CN105677484A - A multi-core CPU real-time data processing method with automatic load balancing - Google Patents

Abstract

The invention provides an automatic load balancing multi-core CPU real-time data processing method, which comprises the following steps: receiving real-time data to be processed of a multi-core CPU; and distributing the real-time data to be processed to the corresponding task interfaces according to the states of all the current task interfaces, the task running states of the CPU cores connected with the task interfaces and the attributes of the real-time data to be processed so as to process the data for the CPU cores connected with the task interfaces. The invention can ensure the reasonable and full utilization of CPU resources and the real-time performance of data processing.

Description

A multi-core CPU real-time data processing method with automatic load balancing

technical field

The invention relates to the field of data processing, in particular to a multi-core CPU real-time data processing method with automatic load balancing.

Background technique

During the rapid development of today's microelectronics technology and computer technology, the clock frequency of a CPU (Central Processing Unit, central processing unit) gradually reaches its limit. In order to further improve CPU processing performance, computer technology tends to develop in the direction of multi-threading, multi-core, multi-dimensional, etc., especially multi-core CPU, which has become the mainstream technology for maintaining Moore's Law.

It is a common method to process large-scale real-time data in parallel, use multi-core CPU, and make full use of the parallelism of multi-processor cores to improve data processing performance. However, how to allocate data streams to each processor core of the multi-core CPU to achieve optimal utilization of the entire system is a difficult point in the entire system.

In addition, the biggest feature of real-time data processing is real-time performance. If the data flow cannot be guaranteed to be processed within a certain period of time, some data will be lost without being processed. How to ensure the real-time performance of data processing has become another difficulty of the whole system.

Contents of the invention

The invention proposes a multi-core CPU real-time data processing method with automatic load balancing to solve one or more deficiencies in the prior art.

The present invention proposes a multi-core CPU real-time data processing method with automatic load balancing, comprising: receiving real-time data to be processed by the multi-core CPU; Attributes of the real-time data to be processed, distributing the real-time data to be processed to a corresponding task interface for the CPU core connected to the task interface to perform data processing.

In one embodiment, according to the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed, the real-time data to be processed is distributed to the corresponding task interface, including : According to the status of all current task interfaces, the task running status of the CPU core connected to each task interface and the attribute identification of the real-time data to be processed, at least one task interface that can currently receive the real-time data to be processed; Pending data is distributed to identified task interfaces.

In one embodiment, before distributing the real-time data to be processed to the corresponding task interface, the method further includes: caching the real-time data to be processed in a shared cache, to be distributed to the task interface.

In one embodiment, before distributing the real-time data to be processed to the identified task interface, it includes: buffering the real-time data to be processed to an independent cache of the identified task interface, so as to smooth the transmission of the real-time data to be processed to The imbalance between the speed of the CPU core and the speed processed by the CPU core.

In one embodiment, it also includes: according to one or more of the state of all current task interfaces, the task running status of the CPU core connected to each task interface and the attributes of the real-time data to be processed, setting the task interface in the CPU core The number and/or properties of the data processed by the CPU processing task.

In one embodiment, it also includes: according to one or more of the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed, setting the shared cache size.

In one embodiment, it also includes: according to one or more of the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed, setting the independent cache size.

In one embodiment, the real-time data to be processed includes a plurality of data particles; according to the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed, the The real-time data to be processed is distributed to the corresponding task interface for the CPU core connected to the task interface to perform data processing, including: according to the status of all current task interfaces, the task running status of the CPU core connected to each task interface and the data attributes of the granules, and distribute each of the data granules to corresponding task interfaces in parallel for the CPU cores connected to the task interfaces to perform data processing.

In one embodiment, it also includes: detecting the change in the number of running tasks of each CPU core; redistributing the real-time data to be processed that has been distributed to the task interface according to the running tasks after the number change, so as to balance the number of running tasks on the multi-core CPU. The data load of the CPU core.

In one embodiment, it also includes: judging whether the resource consumption ratio of a CPU core of the multi-core CPU exceeds a set value; if so, real-time data to be processed in the CPU core whose resource consumption ratio exceeds the set value , allocating to other CPU cores in the multi-core CPU whose resource consumption ratio does not exceed the set value.

The multi-core CPU real-time data processing method with automatic load balancing of the present invention is a multi-core CPU real-time data processing method capable of automatic load balancing. For applications that use multi-core CPUs for multi-task real-time data processing, automatic load balancing can be performed. According to information such as the number of tasks started by the multi-core CPUs, the running status of tasks, the status of task interfaces, and the attributes of data, the data to be processed is automatically distributed. It is allocated to each CPU core to ensure the reasonable and full utilization of CPU resources and the real-time performance of data processing. The method can also use a dynamic shared cache method to cache the data to be processed of each CPU core, so as to improve the real-time performance of data processing, reduce or avoid the loss of unprocessed data, thereby optimizing the performance of the entire system.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work. In the attached picture:

Fig. 1 is the structural representation of the real-time data processing system based on multi-core CPU of an embodiment of the present invention;

Fig. 2 is the schematic flow chart of the multi-core CPU real-time data processing method of the automatic load balancing of the embodiment of the present invention;

Fig. 3 is a schematic flowchart of a data distribution method in an embodiment of the present invention;

Fig. 4 is the schematic flow chart of the multi-core CPU real-time data processing method of the automatic load balancing of another embodiment of the present invention;

FIG. 5 is a schematic flowchart of a data distribution method in another embodiment of the present invention;

6 is a schematic flow diagram of a multi-core CPU real-time data processing method for automatic load balancing in another embodiment of the present invention;

7 is a schematic flow diagram of a multi-core CPU real-time data processing method for automatic load balancing in yet another embodiment of the present invention;

Fig. 8 is a schematic flowchart of a data distribution method in another embodiment of the present invention;

9 is a schematic flow diagram of a multi-core CPU real-time data processing method for automatic load balancing in another embodiment of the present invention;

10 is a schematic flow diagram of a multi-core CPU real-time data processing method for automatic load balancing according to another embodiment of the present invention;

11 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to an embodiment of the present invention;

Fig. 12 is a schematic structural diagram of a real-time data processing unit in an embodiment of the present invention;

13 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to another embodiment of the present invention;

Fig. 14 is a schematic structural diagram of a real-time data processing unit in another embodiment of the present invention;

15 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to yet another embodiment of the present invention;

16 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to another embodiment of the present invention;

Fig. 17 is a schematic structural diagram of a real-time data processing unit in another embodiment of the present invention;

18 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to another embodiment of the present invention;

FIG. 19 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to another embodiment of the present invention.

detailed description

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the embodiments of the present invention will be further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments and descriptions of the present invention are used to explain the present invention, but not to limit the present invention.

The invention proposes a multi-core CPU real-time data processing method, system and device with automatic load balancing. The method of the present invention can be realized based on the system of the present invention, and can also be realized based on systems of other structures as required. This method can perform automatic load balancing for applications that use multi-core CPUs for multi-task real-time data processing. Data is automatically allocated to each CPU core to ensure reasonable and full utilization of CPU resources and real-time data processing.

FIG. 1 is a schematic structural diagram of a real-time data processing system based on a multi-core CPU according to an embodiment of the present invention. As shown in FIG. 1 , the real-time data processing system includes a multi-core CPU 100 and a task interface module 200 . Wherein, each CPU core in the multi-core CPU 100 can run one or more tasks 101; the task interface module 200 includes multiple task interfaces 201; each task 101 is bound to one task interface 201 to receive real-time data.

The above real-time data processing system may further include a load balancing distribution channel 300 and a load balancing scheduling module 400 . Wherein, each load balancing distribution channel 300 corresponds to a task interface 201 . The load balancing distribution channel 300 can be controlled by the load balancing scheduling module 400 to distribute real-time data to each task interface 201 . The task 101 running on the CPU core receives the data transmitted from the load balancing distribution channel 300 through the task interface 201 . Among them, the load balancing scheduling module 400 can control the flow of real-time data and determine which task interface the real-time data is distributed to according to information such as the number of tasks, the running status of each task, the status of the task interface, and the attributes of the real-time data.

The above real-time data processing system may further include a shared cache 500 . In this way, the real-time data does not need to be directly transmitted to the CPU to be processed, but can first enter the shared cache 500 , temporarily store in the shared cache 500 , and then enter the load balancing distribution channel 300 after a certain time interval.

The above-mentioned task interface module 200 is an interface between the load balancing distribution channel 300 and the task 101, and is responsible for transmitting data to each task. An independent cache can also be set in each task interface module 200 to smooth the imbalance between data transmission and task execution speed.

The above real-time data processing system may further include a configuration module 600 . The configuration module 600 is responsible for configuring each module, including configuring the size of the shared cache 500, the scheduling mode of the load balancing scheduling module 400, the size of the independent cache in the task interface module 200, and the number of task interfaces 101, etc.

Task 101 is related to the specific business of the real-time data it processes. Task 101 is generally an application program. The task 101 can interact with the load balancing scheduling module 400 . Whether the task 101 is running normally, the running status of the task 101 and other information can be acquired by the load balancing scheduling module 400 . The attributes of real-time data that can be processed by different tasks 101 may be different. For example, in Ethernet data processing, some tasks 101 process web traffic, and some tasks 101 process P2P audio and video traffic. The load balancing module 400 obtains. The manner of obtaining may be, for example, through automatic interaction, or may be manually configured through the configuration module 600 .

The task interface 201 is an interface between the load balancing distribution channel 300 and the tasks 101 , and is responsible for transmitting data to each task 101 . Since the input speed of the load balancing distribution channel 300 to the task interface 201 and the execution speed of the task 101 to process data must be unbalanced, an independent cache can be set inside the task interface 201 so that the input data to be processed can be buffered. When the speed of task 101 execution is always lower than the rate of input data, the above-mentioned independent data cache will gradually be filled, and the load balancing scheduling module 400 can gradually balance the data to other task interfaces 201 to reduce the number of tasks entering the task interface 201. Data: When the execution speed of the task 101 is always higher than the rate of input data, the independent data cache will gradually become empty, and the load balancing scheduling module 400 can increase the data entering the task interface 201 . Therefore, according to the empty and full state of the independent data cache in the task interface 201 , the load balancing scheduling module 400 can perform effective automatic load balancing.

The task interface 201 logically looks like a data cache queue, but in terms of physical implementation, it also means a data transfer interface for data from the outside to the host memory. Any type of data that is processed in parallel by multiple tasks running on the CPU must be entered into the host memory in real time. From the host peripherals to the host memory, the most common way is a DMA (DirectMemoryAccess, direct memory access) channel that does not require the CPU to participate in the data transfer process.

The input real-time data enters the whole system and is finally processed by multiple tasks 101 running on the multi-core CPU 100 in parallel, and each data particle of the real-time data can have parallelism. The above-mentioned data granularity refers to the granularity when the load balancing module performs data distribution. For example, for Ethernet data processing, each Ethernet data packet is a data particle; for ATM network data processing, each ATM cell is a data particle; for voice signal data processing, each independent voice coded signal is a data particle . Load balancing scheduling is performed on data, and the smallest unit of scheduling is a data particle. For each data granule, either all of it goes through this task interface 201, or all of it enters another task interface 201;

The load balancing load distribution channel 300 can accept the control of the load balancing scheduling module 400 and distribute data to each task interface 201 . The load balancing distribution channel 300 can be connected to the shared cache 500, the load balancing scheduling module 400, and the task interface 201, and can read data from the shared cache 500 module. Information such as running status determines the whereabouts of data particles, that is, the target task interface. If the cache in the current target task interface is not full and data particles can be received, then the data particles are sent to the corresponding task interface. If the cache of the current target task interface is full and cannot receive data granules, the data granules will continue to be stored in the shared cache 500; Data granular readout in 500. That is to say, the load balancing distribution channel 300 transmits data to each task interface 201, and there are two ways to start it, one is that there is real-time data input, and the other is that the cache in the target task interface is full and becomes full.

The load balancing scheduling module 400 can determine which task interface 201 the data granules are distributed to based on information such as the number of tasks 101, the running status of each task 101, the status of the task interface 201, and the attributes of the data granules, and through load balancing The control of the distribution channel 300 transmits data particles to the task interface 201 .

The number of tasks 101 can be used as important information that the balanced scheduling module 400 pays attention to. To evenly schedule data to each task 101, it is necessary to know the number of current tasks. The number of tasks 101 may change dynamically, and the load balancing scheduling module 400 can detect the number of current tasks and adjust dynamically. When the number of tasks 101 increases, the data is allocated to the increased tasks 101; when the number of tasks 101 decreases, the data originally allocated to the tasks that are not running is changed to the tasks that are still running. The method of detecting whether the current task is running can be an automatic detection negotiated with the task, or a manual configuration.

The running state of the task 101 may include information such as whether the task is running normally, CPU resources consumed by the task running, memory resources consumed by the task running, priority of the task running, and data attributes that the task can process. When the task 101 is not running normally, the data can be balanced to other tasks; for the CPU and memory resources consumed by the task running, a threshold can be set, and when it exceeds a certain level, the data can be balanced to other tasks; when the task priority is low, the When the overall CPU and memory resource consumption is large, the data volume of the task 101 can be reduced; when the task 101 has a data attribute that can be processed, it can only transmit data with the same attribute.

The status of the task interface 201 may include the empty and full status of the independent cache in the task interface, the proportion of data in the independent cache, and the like. When there is too much data in the task interface 201, the load balancing scheduling module 400 can schedule the data to other task interfaces 201, so as to avoid the situation that the data accumulates in a certain task interface and is too little in other task interfaces.

The attribute information of the data granule is related to the specific processing content of the task 101 . For example, Ethernet data packets, IPv4/IPv6 data packets, TCP/UDP data packets, quintuples of data packets, TCP flag bits of data packets, application layer protocols of data packets, etc., are all attributes of data particles. Some tasks 101 can only process IPv4 data packets, so the load balancing scheduling module only dispatches IPv4 data packets to this task 101; Mission 101.

The shared cache 500 can be responsible for caching the data flow, and the data flow will eventually be load-balanced and sent to multiple task interfaces. The data belonging to each task interface 201 can share this part of the cache space, so as to improve the utilization rate of the cache. For each task interface 201, there may be an independent cache module. If the processing speed of a certain task is slow, the independent cache in the task interface 201 will gradually fill up. At this time, the shared cache space is needed to temporarily store data. Due to the different execution conditions of each task, some are fast and some are slow, some need to share cache temporary data, and some independent caches in the task interface 201 are sufficient for use and do not need shared caches. The shared cache is designed as a way that each task can be shared. Can maximize the utilization of the cache.

The configuration module 600 can be responsible for the configuration of each module, and can be responsible for providing an interface to the user so that the user can configure various parameters of the entire system. These parameters may include configuring the size of the shared cache, the way of data load balance scheduling, the size of the task interface module cache, the number of task interfaces, and the like.

The scheduling mode of the data load balancing scheduling module 400 may include many configurable parameters. For example, include the data attributes that tasks can process, give priority to providing data to high priority tasks 101, the threshold for automatic load balancing when the independent cache in task interface 201 is about to be full, and automatic load when task 201 consumes too much CPU or memory resources Equilibrium threshold, etc.

FIG. 2 is a schematic flowchart of a multi-core CPU real-time data processing method with automatic load balancing according to an embodiment of the present invention. As shown in Figure 2, the multi-core CPU real-time data processing method of the automatic load balancing of the embodiment of the present invention may comprise steps:

S110: receiving real-time data to be processed by the multi-core CPU;

S120: Distribute the real-time data to be processed to corresponding task interfaces according to the status of all current task interfaces, the task running status of the CPU cores connected to each task interface, and the attributes of the real-time data to be processed, for use with the task The CPU core connected to the interface performs data processing.

The above step S120 can be implemented by the load balancing scheduling module 400 and the load balancing load distribution channel 300 in the multi-core CPU-based real-time data processing system shown in FIG. 1 in combination with other necessary modules (for example, the task interface module 200).

The multi-core CPU real-time data processing method with automatic load balancing in the embodiment of the present invention distributes the real-time pending data according to the status of all current task interfaces, the task running status of the CPU cores connected to each task interface, and the attributes of the real-time pending data. Data processing can make full use of idle resources in some CPU cores and avoid overloading data load in another part of CPU cores, so as to automatically load balance multi-core CPUs to ensure reasonable and full utilization of CPU resources and real-time data processing.

Fig. 3 is a schematic flowchart of a data distribution method in an embodiment of the present invention. As shown in Figure 3, the multi-core CPU real-time data processing method of automatic load balancing shown in Figure 2, in step S120, according to the status of all current task interfaces, the task running status of the CPU core connected with each task interface and the described The attribute of the real-time data to be processed, and the method of distributing the real-time data to be processed to the corresponding task interface may include the steps of:

S121: Identify at least one task interface that can currently receive the real-time data to be processed according to the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed;

S122: Distribute the real-time data to be processed to the identified task interface.

The above step S121 can be realized by the load balancing scheduling module 400 in the above system, and the above step S122 can be realized by combining the necessary modules (such as the task interface module 200 ) through the load balancing load distribution channel 300 in the above system.

In the embodiment of the present invention, by first identifying the task interface capable of receiving the real-time data to be processed, and then distributing the real-time data to be processed to the identified task interface, automatic load balancing of multi-core CPUs can be effectively realized.

FIG. 4 is a schematic flowchart of a multi-core CPU real-time data processing method with automatic load balancing according to another embodiment of the present invention. As shown in Figure 4, the multi-core CPU real-time data processing method of automatic load balancing shown in Figure 2, before step S120, before the described real-time data to be processed is distributed to corresponding task interface, the multi-core CPU of this automatic load balancing real-time The data processing method may also include the steps of:

S130: Cache the real-time data to be processed in a shared cache, to be distributed to the task interface.

The above step S130 can be implemented by the shared cache 500 of the above system.

In the embodiment of the present invention, the above-mentioned shared cache can be shared and used by various tasks, which can maximize the utilization rate of the cache. By using the method of dynamic shared cache to cache the data to be processed of each CPU core, the real-time performance of data processing can be improved, and the loss of unprocessed data can be reduced or avoided.

Fig. 5 is a schematic flowchart of a data distribution method in another embodiment of the present invention. As shown in Figure 5, the multi-core CPU real-time data processing method of automatic load balancing shown in Figure 3, before the step S122, before the real-time data to be processed is distributed to the identified task interface, may include the steps:

S123: Cache the real-time data to be processed to an independent cache of the identified task interface, so as to smooth the imbalance between the speed at which the real-time data to be processed is transmitted to the CPU core and the speed at which it is processed by the CPU core.

The above step S123 can be realized through the independent cache set in the task interface of the above system.

In the embodiment of the present invention, the data in the task interface can be buffered through the independent cache in the task interface, so as to smooth the speed at which the data is transmitted to the CPU core and the speed at which the data is processed by the CPU core.

FIG. 6 is a schematic flowchart of a multi-core CPU real-time data processing method with automatic load balancing according to another embodiment of the present invention. As shown in Figure 6, the multi-core CPU real-time data processing method of automatic load balancing shown in Figure 2, before step S120, before the described real-time data to be processed is distributed to corresponding task interface, the multi-core CPU of this automatic load balancing real-time The data processing method may also include the steps of:

S140: Set the number of task interfaces in the CPU core and/or the CPU according to one or more of the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed The properties of the data processed by the processing task.

The above-mentioned step S140 can be realized by the configuration module 600 in the above-mentioned system, and an interface can be provided to the user to enable the user to set the number of task interfaces in the CPU core and/or the attributes of the data processed by the CPU processing task.

In the embodiment of the present invention, by setting the number of task interfaces in the CPU core and/or the attributes of the data processed by the CPU processing task, the load of the CPU core can be further balanced.

FIG. 7 is a schematic flowchart of a multi-core CPU real-time data processing method with automatic load balancing according to yet another embodiment of the present invention. As shown in Figure 7, the multi-core CPU real-time data processing method of automatic load balancing shown in Figure 4, before step S130, before the described real-time data to be processed is cached to a shared cache, the multi-core CPU real-time data of this automatic load balancing The processing method may also include the steps of:

S150: Set the size of the shared cache according to one or more of the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed.

The above-mentioned step S150 can be realized by the configuration module 600 in the above-mentioned system. By providing an interface to the user, the user can set the shared cache according to one or more of the status of the task interface, the task running status and the attributes of the real-time data to be processed. size.

In the embodiment of the present invention, by setting the size of the shared cache, it is possible to further prevent the loss of data to be processed when there is a lot of data accumulation.

Fig. 8 is a schematic flowchart of a data distribution method in still another embodiment of the present invention. As shown in Figure 8, the multi-core CPU real-time data processing method of automatic load balancing shown in Figure 5, before step S123, before the independent cache of described real-time data to be processed is cached to the task interface of described identification, can comprise the step :

S124: Set the size of the independent cache according to one or more of the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed.

The above-mentioned step S124 can be realized by the configuration module 600 in the above-mentioned system. By providing an interface to the user, the user can set the independent buffer according to one or more of the status of the task interface, the task running status and the attributes of the real-time data to be processed. size.

In the embodiment of the present invention, by setting the size of the independent cache, the utilization rate of the cache can be further improved, and the unbalanced smoothing effect between the speed at which real-time data to be processed is transmitted to the CPU core and the speed at which it is processed by the CPU core can be improved.

In one embodiment, the real-time data to be processed includes multiple data particles. The multi-core CPU real-time data processing method of automatic load balancing shown in Figure 2, step S120, according to the status of all current task interfaces, the task running status of the CPU core connected with each task interface and the attributes of the real-time data to be processed, the The real-time data to be processed is distributed to the corresponding task interface for data processing by the CPU core connected to the task interface, which may include steps:

S1221: According to the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the data particles, distribute each of the data particles to the corresponding task interface in parallel for connection with the task interface CPU cores for data processing.

In the embodiment of the present invention, data is distributed to corresponding task interfaces in units of data particles, which can effectively realize data distribution, and can process data particles with the same attribute in parallel, thereby increasing the speed of real-time data processing.

FIG. 9 is a schematic flowchart of a multi-core CPU real-time data processing method with automatic load balancing according to another embodiment of the present invention. As shown in Figure 9, the multi-core CPU real-time data processing method of automatic load balancing shown in Figure 2 may also include steps:

S160: Detect changes in the number of running tasks of each CPU core;

S170: Redistributing the real-time data to be processed that has been distributed to the task interface according to the number of running tasks changed, so as to balance the data load of each CPU core on the multi-core CPU.

The above step S160 can be implemented by the load balancing scheduling module 400 in the above system. The above step S170 can be realized through the load balancing scheduling module 400 and the load balancing load distribution channel 300 in combination with other necessary modules (eg, the task interface module 200 ).

In the embodiment of the present invention, by detecting the change in the number of running tasks of each CPU core, and dynamically adjusting the CPU load, when the number of tasks increases, the data is allocated to the increased tasks; when the number of tasks decreases, the original allocation The data from tasks that are no longer running are allocated to tasks that are still running, which can adjust the CPU load in time to prevent CPU load imbalance due to changes in the number of tasks.

FIG. 10 is a schematic flowchart of a multi-core CPU real-time data processing method with automatic load balancing according to another embodiment of the present invention. As shown in Figure 10, the multi-core CPU real-time data processing method of automatic load balancing shown in Figure 2 may also include steps:

S180: Determine whether the resource consumption ratio of a CPU core of the multi-core CPU exceeds a set value;

S190: If yes, allocate the real-time data to be processed in the CPU core whose resource consumption ratio exceeds the set value to other CPU cores in the multi-core CPU whose resource consumption ratio does not exceed the set value.

The above step S180 can be implemented by the load balancing scheduling module 400 in the above system.

In the embodiment of the present invention, a threshold is set for the CPU resources consumed by task running, and the data can be balanced to other tasks when the CPU resources exceed a certain level. By redistributing the data to be processed in the CPU core that consumes more resources to the CPU core that consumes less resources, the data load of the CPU can be more balanced.

In a specific embodiment, the above real-time data to be processed is, for example, Ethernet data. The Ethernet data is processed by using the method of the embodiment of the present invention. Taking 8 Ethernet data processing tasks as an example, the data packet attributes and task priorities that can be processed by the tasks are shown in Table 1. The multi-core processor is, for example, 4 CPUs, and the memory is 16 GB. Table 2 shows the CPU occupied by Ethernet traffic processing. The input real-time Ethernet data traffic can be 10Gbps, and its data packet attributes and proportion are shown in Table 3.

task number Corresponding task interface number Processable Packet Properties task priority 1 1 HTTP traffic high 2 2 HTTP traffic high 3 3 HTTP traffic high 4 4 HTTP traffic high 5 5 MAIL traffic Low 6 6 MAIL traffic Low 7 7 P2P audio and video traffic lowest 8 8 P2P audio and video traffic lowest

Table 1 Ethernet packet attributes and task priorities

serial number packet properties CPU usage per 1Gb traffic Memory usage per 1Gb traffic 1 HTTP traffic 20% 1GB 2 MAIL traffic 30% 1GB 3 P2P audio and video traffic 40% 1GB

Table 2 CPU usage when processing Ethernet data traffic

serial number packet properties The proportion flow 1 HTTP traffic 40% 4Gbps 2 MAIL traffic 20% 2Gbps 3 P2P audio and video traffic 30% 3Gbps 4 other traffic 10% 1Gbps

Table 3 Ethernet packet attributes and their proportions

The load balancing is performed by using the method of the embodiment of the present invention, and the scheduling result is obtained, as shown in Table 4. 10% of other traffic is directly discarded by the automatic load balancing module; 40% of HTTP traffic is balanced to task interfaces 1-4, each task interface handles 10% of traffic, and each task consumes 20% of CPU resources. The memory is 1GB; 20% of the MAIL traffic is load-balanced to task interfaces 5-6, each task interface handles 10% of the traffic, each task consumes 20% of CPU resources, and the memory is 1GB; 30% of P2P audio and video Traffic is load-balanced to task interfaces 7-8, each task interface handles 15% of traffic, each task consumes 60% of CPU resources, and memory is 1.5GB. The total consumption of CPU resources is 260%, and the memory is 9GB. The multi-core processor is 4 CPUs, the total resource of the CPU is 400%, and the total resource of the memory is 16GB, which can completely handle the above traffic. The load balancing module has achieved good results.

task interface number Scheduled Packet Properties flow ratio flow CPU usage memory usage 1 HTTP traffic 10% 1Gbps 20% 1GB 2 HTTP traffic 10% 1Gbps 20% 1GB 3 HTTP traffic 10% 1Gbps 20% 1GB 4 HTTP traffic 10% 1Gbps 20% 1GB 5 MAIL traffic 10% 1Gbps 30% 1GB 6 MAIL traffic 10% 1Gbps 30% 1GB 7 P2P audio and video traffic 15% 1.5Gbps 60% 1.5GB 8 P2P audio and video traffic 15% 1.5Gbps 60% 1.5GB total 90% 9Gbps 260% 9GB

Table 4 Scheduling results of the load balancing module

If the situation of the Ethernet data traffic changes, assuming that the input traffic becomes 20Gbps, then the above scheduling changes are shown in Table 5. The total consumption of CPU resources is 520%, and the memory is 18GB, which has exceeded the maximum capacity of multi-core CPU and memory resources. According to the method of the embodiment of the present invention, load balancing can be performed according to the priority of tasks. For example, since the P2P audio and video processing task has the lowest priority, the Ethernet traffic data for P2P audio and video processing will be reduced first, and the scheduling changes are shown in Table 6. After scheduling through the method of the embodiment of the present invention, the P2P audio and video traffic is reduced, 15% of the traffic of 3Gbps in total is discarded, and 7.5% of the traffic of 1.5Gbps is given to each task of task interfaces 7-8, so that the CPU resources of the entire system are occupied is 400%, and the memory usage is 15GB, reaching the maximum processing capacity of the system.

task interface number Scheduled Packet Properties flow ratio flow CPU usage memory usage 1 HTTP traffic 10% 2Gbps 40% 2GB 2 HTTP traffic 10% 2Gbps 40% 2GB 3 HTTP traffic 10% 2Gbps 40% 2GB 4 HTTP traffic 10% 2Gbps 40% 2GB 5 MAIL traffic 10% 2Gbps 60% 2GB 6 MAIL traffic 10% 2Gbps 60% 2GB 7 P2P audio and video traffic 15% 3Gbps 120% 3GB 8 P2P audio and video traffic 15% 3Gbps 120% 3GB total 90% 18Gbps 520% 18GB

Table 5 Scheduling change results

Table 6 Scheduling change results

The multi-core CPU real-time data processing method of automatic load balancing of the present invention can perform automatic load balancing for the application of multi-task real-time data processing using a multi-core CPU, according to the number of tasks started by the multi-core CPU, the running state of the task, and the state of the task interface , and data attributes and other information, automatically distribute the data to be processed to each CPU core, so as to ensure the reasonable and full utilization of CPU resources and the real-time performance of data processing. The method can also use a dynamic shared cache method to cache the data to be processed by each CPU core, so as to improve the real-time performance of data processing and reduce or avoid the loss of unprocessed data.

Based on the same inventive concept as the automatic load balancing multi-core CPU real-time data processing method shown in FIG. 2 , the embodiment of the present application also provides an automatic load balancing multi-core CPU real-time data processing device, as described in the following embodiments. Since the problem-solving principle of the automatic load balancing multi-core CPU real-time data processing device is similar to the automatic load balancing multi-core CPU real-time data processing method, the implementation of the automatic load balancing multi-core CPU real-time data processing device can be found in the automatic load balancing multi-core The implementation of the CPU real-time data processing method will not be repeated.

FIG. 11 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to an embodiment of the present invention. As shown in FIG. 11 , the automatic load balancing multi-core CPU real-time data processing device may include: a real-time data receiving unit 210 and a real-time data processing unit 220 , which are connected to each other.

The real-time data receiving unit 210 is configured to receive real-time data to be processed of the multi-core CPU.

The real-time data processing unit 220 is used for distributing the real-time data to be processed to corresponding task interfaces according to the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed , for the CPU core connected to the task interface to perform data processing.

The multi-core CPU real-time data processing device with automatic load balancing in the embodiment of the present invention distributes real-time pending data according to the status of all current task interfaces, the task running status of the CPU cores connected to each task interface, and the attributes of the real-time pending data. Data processing can make full use of idle resources in some CPU cores and avoid overloading data load in another part of CPU cores, so as to automatically load balance multi-core CPUs to ensure reasonable and full utilization of CPU resources and real-time data processing.

Fig. 12 is a schematic structural diagram of a real-time data processing unit in an embodiment of the present invention. As shown in FIG. 12 , the real-time data processing unit 220 may include: a real-time data flow identification module 221 and a real-time data distribution module 222 , which are connected to each other.

The real-time data flow identification module 221 is used to identify at least one currently capable of receiving the real-time data to be processed according to the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed. task interface.

The real-time data distribution module 222 is used to distribute the real-time data to be processed to the identified task interface.

In the embodiment of the present invention, the real-time data flow identification module first identifies the task interface that can currently receive the real-time data to be processed, and then distributes the real-time data to be processed to the identified task interface through the real-time data distribution module, which can effectively implement multi-core Automatic load balancing of CPUs.

FIG. 13 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to another embodiment of the present invention. As shown in FIG. 13 , the automatic load balancing multi-core CPU real-time data processing device shown in FIG. 11 may further include a shared cache module 230 connected between the real-time data receiving unit 210 and the real-time data processing unit 220 .

The shared cache module 230 is used to cache the real-time data to be processed in a shared cache, to be distributed to the task interface.

In the embodiment of the present invention, the cache space in the above-mentioned shared cache module can be shared by various tasks, which can maximize the utilization rate of the cache. By using the method of dynamic shared cache to cache the data to be processed of each CPU core, the real-time performance of data processing can be improved, and the loss of unprocessed data can be reduced or avoided.

Fig. 14 is a schematic structural diagram of a real-time data processing unit in another embodiment of the present invention. As shown in FIG. 14 , the real-time data processing unit 220 shown in FIG. 12 may further include: an independent cache module 223 connected between the real-time data flow identification module 221 and the real-time data distribution module 222 .

The independent cache module 223 is used to cache the real-time data to be processed to the independent cache of the identified task interface, so as to smooth the imbalance between the speed at which the real-time data to be processed is transmitted to the CPU core and the speed at which it is processed by the CPU core .

In the embodiment of the present invention, the data in the task interface can be buffered by the independent cache module, so as to smooth the speed at which the data is transmitted to the CPU core and the speed at which the data is processed by the CPU core.

FIG. 15 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to yet another embodiment of the present invention. As shown in FIG. 15 , the automatic load balancing multi-core CPU real-time data processing device shown in FIG. 11 may further include: a CPU task configuration unit 240 connected to the real-time data processing unit 220 .

The CPU task configuration unit 240 is used to set the task interface in the CPU core according to one or more of the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed. The number and/or attributes of the data processed by the CPU processing task.

In the embodiment of the present invention, the CPU core load can be further balanced by setting the number of task interfaces in the CPU core and/or the attributes of the data processed by the CPU processing task through the CPU task configuration unit.

FIG. 16 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to another embodiment of the present invention. As shown in FIG. 16 , the automatic load balancing multi-core CPU real-time data processing device shown in FIG. 13 may further include: a shared cache configuration module 250 connected to the shared cache module 230 .

The shared cache configuration module 250 is used to set the size of the shared cache according to one or more of the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed .

In the embodiment of the present invention, the size of the shared cache is set by the shared cache configuration module, which can further prevent the loss of data to be processed when there is a lot of data accumulation.

Fig. 17 is a schematic structural diagram of a real-time data processing unit in another embodiment of the present invention. As shown in FIG. 17 , the real-time data processing unit 220 shown in FIG. 14 may further include an independent cache configuration module 224 connected to the above-mentioned independent cache module 223 .

The independent cache configuration module 224 is used to set the size of the independent cache according to one or more of the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed .

In the embodiment of the present invention, by setting the size of the independent cache through the independent cache configuration module, the utilization rate of the cache can be further improved, and the imbalance between the speed at which real-time data to be processed is transmitted to the CPU core and the speed at which it is processed by the CPU core can be improved. smoothing effect.

In one embodiment, the real-time data to be processed includes multiple data particles. In the automatic load balancing multi-core CPU real-time data processing device shown in FIG. 11 , the real-time data processing unit 220 may include: a data granule processing module. The methods in the above embodiments can be processed on data granules. For example, data granules replace the aforementioned real-time data to be processed.

The data granule processing module 2221 is used to distribute the data granules to corresponding task interfaces in parallel according to the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the data granules. For data processing by the CPU core connected to the task interface.

In the embodiment of the present invention, the data granule processing module distributes data to corresponding task interfaces in units of data granules, which can effectively realize data distribution, and can process data granules with the same attribute in parallel, thereby improving the speed of real-time data processing.

FIG. 18 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to another embodiment of the present invention. As shown in Figure 18, the multi-core CPU real-time data processing device of automatic load balancing shown in Figure 11 can also include: a task state detection unit 260 and a first data redistribution unit 270, the two are connected to each other, and the task state detection unit 260 It is connected with the above-mentioned real-time data processing unit 220.

The task state detection unit 260 is used to detect the change in the number of running tasks of each CPU core.

The first data redistribution unit 270 is configured to redistribute the real-time data to be processed that has been distributed to the task interface according to the changed number of running tasks, so as to balance the data load of each CPU core on the multi-core CPU.

In the embodiment of the present invention, the change in the number of running tasks of each CPU core is detected by the task state detection unit, and the CPU load is dynamically adjusted by the first data redistribution unit. When the number of tasks increases, the data is allocated to the increased tasks. When the number of tasks decreases, the data originally allocated to tasks that are not running can be allocated to tasks that are still running, so that the CPU load can be adjusted in time to prevent CPU load imbalance due to changes in the number of tasks.

FIG. 19 is a schematic structural diagram of a multi-core CPU real-time data processing device with automatic load balancing according to another embodiment of the present invention. As shown in Figure 19, the multi-core CPU real-time data processing device of automatic load balancing shown in Figure 11 can also include: CPU resource status detection unit 280 and the second data redistribution unit 290, the two are connected to each other, CPU resource status detection The unit 280 is connected with the above-mentioned real-time data processing unit 220 .

The CPU resource state detection unit 280 is used to determine whether the resource consumption ratio of a CPU core of the multi-core CPU exceeds a set value.

The second data redistribution unit 290 is configured to, if so, distribute the real-time data to be processed in the CPU core whose resource consumption ratio exceeds the set value to the multi-core CPU whose resource consumption ratio does not exceed the set value other CPU cores.

In the embodiment of the present invention, through the CPU resource state detection unit, a threshold can be set for the CPU resources consumed by task operation, and the data can be balanced to other tasks when the CPU resources exceed a certain level. By redistributing the data to be processed in the CPU core that consumes more resources to the CPU core that consumes less resources, the data load of the CPU can be more balanced.

The multi-core CPU real-time data processing device with automatic load balancing of the present invention can perform automatic load balancing for the application of multi-task real-time data processing using a multi-core CPU, and can perform automatic load balancing according to the number of tasks started by the multi-core CPU, the running status of the tasks, and the status of the task interface , and data attributes and other information, automatically distribute the data to be processed to each CPU core, so as to ensure the reasonable and full utilization of CPU resources and the real-time performance of data processing. The device can also use a dynamic shared cache module to cache the data to be processed by each CPU core, so as to improve the real-time performance of data processing and reduce or avoid the loss of unprocessed data.

Those skilled in the art should understand that the embodiments of the present invention may be provided as methods, systems, or computer program products. Accordingly, the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the scope of the present invention. Protection scope, within the spirit and principles of the present invention, any modification, equivalent replacement, improvement, etc., shall be included in the protection scope of the present invention.

Claims (10)

1. a kind of multi-core CPU real-time data processing method of automatic load balancing, it is characterized in that, comprising: Receive real-time data to be processed by multi-core CPU; According to the status of all current task interfaces, the task running status of the CPU core connected to each task interface and the attributes of the real-time data to be processed, the real-time data to be processed is distributed to the corresponding task interface for connection with the task interface CPU cores for data processing. 2. the multi-core CPU real-time data processing method of automatic load balancing as claimed in claim 1, is characterized in that, according to the state of all current task interfaces, the task running state of the CPU core that is connected with each task interface and described real-time to be processed The attributes of the data, and distribute the real-time data to be processed to the corresponding task interface, including: Identify at least one task interface that can currently receive the real-time data to be processed according to the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed; Distributing the real-time pending data to identified task interfaces. 3. the multi-core CPU real-time data processing method of automatic load balancing as claimed in claim 1, is characterized in that, before described real-time data to be processed is distributed to corresponding task interface, also comprises: The real-time data to be processed is cached in a shared cache, ready to be distributed to the task interface. 4. the multi-core CPU real-time data processing method of automatic load balancing as claimed in claim 2, is characterized in that, before the task interface that described real-time data to be processed is distributed to identification, comprises: The real-time data to be processed is cached in an independent cache of the identified task interface, so as to smooth the imbalance between the speed at which the real-time data to be processed is transmitted to the CPU core and the speed at which it is processed by the CPU core. 5. the multi-core CPU real-time data processing method of automatic load balancing as claimed in claim 1, is characterized in that, also comprises: Set the number of task interfaces in the CPU core and/or CPU processing tasks according to one or more of the status of all current task interfaces, the task running status of the CPU cores connected to each task interface, and the attributes of the real-time data to be processed Attributes of the processed data. 6. the multi-core CPU real-time data processing method of automatic load balancing as claimed in claim 3, is characterized in that, also comprises: The size of the shared cache is set according to one or more of the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed. 7. the multi-core CPU real-time data processing method of automatic load balancing as claimed in claim 4, is characterized in that, also comprises: The size of the independent cache is set according to one or more of the status of all current task interfaces, the task running status of the CPU core connected to each task interface, and the attributes of the real-time data to be processed. 8. the multi-core CPU real-time data processing method of automatic load balancing as claimed in claim 1, is characterized in that, described real-time data to be processed comprises a plurality of data particles; According to the status of all current task interfaces, the task running status of the CPU core connected to each task interface and the attributes of the real-time data to be processed, the real-time data to be processed is distributed to the corresponding task interface for connection with the task interface CPU core for data processing, including: According to the status of all current task interfaces, the task running status of the CPU core connected to each task interface and the attributes of the data particles, distribute each of the data particles to the corresponding task interface in parallel for the CPU connected to the task interface The core performs data processing. 9. the multi-core CPU real-time data processing method of automatic load balancing as claimed in claim 1, is characterized in that, also comprises: Detect changes in the number of running tasks of each CPU core; According to the number of running tasks after the change, the real-time data to be processed that has been distributed to the task interface is redistributed to balance the data load of each CPU core on the multi-core CPU. 10. the multi-core CPU real-time data processing method of automatic load balancing as claimed in claim 1, is characterized in that, also comprises: judging whether the resource consumption ratio of a CPU core of the multi-core CPU exceeds a set value; If so, allocate the real-time data to be processed in the CPU core whose resource consumption ratio exceeds the set value to other CPU cores in the multi-core CPU whose resource consumption ratio does not exceed the set value.