TWI644540B - Flow meter flexible cutting system for virtual network in multi-tenant software-defined network - Google Patents

Abstract

A flexible flow table cutting system for a virtual network in a multi-tenant software-defined network. It mainly manages flow table resources in a multi-tenant SDN environment. SPRM (Soft Partitioning Resource Manager) is mainly solved by the following three types: Methods for management: (1) Mixed mode: SPRM uses two modes for mixed processing in the management process: the supervision process checks and processes the flow bar before the intermediary controller loads it, and the side recording process loads the flow bar first. Subsequent release processes are used to reduce management delays. (2) LRU-PTR (Least recently used-Partial Timeout Reset) implementation module: SPRM uses the method of cutting the flow bar to let the switch actively report the idle state of the flow bar to Select the target that needs to be replaced. (3) Configure the tenant resource dynamic cap module: SPRM sets a dynamic cap value for each tenant.

Description

   Flow meter flexible cutting system for virtual network in multi-tenant software-defined network   

The invention relates to a flow table elastic cutting system for a virtual network in a multi-tenant software-defined network that can reduce a large number of packet_in packets and improve resource utilization, without the problem of resource snatch.

SDN (Software-defined networking, software-defined networking) is an emerging network architecture in recent years. Its concept is to separate the control-plane function of the network from the data-plane function. In the data layer, how packets are forwarded will perform corresponding actions through the corresponding flow bar, and Flow table resources are limited. If the flow table resources are not cut and managed, it will cause the problem of resource grabbing, and then generate a lot of packet_in events (packet_in events) and cause a lot of network delays. If the flow table resources are managed in a hard cut manner, the problem of resource waste will be caused, which will reduce the flow table usage, and then generate a large number of packet_in events and cause a lot of network delays.

However, most of the previous applications in the ordinary SDN network environment are managed by the use of flow tables by multiple applications. Similar concepts are mentioned in the following previous patented technologies: US 9065724 (Managing a flow table). On the other hand, most of the resource management in the multi-tenant SDN environment is to manage and control network traffic or bandwidth resources. If there is a method to manage the flow table, it is often handled in a hard cut way, but There is a problem of waste of resources in hard cutting management.

It can be seen that there are still many shortcomings in the above-mentioned customary methods. It is not a good design, and it needs to be improved.

In view of the various shortcomings derived from the above-mentioned conventional methods, the inventor of this case has been eager to improve and innovate. After years of painstaking and meticulous research, he has finally successfully developed and applied this multi-tenant (multi-tenant) SDN resource elastic management mechanism.

The purpose of the present invention is to design a resource management method for a Multi-tenant SDN network environment. In such an environment, multiple tenants share the same physical resource. If multiple users want to set a traffic bar, they need to go through an intermediary control. If the size of the physical resource flow table is limited, if the flow table is full, the following requirements will be rejected by the intermediary controller. At this time, users who issue this request will generate a large number of packet_in because they cannot set the flow bar. Causes a lot of burden on the tenant controller and creates a lot of network latency. If the reason that the uncut resources are shared and the flow table is full is due to the heavy use of a user, we call it resource seizure. If the hard cutting method is used, a large number of users cannot use the empty resources. Resources are wasted. Therefore, the above-mentioned problem can be effectively solved by the flow table flexible cutting system of the virtual network in the tenant software-defined network of the present invention.

In order to achieve the above-mentioned object of the invention, the present invention provides a flexible flow table cutting system for a virtual network in a tenant software-defined network of the present invention. Before the intermediary controller sets the requirements for the tenant controller, the requirements are transferred to the inspection module for inspection. Select the corresponding processing method for the usage status of the flow bar and the existing resource utilization. After checking by this management mechanism, the corresponding processing action set is turned into a command, and then this command is given to the intermediary controller for execution into the physical resource.

The present invention mainly includes three modules: (1) a hybrid mode switching module: SPRM (Soft Partitioning Resource Manager) uses two modes for hybrid processing in the management process: the supervision process is in the intermediary controller Before the flow bar is checked, the inspection process is performed first, and the profile recording process is performed after the flow bar is lowered and then the subsequent release process is performed to reduce the management delay time. (2) LRU-PTR (Least recently used-Partial Timeout Reset) implementation module: SPRM uses a flow bar to cut the flow bar to allow the switch to actively report the idle state of the flow entry. Select the target that needs to be replaced. (3) Configure the tenant resource dynamic cap module: SPRM sets a dynamic cap value for each tenant.

10‧‧‧ Management

20‧‧‧SPRM Management Module

30‧‧‧Tenant Controller

40‧‧‧ intermediate controller

50‧‧‧ physical resources

60‧‧‧Mixed mode switching module

70‧‧‧Configure tenant resource dynamic cap module

80‧‧‧LRU-PTR Implementation Module

101‧‧‧Tenant Management Interface

102‧‧‧Root Management Interface

201‧‧‧ Comparison of remaining resources

202‧‧‧Comparison of Tenant Usage and Available Resources

203‧‧‧Comparison of Flexible Space

204‧‧‧ has been replaced by space user settings

205‧‧‧ send remove flow bar command

206‧‧‧ Send Set Flow Bar Command

601‧‧‧ supervision process

602‧‧‧Side recording process

6011‧‧‧Tenant controller requirements

6012‧‧‧Forwarding demand

6013‧‧‧Notice of results

6014‧‧‧Transfer order

6015‧‧‧ execute order

6021‧‧‧Tenant controller requirements

6022‧‧‧Forwarding demand

6023‧‧‧Implement tenant demand

6024‧‧‧Notice of results

6025‧‧‧Transfer order

6026‧‧‧ execute order

701‧‧‧ Reserve tenant space calculation

702‧‧‧Remaining space calculation

703‧‧‧Flexible upper limit setting

7011‧‧‧Comparison of Tenant's Current Usage and Available Resources

7012‧‧‧Comparison of Tenant's Current Usage and Average Usage

7013‧‧‧ Average Tenant Usage vs. Available Resources

7014‧‧‧ Reasonable comparison of reserved space

801‧‧‧ Traffic bar idle judgment

802‧‧‧ Total idle time compared with original demand

803‧‧‧Select the flow bar for maximum idle time

Fig. 1 is a system architecture diagram of the present invention applied to a multi-tenant SDN resource elastic management mechanism; Fig. 2 is a structural diagram of a hybrid mode switching module of the present invention; Fig. 3 is an operation flowchart of a supervision process of the present invention; The operation flow chart of the recording process of the invention; Figure 5 is a flowchart of the SPRM management module of the invention; Figure 6 is a flowchart of the tenant resource dynamic cap module configuration of the invention; and Figure 7 is a reserved tenant space of the invention Calculation flowchart; Figure 8 is a flowchart of the LRU-PTR implementation module of the present invention.

The invention provides a flexible flow table cutting system for a virtual network in a tenant software-defined network of the present invention. It provides a multi-tenant virtual network service in a software-defined network (Software Defined Networks) environment to utilize OpenFlow switch (switch) flow table resource cutting method. When the tenant controller issues a request, the mechanism of the flow table resource management through the Soft Partitioning Resource Manager (SPRM) determines the range of resources available to each tenant and the subsequent release and return of resources. To achieve the purpose of resource sharing.

The SPRM of the present invention performs resource management in a Soft Partitioning mode, and determines whether the environment is in multiple states through the Hybrid Mode switching module, and then decides whether to use a tenant resource dynamic cap module and LRU-PTR implementation module. The hybrid mode switching module determines the management process to be performed from the monitoring process (In-line mode) or the side recording process (Sniff mode) according to the resource usage status.

Among them, the supervision process (In-line mode) is that the requirements issued by the tenant controller will be checked by the management layer, and after the availability is determined, the management layer will issue an order to inform the intermediary controller that it can correctly handle the requirements of the tenant controller. The sniff mode is a request from the tenant controller to be recorded by the management for inspection. If it is found to be illegal, the management then issues an order to notify the intermediary controller to make up after action.

The above Hybrid Mode module determines which of the various states the environment is in. The various states include: a. If the tenant is in normal use and the physical resources are sufficient, the profile recording process is used as the management process and the tenant can use it normally. The layer also does not need to make up actions; b. If the tenant wants to overuse and the physical resources are sufficient, the profile recording process is adopted and the dynamic upper limit of the tenant resource is configured; if the tenant usage exceeds the upper limit, the tenant is replaced by himself. The space method is processed, and then the management controller issues an order to request the intermediary controller to handle it, otherwise it will be used normally; c. If the tenant is normally used but the physical resources are insufficient, use the supervision process and select the tenant with the current maximum overuse ratio to be replaced by the replaced tenant Space method processing; and d. If the tenant wants to overuse but the physical resources are insufficient, use the supervision process and use the replacement space method as the replaced tenant.

The method for configuring the dynamic upper limit value of the tenant resource in the present invention is set in a dynamically changing manner, and a variety of tenant information is used to determine the over-use ratio as the upper limit value for the tenant's resource use. Its various tenant information includes: a. The average past tenant's usage; b. The lower limit of available resources; and c. The current resource usage.

The replacement space method of the present invention uses an algorithm to determine the flow bar to be replaced, and then deletes this flow bar. The algorithm for determining the flow bar to be replaced, called Least Recently Used-Partial Timeout Reset (LRU-PTR), includes calculating the idle time of each flow bar, and then selecting the flow bar with the maximum idle time as the flow bar to be replaced. And calculating the idle time of each flow bar includes methods such as cutting idle_timeout value and using the flag to make the switch automatically report the status of the flow bar.

Please refer to FIG. 1, which is a system architecture diagram of the present invention, which mainly includes a management layer 10, an SPRM management module 20, a tenant controller 30, an intermediary controller 40, a physical resource 50, and a mixed mode switching module 60.

Management layer 10 is mainly used as a level of management resources, including: tenant management interface 101, Root management interface 102, and SPRM management module 20; SPRM management module 20 is mainly used as a management process to check whether resources are available Modules, which include: the configuration of the tenant resource dynamic cap module 70 and the LRU-PTR implementation module 80; the tenant controller 30 is mainly used as a controller owned by each tenant in a software-defined network environment to control its own Software-defined network environment; the intermediary controller 40 is mainly used as a network virtualization intermediary layer in a multi-tenant software-defined network environment. It is responsible for virtualizing the underlying physical resources 50 to provide one or more virtual tenant controllers 30. Software-defined network environment; the physical resource 50 is mainly used as the physical network resource, which is generally referred to as the OpenFlow switch; the mixed mode switching module 60 is mainly used as a module to determine which management process to use.

Please refer to Figure 2. The mixed mode switching module 60 is composed of a supervision process 601 and a side recording process 602. The supervision process 601 is more secure when the resource utilization rate is too high. After the controller 30 issues the demand and then manages it through the supervision process 601, it actually completes the demand into the physical resource 50; and the side recording process 602. When the resource utilization rate is low, the side recording process 602 is used to reduce the demand delay time. After the controller 30 issues a request, the controller 30 executes the management check through the profile recording process 602 and then executes the management check to the physical resource 50.

Please refer to FIG. 3, which is an operation flowchart of the supervision process 601 applied to the multi-tenant SDN resource elastic management mechanism according to the present invention. 6015 is composed of the tenant controller requirements 6011. The tenant controller requests 6011 to the intermediary controller 40, and then determines whether to perform the supervision process 601 through the hybrid mode switching module 60. If the hybrid mode switching module confirms the supervision process 601, Then, it forwards the demand 6012 and the tenant controller needs to the SPRM management module 20 for inspection; the result notification 6013 is used to notify the tenant management interface 101 of the inspection result and send the command 6014 to the intermediary controller 40 at the same time; and finally execute the command 6015 After receiving the command, the intermediary controller 40 executes a required execution action to the physical resource 50.

Please refer to FIG. 4, which is an operation flowchart of the side recording process 602 applied to the multi-tenant SDN resource elastic management mechanism according to the present invention, which is performed by the tenant controller request 6021, the transfer request 6022, the execution of the tenant request 6023, the result notification 6024, It is composed of sending command 6025 and executing command 6026. The tenant controller requirement 6021 is issued by the tenant controller to the intermediary controller 40, and then the mixed mode switching module 60 is used to determine whether to perform the side recording process 602. The group confirms that the side recording process 602 is performed, and then the requirement 6022 is forwarded to the SPRM management module 20 for inspection and the tenant controller requirement 6021 is executed to the physical resource 50 by executing the tenant requirement 6023; the result notification 6024 is used to transfer The inspection result informs the tenant management interface 101, and at the same time, the necessary remedial actions are sent to the intermediary controller 40 via the transmission command 6025; and the execution command 6026 is used as the final remedial action of the side recording process 602, which is mostly a command to delete the traffic bar If there is no need to make up for it, there is no process of executing order 6026.

Please refer to FIG. 5, which is a flowchart of the SPRM management module 20 applied to the multi-tenant SDN resource elastic management mechanism according to the present invention. The remaining resources are compared 201, the usage of tenants is compared with the available resources 202, and the flexible space is compared 203. , Replaced by space user setting 204, transmitting remove flow bar command 205, transmitting set flow bar command 206, configuring tenant resource dynamic cap module 70 and LRU-PTR implementation module 80, among which: remaining resource comparison 201, using To check whether there are remaining physical resources and then perform subsequent processing, if so, it means that there is still a large amount of remaining resources available for the physical resource 50; otherwise, it means that the remaining resources are insufficient; the tenant usage is compared with the available resources 202 to check whether the tenant exceeds the available resources Lower limit, if it is, the tenant is considered as overused, otherwise it is not overused.

The over-used flexible space comparison 203 is used to check whether the tenant has over-used the upper limit of the available resources. If so, the replaced space user is set to 204 as the user, and the LRU-PTR implementation module 80 is executed, and then the removal traffic is transmitted. Command 205 to the intermediary controller 40, if not, do nothing; configure the tenant resource dynamic cap module 70 to calculate the configurable flexible usage to give the tenant temporary resources, only when the remaining resource comparison 201 judges to be yes, It is used only when the usage of the tenant is compared with the available resources. 202 It is judged that it is overused; and the LRU-PTR implementation module 80 is used when it is determined that the space needs to be replaced. The purpose is to determine the traffic bar that has been idle for too long, and then Do the transfer to remove the flow bar command 205 to release the space, and if necessary, use the transfer setting flow bar command 206 to notify the intermediary controller 40 to set the tenant demand flow bar in the released space.

Please refer to FIG. 6. The flowchart for configuring the tenant resource dynamic cap module 70 of the present invention applied to the multi-tenant SDN resource elastic management mechanism is based on the reserved tenant space calculation 701, the remaining space calculation 702, and the elastic upper limit setting 703. It consists of the reserved tenant space calculation 701 to help other tenants reserve a small amount of space to protect the use rights of other tenants and reduce the burden of using space replacement methods later; the remaining space calculation 702 will reserve tenant space 701 adds up the space reserved by each tenant, and uses the remaining resources of the system to deduct the total reserved space, which is configurable space; and the flexible upper limit setting 703, which is used to configure the result of the remaining space calculation 702, and adds this value The upper tenant lower limit is used as the flexible upper limit.

The present invention is mainly used when the tenant wants to overuse and the resources are sufficient. In order to avoid waste of resources, an elastic upper limit is set for the tenant who wants to overuse for use. This flexible upper limit value changes every time the tenant sends out a demand in accordance with the current system status and improves resource utilization.

Referring to FIG. 7, the present invention is applied to a multi-tenant SDN resource elasticity management mechanism configured with an elastic resource upper limit for the reserved tenant space calculation method 701. The method steps include:

a. The current usage of the tenant is compared with the available resources in step 7011. When the current usage of the tenant exceeds its available resources, it means that this tenant is currently in an over-used state, and no space is reserved. Otherwise, it proceeds to step b;

b. Compare the current usage and average usage of the tenant to step 7012. If the result of step a is NO, compare the current usage with the average usage. If the current usage has exceeded the average usage, we consider that this tenant will have space requirements in the future and make d. The comparison of the reasonableness of the reserved space in step 7014, otherwise, we enter c. step and continue to compare the average usage with the available resources in step 7013;

c. The average usage of the tenant is compared with the available resources in step 7013. If the result of step b. is no, it is determined that the current usage is less than the available resources and average usage. At this time, the average usage and available resources are compared. If the average usage is less than the available resources, then Set the reserved space to the average usage minus the current usage, otherwise set to the available resource minus the current resource.

d. The reasonableness of the reserved space is compared with step 7014. If the result of step b. is yes, it is determined that the current usage is less than the available resources and the current usage is greater than the average usage. At this time, if the allocation of reserved space is the current usage minus the average usage, It may be unreasonable to exceed the amount of the original available resources, so a reasonable comparison is made here. If the result is yes (exceeds the original available resources), the available resources are subtracted from the current usage, otherwise the current resources are reserved. Dosage minus average.

Please refer to FIG. 8, which is a flowchart of the LRU-PTR implementation module 80 applied to the multi-tenant SDN resource elasticity management mechanism according to the present invention, including, a. Step 801 for determining the idleness of the traffic bar, and the intermediate controller 40 receives The current timestamp is immediately recorded when the flow bar is removed from the physical resource. If the current timestamp minus the last update timestamp is greater than the fragment idle time removal setting, it is determined that the flow bar has been used, and it is marked as yes. And set the update timestamp to the current timestamp and the idle time to the fragment idle time setting; otherwise, set the update timestamp to the current timestamp, the total idle time plus the fragment idle time setting and remove Increase the number of times by one and proceed to the next step; b. Compare the total idle time with the original demand in step 802, multiply the number of removals in step a. By the fragment idle time removal setting value and compare the original idle time removal setting value, if Greater than or equal to the original idle time to remove the set value, it is determined that this flow bar has reached the original idle time, then this flow bar is removed, otherwise, the secondary flow Reset the real-time resource to the physical resource; c. Select the flow bar with the maximum idle time in step 803, and select the flow bar with the maximum idle time in the entire data table as the flow bar that needs to be replaced. This reflects the LRU algorithm. Spirit, replace the longest idle traffic bar.

The technologies that were previously applied to manage multi-tenant software-defined networks are mostly for bandwidth management or hard cutting. It does not consider that hard cutting and reprocessing the management flow table will have a waste of resources and affect Tenant network performance. Compared with other conventional technologies, the multi-tenant SDN management mechanism provided by the present invention has the following advantages:

1. The present invention utilizes a hybrid mode switching module, which can reduce the delay in executing a large number of requirements.

2. The present invention can increase the resource utilization rate of the flow table by using a dynamic cap module for configuring tenant resources.

3. The present invention utilizes the LRU-PTR implementation module, which can reduce the number of tenants' demands and greatly reduce network delay.

The above detailed description is a specific description of a feasible embodiment of the present invention, but this embodiment is not intended to limit the patent scope of the present invention. Any equivalent implementation or change that does not depart from the technical spirit of the present invention should be included in Within the scope of the patent in this case.

To sum up, this case is not only innovative in terms of technical ideas, but also has many of the above-mentioned effects that are not used by traditional methods. It has fully met the requirements of statutory invention patents that are novel and progressive. To approve this invention patent application, to encourage invention, to the utmost convenience.

Claims (8)

A flexible flow table cutting system for a virtual network in a multi-tenant software-defined network. The system mainly includes: a management layer for resource management. The management layer includes at least a tenant management interface and SPRM (Soft Partitioning Resource Manager). Server) management module, the SPRM management module mainly determines the range of resources available to each tenant and subsequent resource release and return processing; the tenant controller, as a software-defined network environment, each tenant has its own controller to control Own software-defined network environment; the intermediary controller, as a network virtualization intermediary layer in a multi-tenant software-defined network environment, is responsible for virtualizing the underlying physical resources and providing the upper-level tenant controller with one or more virtual software Define the network environment; the mixed mode switching module is used to perform the supervision process and the side recording process. The supervision process is the requirement issued by the tenant controller will be checked by the management first, and the correctness of its content will be determined by The management issued an order to inform the intermediary controller that it can correctly handle the needs of the tenant controller, and recorded the process. Skimming demand from tenants controller issued a check handed over to management, if the check out illegal, and then issue a command to inform the agency management controller to make up after work action. The flexible table cutting system for a virtual network in a multi-tenant software-defined network as described in claim 1, wherein the physical resource is the physical network resource of the network. The flexible cutting system for the flow table of the virtual network in the multi-tenant software-defined network as described in claim 1, wherein the monitoring process is composed of the requirements of the tenant controller, the transfer requirements, the result notification, the transmission command, and the execution command, of which the tenant The controller request is issued by the tenant controller to the intermediary controller, and then the hybrid mode switching module is used to determine whether to perform the supervision process. If the hybrid mode switching module confirms the supervision process, the request is forwarded to the tenant controller. The SPRM management module performs the inspection, and the result notification is used to notify the tenant management interface of the inspection result and send a command to the intermediary controller at the same time, and the execution command is to execute the required execution action to the physical resource after the intermediary controller receives the command. The flexible cutting system for the flow table of the virtual network in the multi-tenant software-defined network as described in claim 1, wherein the recording process is performed by the tenant controller request, the transfer request, the execution of the tenant request, the result notification, the transmission command, and the execution command. It consists of the requirements of the tenant controller. The tenant controller sends requirements to the intermediary controller, and then determines whether to perform the side recording process through the mixed mode switching module. If the mixed mode switching module confirms the side recording process, the requirements are forwarded. The requirements are forwarded to the SPRM management module for inspection, and the tenant controller requirements are executed to the physical resources by executing the tenant requirements. The result notification is used to notify the tenant management interface of the inspection results, and at the same time, the necessary remedial actions are transmitted via the command. It is sent to the intermediary controller, and the execution command is used as the final remedial action of the profile recording process. Most of them are commands to delete the flow bar. If there is no need to make any remedy, there is no process of executing the command. The flexible cutting system for the flow table of the virtual network in the multi-tenant software-defined network as described in claim 1, wherein the SPRM management module is compared with the remaining resources, the tenant usage and the available resources, the super-used flexible space comparison, and is replaced Space user setting, sending remove flow bar command, sending set flow bar command, configuring tenant resource dynamic cap module and LRU-PTR (Least recently used-Partial Timeout Reset, least recently used-partial timeout reset algorithm) implementation model The remaining resource comparison is used to check whether there are remaining physical resources for subsequent processing. If there are remaining physical resources for subsequent processing, it means that there are still a lot of remaining resources available for the physical resources. The follow-up processing means that the remaining resources are insufficient. The comparison of the usage of the tenant with the available resources is used to check whether the tenant exceeds the lower limit of the available resources. If the lower limit of the available resources is exceeded, the tenant is considered to be overused. If the available resources are not exceeded The lower limit is for non-overcommitment. The overcompletion flexible space comparison is used to check whether the tenant is overcommitted. The flexible upper limit value of the source. If the tenant exceeds the flexible upper limit value of the available resources, it will be replaced by the space user as the user itself, and the LRU-PTR implementation module will be executed, and then the remove flow bar command will be sent to the intermediary controller. If the tenant does not exceed the flexible upper limit of the available resources, no processing is performed. The tenant resource dynamic cap module is configured to calculate the configurable flexible usage to give the tenant temporary resources. Only when the remaining resources are judged to be YES, and It is only used when judging the usage of the tenant and the available resources as being over-used. The LRU-PTR implementation module is used when it is determined that the space needs to be replaced. The purpose is to determine the flow bar that has been idle for a long time, and then transfer it. In addition to the free flow bar command, the space is released, and if necessary, the current demand is set to send the set flow bar command to the intermediary controller to set the tenant demand flow bar in the released space. The flexible cutting system for the flow table of the virtual network in the multi-tenant software-defined network as described in claim 5, wherein the configuration of the dynamic resource cap module of the tenant is determined by the reserved tenant space calculation, remaining space calculation, and flexible upper limit setting. Composition, in which the reserved tenant space calculation is used to help other tenants to reserve a small amount of space to protect the use rights of other tenants and reduce the burden of using space replacement methods in the future. The remaining space calculation is to reserve tenant space to help each tenant The reserved space is added up, and the remaining reserved resources are used to deduct the total reserved space, which is configurable space. The elastic upper limit setting is used to configure the calculation result of the remaining space. Add this value to the tenant lower limit value as Upper elastic limit. The flexible cutting system for the flow table of the multi-tenant software-defined network in the multi-tenant software-defined network as described in the request item 6, wherein the calculation of the reserved tenant space includes the following steps: a. The step of comparing the current usage of the tenant with the available resources. When the usage exceeds its available resources, it means that the tenant is currently in an over-used state, so no space is reserved, otherwise it goes to step b; b. The tenant's current usage is compared with the average usage. If a. The result of the step is no, then the current comparison is compared. Usage and average usage. If the current usage has exceeded the average usage, we determine that the tenant will have space requirements in the future, and make a comparison of the reasonableness of the reserved space in step d. Otherwise, we enter step c. Continue to make the average usage and Available resources comparison step; c. Tenant average usage and available resources comparison step. If the result of step b. Is no, then it is determined that the current usage is less than the available resources and average usage. At this time, the average usage and available resources are compared. Less than the available resources, the reserved space is set to the average usage minus the current usage, otherwise it is set to the available resources minus the current resources D. The comparison step of the rationality of the reserved space. If the result of step b. Is yes, then it is determined that the current usage is less than the available resources and the current usage is greater than the average usage. At this time, if the allocation of reserved space is the current usage minus the average usage , It may be unreasonable to exceed the original available resource usage, so a reasonable comparison is made here. If the result is more than the original available resource, the available resource is subtracted from the current usage, otherwise the current usage is reserved. Subtract the average amount. The flexible cutting system for the flow meter of the virtual network in the multi-tenant software-defined network as described in claim 5, wherein the steps of the LRU-PTR implementation module to determine the need to replace the space include: a. The flow bar idle determination step, the intermediary control The server immediately records the current timestamp when it receives the message that the traffic bar is removed from the physical resource. If the current timestamp minus the last update timestamp is greater than the fragment idle time removal setting, it is determined that this traffic bar has been used. Mark as yes, and set the update timestamp to the current timestamp and the idle time to the fragment idle time setting value; otherwise, set the update timestamp to the current timestamp, the total idle time plus the fragment idle time setting value and Increase the number of removals by one and proceed to the next step; b. Compare the total idle time with the original demand step, multiply the number of removals in step a. By the fragment idle time removal setting value and compare with the original idle time removal setting value , If it is greater than or equal to the original idle time removal setting value, it is determined that this traffic bar has reached the original idle time, then this traffic bar is removed, and vice versa Reset the secondary flow bar to the physical resource; c. Select the flow bar step with the maximum idle time, select the flow bar with the maximum idle time in the entire data table as the flow bar to be replaced, and set the longest idle bar The flow bar is replaced.