Resource management system and resource management method

Abstract

A resource management system and a resource management method. The resource management system includes a plurality of worker nodes and a master node that are communicatively connected to each other, and each of the worker nodes and the master node includes a resource pool, a resource monitor, a collector agent module, a traffic guidance reasoning module, a request processing module, a permission controller, and a critical value adjuster. The master node also includes a data collector, an artificial intelligence computing module, and a load balancer. The resource management system and resource management method provided by the present invention have an innovative distributed resource management architecture, can mix and apply centralized and distributed resource management mechanisms, can be compatible with existing centralized resource management mechanisms, and can further avoid delays that may occur in non-network transmission environments and centralized decision-making mechanisms, and achieve the purpose of load balancing and resource utilization optimization.

Description

Technical Field

[0001] The present invention relates to a resource management system and a resource management method, and in particular to a resource management system and a resource management method capable of hybrid application of centralized and distributed resource management mechanisms. Background Art

[0002] With the rise of 5G private networks, hybrid cloud computing has become mainstream. Edge computing devices in computing devices are often built in the application field in a decentralized manner, and then connected to private clouds or public clouds through the network. In order to effectively balance the use of computing or network resources, the current "cluster" method is mostly used to select a master node (MN) to manage the work demand dispatch of the work point (WN).

[0003] However, network transmission delay and data compression delay make it impossible to meet the service quality requirements of end users, and cause resource waste and power consumption. Although many distributed resource management mechanism inventions have been proposed, there are still the following problems:

[0004] 1. The existing distributed resource management mechanism cannot coexist with the centralized cluster resource management mechanism.

[0005] 2. In the existing distributed resource management mechanism, in order to achieve fast dispatch of load balancing, the resource status of each working node needs to be shared, resulting in additional network traffic and decision delays; the balancing effect will also be affected by the transfer of work hotspots and the long convergence time, which is prone to a pendulum effect.

[0006] 3. The existing distributed resource management mechanism partially adopts the reserved "shareable" / "non-shareable" resource policy for distributed management. Its fixed and inflexible reservation mechanism leads to waste of resource use and lower efficiency.

[0007] 4. In many Mesh Edge Cloud solutions that use wireless relay interconnection, either purely centralized or purely distributed resource management is not conducive to the quality and efficiency of resource utilization services.

[0008] Therefore, how to overcome the above-mentioned defects by improving the resource management mechanism has become one of the important issues that this business wants to solve. Summary of the invention

[0009] The technical problem to be solved by the present invention is to provide a resource management system and a resource management method capable of hybrid application of centralized and distributed resource management mechanisms in view of the deficiencies in the prior art.

[0010] In order to solve the above technical problems, one of the technical solutions adopted by the present invention is to provide a resource management system, which includes a plurality of worker nodes and a master node, wherein the worker nodes and the master node are connected to each other in communication, wherein the worker nodes and the master node each include a resource pool, a resource monitor, a collector agent module, a traffic guidance reasoning module, a request processing module, a permission controller and a critical value adjuster. The resource pool divides the corresponding system resources into sharable resources and non-sharable resources. The resource monitor records the usage status of the resource pool and generates resource usage monitoring data. The collector agent module is used to transmit the generated resource usage monitoring data. The traffic guidance reasoning module receives the resource usage monitoring data and calculates the weight ratio of non-sharable resources and sharable resources based on the resource usage monitoring data to generate a permission control policy. The request processing module receives a plurality of resource usage requests from a plurality of electronic devices. The permission controller allocates the resource usage requests to non-sharable resources and sharable resources according to the permission control policy. The critical value adjuster is used to adjust the allocation ratio of sharable resources and non-sharable resources. Among them, the master node also includes a data collector, an artificial intelligence computing module and a load balancer. The data collector receives resource usage monitoring data of the worker nodes and the master node from the collector agent modules of the worker nodes and the master node to generate resource usage statistics. The artificial intelligence calculation module is configured to input the resource usage statistics into the trained artificial intelligence model to generate a plurality of critical value data for respectively allocating the resource pools of the worker nodes and the master node. Among them, each critical value data defines the allocation ratio in each resource pool, and the traffic guidance reasoning modules of the worker nodes and the master node are configured to respectively control the corresponding critical value adjusters to adjust the sharable resources and non-sharable resources of each resource pool according to the optimized critical value data. The load balancer is configured to balance the resource usage requests allocated to the sharable resources in the worker nodes and the master node to the sharable resources of the worker nodes and the master node.

[0011] In order to solve the above-mentioned technical problems, another technical solution adopted by the present invention is to provide a resource management method, which is applicable to multiple worker nodes and a main responsible node that are communicatively connected to each other, and these worker nodes and the main responsible node each include a resource pool, a resource monitor, a collector agent module, a traffic guidance reasoning module, a request processing module, a permission controller, and a critical value adjuster. The main responsible node also includes a data collector, an artificial intelligence computing module and a load balancer. The resource management method includes: executing an internal resource management process for these worker nodes and the main responsible node, and the internal resource management process includes: configuring the resource pool to divide the corresponding system resource into a sharable resource and a non-shareable resource; configure the resource monitor to record the usage status of the resource pool and generate a resource usage monitoring data; configure the collector agent module to transmit the generated resource usage monitoring data to the traffic guidance reasoning module; configure the traffic guidance reasoning module to receive the resource usage monitoring data and calculate the weight ratio of the non-shareable resource and the sharable resource according to the resource usage monitoring data to generate a permission control policy; configure the request processing module to receive multiple resource usage requests from multiple electronic devices; and configure the permission controller to allocate the resource usage requests to the non-shareable resource and the sharable resource according to the permission control policy. Wherein, the critical value adjuster is used to adjust an allocation ratio of the sharable resource and the non-shareable resource. The resource management method also includes executing an external resource management process for the worker nodes and the main responsible node, and the external resource management process includes: configuring the data collector to receive the resource usage monitoring data of the worker nodes and the main responsible node from the collector agent modules of the worker nodes and the main responsible node to generate a resource usage statistical data; configuring the artificial intelligence computing module to input the resource usage statistical data into a trained artificial intelligence model to generate a plurality of critical value data for respectively allocating the resource pools of the worker nodes and the main responsible node, wherein each of the critical value data defines the allocation ratio in each of the resource pools; configuring the traffic guidance reasoning modules of the worker nodes and the main responsible node to respectively control the corresponding critical value adjusters to adjust the sharable resources and the non-sharable resources of each of the resource pools according to the optimized critical value data; and configuring the load balancer to balance the resource usage requests allocated to the sharable resources in the worker nodes and the main responsible node to the sharable resources of the worker nodes and the main responsible node.

[0012] One of the beneficial effects of the present invention is that the resource management system and resource management method provided by the present invention have an innovative distributed resource management architecture, which can mix and match centralized and distributed resource management mechanisms. Among them, the internal resource management process realizes a fast and low-latency distributed resource management architecture and a pipelined synchronous execution mechanism. At the same time, based on the architecture of the resource pool, the dynamic ratio adjustment of adjustable sharable resources and non-sharable resources is realized.

[0013] In addition, the resource management system and resource management method provided by the present invention integrate the sharable resources of each worker node to form a sharable resource pool architecture, which is compatible with the existing centralized resource management mechanism. On the other hand, the external resource management process uses artificial intelligence big data analysis to further realize a dynamic adjustment mechanism for resource critical values, which can adjust the allocation ratio of sharable resources and non-sharable resources.

[0014] In terms of performance and resource utilization, the resource management system and resource management method provided by the present invention have a distributed internal resource management process and a pipelined synchronous execution mechanism, so that local requests can immediately use local resources, which can further avoid delays that may occur in a network-free transmission environment and a centralized decision-making mechanism.

[0015] Furthermore, the dynamic adjustment mechanism of resource critical values ​​achieved through big data analysis based on artificial intelligence can help work nodes with heavier loads achieve load balancing and resource utilization optimization through vAPP transfer (virtual App Migration) of the program execution location and increasing the proportion of non-shared resources.

[0016] To further understand the features and technical contents of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are only for reference and description and are not intended to limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 FIG. 4 is a block diagram of a resource management system according to an embodiment of the present invention.

[0018] Figure 2 Schematic diagram of the internal resource management process according to an embodiment of the present invention.

[0019] Figure 3 FIG. 4 is a flow chart of the internal resource management process according to the present invention.

[0020] Figure 4 Schematic diagram of a neural network-like model according to an embodiment of the present invention.

[0021] Figure 5Schematic diagram of an external resource management process according to an embodiment of the present invention.

[0022] Figure 6 Flow chart of an external resource management process according to an embodiment of the present invention.

[0023] Main component symbols:

[0024] 1 Resource Management System

[0025] AC0, AC1, ..., ACi License Controller

[0026] AI artificial intelligence computing module

[0027] AIM Artificial Intelligence Model

[0028] C1~C3 time information

[0029] CA0, CA1, …, CAi Collector Agent Modules

[0030] d1~d3 User demand resources

[0031] DC Data Collector

[0032] Hid1, hid2, hid3 hidden layers

[0033] HIS historical resource usage statistics

[0034] Inp Input layer

[0035] LB Load Balancer

[0036] Nb1~Nb3 Non-shareable resources Remaining resources

[0037] NSR0, NSR1, ..., NSRi non-shareable resources

[0038] O1-Tx~O3-Tx resource usage status

[0039] Out Output Layer

[0040] RD0, RD1, ..., RDi resource usage monitoring data

[0041] RH0, RH1, ..., RHi request processing module

[0042] RM0, RM1, …, RMi resource monitors

[0043] RP0, RP1, ..., RPi resource pool

[0044] RQ0, RQ1, ..., RQi resource usage requests

[0045] RS resource usage statistics

[0046] Steps S30~S35, S60~S63

[0047] Sb1~Sb3 Shareable resources Remaining resources

[0048] SR0, SR1, ..., SRi can share resources

[0049] TA0, TA1, ..., TAi threshold adjusters

[0050] TS0, TS1, ..., Tsi Traffic steering inference module

[0051] MN Master Node

[0052] WN1, ..., WNi worker nodes

[0053] z[1][1]~z[1][5], …, z[3][1]~z[3][5] output vector DETAILED DESCRIPTION

[0054] The following is an explanation of the implementation methods of the "resource management system and resource management method" disclosed in the present invention through specific specific embodiments. Those skilled in the art can understand the advantages and effects of the present invention from the contents disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and the details in this specification can also be modified and changed in various ways based on different viewpoints and applications without departing from the concept of the present invention. In addition, the drawings of the present invention are only simple schematic illustrations and are not depicted in actual size. It is stated in advance. The following implementation methods will further explain the relevant technical content of the present invention in detail, but the disclosed content is not intended to limit the scope of protection of the present invention. In addition, the term "or" used in this article may include any one or more combinations of the associated listed items depending on the actual situation.

[0055] Figure 1 FIG. 1 is a block diagram of a resource management system according to an embodiment of the present invention. Figure 1 As shown, an embodiment of the present invention provides a resource management system 1, which includes a plurality of worker nodes WN1, ..., WNi and a master node MN, which are communicatively connected with each other.

[0056] The worker node WN1 includes a resource pool RP1, a resource monitor RM1, a collector agent module CA1, a traffic steering reasoning module TS1, a request processing module RH1, an admission controller AC1, and a critical value adjuster TA1. The remaining worker nodes and the master node MN basically have the same configuration as the worker node WN1. For example, the worker node WNi includes a resource pool RPi, a resource monitor RMi, a collector agent module CAi, a traffic steering reasoning module TSi, a request processing module RHi, an admission controller ACi, and a critical value adjuster TAi, and the master node MN includes a resource pool RP0, a resource monitor RM0, a collector agent module CA0, a traffic steering reasoning module TS0, a request processing module RH0, an admission controller AC0, and a critical value adjuster TA0. Therefore, if necessary, otherwise it will not be repeated below.

[0057] It should be noted that the resource pool, resource monitor, collector agent module, traffic steering inference module, request processing module, admission controller and threshold adjuster used in the worker node WN1 and the master node MN can realize their functions in the form of hardware, software or solid. When realized in the form of software, the program code for executing the functions of the above elements or modules can be stored in the memory built into the general computer system, and the above elements or modules are realized through the execution of the processor.

[0058] Resource pool RP1 divides the system resources corresponding to worker node WN1 into sharable resources SR1 and non-sharable resources NSR1. Taking worker node WN1 as an example of a common computer system, worker node WN1 may include a central processing unit, a memory, a flash memory, a network transmission interface, a hard disk and a graphics processor. Therefore, the above-mentioned system resources may include the processing capacity of the central processing unit, the available capacity of the memory, the available capacity of the network, the available capacity of the flash memory, the available capacity of the hard disk and the processing capacity of the graphics processor, but the present invention is not limited to this. The present invention mainly integrates the sharable resources SR1, ..., SRi of each worker node WN1, ..., WNi to form a sharable resource pool architecture, which is compatible with the existing centralized resource management mechanism.

[0059] The request processing module RH1 receives multiple resource usage requests RQ1 from multiple electronic devices. The worker node WN1 can be connected to multiple electronic devices, and these multiple electronic devices are local devices relative to the worker node WN1. In some embodiments, the non-sharable resource NSR1 is only used to serve the local device, that is, only used to respond to the request RQ1 from the local device, and the sharable resource SR1 can be used to serve the resource usage requests from the electronic devices on all worker nodes WN1, ..., WNi and the master node MN.

[0060] The resource monitor RM1 is used to record the usage status of the resource pool RP1 and generate resource usage monitoring data RD1. For example, the resource monitor RM1 can continuously collect resource usage requests RQ1 of multiple electronic devices ED1 (ie, local devices) for a long time to form the resource usage monitoring data RD1.

[0061] The collector agent module CA1 is used to transmit the generated resource usage monitoring data RD1, and the traffic steering reasoning module TS1 receives the resource usage monitoring data RD1, and calculates the weight ratio WNSR1 of the non-sharable resource NSR1 and the weight ratio WSR1 of the sharable resource SR1 according to the resource usage monitoring data RD1, so as to generate a permission control policy. As mentioned above, the resource usage monitoring data RD1 is continuously collected for a long time by the resource monitor RM1, and may include the time point corresponding to the resource usage request collected by the local device. Therefore, the traffic steering reasoning module TS1 can further obtain the current time, and calculate the weight ratio WNSR1 of the non-sharable resource NSR1 and the weight ratio WSR1 of the sharable resource SR1 according to the current time and the number of resource usage requests corresponding to the current time in the resource usage monitoring data RD1.

[0062] For example, if the current time corresponds to a period when the number of resource usage requests of the local device is relatively large, more non-sharable resources NR1 tend to be reserved. Therefore, the weight ratio WNSR1 of the non-sharable resource NSR1 will be greater than the weight ratio WSR1 of the sharable resource SR1. However, it should be noted that the sum of the weight ratio WNSR1 of the non-sharable resource NSR1 and the weight ratio WSR1 of the sharable resource SR1 will be 1. Next, the admission control policy is used to enable the admission controller AC1 to decide how to allocate these resource usage requests RQ1 to the non-sharable resource NSR1 and the sharable resource SR1. In some embodiments, the traffic steering reasoning module TS1 calculates the load of the non-sharable resource NSR1 and the sharable resource SR1 based on the resource usage monitoring data RD1 to generate a admission control policy. For example, the admission control policy may include the following items:

[0063] 1. If the load of the non-sharable resource NSR1 is less than the overload threshold, the resource usage requests RQ1 are allocated to the non-sharable resource NSR1.

[0064] 2. If the load of the non-sharable resource NSR1 is greater than or equal to the overload threshold, and the random value associated with the resource usage requests RQ1 is less than or equal to the weight ratio WNSR1 of the non-sharable resource NSR1, the resource usage requests RQ1 are allocated to the non-sharable resource NSR1. In other words, in the case of overload, the resource usage request RQ1 to be determined is associated with a random value between 0 and 1, and the random value is compared with the weight ratio WNSR1 of the non-sharable resource NSR1.

[0065] 3. If the load of the non-sharable resource NSR1 is greater than or equal to the overload threshold, and the random value associated with the resource usage requests RQ1 is greater than the weight ratio WNSR1 of the non-sharable resource NSR1, the resource usage requests RQ1 are allocated to the sharable resource SR1.

[0066] 4. If the number of resource usage requests RQ1 is greater than the current available resources in resource pool RP1, reject these resource usage requests.

[0067] Therefore, the admission controller AC1 allocates the resource usage requests RQ1 to the non-sharable resource NSR1 and the sharable resource SR1 according to the admission control policy. Figure 1 The threshold adjuster TA1 is used to adjust the allocation ratio of the sharable resource SR1 and the non-sharable resource NSR1 . However, in the internal resource management process to be described below, the threshold adjuster TA1 will not be used.

[0068] Please refer to Figure 2 and Figure 3 , Figure 2 is a schematic diagram of an internal resource management process according to an embodiment of the present invention, Figure 3 Flow chart of the internal resource management process according to the present invention. As shown in the figure, under this configuration, an internal resource management process can be executed for modules such as resource pool RP1, resource monitor RM1, collector agent module CA1, traffic steering reasoning module TS1, request processing module RH1, admission controller AC1 and threshold adjuster TA1 of worker node WN1 to pipeline resource management and load balancing mechanisms. The internal resource management process includes:

[0069] Step S30: configuring the resource pool RP1 to divide the corresponding system resources into sharable resources SR1 and non-sharable resources NSR1.

[0070] Step S31: configure the resource monitor RM1 to record the usage status of the resource pool RP1 and generate resource usage monitoring data RD1.

[0071] Step S32: Configure the collector agent module CA1 to transmit the resource usage monitoring data RD1 to the traffic steering reasoning module TS1.

[0072] Step S33: configure the traffic steering inference module TS1 to receive the resource usage monitoring data RD1, and calculate the weight ratio of the non-sharable resource NSR1 and the sharable resource SR1 according to the resource usage monitoring data RD1 to generate a permission control policy.

[0073] Step S34: The configuration request processing module receives multiple resource usage requests from multiple electronic devices.

[0074] Step S35: configure the admission controller to allocate the resource usage requests to the non-sharable resources and the sharable resources according to the admission control policy.

[0075] Therefore, the internal resource management process can achieve a fast and low-latency distributed resource management architecture and a pipelined synchronous execution mechanism, so that local requests can immediately use local resources, further avoiding delays that may occur in a non-network transmission environment and a centralized decision-making mechanism.

[0076] Please refer to Figure 1 The following further describes the master node MN of the resource management system 1 of the present invention. Figure 1 As shown, the master node MN has the same architecture as the worker nodes WN1, ..., WNi, and can therefore be used to process local requests relative to the master node MN. In addition, the master node MN also includes a data collector DC, an artificial intelligence computing module AI, and a load balancer LB.

[0077] The data collector DC receives resource usage monitoring data RD0, RD1, ..., RDi from the collector agent modules CA1, ..., CAi of the worker nodes WN1, ..., WNi and the collector agent module CA0 of the master node to generate resource usage statistical data RS.

[0078] The artificial intelligence calculation module AI is configured to input the resource usage statistics RS into the trained artificial intelligence model to generate critical value data for resource pools RP1, ..., RPi of worker nodes WN1, ..., WNi and resource pool RP0 of the master node MN. The trained artificial intelligence model is generated by training the artificial intelligence model AIM with the historical resource usage statistics HIS. In the embodiment of the present invention, the artificial intelligence model AIM can be a neural network model.

[0079] Please refer to Figure 4 , which is a schematic diagram of a neural network model according to an embodiment of the present invention. Figure 4As shown, the neural network model includes an input layer Inp, an output layer Out, and multiple hidden layers hid1, hid2, hid3, and the input layer Inp, the output layer Out, and the hidden layers hid1, hid2, hid3 each include multiple neurons. Figure 4 For example, the input layer Inp and the output layer Out each include 15 neurons, and the input historical resource usage statistical data HIS include user demand resources d1~d3, non-sharable resource remaining resources Nb1~Nb3, sharable resource remaining resources Sb1~Sb3, time information C1~C3, and resource usage status O1-Tx~O3-Tx at a specific time Tx.

[0080] In other words, three pieces of data can be taken from each field of the historical resource usage statistical data HIS as the aforementioned user demand resources d1~d3, non-shareable resource remaining resources Nb1~Nb3, shareable resource remaining resources Sb1~Sb3, time information C1~C3, and resource usage status O1-Tx~O3-Tx at a specific time (Tx) to train the neural network model, and give the neural network model a corresponding optimized computing allocation method (such as Figure 4 The output vectors z[1][1]~z[1][5], …, z[3][1], …, z[3][5]) in the training dataset can eventually produce a trained artificial intelligence model.

[0081] Then, the resource usage statistics RS generated by long-term data collection is input into Figure 4 The multi-layer neural network and the selected artificial intelligence / or machine learning model, the output critical value data can be applied to the resource pools RP0, RP1, ..., RPi of the main node MN and the worker nodes WN1, ..., WNi, for example, the critical value adjusters TA0, TA1, ..., TAi are adjusted according to the different allocation ratios between the non-shareable resources NSR and the shareable resources SR after a specific time. It should be noted that, taking the worker node WN1 as an example, the allocation ratio mentioned above refers to the ratio of the shareable resource SR0 to the non-shareable resource NSR0 obtained after the critical value adjuster TA1 adjusts the resource pool RP1 according to the critical value data output by the artificial intelligence / or machine learning model, and the allocation ratio is different from the weight ratio mentioned above. The weight ratio WNSR1 of the non-shareable resource NSR1 and the weight ratio WSR1 of the shareable resource SR1 can be used to enable the admission controller AC1 to decide how to allocate the multiple resource usage requests RQ1 received by the admission controller AC1 to one of the shareable resource SR0 and the non-shareable resource NSR0.

[0082] Therefore, the dynamic adjustment mechanism of resource critical values ​​achieved through big data analysis of artificial intelligence can help work nodes with heavy loads to achieve load balancing and resource utilization optimization through vAPP transfer (virtual App Migration) of the program execution location and increasing the proportion of non-shared resources.

[0083] After determining the allocation ratio of non-shareable resources and shareable resources in resource pools RP0, RP1, …, RPi based on the critical value data, the load balancer LB can further balance the resource usage requests RQ0, RQ1, …, RQi in the master node MN and the worker nodes WN1, …, WNi that are allocated to the shareable resources SR0, SR1, …, SRi in the aforementioned internal resource management process to the shareable resources SR0, SR1, …, SRi of the master node WN and the worker nodes WN1, …, WNi.

[0084] Therefore, under the above architecture, an external resource management process can be executed for the master node MN and the worker nodes WN1, ..., WNi. Figure 5 and Figure 6 , Figure 5 is a schematic diagram of an external resource management process according to an embodiment of the present invention, Figure 6 This is a flow chart of an external resource management process according to an embodiment of the present invention. As shown in the figure, the external resource management process includes:

[0085] Step S60: Configure the data collector DC to receive resource usage monitoring data RD0, RD1, ..., RDi from the collector agent modules CA0, CA1, ..., CAi of the master node MN and the worker nodes WN1, ..., WNi to generate resource usage statistics RS.

[0086] Step S61: Configure the artificial intelligence computing module AI to input the resource usage statistics RS into the trained artificial intelligence model to generate optimized critical value data for resource pools RP0, RP1, ..., RPi for allocating the master node MN and the worker nodes WN1, ..., WNi respectively.

[0087] Step S62: Configure the traffic steering reasoning modules TS0, TS1, ..., TSi of the master node MN and the worker nodes WN1, ..., WNi to control the corresponding critical value adjusters TA0, TA1, ..., TAi to adjust the allocation ratios of the shareable resources SR0, SR1, ..., SRi and the non-shareable resources NSR0, NSR1, ..., NSRi of the resource pools RP0, RP1, ..., RPi according to the optimized critical value data.

[0088] Step S63: Configure the load balancer LB to balance the resource usage requests RQ0, RQ1, ..., RQi allocated to the sharable resources SR0, SR1, ..., SRi in the master node MN and the worker nodes WN1, ..., WNi to the sharable resources SR0, SR1, ..., SRi of the master node WN and the worker nodes WN1, ..., WNi.

[0089] [Beneficial Effects of Embodiments]

[0090] One of the beneficial effects of the present invention is that the resource management system and resource management method provided by the present invention have an innovative distributed resource management architecture, which can mix and match centralized and distributed resource management mechanisms. Among them, the internal resource management process realizes a fast and low-latency distributed resource management architecture and a pipelined synchronous execution mechanism. At the same time, based on the architecture of the resource pool, the dynamic ratio adjustment of adjustable sharable resources and non-sharable resources is realized.

[0091] In addition, the resource management system and resource management method provided by the present invention integrate the sharable resources of each worker node to form a sharable resource pool architecture, which is compatible with the existing centralized resource management mechanism. On the other hand, the external resource management process uses artificial intelligence big data analysis to further realize a dynamic adjustment mechanism for resource critical values, which can adjust the allocation ratio of sharable resources and non-sharable resources.

[0092] In terms of performance and resource utilization, the resource management system and resource management method provided by the present invention have a distributed internal resource management process and a pipelined synchronous execution mechanism, so that local requests can immediately use local resources, which can further avoid delays that may occur in a network-free transmission environment and a centralized decision-making mechanism.

[0093] Furthermore, the dynamic adjustment mechanism of resource critical values ​​achieved through big data analysis based on artificial intelligence can help work nodes with heavier loads achieve load balancing and resource utilization optimization through vAPP transfer (virtual App Migration) of the program execution location and increasing the proportion of non-shared resources.

[0094] The contents disclosed above are only preferred feasible embodiments of the present invention, and are not intended to limit the scope of the claims of the present invention. Therefore, all equivalent technical changes made using the contents of the present invention description and drawings are included in the scope of the claims of the present invention.

Claims

1. A resource management system, comprising: A plurality of worker nodes and a master node, wherein the worker nodes and the master node are communicatively connected to each other, wherein the worker nodes and the master node each include: a resource pool, wherein the resource pool divides a corresponding system resource into a sharable resource and a non-sharable resource; a resource monitor, the resource monitor records the usage status of the resource pool and generates resource usage monitoring data; a collector agent module, the collector agent module being used to transmit the generated resource usage monitoring data; a flow guidance reasoning module, the flow guidance reasoning module receiving the resource usage monitoring data, and calculating a weight ratio of the non-sharable resource and the sharable resource according to the resource usage monitoring data to generate a permission control policy; a request processing module, the request processing module receiving a plurality of resource usage requests from a plurality of electronic devices; an admission controller that allocates the resource usage requests to the non-sharable resource and the sharable resource according to the admission control policy; and a threshold value adjuster, the threshold value adjuster is used to adjust an allocation ratio of the sharable resource and the non-sharable resource; The main responsible nodes also include: a data collector, which receives the resource usage monitoring data of the worker nodes and the master node from the collector agent modules of the worker nodes and the master node to generate resource usage statistics; an artificial intelligence computing module, the artificial intelligence computing module being configured to input the resource usage statistics data into a trained artificial intelligence model to generate a plurality of critical value data respectively used to allocate the resource pools of the worker nodes and the master node, wherein each of the critical value data defines the allocation ratio in each of the resource pools, and the traffic steering reasoning modules of the worker nodes and the master node are configured to control the corresponding critical value adjusters to adjust the sharable resources and the non-sharable resources of each of the resource pools according to the optimized critical value data; and A load balancer is configured to balance the resource usage requests allocated to the sharable resources in the worker nodes and the master node to the sharable resources in the worker nodes and the master node.

2. A resource management system as described in claim 1, wherein the traffic steering reasoning module also calculates the load of the non-sharable resources and the sharable resources based on the resource usage monitoring data to generate the permission control policy, and the weight ratio of the non-sharable resources and the sharable resources of each resource pool is added up to 1.

3. The resource management system as claimed in claim 2, wherein the admission control policy comprises: If the load of the non-sharable resource is less than an overload threshold, allocating the resource usage requests to the non-sharable resource; If the load of the non-sharable resource is greater than or equal to the overload threshold and a random value associated with the resource usage requests is less than or equal to the weight ratio of the non-sharable resource, allocating the resource usage requests to the non-sharable resource; and If the load of the non-sharable resource is greater than or equal to the overload threshold and the random value associated with the resource use requests is greater than the weight ratio of the non-sharable resource, the resource use requests are allocated to the sharable resource.

4. The resource management system as claimed in claim 1, wherein the admission control policy comprises: If the number of resource usage requests is greater than the current available resource amount of the resource pool, the resource usage requests are rejected, and the non-shareable resources of each resource pool are only used by the resource usage requests of the electronic devices local to the worker nodes and the master node.

5. The resource management system as claimed in claim 1, wherein the trained artificial intelligence model is generated by training an artificial intelligence model with historical resource usage statistical data.

6. A resource management system as described in claim 5, wherein the artificial intelligence model is a type of neural network model, which includes an input layer, an output layer and multiple hidden layers, and the input layer, the output layer and the hidden layers each include multiple neurons.

7. The resource management system as claimed in claim 5, wherein the historical resource usage statistics include user demand resources, remaining resources of non-sharable resources, remaining resources of sharable resources, time information, and resource usage status at a specific time.

8. The resource management system as claimed in claim 1, wherein the system resources include CPU processing capacity, memory available capacity, network available capacity, flash memory available capacity, hard disk available capacity and graphics processor processing capacity.

9. A resource management method, the resource management method is applicable to a plurality of worker nodes and a master node that are communicatively connected to each other, and the worker nodes and the master node each include a resource pool, a resource monitor, a collector agent module, a traffic guidance reasoning module, a request processing module, a permission controller and a threshold adjuster, the master node also includes a data collector, an artificial intelligence computing module and a load balancer, the resource management method includes: An internal resource management process is executed for the worker nodes and the master node, and the internal resource management process includes: Configuring the resource pool to divide a corresponding system resource into a sharable resource and a non-sharable resource; Configuring the resource monitor to record the usage status of the resource pool and generate resource usage monitoring data; configuring the collector agent module to transmit the generated resource usage monitoring data to the traffic steering reasoning module; The traffic guidance reasoning module is configured to receive the resource usage monitoring data, and calculate the weight ratio of the non-sharable resource and the sharable resource according to the resource usage monitoring data to generate a permission control policy; configuring the request processing module to receive a plurality of resource usage requests from a plurality of electronic devices; and configuring the admission controller to allocate the resource usage requests to the non-sharable resource and the sharable resource according to the admission control policy, The threshold value adjuster is used to adjust an allocation ratio of the sharable resource and the non-sharable resource; and An external resource management process is executed for the worker nodes and the master node, and the external resource management process includes: The data collector is configured to receive the resource usage monitoring data of the worker nodes and the master node from the collector agent modules of the worker nodes and the master node to generate a resource usage statistical data; The artificial intelligence computing module is configured to input the resource usage statistics into a trained artificial intelligence model to generate a plurality of critical value data for respectively allocating the resource pools of the worker nodes and the master node, wherein each of the critical value data defines the allocation ratio in each of the resource pools; The traffic guiding reasoning modules of the worker nodes and the master node are configured to control the corresponding threshold adjusters to adjust the sharable resources and the non-sharable resources of each of the resource pools according to the optimization threshold data; and The load balancer is configured to balance the resource usage requests allocated to the sharable resources in the worker nodes and the master node to the sharable resources in the worker nodes and the master node.

10. The resource management method according to claim 9, further comprising: The traffic steering reasoning module is also configured to calculate the load of the non-sharable resources and the sharable resources based on the resource usage monitoring data to generate the permission control policy, wherein the weight ratios of the non-sharable resources and the sharable resources of each resource pool add up to 1.

11. The resource management method as claimed in claim 10, wherein the admission control policy comprises: If the load of the non-sharable resource is less than an overload threshold, allocating the resource usage requests to the non-sharable resource; If the load of the non-sharable resource is greater than or equal to the overload threshold and a random value associated with the resource usage requests is less than or equal to the weight ratio of the non-sharable resource, allocating the resource usage requests to the non-sharable resource; and If the load of the non-sharable resource is greater than or equal to the overload threshold and the random value associated with the resource use requests is greater than the weight ratio of the non-sharable resource, the resource use requests are allocated to the sharable resource.

12. The resource management method as claimed in claim 9, wherein the admission control policy comprises: If the number of resource usage requests is greater than the current available resource amount of the resource pool, the resource usage requests are rejected, wherein the non-shareable resources of each resource pool are only used by the resource usage requests of the electronic devices local to the worker nodes and the master node.

13. The resource management method of claim 9, wherein the trained artificial intelligence model is generated by training an artificial intelligence model with historical resource usage statistical data.

14. The resource management method as described in claim 13, wherein the artificial intelligence model is a type of neural network model, the type of neural network model includes an input layer, an output layer and multiple hidden layers, and the input layer, the output layer and the hidden layers each include multiple neurons.

15. The resource management method as claimed in claim 13, wherein the historical resource usage statistics include user demand resources, remaining resources of non-sharable resources, remaining resources of sharable resources, time information, and resource usage status at a specific time.

16. The resource management method of claim 11, wherein the system resources include CPU processing capacity, memory available capacity, network available capacity, flash memory available capacity, hard disk available capacity, and graphics processor processing capacity.

Images