A random uninstallation method and system for MEC servers

Abstract

The present invention relates to a random unloading method and system for MEC servers. The method includes the following steps: S1, the node obtains the basic configuration information of the network and organizes it into a known parameter information library; S2, with the goal of minimizing the total system delay and the queue stability as a constraint, a system optimization model is established; S3, based on the known parameter information library, the system optimization model is solved by the golden section method, and the random processing probability at the MEC is obtained after several iterations; S4, based on the solved random processing probability, the MEC processes the data by itself or submits it to the cloud for processing. The present invention can minimize the average waiting delay of users and achieve efficient use of resources while ensuring the stability of the actual system queue; the solution method of the optimization target of the present invention can be implemented based on classic convex optimization algorithms such as the golden section method, the required parameter information is easy to obtain, and the overall algorithm is simple and easy to expand.

Description

Technical Field

[0001] The present invention belongs to the field of information and communication engineering technology, and specifically relates to a random unloading method and system for an MEC server. Background Art

[0002] With the widespread application of multimedia technologies such as terminal games, augmented reality, and high-definition online videos in mobile communications, more and more applications require persistent high-performance computing on wireless devices to perform complex tasks with ultra-low latency. In the past decade, the large-scale deployment of cloud computing servers has enabled wireless devices to offload intensive computing data streams to remote cloud servers with abundant computing resources. However, the remote return of data streams brings a large latency problem. Based on this, mobile edge computing (MEC) technology has been proposed. Pushing publicly accessible computing resources to the edge of wireless access networks, such as cellular base stations and access points, allows users to quickly offload computing tasks to edge servers near them, thereby reducing latency.

[0003] However, reducing latency will also reduce system queue stability. Therefore, there is an urgent need for a solution that can reduce system latency while ensuring system queue stability. Summary of the invention

[0004] In view of the above problems existing in the prior art, the present invention provides a random unloading method and system for MEC servers, which can reduce system latency while ensuring the stability of the system queue.

[0005] The present invention adopts the following technical solutions:

[0006] A random unloading method for a MEC server includes the following steps:

[0007] S1. The node obtains the basic configuration information of the network and organizes it into a known parameter information library;

[0008] S2. Establish a system optimization model with the goal of minimizing the total system delay and queue stability as a constraint;

[0009] S3, based on the known parameter information library, the system optimization model is solved by the golden section method, and the random processing probability at the MEC is obtained after several iterations;

[0010] S4. MEC processes the data itself or submits it to the cloud for processing based on the obtained random processing probability.

[0011] As a preferred solution, the total system delay in step S2 includes MEC processing time and cloud processing time, and the queue stability constraint conditions include MEC processing queue stability constraint conditions and cloud processing queue stability constraint conditions.

[0012] As a preferred solution, the calculation formula of the total system delay T is:

[0013] T=pt+(1-p)t′ (7)

[0014] Where pt represents the MEC processing time, (1-p)t′ represents the cloud processing time, p represents the random processing probability at the MEC, t represents the average waiting time for the MEC to process data, and t′ represents the average waiting time for the cloud to process data.

[0015] As a preferred solution, the calculation formula for the average waiting time t for MEC to process data is as follows:

[0016] t=dλx+C MEC +d′λx (3),

[0017] Where d represents the average transmission delay of a unit-length task from the user to the MEC, λ represents the average arrival rate of the user data stream, x represents the average length of the data stream, and C MEC represents the average computational delay of MEC processing data, and d′ represents the average transmission delay of unit-length tasks from MEC to users.

[0018] As a preferred solution, C MEC The calculation formula is:

[0019]

[0020] Among them, μ MEC It is the service rate of MEC.

[0021] As a preferred solution, μ MEC The calculation formula is:

[0022] μ MEC =f MEC K1 (2)

[0023] Among them, f MEC represents the frequency of the MEC computer, and K1 represents the average data flow size that the MEC computer can process per MHz.

[0024] As a preferred solution, the average waiting time t′ for cloud data processing is calculated as follows:

[0025]

[0026] in, represents the average transmission delay of unit-length tasks from MEC to the cloud, C cloud The average computational latency of processing data in the cloud, is the average transmission delay of unit-length tasks from the cloud to MEC.

[0027] As a preferred solution, the average computing delay C of cloud data processing cloud The calculation formula is:

[0028]

[0029] Where c(·) is the waiting formula, n represents the number of computers equipped in the cloud, μ cloud Indicates the service rate of the cloud server;

[0030] The service rate of the cloud server μ cloud The calculation formula is:

[0031] μ cloud =nf cloud K2 (4)

[0032] f cloud represents the frequency of the cloud computer, and K2 represents the average data flow size that the cloud computer can process per MHz.

[0033] As a preferred solution, in step S2, the optimization model is:

[0034]

[0035] stf MEC K1>λpx (8-1)

[0036] nf cloud K2>(1-p)λx (8-2)

[0037] Among them, formula (8) represents the minimization of the total system delay, formula (8-1) represents the MEC processing queue stability constraint, and formula (8-2) represents the cloud processing queue stability constraint.

[0038] A random unloading system for MEC servers is also provided, which is based on the above unloading method and includes an information acquisition module, a model building module, a solution module, and a data unloading module connected in sequence;

[0039] The information acquisition module is used to obtain basic configuration information of the network and organize it into a known parameter information library;

[0040] The model building module is used to build a system optimization model with the goal of minimizing the total system delay and queue stability as a constraint;

[0041] The solution module is used to solve the system optimization model based on the known parameter information library and the golden section method, and obtain the random processing probability at the MEC after several iterations;

[0042] The data unloading module is located at the MEC and is used to control the MEC to process the data itself or submit it to the cloud for processing based on the obtained random processing probability.

[0043] The beneficial effects of the present invention are:

[0044] The present invention can minimize the average waiting delay of users while ensuring the stability of the actual system queue, thereby achieving efficient utilization of resources.

[0045] The method for solving the optimization target of the present invention can be implemented based on classical convex optimization algorithms such as the golden section method, the required parameter information is easy to obtain, and the overall algorithm is simple and easy to expand. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 It is a flow chart of a random unloading method for a MEC server according to the present invention;

[0048] Figure 2 It is a schematic diagram of the network structure;

[0049] Figure 3 This is a simulation diagram of the average waiting time changing with the probability size when the average arrival rate is large;

[0050] Figure 4 It is a structural schematic diagram of a random unloading system for MEC servers described in the present invention. DETAILED DESCRIPTION

[0051] The following describes the embodiments of the present invention through specific embodiments, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0052] Embodiment 1:

[0053] Reference Figure 1 , this embodiment provides a random unloading method for a MEC server, comprising the steps of:

[0054] S1. The node obtains the basic configuration information of the network and organizes it into a known parameter information library;

[0055] S2. Establish a system optimization model with the goal of minimizing the total system delay and queue stability as a constraint;

[0056] S3, based on the known parameter information library, the system optimization model is solved by the golden section method, and the random processing probability at the MEC is obtained after several iterations;

[0057] S4. MEC processes the data itself or submits it to the cloud for processing based on the obtained random processing probability.

[0058] This probability can minimize the average waiting time of the system. Specifically, two points are appropriately inserted into the search interval through the golden ratio of 0.618 and their function values ​​are calculated. The two points will divide the interval into three segments. Since the function has a global optimal solution, one segment can be deleted by comparing the function value to reduce the search interval. Then the same process is performed on the remaining interval. In this way, the search interval will be infinitely reduced. When the required accuracy is reached, the solution obtained will be infinitely close to the true solution.

[0059] It can be seen that the present invention can minimize the average waiting delay of users while ensuring the stability of the actual system queue, thereby achieving efficient utilization of resources.

[0060] The method for solving the optimization target of the present invention can be implemented based on classical convex optimization algorithms such as the golden section method, the required parameter information is easy to obtain, and the overall algorithm is simple and easy to expand.

[0061] This embodiment considers Figure 2 The figure shows a three-layer cloud computing network, which includes a user without computing power, a MEC server, and a cloud server equipped with n computers to provide computing functions. It is assumed that the arrival of user data streams follows a Poisson point process with an average arrival rate of λ, the length of the data stream follows an exponential distribution with an average length of x, and the data stream is indivisible.

[0062] Based on this, the average transmission delay from user to MEC is dλx, where d is the average transmission delay of unit-length tasks from user to MEC.

[0063] When the data stream reaches the MEC, the MEC processes it with probability p, and at the same time, there is a probability of (1-p) that it will be handed over to the cloud for processing. If it is processed locally in the MEC, since there is only one computer providing computing services for it, this part of the system can be regarded as an M / M / 1 queuing system. Since the departure of a stable M / M / 1 system is also a Poisson process, and the average arrival rate of departure is equal to the average arrival rate of arrival, and when the Poisson process splits into two processes, both processes are Poisson processes, and the average arrival rate is proportional to the split probability, the average computing delay of processing data locally by the MEC is

[0064]

[0065] μ MEC It is the service rate of MEC, which is determined by the performance of the computer deployed by MEC.

[0066] μ MEC =f MEC K1 (2)

[0067] where f MEC represents the frequency of the MEC computer, and K1 represents the average data flow size that the MEC computer can process per MHz.

[0068] After the task processing is completed, MEC sends the result to the user. The average transmission delay of the data stream from MEC to the user is d′λx, where d′ represents the average transmission delay of the unit length task from MEC to the user.

[0069] In summary, if the data is processed by MEC, the average waiting time t of the system from the time the user data is sent out to the time it is processed and returned is:

[0070] t=dλx+C MEC +d′λx (3)

[0071] If data is submitted to the cloud for processing with a probability of (1-p), since users leave the M / M / 1 system, the average arrival rate of departure is (1-p)λ. At this time, the data transmission delay from MEC to the cloud server is in Represents the average transmission delay of unit-length tasks from MEC to the cloud.

[0072] When the data stream arrives at the cloud server, the average arrival rate is (1-p)λ. Since the cloud consists of multiple computers, the queue system can be regarded as an M / M / n queue system. If the computer frequency of all computers in the cloud is f Cloud , then the service rate of the cloud server is:

[0073] μ cloud=nf cloud K2 (4)

[0074] Where n is the number of computers enabled by the cloud server, and K2 represents the average data flow size that the cloud computer can process per MHz. From the M / M / n queuing system, we can further know that the average calculation delay of the data flow calculated by the cloud server is:

[0075]

[0076] where c(·) is the Erlang waiting formula, which gives the probability that a task needs to wait when it arrives at the cloud.

[0077] When the task processing is completed, the cloud server first sends the result to MEC, and then MEC forwards it to the user. The transmission delay from the cloud server to MEC is in is the average transmission delay of unit-length tasks from the cloud server to MEC.

[0078] The average waiting time for cloud servers to perform calculations is:

[0079]

[0080] Therefore, the total average waiting time of the system (i.e., total delay) is composed of the time that is processed by MEC alone with probability p and the time that is processed by the cloud via MEC with probability (1-p), which can be calculated by the following formula:

[0081] T=pt+(1-p)t′ (7)

[0082] The overall problem model is:

[0083]

[0084] stf MEC K1>λpx (8-1)

[0085] nf cloud K2>(1-p)λx (8-2)

[0086] The objective function is to minimize the total delay of the system. Formula (8-1) ensures that the queuing system can remain stable when tasks are processed by MEC, and formula (8-2) ensures that the queuing system can remain stable when tasks are processed by the cloud.

[0087] Reference Figure 3 It simulates that when the average arrival rate is large, the average waiting time of the system is the smallest when the random probability is set to 0.79. Compared with simply offloading all tasks to MEC or cloud, that is, the probability is set to 1 or 0, it is obvious that the average waiting time is greatly reduced.

[0088] Embodiment 2:

[0089] Reference Figure 4 ,This embodiment provides a random unloading system for MEC servers, based on the unloading method described in Embodiment 1, including an information acquisition module, a model building module, a solution module, and a data unloading module connected in sequence;

[0090] The information acquisition module is used to obtain basic configuration information of the network and organize it into a known parameter information library;

[0091] The model building module is used to build a system optimization model with the goal of minimizing the total system delay and queue stability as a constraint;

[0092] The solution module is used to solve the system optimization model based on the known parameter information library and the golden section method, and obtain the random processing probability at the MEC after several iterations;

[0093] The data unloading module is located at the MEC and is used to control the MEC to process the data itself or submit it to the cloud for processing based on the obtained random processing probability.

[0094] It should be noted that the random unloading system for MEC servers provided in this embodiment is similar to that in Embodiment 1 and will not be described in detail here.

[0095] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope of the present invention.

Claims

1. A random unloading method for a MEC server, characterized in that: Includes steps: S1. The node obtains the basic configuration information of the network and organizes it into a known parameter information library; S2. Establish a system optimization model with the goal of minimizing the total system delay and queue stability as a constraint; S3, based on the known parameter information library, the system optimization model is solved by the golden section method, and the random processing probability at the MEC is obtained after several iterations; S4, MEC processes the data itself or submits it to the cloud for processing based on the random processing probability obtained; In step S2, the total system delay includes MEC processing time and cloud processing time, and the queue stability constraint includes MEC processing queue stability constraint and cloud processing queue stability constraint.

2. A random unloading method for a MEC server according to claim 1, characterized in that ,The calculation formula of the total system delay T is: T=pt+(1-p)t′ (7) Where pt represents the MEC processing time, (1-p)t′ represents the cloud processing time, p represents the random processing probability at the MEC, t represents the average waiting time for the MEC to process data, and t′ represents the average waiting time for the cloud to process data.

3. A random unloading method for a MEC server according to claim 2, characterized in that: The calculation formula for the average waiting time t for MEC to process data is as follows: t=dλx+C MEC +d′λx (3), Where d represents the average transmission delay of a unit-length task from the user to the MEC, λ represents the average arrival rate of the user data stream, x represents the average length of the data stream, and C MEC represents the average computational delay of MEC processing data, and d′ represents the average transmission delay of unit-length tasks from MEC to users.

4. A random unloading method for a MEC server according to claim 3, characterized in that: C MEC The calculation formula is: Among them, μ MEC It is the service rate of MEC.

5. A random uninstallation method for a MEC server according to claim 4, characterized in that: μ MEC The calculation formula is: m MEC =f MEC K1 (2) Among them, f MEC represents the frequency of the MEC computer, and K1 represents the average data flow size that the MEC computer can process per MHz.

6. A random unloading method for a MEC server according to claim 5, characterized in that: The average waiting time t′ for cloud data processing is calculated as follows: in, represents the average transmission delay of unit-length tasks from MEC to the cloud, C cloud is the average computing latency for processing data in the cloud. is the average transmission delay of unit-length tasks from the cloud to MEC.

7. A random unloading method for a MEC server according to claim 6, characterized in that: The average computing delay C of cloud data processing cloud The calculation formula is: Where c(·) is the waiting formula, n represents the number of computers equipped in the cloud, μ cloud Indicates the service rate of the cloud server; The service rate of the cloud server μ cloud The calculation formula is: m cloud =nf cloud K2 (4) f cloud represents the frequency of the cloud computer, and K2 represents the average data flow size that the cloud computer can process per MHz.

8. A random uninstallation method for a MEC server according to claim 7, characterized in that: In step S2, the optimization model is: s.t. f MEC K1>λpx (8-1) nf cloud K2>(1-p)λx (8-2) Among them, formula (8) represents the minimization of the total system delay, formula (8-1) represents the MEC processing queue stability constraint, and formula (8-2) represents the cloud processing queue stability constraint.

9. A random unloading system for a MEC server, used to implement the unloading method according to any one of claims 1 to 8, characterized in that: It includes an information acquisition module, a model building module, a solution module, and a data unloading module which are connected in sequence; The information acquisition module is used to obtain basic configuration information of the network and organize it into a known parameter information library; The model building module is used to build a system optimization model with the goal of minimizing the total system delay and queue stability as a constraint; The solution module is used to solve the system optimization model based on the known parameter information library and the golden section method, and obtain the random processing probability at the MEC after several iterations; The data unloading module is located at the MEC and is used to control the MEC to process the data itself or submit it to the cloud for processing based on the random processing probability obtained.

Images