A 5G fairness scheduling method based on time quantity

Abstract

The present invention discloses a 5G fairness scheduling method based on time quantity, which solves the problem in the prior art that 5G scenarios are greatly differentiated and general scheduling methods cannot meet the dynamic needs of multiple services in differentiated scenarios. The method obtains user service data and extracts parameters related to time quantity; for the current moment service, a positive and negative time quantity service set is constructed according to the time quantity relationship, and a reward and punishment regression judgment is performed on each service; according to the time quantity related parameters, the weight of each service is initialized; a reward adjustment coefficient is set for the positive time quantity related service, and a penalty adjustment coefficient is set for the negative time quantity related service; the priority is updated according to the reward and penalty adjustment coefficient; and the service is sent to the 5G scheduler for scheduling according to the updated priority. The present invention sorts the service data from the dimension of time quantity, and updates the priority of the service in combination with the filtering system and the reward and punishment assessment adjustment coefficient to meet the dynamic needs of different service indicators in different scenarios.

Description

Technical Field

[0001] The present invention relates to the field of mobile communication technologies, and in particular to a 5G fairness scheduling method based on time quantity. Background Art

[0002] Unlike previous 3G and 4G systems, 5G addresses a wider range of application scenarios and demands finer-grained service metrics, leading to the emergence of different slicing within the 5G landscape. 5G scheduling is designed to better match the implementation needs of different services. Currently, 5G scenarios are highly diverse, and conventional scheduling methods are unable to meet the dynamic demands of multiple services in these diverse scenarios. Scheduling often fails to optimize service metrics and can even penalize services that are over-optimized. Summary of the Invention

[0003] The present invention mainly solves the problem in the existing technology that 5G scenarios are highly differentiated and general scheduling methods cannot meet the dynamic needs of multiple services in differentiated scenarios, and provides a 5G fairness scheduling method based on time quantity. Starting from the service attributes, under the premise of ensuring fairness, it can automatically implement different levels of reward and punishment mechanisms according to the positive and negative properties of the service time quantity. By setting the filter coefficient, the weight matching of positive and negative time quantities can be achieved, thereby ensuring the 5QI (5GQoS Identifier) ​​indicator requirements of different services.

[0004] The above technical problems of the present invention are mainly solved by the following technical solutions: a 5G fairness scheduling method based on time quantity, comprising the following steps:

[0005] Step 1: Obtain user service data and extract time-related parameters;

[0006] Step 2: Update business priorities by combining the filtering system and the reward and punishment assessment mechanism;

[0007] (2-1): For the current business, a positive and negative time amount business set is constructed based on the time amount relationship, and reward and punishment regression judgment is performed on each business;

[0008] (2-2): Initialize the weight of each business according to the time-related parameters;

[0009] (2-3): Set a penalty adjustment coefficient for negative time-related businesses, and set a reward adjustment coefficient for positive time-related businesses;

[0010] (2-4): Update the priority based on the reward and penalty adjustment coefficient;

[0011] Step 3: Send the service to the 5G scheduler for scheduling based on the updated priority.

[0012] This invention is a time-based 5G fairness scheduling method. It organizes service data from the perspective of time and updates service priorities by combining a filtering system and reward and punishment adjustment coefficients, thereby meeting the dynamic needs of different service indicators in different scenarios. This method is simpler, more robust, and more applicable, meeting differentiated individual needs.

[0013] As a preferred solution, the service in step (2-1) includes n services Svr T ={Svr1, Svr2, ..., Svr n}, the extracted parameters include the current service time {Tusd1, Tusd2,…, Tusd n}, the total service time required {Tall1,Tall2,…,Tall n}, current service delay {Dly1, Dly2, …, Dly n}, the corresponding delay tolerance {Dlm1,Dlm2,…,Dlm n}, the number of rewards and punishments for each business {Nm1, Nm2,…, Nm n}, reward and punishment coefficients {WP1, WP2,…, WP n}.

[0014] As a preferred solution, the specific process of step (2-1) includes:

[0015] (2-1-1): Set up a service data table. The service data table includes the service name, current service time, total service time required, service delay, delay tolerance, current number of rewards and punishments, and service attributes. Based on the positive and negative correlation with the amount of time, the service attributes include positive time services and negative time services. Negative time services represent services that are negatively correlated with the amount of time, such as services that are sensitive to service delays; positive time services represent services that are positively correlated with the amount of time, such as services that are sensitive to service time.

[0016] (2-1-2): According to the service attributes, the services are divided into negative time services and positive time services; thus, the negative time service set Mset = {Svr j ,j=1,2…,m}, positive time quantity service set Aset={Svr k ,k=1,2,…,nm}; wherein negative time quantity business indicates business that is negatively correlated with time quantity, such as business that is sensitive to business delay; positive time quantity business indicates business that is positively correlated with time quantity, such as business that is sensitive to service time.

[0017] (2-1-3): Set the reward and punishment threshold Thr PW , for the current moment Svr T ={Svr1, Svr2, ..., Svr nEach business Svr in i , check Nm i , if Nm i ≥Thr PW , then Nm i =0,WP i =0.

[0018] As a preferred solution, the specific process of step (2-2) includes:

[0019] (2-2-1): Set the filter coefficient Ft k , service time threshold Thr usd , service delay threshold Thr dly ; Among them, the filter coefficient, service time threshold, and service delay threshold are business basic data, and the filter coefficient Ft k ∈(4,∞), indicating that the positive time quantity is filtered more, in Ft k The larger the value is within the range of ∈(4,∞), the more negative time quantities are filtered; the service time threshold and service delay threshold meet the condition Thr usd ∈(0,1),Thr dly ∈(0, 1).

[0020] (2-2-2): For each business Svr i , calculate the service time ratio Calculating the delay ratio

[0021] Calculate the positive time weight α separately i =x i / smx, negative time weight β i =y i / smy;

[0022] in

[0023] (2-2-3): For each business Svr i , calculate the initial weight The weights of positive and negative time quantities are matched through the filter coefficient. By setting the filter coefficient, the weight matching of positive and negative time quantities is achieved, thereby ensuring the 5QI indicator requirements of different businesses.

[0024] As a preferred solution, the process of setting the reward adjustment coefficient for the negative time-related business in step (2-3) includes:

[0025] (2-3-01): For the negative time service set Mset = {Svr j ,j=1,2…,m}, calculate the intermediate reward value MIWD j=y j / n ;

[0026] (2-3-02): Filter out Dly that meets Case 1 conditions in Mset j ≥Thr dly *Dlm j Subset of And satisfy Case2 condition Dly j <Thr dly *Dlm j Subset of Calculate their reward and punishment coefficients respectively

[0027]

[0028] As a preferred solution, the process of setting the penalty adjustment coefficient for the positive time-related business in step (2-3) includes:

[0029] (2-3-11): For the positive time amount service set Aset = {Svr k ,k=1,2,…,nm}, calculate the intermediate penalty value MIPN k =x k / n;

[0030] (2-3-12): Filter out Tusd that meets Case 3 conditions in Aset k ≥Thr usd *Tall k Subset of And satisfy Case 4 condition Tusd k <Thr usd *Tall k Subset of Calculate their reward and punishment coefficients respectively

[0031]

[0032] As a preferred solution, the specific process of steps (2-4) includes:

[0033] (2-4-1): For each service in the negative time service set Mset, calculate the priority adjustment coefficient Δ j =1+WP j , and update the priority γ j =Δ j *γ j ,

[0034] For each service in the positive time service set Aset, calculate the priority adjustment coefficient Δ k =1-WP k, and update the priority γ k =Δ k *γ k ;

[0035] (2-4-2): Get the updated priority {γ i}={γ j}∪{γ k}, i∈{1,2..n}.

[0036] Therefore, the advantages of the present invention are:

[0037] 1. Business data is sorted from the perspective of time, and business priorities are updated using a filtering system and reward and punishment adjustment coefficients. This prevents excessive rewards and penalties for businesses. Business indicators can be optimized within the threshold range of positive and negative time quantities. By matching the weights between positive and negative time quantities, the dynamic needs of different business indicators in different scenarios can be further met.

[0038] 2. The method is simpler, more robust, more applicable, and can meet differentiated individual needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flow chart of the present invention;

[0040] Figure 2 This is a comparison chart of throughput / positive time between the present invention and other algorithms;

[0041] Figure 3 This is a comparison chart of the time delay / negative time amount of the present invention and other algorithms;

[0042] Figure 4 This is a comparison chart of the scheduling fairness of the present invention and other algorithms. DETAILED DESCRIPTION

[0043] The technical solution of the present invention will be further specifically described below through embodiments and in conjunction with the accompanying drawings.

[0044] Example:

[0045] This embodiment provides a 5G fairness scheduling method based on time quantity, such as Figure 1 As shown, the following steps are included:

[0046] Step 1: Obtain user service data and extract time-related parameters;

[0047] The business includes n business Svr T ={Svr1, Svr2, ..., Svr n}, the extracted parameters include the current service time {Tusd1, Tusd2,…, Tusd n}, the total service time required {Tall1,Tall2,…,Tall n}, current service delay {Dly1, Dly2, …, Dly n}, the corresponding delay tolerance {Dlm1,Dlm2,…,Dlm n}, the number of rewards and punishments for each business {Nm1, Nm2,…, Nm n}, reward and punishment coefficients {WP1, WP2,…, WP n}.

[0048] Step 2: Update business priorities by combining the filtering system and the reward and punishment assessment mechanism;

[0049] (2-1): For the current business, a positive and negative time amount business set is constructed based on the time amount relationship, and reward and punishment regression judgment is performed on each business; the specific process includes:

[0050] (2-1-1): Set up a service data table. The service data table includes the service name, current service time, total service time required, service delay, delay tolerance, current number of rewards and punishments, and service attributes. Based on the positive and negative correlation with the amount of time, the service attributes include positive time services and negative time services. Negative time services represent services that are negatively correlated with the amount of time, such as services that are sensitive to service delays; positive time services represent services that are positively correlated with the amount of time, such as services that are sensitive to service time.

[0051] (2-1-2): According to the service attributes, the services are divided into negative time services and positive time services; thus, the negative time service set Mset = {Svr j ,j=1,2…,m}, positive time quantity service set Aset={Svr k ,k=1,2,…,nm}; wherein negative time quantity business indicates business that is negatively correlated with time quantity, such as business that is sensitive to business delay; positive time quantity business indicates business that is positively correlated with time quantity, such as business that is sensitive to service time.

[0052] (2-1-3): Set the reward and punishment threshold Thr PW , for the current moment Svr T ={Svr1, Svr2, ..., Svr n Each business Svr in i , check Nm i , if Nm i ≥Thr PW , then Nm i =0,WP i =0.

[0053] (2-2): Initialize the weight of each service based on the time-related parameters. The specific process includes:

[0054] (2-2-1): Set the filter coefficient Ft k , service time threshold Thr usd , service delay threshold Thr dly ; Among them, the filter coefficient, service time threshold, and service delay threshold are business basic data, and the filter coefficient Ft k ∈(4,∞), indicating that the positive time quantity is filtered more, in Ft k The larger the value is within the range of ∈(4,∞), the more negative time quantities are filtered; the service time threshold and service delay threshold meet the condition Thr usd ∈(0,1),Thr dly ∈(0, 1).

[0055] (2-2-2): For each business Svr i , calculate the service time ratio Calculating the delay ratio

[0056] Calculate the positive time weight α separately i =x i / smx, negative time weight β i =y i / smy;

[0057] in

[0058] (2-2-3): For each business Svr i , calculate the initial weight The weights of positive and negative time quantities are matched through the filter coefficient. By setting the filter coefficient, the weight matching of positive and negative time quantities is achieved, thereby ensuring the 5QI indicator requirements of different businesses.

[0059] (2-3): Set a penalty adjustment coefficient for negative time-related services. The process includes:

[0060] (2-3-01): For the negative time service set Mset = {Svr j ,j=1,2…,m}, calculate the intermediate reward value MIWD j =y j / n ;

[0061] (2-3-02): Filter out Dly that meets Case 1 conditions in Mset j ≥Thr dly *Dlm j Subset of And satisfy Case2 condition Dly j<Thr dly *Dlm j Subset of Calculate their reward and punishment coefficients respectively

[0062]

[0063] Set reward adjustment coefficients for positive time-related business; the process includes:

[0064] (2-3-11): For the positive time amount service set Aset = {Svr k ,k=1,2,…,nm}, calculate the intermediate penalty value MIPN k =x k / n;

[0065] (2-3-12): Filter out Tusd that meets Case 3 conditions in Aset k ≥Thr usd *Tall k Subset of And satisfy Case 4 condition Tusd k <Thr usd *Tall k Subset of Calculate their reward and punishment coefficients respectively

[0066]

[0067] (2-4): Update the priority based on the reward and penalty adjustment coefficient; the specific process includes:

[0068] (2-4-1): For each service in the negative time service set Mset, calculate the priority adjustment coefficient Δ j =1+WP j , and update the priority γ j =Δ j *γ j ,

[0069] For each service in the positive time service set Aset, calculate the priority adjustment coefficient Δ k =1-WP k , and update the priority γ k =Δ k *γ k ;

[0070] (2-4-2): Get the updated priority {γ i}={γ j}∪{γ k}, i∈{1,2..n}.

[0071] Step 3: Send the service to the 5G scheduler for scheduling based on the updated priority.

[0072] The present embodiment is described below using a specific case. Taking n=6 as an example, the services currently being carried out by the 5G system are shown in Table 1:

[0073]

[0074] Table 1

[0075] The basic data are shown in Table 2:

[0076]

[0077]

[0078] Table 2

[0079] A 5G fairness scheduling method based on time quantity includes the following steps:

[0080] Step 1: Get user service data and extract time-related parameters; the current service is Svr T ={Svr1, Svr2,…, Svr6}.

[0081] Step 2: Update business priorities by combining the filtering system and the reward and punishment assessment mechanism, including:

[0082] (2-1): For the current business, a positive and negative time amount business set is constructed based on the time amount relationship, and reward and punishment regression judgment is performed on each business;

[0083] For the current business Svr T = {Svr1, Svr2, ..., Svr6}, construct the negative time service set Mset = {Svr1, Svr2, Svr3} and the positive time service set Aset = {Svr4, Svr5, Svr6} according to the service attributes. The number of rewards and punishments in the initial state is 0, and the reward and punishment threshold Thr is set. PW , for each business Svr i Verify Nm i , there is no Nm that satisfies the condition i ≥Thr PW business, no need to return.

[0084] (2-2): Initialize the weight of each business according to the time-related parameters;

[0085] (2-2-1): Set the filter coefficient Ft k , service time threshold Thr usd , service delay threshold Thr dly ;

[0086] (2-2-2): For each business Svr i , calculate the service time ratio

[0087]

[0088] Calculating the delay ratio

[0089]

[0090] Sum

[0091] Calculating positive time quantity weights

[0092] α i =x i / smx={0.16,0.15,0.21,0.22,0.09,0.17},

[0093] Calculating negative time weights

[0094] β i =y i / smy={0.19,0.09,0.21,0.19,0.17,0.16}.

[0095] (2-2-3): For each business Svr i , calculate the initial weight

[0096]

[0097] (2-3) Setting a penalty adjustment coefficient for negative time-related services includes:

[0098] (2-3-01): For each service in the negative time service set Mset = {Svr1, Svr2, Svr3}, calculate the intermediate reward value MIWD j =y j / n={0.13,0.07,0.15};

[0099] (2-3-02): Filter out Dly that meets Case 1 conditions in Mset j ≥Thr dly *Dlm j The subset Mset1={Svr1,Svr3}, and the subset Dly that satisfies Case2 condition j <Thr dly *Dlm j The subset Mset2 = {Svr2}, calculate its reward and punishment coefficients respectively

[0100] WP1=MIWD1=y1 / 6=0.13, WP3=0.15, WP2=0.

[0101] Set reward adjustment coefficients for positive time-related business; the process includes:

[0102] (2-3-11): For each service in the positive time service set Aset = {Svr4, Svr5, Svr6}, calculate the intermediate penalty value MIPN k =x k / n={0.14,0.06,0.11};

[0103] (2-3-12): Filter out Tusd that meets Case 3 conditions in Aset k ≥Thr usd *Tall k The subset Aset1={Svr4,Svr6}, and the condition Tusd that satisfies Case4 k <Thr usd *Tall k The subset Aset2={Svr5}, calculate its reward and punishment coefficients respectively

[0104] WP4=MIPN4=x4 / 6=0.14, WP6=0.11, WP5=0.

[0105] (2-4) Update the priority based on the reward and penalty adjustment coefficient; the specific process includes:

[0106] (2-4-1): For each service in the negative time service set Mset, calculate the priority adjustment coefficient

[0107] Δ j =1+WP j ={1.13,1,1.15},

[0108] and update the priority

[0109] γ j =Δ j *γ j ={0.19,0.13,0.24},

[0110] For each service in the positive time service set Aset, calculate the priority adjustment coefficient

[0111] Δ k =1-WP k ={0.86,1,0.89},

[0112] and update the priority

[0113] γ k=Δ k *γ k ={0.18,0.12,0.15};

[0114] (2-4-2): Get the updated priority

[0115] {γ i}={γ j}∪{γ k}={0.19,0.13,0.24,0.18,0.12,0.15}.

[0116] Step 3: Send the service to the 5G scheduler for scheduling based on the updated priority.

[0117] The simulation experiment is explained:

[0118] Four methods, including TDFS (the method of the present invention), RR (Round Robin), SJF (Short Job First), and HRRN (Highest Response Ratio Next), were simulated on the MATLAB platform. A certain number of users were randomly introduced and random services were configured. The overall load of the 5G base station gNB was maintained at around 50%. Reward and penalty regression was forced on several of the services. The throughput / positive time quantity, latency / negative time quantity, and fairness were obtained, respectively. Figures 2 to 4 shown.

[0119] like Figure 2 As shown in the figure, TDFS will try its best to schedule services that have not reached the threshold value, thereby improving positive time quantity indicators such as throughput, adjusting the filter coefficient, and moderately increasing the weight of throughput. The simulation results show that TDFS throughput is particularly outstanding among the four algorithms; SJF is a short job priority method, and due to its mechanism, its throughput is the lowest.

[0120] like Figure 3 As shown in the figure, in terms of delay control of negative time quantities, TDFS and HRRN with high response ratio priority are roughly similar, and both have a similar mechanism of pursuing timeout rewards; and as the number of simulated users increases, the delay of non-preemptive SJF will become larger and larger.

[0121] like Figure 4 As shown in the figure, RR still outperforms the other algorithms in terms of fairness, followed by the TDFS algorithm of our invention, followed by HRRN and SJF. Unlike HRRN, TDFS's fairness is flexible and can be dynamically adjusted through reward and penalty thresholds and filter coefficients.

[0122] The specific embodiments described herein are merely illustrative of the spirit of the present invention.

[0123] Those skilled in the art may make various modifications or additions to the described embodiments or replace them with similar methods without departing from the spirit of the present invention or exceeding the scope defined by the appended claims.

Claims

1. A 5G fairness scheduling method based on time quantity, characterized by: The following steps are involved: Step 1: Obtain user service data and extract time-related parameters; Step 2: Update business priorities by combining the filtering system and the reward and punishment assessment mechanism; (2-1): For the current business, a positive and negative time amount business set is constructed based on the time amount relationship, and reward and punishment regression judgment is performed on each business based on the reward and punishment threshold; Negative time volume business is negatively correlated with time volume, and positive time volume business is positively correlated with time volume; (2-2): Initialize the weight of each business according to the time-related parameters; Calculate the service time ratio of each business = service time / total service time, delay ratio = business delay / delay tolerance, Calculate the weight of each service's positive time as the service time ratio, and the weight of the negative time as the delay ratio. Obtain the initial weight of each service through filter coefficient weight matching. (2-3): For services related to negative time quantity, the reward and penalty coefficients are set based on the intermediate reward value obtained based on the delay ratio and the delay threshold; For positive time-related services, the reward and penalty coefficients are set based on the intermediate penalty value obtained based on the service time ratio and the service time threshold; (2-4): Update the priority based on the reward and punishment coefficient; The service priority of a negative time service set is (1 + reward and punishment coefficient) * its own priority, and the service priority of a positive time service set is (1 - reward and punishment coefficient) * its own priority; Step 3: Send the service to the 5G scheduler for scheduling based on the updated priority.

2. A 5G fairness scheduling method based on time quantity according to claim 1, characterized in that The business in step 1 includes n business Svr T ={Svr1, Svr2, ..., Svr n }, the extracted parameters include the current service time {Tusd1, Tusd2,…, Tusd n }, the total service time required {Tall1,Tall2,…,Tall n }, current service delay {Dly1, Dly2, …, Dly n }, the corresponding delay tolerance {Dlm1,Dlm2,…,Dlm n }, the number of rewards and punishments for each business {Nm1, Nm2,…, Nm n }, reward and punishment coefficients {WP1, WP2,…, WP n }.

3. A 5G fairness scheduling method based on time quantity according to claim 2, characterized in that The specific process of step (2-1) includes: (2-1-1): Set up a service data table. The service data table includes the service name, current service time, required total service time, service delay, delay tolerance, current number of rewards and penalties, and service attributes. Based on the positive and negative correlation with time, service attributes include positive time-volume services and negative time-volume services. (2-1-2): According to the service attributes, the services are divided into negative time services and positive time services; thus, the negative time service set Mset = {Svr j ,j=1,2…,m}, positive time quantity service set Aset={Svr k ,k=1,2,…,nm}; (2-1-3): Set the reward and punishment threshold Thr PW , for the current moment Svr T ={Svr1, Svr2, ..., Svr n Each business Svr in i , check Nm i , if Nm i ≥Thr PW , then Nm i =0,WP i =0.

4. The 5G fairness scheduling method based on time quantity according to claim 2 is characterized in that The specific process of step (2-2) includes: (2-2-1): Set the filter coefficient Ft k , service time threshold Thr usd , service delay threshold Thr dly ; (2-2-2): For each business Svr i , calculate the service time ratio Calculating the delay ratio Calculate the positive time weight α separately i =x i / smx, negative time weight β i =y i / smy; (2-2-3): For each business Svr i , calculate the initial weight 5. A 5G fairness scheduling method based on time quantity according to claim 4, characterized in that step( The process of setting reward and penalty coefficients for negative time-related businesses in 2-3) includes: (2-3-01): For the negative time service set Mset = {Svr j ,j=1,2…,m}, calculate the intermediate reward value MIWD j =y j / n; (2-3-02): Filter out Dly that meets Case 1 conditions in Mset j ≥Thr dly *Dlm j Subset of And satisfy Case2 condition Dly j <Thr dly *Dlm j Subset of Calculate their reward and punishment coefficients respectively 6. A 5G fairness scheduling method based on time quantity according to claim 5, characterized in that step( The process of setting the reward and penalty coefficient for the positive time-related business in 2-3) includes: (2-3-11): For the positive time amount service set Aset = {Svr k ,k=1,2,…,nm}, calculate the intermediate penalty value MIPN k =x k / n; (2-3-12): Filter out Tusd that meets Case 3 conditions in Aset k ≥Thr usd *Tall k Subset of And satisfy Case 4 condition Tusd k <Thr usd *Tall k Subset of Calculate their reward and punishment coefficients respectively 7. The 5G fairness scheduling method based on time quantity according to claim 6 is characterized in that The specific process of steps (2-4) includes: (2-4-1): For each service in the negative time service set Mset, calculate the priority adjustment coefficient Δ j =1+WP j , and update the priority γ j =Δ j *γ j , For each service in the positive time service set Aset, calculate the priority adjustment coefficient Δ k =1-WP k , and update the priority γ k =Δ k *γ k ; (2-4-2): Get the updated priority {γ i }={γ j }∪{γ k }, i∈{1,2…n}.

Images