optional implementation, the order The bucket filling cycle of the brand bucket is synchronized with the scheduling cycle of the scheduler. It can be understood that this can make the scheduling (along with buckling buckets) highly match the bucket filling, that is, the bucket filling is completed once when the scheduling is completed once per clock cycle, which can minimize the bucket depth.

With reference to the first aspect, or any of the foregoing possible implementations of the first aspect, or the second aspect, or any of the foregoing possible implementations of the second aspect, in an optional implementation, the first The number of clock cycles h occupied by a port in the scheduling period satisfies the following formula:

Where H is the number of clock cycles included in the scheduling period, s1 is the bandwidth of the first port, and S is the total output bandwidth.

In this implementation, the scheduling period occupied by the first port is determined according to the ratio of the bandwidth of the first port to the total output bandwidth. In this way, the number of clock cycles and bandwidth occupied by each output port during scheduling can be made The actual needs are matched as much as possible and the output ports are relatively balanced, which maximizes the overall bandwidth requirements.

With reference to the first aspect, or any one of the foregoing possible implementations of the first aspect, or the second aspect, or any one of the foregoing possible implementations of the second aspect, in an optional implementation, the During the h clock cycles that the first port has undergone scheduling, the total amount of tokens m filled into the token bucket satisfies the following formula:

m＝s1*H/f

Where f is the working clock frequency of the device.

In this implementation, the total number of tokens required for the first port to fill the bucket during the scheduling period is determined according to the bandwidth of the first port. In this way, each output port can fill the tokens in the bucket during the scheduling process The volume and its actual bandwidth requirements are matched as closely as possible and the output ports are relatively balanced, which maximizes the overall bandwidth requirements.

With reference to the first aspect, or any one of the foregoing possible implementations of the first aspect, or the second aspect, or any one of the foregoing possible implementations of the second aspect, in an optional implementation, the scheduling The number of clock cycles H included in the cycle satisfies the following formula:

Where S is the total output bandwidth of the device, t is the least common divisor of the bandwidth of each output port of the device, p is the total number of ports of each output port of the device, and f is the operating clock frequency of the device .

In this implementation, the first term in the formula minimizes the waste of output port bandwidth, the second term ensures that each port can have a scheduling opportunity in each scheduling cycle, and the third term can ensure that each time The amount of bucket tokens is an integer, which can be realized by a simple circuit, that is, the implementation cost is reduced.

In a third aspect, an embodiment of the present application provides a device including a processor, a storage device, an output device, and a motherboard, each device is connected to the motherboard through a bus or soldering or other electrical connection, and the output device includes a scheduler . The processor controls the output device to perform traffic shaping and scheduling by calling the program stored in the memory.

According to a fourth aspect, an embodiment of the present application provides a computer-readable storage medium that stores program instructions, and when the program instructions run on a processor, implement the first aspect, or the first The method described in any possible implementation of aspects.

BRIEF DESCRIPTION

1A is a schematic diagram of a principle of a token bucket provided by an embodiment of the present invention;

1B is a schematic diagram of another principle of a token bucket provided by an embodiment of the present invention;

2 is a schematic flowchart of a data stream transmission provided by an embodiment of the present invention;

3 is a schematic flowchart of a data flow scheduling and shaping provided by an embodiment of the present invention;

4 is a schematic structural diagram of a traffic shaping device provided by an embodiment of the present invention;

5 is a schematic flowchart of a traffic shaping method according to an embodiment of the present invention;

6 is a schematic structural diagram of yet another flow shaping device provided by an embodiment of the present invention.

detailed description

The following describes the embodiments of the present invention with reference to the drawings in the embodiments of the present invention.

The invention is applied to network equipment, and the network equipment mainly completes the functions of transmission and processing of network data frames. Taking a network switching device (such as a switch, a router, etc.) as an example, as shown in FIG. 2, the network device needs to complete functions such as data receiving input, forwarding, switching, traffic management, and traffic output. In the "output" link of data, a typical process is shown in Figure 3. Specifically, considering the length of data packets and the shortness of a data packet, and a device with multiple output ports of different rates at the same time, it is necessary to First, the longer data packets are sliced according to the internal processing data bit width of the device. After slicing, they enter different egress port queues according to different exits. During the period, it also involves scheduling and shaping the port queues. Correspondingly, the scheduler sends the sliced data to each output port evenly according to the bandwidth of each output port, and finally the output circuit of the output port edits the data into a data frame that meets the output port protocol (that is, the exit is framed), and then sends To the downstream. The embodiment of the present application provides a traffic shaping method, which mainly optimizes the scheduling shaping link in FIG. 3.

Please refer to FIG. 4. FIG. 4 is a device 40 provided by an embodiment of the present invention. The device 40 can perform data transmission, and the processes shown in FIGS. 2 and 3 need to be performed during data transmission. The device 40 includes a processor 401, a storage device 402, a random access memory (random access memory, RAM) 403, an output device 404, an input device 405, and a motherboard 406, wherein each device is connected by a bus or welding or other point connection On the motherboard 406.

The storage device 402 includes but is not limited to a read-only memory (read-only memory, ROM), an erasable programmable read-only memory (erasable programmable read only memory, EPROM), or a portable read-only memory (compact disc read-only memory) , CD-ROM), the storage device 402 is used for related instructions and data.

RAM403 is used to cache the data used in the traffic shaping and scheduling process. For example, there will be a frame data cache in the exit framing link shown in FIG. 3, and these data can be cached in RAM.

The output component 404 is used to perform traffic scheduling and traffic shaping operations under the control of the processor 401. Optionally, the output component may include a scheduler. The scheduling algorithm used by the scheduler may be configured in the scheduler or in the processor 401. in;

The input component 405 is used to receive data transmitted from outside under the control of the processor 401.

The processor 401 may be one or more central processing units (CPU), or one or more field programmable logic gate arrays (FPGA), or one or more application specific integrated circuits (Application Specific (Integrated Circuit, ASIC), when the processor 401 is a CPU, the CPU may be a single-core CPU or a multi-core CPU. The processor 401 is used to read the program codes stored in the storage device 402 and RAM to control the operation of each device, for example, to control the output component 404 and the input component 405 to perform data input and data output.

Optionally, the device 40 may further include a user interface 407, and the user interface 407 may be used to access a keyboard, a mouse, a touch screen, etc., thereby supporting user interaction with the device 40.

For a more detailed introduction of traffic shaping by the device 40, reference may be made to the method embodiment shown in FIG. 5.

Please refer to FIG. 5. FIG. 5 is a traffic shaping method provided by an embodiment of the present invention. The method may be implemented based on the device shown in FIG. 4, or may be implemented based on other architectures. The method includes but is not limited to the following steps:

Step S501: The device acquires the scheduling period of the scheduler and the bucket filling period of the token bucket.

Specifically, taking the time-division multiplex calendar scheduler as an example, a scheduling cycle usually includes one or more fixed clock cycles, and the scheduler can perform scheduling once per clock cycle, for example, if the scheduling cycle includes 3 clocks Period, then the scheduler can perform scheduling 3 times in the scheduling period. Similarly, a bucket filling cycle contains one or more clock cycles, in which the token bucket can be filled once per clock cycle. For example, if the bucket filling cycle includes 4 clock cycles, then within the bucket filling cycle The token bucket can be filled 4 times. For a device with multiple output ports, multiple schedules within a scheduling period correspond to multiple ports, and similarly multiple bucket fillings also correspond to token buckets assigned to multiple ports.

In the embodiment of the present application, the number of clock cycles included in the bucket filling period of the token bucket is equal to the number of clock cycles included in the scheduling period of the scheduler, and is subsequently collectively referred to as a logical period. For example, if the filling cycle of the token bucket is 3 clock cycles, then the scheduling cycle of the scheduler is also 3 clock cycles, if the scheduling cycle of the scheduler is 10 clock cycles, then the filling of the token bucket The bucket cycle is also 10 clock cycles. Optionally, the scheduling logic cycle in the embodiment of the present application is synchronized with the bucket filling logic cycle, and the distribution results of the output ports corresponding to the clock cycles in the logic cycle are also synchronized, that is, each clock cycle in the logic cycle corresponds to a specific port Scheduling and each scheduling is accompanied by a bucket filling, then considering the multiple scheduling opportunities for a certain port scheduled by calendar in a logical cycle, the distribution strategy in multiple clock cycles is evenly distributed, and the token bucket is filled evenly , Token scheduling queue is also uniform, so it can minimize the barrel depth, thereby reducing traffic bursts. For example, the first clock cycle of the scheduling cycle and the first clock cycle of the bucket filling cycle are the same clock cycle, and the last clock cycle of the scheduling cycle is the same as the last clock cycle of the bucket filling cycle. One clock cycle, and so on for the rest.

It should be noted that the number of clock cycles included in the bucket filling period and the scheduling period in this embodiment of the present application is not an arbitrary period, but needs to meet certain conditions. The following uses Formula 1 to illustrate several conditions to facilitate understanding.

In Equation 1, H represents the number of clock cycles included in the scheduling period, S represents the total output bandwidth of the device, t is the least common divisor of the bandwidth of each output port of the device, p is the total number of ports of each output port of the device, and f is The operating clock frequency of the device.

The first term in Equation 1 indicates that the scheduling period is greater than or equal to the total output bandwidth/minimum port granularity. It can be understood that the minimum port granularity is equal to t*1G. For example, if the total output bandwidth of the device is 1200G and the bandwidth of the output port supported by the device is N*5G, the minimum common divisor of the bandwidth of each output port is 5, that is, the minimum port granularity is 5G, so according to the requirements of the first item , The scheduling cycle needs to be greater than or equal to 240 (that is, 240 clock cycles). As another example, if the total output bandwidth of the device is 1200G, and the bandwidth of the output port supported by the device is 5G, start at a 1G granularity step, then the minimum common divisor of the bandwidth of each output port is 1, that is, the minimum port granularity is 1G, so According to the requirements of the first item, the scheduling period needs to be greater than or equal to 1200. N is a positive integer. This first item minimizes the waste of bandwidth at the output port.

The second term in Equation 1 indicates that the number of clock cycles included in the scheduling cycle is greater than or equal to the total number of output ports of the device. For example, when the total number of ports is 10, the scheduling period is at least 10 (that is, 10 clock cycles), so as to ensure that each port can have a scheduling opportunity in each scheduling cycle (one clock cycle corresponds to one scheduling opportunity).

The third term in Equation 1 indicates that the bucket filling period should meet the token bucket filling requirement of 1G accuracy. For example, if the total output bandwidth of the device is 1200G and the working clock of the device is 1.2GHz, if the L*1G bandwidth needs to be accurately configured, an optional bucket filling period is 1350. The value is L*1125bit. If the period is slightly larger or slightly smaller than 1350, there is no guarantee that each bucket filling value is an integer, and the circuit is difficult to implement. Similarly, if the L*1G bandwidth needs to be accurately configured and the bucket filling period is 12, then the value of each bucket filling is L*10bit, and the obtained 10 is an integer, so it meets the accuracy requirements, so it will be The barrel filling period is set to 12 to meet the requirements of the third item above.

Optionally, the scheduling period in the embodiment of the present application needs to satisfy one or more of the three conditions in Formula 1 above. For example, all three items are satisfied. If these three items are required to be satisfied, then the same As an example to illustrate, the scheduling period of 1350 can meet the above three requirements, and the scheduling period of 1206 can also meet the above three requirements. It should be noted that, since the number of clock cycles included in the scheduling cycle and the number of clock cycles included in the bucket filling cycle in the embodiment of the present application are equal, if the scheduling cycle meets the requirements of the above three items, the bucket filling cycle also meets the above three Item requirements.

There are many ways for the device to obtain the bucket filling period and the scheduling period. For example, in one solution, the device automatically calculates the scheduling period and the bucket filling period according to the above conditions, and in another solution, the scheduling period and the bucket filling period are calculated manually according to the above conditions. And then configure it on the device. Of course, there are other solutions, no examples are given here.

Step S502: The device schedules the output port of the device according to the scheduling period and fills the token bucket according to the bucket filling period.

Specifically, the number of clock cycles occupied by the first port in the scheduling period is positively correlated with the proportion of the bandwidth of the first port in the total output bandwidth of the device. The first port is the Any one of the output ports of the device. Optionally, if the device has only the first port, the bandwidth of the first port is equal to the total output bandwidth of the device, and the first port occupies all clock cycles in the scheduling period.

In an optional solution, the number h of clock cycles occupied by the first port in the scheduling period satisfies Formula 2:

Where H is the number of clock cycles included in the scheduling period, s1 is the bandwidth of the first port, and S is the total output bandwidth.

For example, assume that the total output bandwidth of the device is 1200G, and the total output bandwidth is composed of 10*100G+20*10G. That is, 10 ports with 100G bandwidth and 20 ports with 10G bandwidth, and the scheduling period is 1350.

Then the number of clock cycles h occupied by each 100G bandwidth port in this scheduling cycle is:

Then the number of clock cycles h that each 10G bandwidth port occupies in this scheduling cycle is:

In short, the sum of the clock cycles occupied by these 30 ports must not exceed one scheduling cycle. According to the above example, in the scheduling period of 1350, each 100G bandwidth port occupies 112 clock cycles, each 10G bandwidth port occupies 11 clock cycles, and is used to schedule the port in the clock cycle occupied by each port data.

The above describes the clock period h occupied by each port during the scheduling period. In this clock period h, the token bucket needs to be filled. The amount of tokens filled in the bucket needs to meet the corresponding conditions. The following uses the first port as Examples to illustrate. Optionally, the total number of tokens m filled into the token bucket within the h clock cycles experienced by scheduling the first port satisfies formula 3:

m＝s1*H/f Formula 3

In Formula 3, f is the working clock frequency of the device, s1 is the bandwidth of the first port, and H is the scheduling period.

For example, if the bandwidth s1 of the first port is 100G, the scheduling period of the scheduler is 1350, and f is 1.2 GHz, then in the 112 clock cycles that the first port performs scheduling, the token bucket is filled. The total number of tokens m is 112500bit. For example, in 11 of the 112 clock cycles, each clock cycle is filled with 1000 bits in a bucket, and in addition to these 111 clock cycles, the bucket is filled with 1500 bits in 1 clock cycle.

For example, if the bandwidth s1 of the first port is 10G, the scheduling period of the scheduler is 1350, and f is 1.2 GHz, then the token bucket is filled into the token bucket within 11 clock cycles of the first port. The total number of tokens m is 11250bit. For example, in 10 clock cycles out of these 11 clock cycles, each clock cycle fills 1016 bits in a bucket, and in 10 clock cycles other than these 10 clock cycles fills 1090 bits in a bucket.

Taking the first port as an example, the clock period h occupied by each output port during the scheduling period and the total amount m of tokens to be filled in the clock period occupied by each port are introduced. Based on this, the first port is used as an example to describe how to schedule and fill a bucket.

In an optional scheduling and bucket filling scheme, first, the device configures a token parameter for the first port, the token parameter includes a first token parameter, and the first token parameter includes a remaining token The amount K, the first token amount n1, and the marker bit parameter, the marker bit parameter is used to mark one clock cycle within the h clock cycles; the initial value of the remaining token amount is equal to the total amount of tokens m. Then, it turns to schedule the first port when scheduling the first port within the scheduling period, and judge whether the current clock period is the clock period marked by the flag bit in the token parameter; If not, fill the token bucket with the first token amount n1 and update the remaining token amount K in the token parameter, where the updated remaining token amount is equal to the remaining order before the update The card amount minus the first token amount n1; if yes, fill the token bucket with the current token amount K, and initialize the remaining token amount K in the token parameters to the token The total amount m. It can be understood that the first token amount n1 is a preset integer, and the integer can be made as close to m/h as possible. In this way, only the remaining token amount K, the first token amount n1 and the marking parameters need to be configured for the first port, without separately configuring each corresponding clock cycle of the first port, which reduces the configuration Complexity, and this process is relatively simple to implement, which can save hardware and software resources.

It can be understood that after the token parameters are configured according to the first step above, each subsequent scheduling cycle may be scheduled according to the token parameters, and it is not necessary to configure each scheduling cycle.

Optionally, the token parameter further includes a second token parameter, wherein the second token parameter includes a remaining token amount K, a first token amount n1, and a mark bit parameter, wherein the second The initial value of the remaining token amount K in the token is different from the initial value of the remaining token amount K in the first token, and the first token amount n1 in the second token is different from the first The first token amount n1 in the token is different; the first token parameter is used as a basis for the token bucket to fill the token when the bandwidth of the first port has not changed, the second token parameter It is used as a basis for the token bucket to fill the bucket token after the bandwidth of the first port is changed. That is to say, when configuring token parameters, configure multiple sets of parameters, because the port bandwidth may change. If the scheduling and filling of the bucket are performed after the change, there may be problems. If the configuration is temporarily changed after the change, It will cause some delay, that is, the scheduling and bucket filling operations cannot be performed immediately after the port bandwidth changes. This optional solution predicts in advance what kind of changes may occur on the port, and then pre-configures a set of spare token parameters according to this, namely the second token parameter, the parameter type in the second token parameter and the first order The parameter types of the card parameters are the same, that is, the specific values may be different. The parameters in the second token parameters can meet the scheduling and bucket filling requirements after the port bandwidth changes.

In yet another optional scheduling and bucket filling scheme, first, the device maps a value for token size for each clock cycle in the scheduling cycle; wherein, the h clock cycles correspond to The sum of the values is equal to the total amount of tokens m. Then, the device schedules the first port during the h clock cycles in the scheduling cycle, and then performs the token on the token according to the value corresponding to each clock cycle in the h clock cycles Bucket filling.

For example, if the total number of tokens that need to be filled in buckets for the h scheduling cycles corresponding to the first port is 100 bits, and the h clock cycles include the first, third, sixth, and tenth in the scheduling cycle Clock cycles, you can map the value 20 for the first clock cycle in the scheduling cycle, the value 30 for the third clock cycle, the value 40 for the sixth clock cycle, and the value for the 10th clock cycle The value 10, that is, the sum of the values mapped in these four clock cycles is 100. Later, when scheduling to the first clock cycle of the scheduling cycle, the 20-bit token is filled according to the mapped value of 20, and when scheduling to the third clock cycle of the scheduling cycle, the 30-bit bucket is filled according to the mapped value of 30. Cards, when scheduling to the 6th clock cycle of the scheduling cycle, fill the bucket with 40bit tokens according to the mapped value of 40, and when scheduling to the 10th clock cycle of the scheduling cycle, fill the bucket with 10bits according to the mapped value of 10. brand.

In the method described in FIG. 5, the scheduling period of the scheduler and the bucket filling period of the token bucket are set to be the same, so that the speed of the scheduling data matches the speed of the bucket filling token, so the token bucket does not need to be configured. The large bucket depth reduces the implementation cost of configuring a larger bucket depth. Specifically, since the scheduling period and the token bucket filling period are set to be the same, bucket deduction and bucket filling occur at the same time every clock cycle, so Buckle filling and bucket filling can be combined into one read and write, instead of one reading and writing of bucket filling and filling, which significantly reduces the complexity of reading and writing, that is, reduces the cache performance requirements of the memory used; on the other hand , Which reduces traffic bursts and improves the transmission performance of the device.

The method of the embodiment of the present invention is described in detail above, and the device of the embodiment of the present invention is provided below. The above introduces the device from the perspective of hardware structure, and the following introduces the device from the perspective of software functions according to functional modules.

Please refer to FIG. 6. FIG. 6 is a schematic structural diagram of a traffic shaping device 60 according to an embodiment of the present invention. The traffic shaping device 60 may include a processing unit 601, a scheduler 602, and X output ports 603 (FIG. 6 shows output ports 603 shows three examples), where:

The processing unit 601 is used to obtain the scheduling period of the scheduler 602 and the bucket filling period of the token bucket, wherein the number of clock cycles included in the token bucket filling period and the clock included in the scheduling period of the scheduler The number of cycles is equal; the processing unit 601 may be one or more processors, and the processor may be a central processing unit CPU, or a field programmable logic gate array FPGA, or an application specific integrated circuit ASIC, and so on.

The scheduler 602 is configured to schedule the X output ports according to the scheduling period and fill the token bucket according to the bucket filling period, where the first port occupies the scheduling period The number of clock cycles is positively related to the ratio of the bandwidth of the first port to the total output bandwidth of the device. The first port is any one of the output ports of the device, and X is greater than or A positive integer equal to 1.

The X output ports may be X physical data transmission circuits respectively, or may be X logical data transmission circuits obtained by logically dividing one physical data transmission circuit.

By running the above unit, the scheduling period of the scheduler and the bucket filling period of the token bucket are set to be the same, so that the speed of the scheduling data matches the speed of the bucket filling token, so the token bucket does not need to be configured with a large bucket depth On the one hand, the implementation cost of configuring a larger bucket depth is reduced. Specifically, since the scheduling period and the token bucket filling period are set to be the same, the bucket deduction and bucket filling occur at the same time every clock cycle, so the bucket deduction and filling Buckets can be combined into one read and write, instead of deducting and filling buckets, which significantly reduces read and write complexity, that is, reduces the cache performance requirements of the memory used; on the other hand, reduces traffic Burst improves the transmission performance of the device.

In a possible implementation manner, the scheduler, configured to schedule the X output ports according to the scheduling period and fill the token bucket according to the bucket filling period, includes:

Configure a token parameter for the first port, the token parameter includes a first token parameter, the first token parameter includes a remaining token amount K, a first token amount n1, and a mark bit parameter, the The marker bit parameter is used to mark one clock cycle within the h clock cycles; the initial value of the remaining token amount is equal to the total number of tokens m;

When it is in turn to schedule the first port when scheduling the first port, and determine whether the current clock cycle is the clock cycle marked by the flag bit in the token parameter;

If not, fill the token bucket with the first token amount n1 and update the remaining token amount K in the token parameter, where the updated remaining token amount is equal to the remaining order before the update Card amount minus the first token amount n1;

If yes, fill the token bucket with the current token amount K, and initialize the remaining token amount K in the token parameters to the total token amount m.

In this implementation, only the remaining token amount K, the first token amount n1, and the marking parameters need to be configured for the first port, and there is no need to separately configure each corresponding clock cycle of the first port, which reduces The complexity of the configuration, and this process is relatively simple to implement, which can save hardware and software resources.

In yet another possible implementation manner, the token parameter further includes a second token parameter, wherein the second token parameter includes a remaining token amount K, a first token amount n1, and a mark bit parameter, where , The initial value of the remaining token amount K in the second token is different from the initial value of the remaining token amount K in the first token, the first token amount n1 in the second token Different from the first token amount n1 in the first token; the first token parameter is used as a basis for the token bucket to fill the bucket token when the bandwidth of the first port has not changed, the The second token parameter is used as a basis for the token bucket to fill the bucket token after the bandwidth of the first port is changed.

In this implementation, multiple sets of parameters are configured when configuring the token parameters, because the port bandwidth may change. If the configuration is changed and the bucket is filled according to the configuration before the change, there may be problems. If it is temporarily changed after the change The configuration will also cause some delay, that is, the scheduling and bucket filling operations cannot be performed immediately after the port bandwidth changes. This optional solution predicts in advance what kind of changes may occur on the port, and then pre-configures a set of spare token parameters according to this, namely the second token parameter, the parameter type in the second token parameter and the first order The parameter types of the card parameters are the same, that is, the specific values may be different. The parameters in the second token parameters can meet the scheduling and bucket filling requirements after the port bandwidth changes.

In yet another possible implementation manner, the scheduler, configured to schedule the X output ports according to the scheduling period and fill the token bucket according to the bucket filling period, includes:

Mapping a value representing the amount of tokens for each clock cycle within the scheduling cycle; wherein the sum of the values corresponding to the h clock cycles is equal to the total number of tokens m;

Scheduling the first port in the h clock cycles in the scheduling cycle, and filling the token bucket according to the value corresponding to each clock cycle in the h clock cycles .

In this implementation, a value is mapped in advance for each clock cycle in the scheduling cycle, and each subsequent scheduling and bucket filling to each clock cycle will fill the bucket according to the value corresponding to the clock cycle, because each clock cycle The values are independent of each other and do not affect each other, so even in which clock cycle an accident occurs will not affect the bucket filling in other subsequent clock cycles.

In yet another possible implementation manner, the bucket filling period of the token bucket is synchronized with the scheduling period of the scheduler. It can be understood that this can make the scheduling (along with buckling buckets) highly match the bucket filling, that is, the bucket filling is completed once when the scheduling is completed once per clock cycle, which can minimize the bucket depth.

In yet another possible implementation manner, the number h of clock cycles occupied by the first port in the scheduling period satisfies the following formula:

Where H is the number of clock cycles included in the scheduling period, s1 is the bandwidth of the first port, and S is the total output bandwidth.

In this implementation, the scheduling period occupied by the first port is determined according to the ratio of the bandwidth of the first port to the total output bandwidth. In this way, the number of clock cycles and bandwidth occupied by each output port during scheduling can be made The actual needs are matched as much as possible and the output ports are relatively balanced, which maximizes the overall bandwidth requirements.

In yet another possible implementation manner, within h clock cycles of scheduling the first port, the total number of tokens m filled into the token bucket satisfies the following formula:

m＝s1*H/f

Where f is the working clock frequency of the device.

In this implementation, the total number of tokens required for the first port to fill the bucket during the scheduling period is determined according to the bandwidth of the first port. In this way, each output port can fill the tokens in the bucket during the scheduling process The volume and its actual bandwidth requirements are matched as closely as possible and the output ports are relatively balanced, which maximizes the overall bandwidth requirements.

In yet another optional implementation manner, the number of clock cycles H included in the scheduling period satisfies the following formula:

Where S is the total output bandwidth of the device, t is the least common divisor of the bandwidth of each output port of the device, p is the total number of ports of each output port of the device, and f is the operating clock frequency of the device .

In this implementation, the first term in the formula minimizes the waste of output port bandwidth, the second term ensures that each port can have a scheduling opportunity in each scheduling cycle, and the third term can ensure that each time The amount of bucket tokens is an integer, which can be realized by a simple circuit, that is, the implementation cost is reduced.

It should be noted that the implementation and beneficial effects of each unit may also correspond to the corresponding description of the method embodiment shown in FIG. 4.

An embodiment of the present invention further provides a chip system. The chip system includes at least one processor, a memory, and an interface circuit. The memory, the transceiver, and the at least one processor are interconnected by a line. The at least one memory There are instructions stored in it; when the instructions are executed by the processor, the method flow shown in FIG. 5 is realized.

An embodiment of the present invention further provides a computer-readable storage medium, in which instructions are stored, and when it runs on a processor, the method flow shown in FIG. 5 is implemented.

An embodiment of the present invention also provides a computer program product. When the computer program product runs on a processor, the method flow shown in FIG. 5 is realized.

A person of ordinary skill in the art may understand that all or part of the process in the method of the foregoing embodiments may be implemented by a computer program instructing related hardware. The program may be stored in a computer-readable storage medium, and when the program is executed , May include the process of the N method embodiment described above. The foregoing storage media include: ROM or random storage memory RAM, magnetic disks or optical disks and other N types of media that can store program codes.

Claims (17)

1. A traffic shaping method, which is characterized by including:

   The device obtains the scheduling cycle of the scheduler and the bucket filling cycle of the token bucket, where the number of clock cycles included in the token bucket filling cycle is equal to the number of clock cycles included in the scheduler's scheduling cycle;

   The device schedules the X output ports of the device according to the scheduling period and fills the token bucket according to the bucket filling period, wherein the first port occupies the scheduling period The number of clock cycles is positively correlated with the ratio of the bandwidth of the first port to the total output bandwidth of the device, the first port is any one of the X output ports, and X is greater than or A positive integer equal to 1.
2. The method according to claim 1, wherein the bucket filling period of the token bucket is synchronized with the scheduling period of the scheduler.
3. The method according to claim 1 or 2, wherein the number h of clock cycles occupied by the first port in the scheduling period satisfies the following formula:

   

   Where H is the number of clock cycles included in the scheduling period, s1 is the bandwidth of the first port, and S is the total output bandwidth.
4. The method according to claim 3, wherein the total number of tokens m filled into the token bucket within the h clock cycles experienced by scheduling the first port satisfies the following formula:

   m＝s1*H/f

   Where f is the working clock frequency of the device.
5. The method according to claim 4, wherein the device schedules the X output ports of the device according to the scheduling period and fills the token bucket according to the bucket filling period, including :

   Configure a token parameter for the first port, the token parameter includes a first token parameter, the first token parameter includes a remaining token amount K, a first token amount n1, and a mark bit parameter, the The marker bit parameter is used to mark one clock cycle within the h clock cycles; the initial value of the remaining token amount is equal to the total number of tokens m;

   When it is in turn to schedule the first port when scheduling the first port, and determine whether the current clock cycle is the clock cycle marked by the flag bit in the token parameter;

   If not, fill the token bucket with the first token amount n1 and update the remaining token amount K in the token parameter, where the updated remaining token amount is equal to the remaining order before the update Card amount minus the first token amount n1;

   If yes, fill the token bucket with the current token amount K, and initialize the remaining token amount K in the token parameters to the total token amount m.
6. The method according to claim 5, wherein the token parameter further includes a second token parameter, wherein the second token parameter includes a remaining token amount K, a first token amount n1, and a token Bit parameter, wherein the initial value of the remaining token amount K in the second token is different from the initial value of the residual token amount K in the first token, the first in the second token The token amount n1 is different from the first token amount n1 in the first token; the first token parameter is used to fill the bucket token as a token bucket when the bandwidth of the first port has not changed According to this, the second token parameter is used as a basis for the token bucket to fill the bucket token after the bandwidth of the first port is changed.
7. The method according to claim 4, wherein the device schedules the X output ports of the device according to the scheduling period and fills the token bucket according to the bucket filling period, including :

   Mapping a value representing the amount of tokens for each clock cycle within the scheduling cycle; wherein the sum of the values corresponding to the h clock cycles is equal to the total number of tokens m;

   Scheduling the first port in the h clock cycles in the scheduling cycle, and filling the token bucket according to the value corresponding to each clock cycle in the h clock cycles .
8. The method according to any one of claims 1-7, wherein the number H of clock cycles included in the scheduling period satisfies the following formula:

   

   Where S is the total output bandwidth of the device, t is the least common divisor of the bandwidth of each output port of the device, p is the total number of ports of each output port of the device, and f is the operating clock frequency of the device .
9. A traffic shaping device, characterized by comprising a processing unit, a scheduler and X output ports, X is a positive integer greater than or equal to 1, wherein:

   The processing unit is configured to obtain a scheduling period of the scheduler and a bucket filling period of the token bucket, wherein the number of clock cycles included in the token bucket filling period and the scheduling period of the scheduler include The number of clock cycles is equal;

   The scheduler is configured to schedule the X output ports according to the scheduling period and fill the token bucket according to the bucket filling period, wherein the first port is set in the scheduling period The number of clock cycles occupied is positively related to the proportion of the bandwidth of the first port in the total output bandwidth of the device, and the first port is any one of the X output ports.
10. The device according to claim 9, wherein the token bucket filling cycle is synchronized with the scheduling cycle of the scheduler.
11. The device according to claim 9 or 10, wherein the number h of clock cycles occupied by the first port in the scheduling period satisfies the following formula:

    

    Where H is the number of clock cycles included in the scheduling period, s1 is the bandwidth of the first port, and S is the total output bandwidth.
12. The device according to claim 11, wherein the total number of tokens m filled into the token bucket in the h clock cycles experienced by scheduling the first port satisfies the following formula:

    m＝s1*H/f

    Where f is the working clock frequency of the device.
13. The device according to claim 12, wherein the scheduler is configured to schedule the X output ports according to the scheduling period and fill the token bucket according to the bucket filling period ,include:

    Configure a token parameter for the first port, the token parameter includes a first token parameter, the first token parameter includes a remaining token amount K, a first token amount n1, and a mark bit parameter, the The marker bit parameter is used to mark one clock cycle within the h clock cycles; the initial value of the remaining token amount is equal to the total number of tokens m;

    When it is in turn to schedule the first port when scheduling the first port, and determine whether the current clock cycle is the clock cycle marked by the flag bit in the token parameter;

    If not, fill the token bucket with the first token amount n1 and update the remaining token amount K in the token parameter, where the updated remaining token amount is equal to the remaining order before the update Card amount minus the first token amount n1;

    If yes, fill the token bucket with the current token amount K, and initialize the remaining token amount K in the token parameters to the total token amount m.
14. The device according to claim 13, wherein the token parameter further includes a second token parameter, wherein the second token parameter includes a remaining token amount K, a first token amount n1, and a token Bit parameter, wherein the initial value of the remaining token amount K in the second token is different from the initial value of the residual token amount K in the first token, the first in the second token The token amount n1 is different from the first token amount n1 in the first token; the first token parameter is used to fill the bucket token as a token bucket when the bandwidth of the first port has not changed According to this, the second token parameter is used as a basis for the token bucket to fill the bucket token after the bandwidth of the first port is changed.
15. The device according to claim 12, wherein the scheduler is configured to schedule the X output ports according to the scheduling period and fill the token bucket according to the bucket filling period ,include:

    Mapping a value representing the amount of tokens for each clock cycle within the scheduling cycle; wherein the sum of the values corresponding to the h clock cycles is equal to the total number of tokens m;

    Scheduling the first port in the h clock cycles in the scheduling cycle, and filling the token bucket according to the value corresponding to each clock cycle in the h clock cycles .
16. The device according to any one of claims 9 to 15, wherein the number H of clock cycles included in the scheduling period satisfies the following formula:

    

    Where S is the total output bandwidth of the device, t is the least common divisor of the bandwidth of each output port of the device, p is the total number of ports of each output port of the device, and f is the working clock frequency of the device .
17. A computer-readable storage medium, characterized in that program instructions are stored in the computer-readable storage medium, and when the program instructions run on a processor, the method according to any one of claims 1-8 is implemented .

Images