CN117950879B - Self-adaptive cloud server distribution method and device and computer equipment - Google Patents
Self-adaptive cloud server distribution method and device and computer equipment Download PDFInfo
- Publication number
- CN117950879B CN117950879B CN202410346154.XA CN202410346154A CN117950879B CN 117950879 B CN117950879 B CN 117950879B CN 202410346154 A CN202410346154 A CN 202410346154A CN 117950879 B CN117950879 B CN 117950879B
- Authority
- CN
- China
- Prior art keywords
- cloud server
- distribution
- demand data
- static
- cloud
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 28
- 230000003068 static effect Effects 0.000 claims abstract description 116
- 238000004364 calculation method Methods 0.000 claims abstract description 115
- 238000013528 artificial neural network Methods 0.000 claims abstract description 31
- 230000006870 function Effects 0.000 claims description 44
- 238000012549 training Methods 0.000 claims description 19
- 230000003044 adaptive effect Effects 0.000 claims description 12
- 238000010276 construction Methods 0.000 claims description 7
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 238000012544 monitoring process Methods 0.000 claims description 3
- 238000011002 quantification Methods 0.000 claims description 3
- 230000006978 adaptation Effects 0.000 abstract description 5
- 238000011161 development Methods 0.000 abstract description 5
- 230000001360 synchronised effect Effects 0.000 description 7
- 238000012821 model calculation Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of cloud computing, in particular to a self-adaptive cloud server distribution method, a self-adaptive cloud server distribution device and computer equipment, which comprise the following steps: acquiring calculation force demand data of each cloud server distribution area in a period of time, and acquiring calculation force demand data sequences of each cloud server distribution area; at each cloud server distribution area, learning the association relation between the calculation force demand data sequence and the cloud server distribution quantity by using a neural network to obtain a cloud server static distribution model; synchronously combining cloud server static distribution models at all cloud server distribution areas to obtain cloud server dynamic distribution models; and carrying out self-adaptive allocation of cloud servers in each cloud server distribution area by utilizing the cloud server dynamic distribution model. According to the invention, after the cloud server layout is completed, the method is suitable for the computing power demand development conditions of all areas, and the self-adaptive adaptation of the cloud server distribution is performed, so that the cloud server resources are reasonably distributed.
Description
Technical Field
The invention relates to the technical field of cloud computing, in particular to a self-adaptive cloud server distribution method, a self-adaptive cloud server distribution device and computer equipment.
Background
The cloud server (Elastic Compute Service, ECS) is a simple, efficient, safe and reliable computing service with elastically scalable processing capabilities. The management mode is simpler and more efficient than that of a physical server. A user can quickly create or release any plurality of cloud servers without purchasing hardware in advance.
The distribution of cloud servers determines the computing power situation of the region where the cloud servers are located. In the prior art, the number of cloud servers in each region is determined by artificial subjectivity, so that the distribution setting of the cloud servers is completed. Therefore, the cloud servers are high in distribution randomness, are difficult to adapt to actual conditions or actual demands of all areas, are not adjusted after the distribution is completed, are more difficult to adapt to development conditions of all areas, and are poor in distribution effect.
Disclosure of Invention
The invention aims to provide a self-adaptive cloud server distribution method, a self-adaptive cloud server distribution device and computer equipment, which are used for solving the technical problems that in the prior art, the distribution randomness of a cloud server is high, the real situation or real requirement of each region is difficult to adapt, after the distribution is completed, the distribution of the cloud server is difficult to adapt to the development situation of each region without adjustment, and the cloud server distribution effect is poor.
In order to solve the technical problems, the invention specifically provides the following technical scheme:
in a first aspect of the present invention, a method for distributing an adaptive cloud server includes the steps of:
acquiring calculation force demand data of each cloud server distribution area in a period of time, and acquiring calculation force demand data sequences of each cloud server distribution area;
At each cloud server distribution area, learning the association relation between the computing power demand data sequence and the cloud server distribution quantity by using a neural network to obtain a cloud server static distribution model for measuring the cloud server distribution quantity according to the computing power demand at each cloud server distribution area;
Synchronously combining cloud server static distribution models at all cloud server distribution areas to obtain cloud server dynamic distribution models for adaptively allocating cloud server distribution according to calculation power requirements at all cloud server distribution areas;
And carrying out self-adaptive allocation of cloud servers in each cloud server distribution area by utilizing the cloud server dynamic distribution model.
As a preferable mode of the present invention, the calculation force demand data sequence is a data sequence with time sequence as attribute, and the expression of the calculation force demand data sequence is: { q t |t ε [ t0, tn ] }, where q t is the calculation force demand data at the time t, t0 is the initial time sequence for acquiring the calculation force demand, and tn is the end time sequence for acquiring the calculation force demand. As a preferable scheme of the invention, the method for constructing the cloud server static distribution model comprises the following steps:
each calculation force demand data in the calculation force demand data sequence of each cloud server distribution area is correspondingly converted into the number of cloud servers according to the calculation force of a single cloud server;
Sequencing the number of cloud servers corresponding to each power demand data in the power demand data sequence according to the sequence of the power demand data sequence to obtain a cloud server number sequence;
on each cloud server distribution area, taking the computing power demand data sequence as an input item of the neural network, and taking the cloud server quantity sequence as an output item of the neural network;
on each cloud server distribution area, learning and training a calculation force demand data sequence and a cloud server number sequence by using a neural network to obtain a cloud server static distribution model;
the expression of the cloud server static distribution model is as follows:
Severs _num k=BP(computility_datak), wherein Severs _num k is a cloud server number sequence of a kth cloud server distribution area, computility _data k is a calculation power demand data sequence of the kth cloud server distribution area, and BP is a neural network.
As a preferable scheme of the invention, the method for constructing the cloud server dynamic distribution model comprises the following steps:
acquiring a loss function of a cloud server static distribution model of each cloud server distribution area;
weighting and combining the loss functions of the cloud server static distribution models of all cloud server distribution areas to obtain combined loss functions;
synchronously training cloud server static distribution models of all cloud server distribution areas by using a combined type loss function to obtain cloud server dynamic distribution models of the cloud server distribution areas;
The model expression of the cloud server dynamic distribution model is as follows: Wherein Severs _num 1、Severs_num2 and Severs _num N are cloud server number sequences of 1 st, 2 nd and N th cloud server distribution areas, computility _data 1、computility_data2 and computility _data N are calculation force demand data sequences of 1 st, 2 nd and N th cloud server distribution areas, and BP is a neural network.
As a preferred embodiment of the present invention, the method for constructing the combined loss function includes:
acquiring the accuracy of a static distribution model of a cloud server and the variability of a calculation power demand data sequence of a distribution area of the cloud server;
Determining the weighting weight of each cloud server static distribution model based on the accuracy of the cloud server static distribution model and the variability of the calculation force demand data sequence;
And weighting and combining the loss functions of the cloud server static distribution models of the cloud server distribution areas by using the weighting weights of the static distribution of each cloud server to obtain a combined loss function, wherein the expression of the combined loss function is as follows: ；
Wherein Loss is a combined type Loss function, loss k is a Loss function of a k cloud server static distribution model, W k is a weighting weight of the k cloud server static distribution model, and N is the total number of the cloud server static distribution models; ;
wherein W k is the weighted weight of the static distribution model of the kth cloud server, H k is the variability of the calculation power demand data sequence of the distribution area of the kth cloud server, and Z k is the accuracy of the static distribution model of the kth cloud server.
As a preferable scheme of the invention, the accuracy of the cloud server static distribution model is the ratio of the correct number of model predictions to the total number of model predictions.
As a preferable scheme of the present invention, the variability of the power demand data sequence in the cloud server distribution area is quantified by the discrete rate of the power demand data sequence, and the quantification formula of the variability is:；
wherein H k is the variability of the power demand data sequence of the kth cloud server distribution area, q kt is the power demand data at time t in the power demand data sequence of the kth cloud server distribution area, t0 is the initial time sequence for acquiring the power demand, and tn is the final time sequence for acquiring the power demand.
As a preferred scheme of the invention, the cloud server self-adaptive allocation method for each cloud server distribution area by utilizing the cloud server dynamic distribution model comprises the following steps:
Obtaining the number of cloud servers in the current time period of each cloud server distribution area according to the calculation power demand data sequence in the current time period by utilizing a cloud server dynamic distribution model;
Comparing the number of cloud servers in the current time period of each cloud server distribution area with the number of cloud servers in the previous time period, wherein,
When the number of cloud servers in the current time period is smaller than that of cloud servers in the previous time period, taking the cloud server distribution area as a server movable area;
when the number of cloud servers in the current time period is larger than that of cloud servers in the previous time period, taking the cloud server distribution area as a server migratable area;
and (5) adaptively allocating the mobile areas of the server according to the mobile areas of the server.
In a second aspect of the present invention, the present invention provides an adaptive cloud server distribution apparatus, including:
the data monitoring unit is used for acquiring the calculation power demand data of each cloud server distribution area in a period of time to obtain a calculation power demand data sequence of each cloud server distribution area;
the model training unit is used for learning the association relation between the computing power demand data sequence and the cloud server distribution quantity by using the neural network at each cloud server distribution area to obtain a cloud server static distribution model for measuring the cloud server distribution quantity according to the computing power demand at each cloud server distribution area; and
The cloud server dynamic distribution model is used for synchronously combining the cloud server static distribution models at all cloud server distribution areas to obtain cloud server dynamic distribution models at all cloud server distribution areas for adaptively allocating cloud server distribution according to calculation power requirements;
and the self-adaptive allocation unit is used for carrying out self-adaptive allocation of the cloud servers on each cloud server distribution area by utilizing the cloud server dynamic distribution model.
In a third aspect of the invention, the invention provides a computer device comprising: at least one processor; and
A memory communicatively coupled to the at least one processor;
Wherein the memory stores instructions executable by the at least one processor to cause a computer device to perform the adaptive cloud server distribution method based on computing power requirements.
Compared with the prior art, the invention has the following beneficial effects:
According to the cloud server distribution method, the cloud server static distribution models of the cloud server distribution quantity are measured and calculated according to the calculation force demand by utilizing the neural network, the cloud server distribution is realized to adapt to the actual situation or actual requirement of each region, then the cloud server static distribution models of all cloud server distribution regions are synchronously combined to obtain the cloud server dynamic distribution model for self-adaptively allocating the cloud server distribution according to the calculation force demand, the cloud server distribution self-adaptive adaptation is realized to adapt to the calculation force demand development situation of all regions after the cloud server distribution is completed, and the cloud server resources are reasonably distributed.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It will be apparent to those of ordinary skill in the art that the drawings in the following description are exemplary only and that other implementations can be obtained from the extensions of the drawings provided without inventive effort.
FIG. 1 is a flowchart of a method for distributing an adaptive cloud server based on computing power requirements according to an embodiment of the present invention;
Fig. 2 is a block diagram of an adaptive cloud server distribution device based on computing power requirements according to an embodiment of the present invention;
fig. 3 is an internal structure diagram of a computer device according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in fig. 1, in a first aspect of the present invention, the present invention provides an adaptive cloud server distribution method, including the following steps:
acquiring calculation force demand data of each cloud server distribution area in a period of time, and acquiring calculation force demand data sequences of each cloud server distribution area;
At each cloud server distribution area, learning the association relation between the computing power demand data sequence and the cloud server distribution quantity by using a neural network to obtain a cloud server static distribution model for measuring the cloud server distribution quantity according to the computing power demand at each cloud server distribution area;
Synchronously combining cloud server static distribution models at all cloud server distribution areas to obtain cloud server dynamic distribution models for adaptively allocating cloud server distribution according to calculation power requirements at all cloud server distribution areas;
And carrying out self-adaptive allocation of cloud servers in each cloud server distribution area by utilizing the cloud server dynamic distribution model.
In order to adapt to the calculation power demand of the area, the cloud server static distribution model is constructed by utilizing the neural network, so that the quantity of the optimal cloud servers can be calculated quantitatively according to the calculation power demand of the cloud server distribution area, the calculation power demand of the area is further ensured to be met, and the redundant distribution of the cloud servers is avoided.
The cloud server static distribution model is obtained by calculation according to static historical data, is fixed and is difficult to adapt to dynamic change of calculation force requirements of cloud server distribution areas, so that the cloud server dynamic distribution model is synchronously combined and trained at each cloud server distribution area to obtain the cloud server dynamic distribution model for adaptively allocating cloud server distribution according to calculation force requirements at each cloud server distribution area, the required cloud server quantity is calculated in real time according to calculation force requirements of each cloud server distribution area, dynamic adjustment is carried out on the cloud server quantity of each cloud server distribution area, self-adaptive allocation of cloud servers is finally achieved, redundant distribution of cloud servers in each area is avoided while calculation force services of each area are guaranteed, and dynamic rationalization allocation of cloud server resources is achieved.
Specifically, the cloud server static distribution model calculation method and the cloud server static distribution model calculation system realize synchronous training by constructing a uniform loss function, so that synchronous combination of cloud server static distribution models at all cloud server distribution areas is realized, the calculation synchronism of all cloud server static distribution models is ensured, the calculation force requirements of all distribution areas can be uniformly calculated, the difference of all distribution areas is eliminated, and the calculation accuracy of the cloud servers is ensured.
In addition, in the loss function, each cloud server static distribution model has a weight, the weight consists of the accuracy of the cloud server static distribution model and the fluctuation rate of the calculation power demand data sequence of the cloud server distribution region, when the fluctuation rate of the calculation power demand data sequence is higher, the fluctuation degree of the calculation power demand of the corresponding cloud server distribution region is larger, which means that the accuracy of the cloud server static distribution model of the region is lower, and retraining is needed, therefore, the invention endows the region with higher weight so that all the static distribution models have more learning resources during synchronous training, and the number of the cloud servers which are suitable for the calculation power demand of the region can be calculated fastest. The accuracy of the cloud server static distribution model is used as a weight adjustment coefficient, so that the lower the accuracy of the cloud server static distribution model in a region is, the lower the weight is given to the region, and unnecessary learning resources are prevented from being robbed by the static distribution model.
Therefore, the self-adaptive optimization of the static distribution model synchronous training performance is performed by using the weights, so that the optimal distribution of learning training resources can be realized while the accurate adaptation to the change of computing power demands in each region is ensured, and the redundant distribution of cloud servers is avoided.
The calculation force demand data sequence is a data sequence taking time sequence as an attribute, and the expression of the calculation force demand data sequence is as follows: { q t |t ε [ t0, tn ] }, where q t is the calculation force demand data at the time t, t0 is the initial time sequence for acquiring the calculation force demand, and tn is the end time sequence for acquiring the calculation force demand. In order to adapt to the calculation power demand of the area, the invention utilizes the neural network to construct the cloud server static distribution model, can realize the quantitative calculation of the quantity of the optimal cloud servers according to the calculation power demand of the cloud server distribution area, further ensures that the calculation power demand of the area is met, and simultaneously avoids the redundant distribution of the cloud servers, and is specific:
The construction method of the cloud server static distribution model comprises the following steps:
each calculation force demand data in the calculation force demand data sequence of each cloud server distribution area is correspondingly converted into the number of cloud servers according to the calculation force of a single cloud server;
Sequencing the number of cloud servers corresponding to each power demand data in the power demand data sequence according to the sequence of the power demand data sequence to obtain a cloud server number sequence;
on each cloud server distribution area, taking the computing power demand data sequence as an input item of the neural network, and taking the cloud server quantity sequence as an output item of the neural network;
on each cloud server distribution area, learning and training the calculation force demand data sequence and the cloud server number sequence by using a neural network to obtain a cloud server static distribution model;
the expression of the cloud server static distribution model is as follows:
Severs _num k=BP(computility_datak), wherein Severs _num k is a cloud server number sequence of a kth cloud server distribution area, computility _data k is a calculation power demand data sequence of the kth cloud server distribution area, and BP is a neural network.
The cloud server static distribution model is obtained by calculation according to static historical data, is fixed and is difficult to adapt to the dynamic change of calculation power demands of cloud server distribution areas, so that the cloud server dynamic distribution model at each cloud server distribution area is synchronously combined and synchronously trained to obtain the cloud server dynamic distribution model for adaptively allocating cloud server distribution according to calculation power demands at each cloud server distribution area, the required cloud server quantity is calculated in real time according to calculation power demand conditions of each cloud server distribution area, the cloud server quantity of each cloud server distribution area is dynamically adjusted, the self-adaptive allocation of cloud servers is finally realized, the calculation power service of each area is ensured, the redundant distribution of cloud servers in each area is avoided, the dynamic rationalization allocation of cloud server resources is realized, and the cloud server static distribution model is particularly:
The construction method of the cloud server dynamic distribution model comprises the following steps:
acquiring a loss function of a cloud server static distribution model of each cloud server distribution area;
weighting and combining the loss functions of the cloud server static distribution models of all cloud server distribution areas to obtain combined loss functions;
Synchronously training the cloud server static distribution models of all cloud server distribution areas by using a combined type loss function to obtain cloud server dynamic distribution models of the cloud server distribution areas;
The model expression of the cloud server dynamic distribution model is as follows: ;
Wherein Severs _num 1、Severs_num2 and Severs _num N are cloud server number sequences of 1 st, 2 nd and N th cloud server distribution areas, computility _data 1、computility_data2 and computility _data N are calculation force demand data sequences of 1 st, 2 nd and N th cloud server distribution areas, and BP is a neural network.
The construction method of the combined loss function comprises the following steps:
acquiring the accuracy of a static distribution model of a cloud server and the variability of a calculation power demand data sequence of a distribution area of the cloud server;
Determining the weighting weight of each cloud server static distribution model based on the accuracy of the cloud server static distribution model and the variability of the calculation force demand data sequence;
and weighting and combining the loss functions of the cloud server static distribution models of the cloud server distribution areas by using the weighting weights of the static distribution of each cloud server to obtain a combined loss function, wherein the expression of the combined loss function is as follows: ; wherein Loss is a combined type Loss function, loss k is a Loss function of a k cloud server static distribution model, W k is a weighting weight of the k cloud server static distribution model, and N is the total number of the cloud server static distribution models; /(I) Wherein, W k is the weight of the static distribution model of the kth cloud server, H k is the variability of the calculation power demand data sequence of the distribution area of the kth cloud server, and Z k is the accuracy of the static distribution model of the kth cloud server.
The accuracy of the cloud server static distribution model is the ratio of the correct number of model predictions to the total number of model predictions.
The change rate of the calculation force demand data sequence in the cloud server distribution area is quantified through the discrete rate of the calculation force demand data sequence, and the quantification formula of the change rate is as follows:；
wherein H k is the variability of the power demand data sequence of the kth cloud server distribution area, q kt is the power demand data at time t in the power demand data sequence of the kth cloud server distribution area, t0 is the initial time sequence for acquiring the power demand, and tn is the final time sequence for acquiring the power demand. According to the cloud server static distribution model calculation method and the cloud server static distribution model calculation system, the unified loss function is built for synchronous training, so that the cloud server static distribution models at the cloud server distribution areas are synchronously combined, the calculation synchronism of the cloud server static distribution models is guaranteed, the calculation power requirements of the cloud server distribution areas can be uniformly calculated, the difference of the cloud server distribution areas is eliminated, and the calculation accuracy of the cloud server is guaranteed.
In addition, in the loss function, each cloud server static distribution model has a weight, the weight consists of the accuracy of the cloud server static distribution model and the fluctuation rate of the calculation power demand data sequence of the cloud server distribution region, when the fluctuation rate of the calculation power demand data sequence is higher, the fluctuation degree of the calculation power demand of the corresponding cloud server distribution region is larger, which means that the accuracy of the cloud server static distribution model of the region is lower, and retraining is needed, therefore, the invention endows the region with higher weight so that all the static distribution models have more learning resources during synchronous training, and the number of the cloud servers which are suitable for the calculation power demand of the region can be calculated fastest. The accuracy of the cloud server static distribution model is used as a weight adjustment coefficient, so that the lower the accuracy of the cloud server static distribution model in a region is, the lower the weight is given to the region, and unnecessary learning resources are prevented from being robbed by the static distribution model.
Therefore, the self-adaptive optimization of the static distribution model synchronous training performance is performed by using the weights, so that the optimal distribution of learning training resources can be realized while the accurate adaptation to the change of computing power demands in each region is ensured, and the redundant distribution of cloud servers is avoided.
The cloud server self-adaptive allocation method for each cloud server distribution area by utilizing the cloud server dynamic distribution model comprises the following steps:
Obtaining the number of cloud servers in the current time period of each cloud server distribution area according to the calculation power demand data sequence in the current time period by utilizing a cloud server dynamic distribution model;
Comparing the number of cloud servers in the current time period of each cloud server distribution area with the number of cloud servers in the previous time period, wherein,
When the number of cloud servers in the current time period is smaller than that of cloud servers in the previous time period, taking the cloud server distribution area as a server movable area;
when the number of cloud servers in the current time period is larger than that of cloud servers in the previous time period, taking the cloud server distribution area as a server migratable area;
and (5) adaptively allocating the mobile areas of the server according to the mobile areas of the server.
As shown in fig. 2, in a second aspect of the present invention, the present invention provides an adaptive cloud server distribution apparatus, including:
the data monitoring unit is used for acquiring the calculation power demand data of each cloud server distribution area in a period of time to obtain a calculation power demand data sequence of each cloud server distribution area;
the model training unit is used for learning the association relation between the computing power demand data sequence and the cloud server distribution quantity by using the neural network at each cloud server distribution area to obtain a cloud server static distribution model for measuring the cloud server distribution quantity according to the computing power demand at each cloud server distribution area; and
The cloud server dynamic distribution model is used for synchronously combining the cloud server static distribution models at all cloud server distribution areas to obtain cloud server dynamic distribution models at all cloud server distribution areas for adaptively allocating cloud server distribution according to calculation power requirements;
and the self-adaptive allocation unit is used for carrying out self-adaptive allocation of the cloud servers on each cloud server distribution area by utilizing the cloud server dynamic distribution model.
As shown in fig. 3, in a third aspect of the present invention, the present invention provides a computer apparatus comprising: at least one processor; and
A memory communicatively coupled to the at least one processor;
The memory stores instructions executable by the at least one processor to cause the computer device to perform an adaptive cloud server distribution method based on computational power requirements.
According to the cloud server distribution method, the cloud server static distribution models of the cloud server distribution quantity are measured and calculated according to the calculation force demand by utilizing the neural network, the cloud server distribution is realized to adapt to the actual situation or actual requirement of each region, then the cloud server static distribution models of all cloud server distribution regions are synchronously combined to obtain the cloud server dynamic distribution model for self-adaptively allocating the cloud server distribution according to the calculation force demand, the cloud server distribution self-adaptive adaptation is realized to adapt to the calculation force demand development situation of all regions after the cloud server distribution is completed, and the cloud server resources are reasonably distributed.
The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application, the scope of which is defined by the claims. Various modifications and equivalent arrangements of this application will occur to those skilled in the art, and are intended to be within the spirit and scope of the application.
Claims (6)
1. The self-adaptive cloud server distribution method is characterized by comprising the following steps of:
acquiring calculation force demand data of each cloud server distribution area in a period of time, and acquiring calculation force demand data sequences of each cloud server distribution area;
At each cloud server distribution area, learning the association relation between the computing power demand data sequence and the cloud server distribution quantity by using a neural network to obtain a cloud server static distribution model for measuring the cloud server distribution quantity according to the computing power demand at each cloud server distribution area;
Synchronously combining cloud server static distribution models at all cloud server distribution areas to obtain cloud server dynamic distribution models for adaptively allocating cloud server distribution according to calculation power requirements at all cloud server distribution areas;
Performing self-adaptive allocation of cloud servers in each cloud server distribution area by utilizing a cloud server dynamic distribution model;
The power demand data sequence is a data sequence taking time sequence as an attribute, and the expression of the power demand data sequence is as follows: { q t |t ε [ t0, tn ] }, wherein q t is the calculation force demand data at time t, t0 is the initial time sequence for acquiring the calculation force demand, and tn is the end time sequence for acquiring the calculation force demand; the method for constructing the cloud server static distribution model comprises the following steps:
each calculation force demand data in the calculation force demand data sequence of each cloud server distribution area is correspondingly converted into the number of cloud servers according to the calculation force of a single cloud server;
Sequencing the number of cloud servers corresponding to each power demand data in the power demand data sequence according to the sequence of the power demand data sequence to obtain a cloud server number sequence;
on each cloud server distribution area, taking the computing power demand data sequence as an input item of the neural network, and taking the cloud server quantity sequence as an output item of the neural network;
on each cloud server distribution area, learning and training a calculation force demand data sequence and a cloud server number sequence by using a neural network to obtain a cloud server static distribution model;
the expression of the cloud server static distribution model is as follows:
Severs_numk=BP(computility_datak);
Wherein Severs _num k is a cloud server number sequence of a kth cloud server distribution area, computility _data k is a computing power demand data sequence of the kth cloud server distribution area, and BP is a neural network;
The construction method of the cloud server dynamic distribution model comprises the following steps:
acquiring a loss function of a cloud server static distribution model of each cloud server distribution area;
weighting and combining the loss functions of the cloud server static distribution models of all cloud server distribution areas to obtain combined loss functions;
synchronously training cloud server static distribution models of all cloud server distribution areas by using a combined type loss function to obtain cloud server dynamic distribution models of the cloud server distribution areas;
The model expression of the cloud server dynamic distribution model is as follows: ;
Wherein Severs _num 1、Severs_num2 and Severs _num N are cloud server number sequences of 1 st, 2 nd and N th cloud server distribution areas, computility _data 1、computility_data2 and computility _data N are calculation force demand data sequences of 1 st, 2 nd and N th cloud server distribution areas, and BP is a neural network;
the construction method of the combined type loss function comprises the following steps:
acquiring the accuracy of a static distribution model of a cloud server and the variability of a calculation power demand data sequence of a distribution area of the cloud server;
Determining the weighting weight of each cloud server static distribution model based on the accuracy of the cloud server static distribution model and the variability of the calculation force demand data sequence;
And weighting and combining the loss functions of the cloud server static distribution models of the cloud server distribution areas by using the weighting weights of the static distribution of each cloud server to obtain a combined loss function, wherein the expression of the combined loss function is as follows: ; wherein Loss is a combined type Loss function, loss k is a Loss function of a k cloud server static distribution model, W k is a weighting weight of the k cloud server static distribution model, and N is the total number of the cloud server static distribution models; /(I) ; Wherein W k is the weighted weight of the static distribution model of the kth cloud server, H k is the variability of the calculation power demand data sequence of the distribution area of the kth cloud server, and Z k is the accuracy of the static distribution model of the kth cloud server.
2. The adaptive cloud server distribution method according to claim 1, wherein: and the accuracy of the cloud server static distribution model is the ratio of the correct number of model predictions to the total number of model predictions.
3. The adaptive cloud server distribution method according to claim 2, wherein: the change rate of the calculation force demand data sequence of the cloud server distribution area is quantified through the discrete rate of the calculation force demand data sequence, and the quantification formula of the change rate is as follows:；
wherein H k is the variability of the power demand data sequence of the kth cloud server distribution area, q kt is the power demand data at time t in the power demand data sequence of the kth cloud server distribution area, t0 is the initial time sequence for acquiring the power demand, and tn is the final time sequence for acquiring the power demand.
4. The adaptive cloud server distribution method according to claim 3, wherein: the cloud server self-adaptive allocation method for each cloud server distribution area by utilizing the cloud server dynamic distribution model comprises the following steps:
Obtaining the number of cloud servers in the current time period of each cloud server distribution area according to the calculation power demand data sequence in the current time period by utilizing a cloud server dynamic distribution model;
Comparing the number of cloud servers in the current time period of each cloud server distribution area with the number of cloud servers in the previous time period, wherein,
When the number of cloud servers in the current time period is smaller than that of cloud servers in the previous time period, taking the cloud server distribution area as a server movable area;
when the number of cloud servers in the current time period is larger than that of cloud servers in the previous time period, taking the cloud server distribution area as a server migratable area;
and (5) adaptively allocating the mobile areas of the server according to the mobile areas of the server.
5. An adaptive cloud server distribution apparatus, comprising:
the data monitoring unit is used for acquiring the calculation power demand data of each cloud server distribution area in a period of time to obtain a calculation power demand data sequence of each cloud server distribution area;
the model training unit is used for learning the association relation between the computing power demand data sequence and the cloud server distribution quantity by using the neural network at each cloud server distribution area to obtain a cloud server static distribution model for measuring the cloud server distribution quantity according to the computing power demand at each cloud server distribution area; and
The cloud server dynamic distribution model is used for synchronously combining the cloud server static distribution models at all cloud server distribution areas to obtain cloud server dynamic distribution models at all cloud server distribution areas for adaptively allocating cloud server distribution according to calculation power requirements;
The self-adaptive allocation unit is used for carrying out self-adaptive allocation of cloud servers in each cloud server distribution area by utilizing the cloud server dynamic distribution model;
The power demand data sequence is a data sequence taking time sequence as an attribute, and the expression of the power demand data sequence is as follows: { q t |t ε [ t0, tn ] }, wherein q t is the calculation force demand data at time t, t0 is the initial time sequence for acquiring the calculation force demand, and tn is the end time sequence for acquiring the calculation force demand;
The method for constructing the cloud server static distribution model comprises the following steps:
each calculation force demand data in the calculation force demand data sequence of each cloud server distribution area is correspondingly converted into the number of cloud servers according to the calculation force of a single cloud server;
Sequencing the number of cloud servers corresponding to each power demand data in the power demand data sequence according to the sequence of the power demand data sequence to obtain a cloud server number sequence;
on each cloud server distribution area, taking the computing power demand data sequence as an input item of the neural network, and taking the cloud server quantity sequence as an output item of the neural network;
on each cloud server distribution area, learning and training a calculation force demand data sequence and a cloud server number sequence by using a neural network to obtain a cloud server static distribution model;
the expression of the cloud server static distribution model is as follows:
Severs_numk=BP(computility_datak);
Wherein Severs _num k is a cloud server number sequence of a kth cloud server distribution area, computility _data k is a computing power demand data sequence of the kth cloud server distribution area, and BP is a neural network;
The construction method of the cloud server dynamic distribution model comprises the following steps:
acquiring a loss function of a cloud server static distribution model of each cloud server distribution area;
weighting and combining the loss functions of the cloud server static distribution models of all cloud server distribution areas to obtain combined loss functions;
synchronously training cloud server static distribution models of all cloud server distribution areas by using a combined type loss function to obtain cloud server dynamic distribution models of the cloud server distribution areas;
The model expression of the cloud server dynamic distribution model is as follows: ;
Wherein Severs _num 1、Severs_num2 and Severs _num N are cloud server number sequences of 1 st, 2 nd and N th cloud server distribution areas, computility _data 1、computility_data2 and computility _data N are calculation force demand data sequences of 1 st, 2 nd and N th cloud server distribution areas, and BP is a neural network;
the construction method of the combined type loss function comprises the following steps:
acquiring the accuracy of a static distribution model of a cloud server and the variability of a calculation power demand data sequence of a distribution area of the cloud server;
Determining the weighting weight of each cloud server static distribution model based on the accuracy of the cloud server static distribution model and the variability of the calculation force demand data sequence;
And weighting and combining the loss functions of the cloud server static distribution models of the cloud server distribution areas by using the weighting weights of the static distribution of each cloud server to obtain a combined loss function, wherein the expression of the combined loss function is as follows: ；
Wherein Loss is a combined type Loss function, loss k is a Loss function of a k cloud server static distribution model, W k is a weighting weight of the k cloud server static distribution model, and N is the total number of the cloud server static distribution models; ; wherein W k is the weighted weight of the static distribution model of the kth cloud server, H k is the variability of the calculation power demand data sequence of the distribution area of the kth cloud server, and Z k is the accuracy of the static distribution model of the kth cloud server.
6. A computer device, comprising: at least one processor; and
A memory communicatively coupled to the at least one processor;
Wherein the memory stores instructions executable by the at least one processor to cause a computer device to perform the method of any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410346154.XA CN117950879B (en) | 2024-03-26 | 2024-03-26 | Self-adaptive cloud server distribution method and device and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410346154.XA CN117950879B (en) | 2024-03-26 | 2024-03-26 | Self-adaptive cloud server distribution method and device and computer equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117950879A CN117950879A (en) | 2024-04-30 |
| CN117950879B true CN117950879B (en) | 2024-06-07 |
Family
ID=90805494
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410346154.XA Active CN117950879B (en) | 2024-03-26 | 2024-03-26 | Self-adaptive cloud server distribution method and device and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117950879B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113435472A (en) * | 2021-05-24 | 2021-09-24 | 西安电子科技大学 | Vehicle-mounted computing power network user demand prediction method, system, device and medium |
| WO2022134965A1 (en) * | 2020-12-25 | 2022-06-30 | 华为技术有限公司 | Configuration method and device for computing resources |
| CN115454654A (en) * | 2022-11-11 | 2022-12-09 | 中诚华隆计算机技术有限公司 | Adaptive resource matching obtaining method and device |
| CN117435335A (en) * | 2023-08-01 | 2024-01-23 | 中国工商银行股份有限公司 | Computing power dispatching method, computing power dispatching device, computer equipment and storage medium |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6207193B2 (en) * | 2013-03-26 | 2017-10-04 | 株式会社日立システムズ | Server number adjusting system, method and program |
| KR101913346B1 (en) * | 2017-07-21 | 2018-10-30 | 주식회사 티맥스 소프트 | Method and apparatus for managing cloud server in cloud environment |
| CN111897399A (en) * | 2019-05-05 | 2020-11-06 | 西安光启未来技术研究院 | Embedded server capable of expanding computational power |
-
2024
- 2024-03-26 CN CN202410346154.XA patent/CN117950879B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022134965A1 (en) * | 2020-12-25 | 2022-06-30 | 华为技术有限公司 | Configuration method and device for computing resources |
| CN114691346A (en) * | 2020-12-25 | 2022-07-01 | 华为技术有限公司 | Configuration method and equipment of computing power resources |
| CN113435472A (en) * | 2021-05-24 | 2021-09-24 | 西安电子科技大学 | Vehicle-mounted computing power network user demand prediction method, system, device and medium |
| CN115454654A (en) * | 2022-11-11 | 2022-12-09 | 中诚华隆计算机技术有限公司 | Adaptive resource matching obtaining method and device |
| CN117435335A (en) * | 2023-08-01 | 2024-01-23 | 中国工商银行股份有限公司 | Computing power dispatching method, computing power dispatching device, computer equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 大规模云计算服务器优化调度问题的最优二元交换算法研究;王万良 等;《通信学报》;20190531;第40卷(第5期);第180-191页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117950879A (en) | 2024-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113705610B (en) | Heterogeneous model aggregation method and system based on federal learning | |
| WO2021057245A1 (en) | Bandwidth prediction method and apparatus, electronic device and storage medium | |
| CN111708640A (en) | Edge calculation-oriented federal learning method and system | |
| CN111629380B (en) | Dynamic resource allocation method for high concurrency multi-service industrial 5G network | |
| CN111726303B (en) | Flow control method and device and computing equipment | |
| CN106453608B (en) | A kind of background request adaptive scheduling algorithm of the mobile application based on cloud | |
| CN110889509A (en) | Joint learning method and device based on gradient momentum acceleration | |
| CN112818025B (en) | Test question generation method, device and system, computer storage medium and program product | |
| CN110502323B (en) | Real-time scheduling method for cloud computing tasks | |
| CN113778691B (en) | Task migration decision method, device and system | |
| CN113206887A (en) | Method for accelerating federal learning aiming at data and equipment isomerism under edge calculation | |
| CN114895773B (en) | Energy consumption optimization method, system and device for heterogeneous multi-core processor and storage medium | |
| CN111507474B (en) | Neural network distributed training method for dynamically adjusting Batch-size | |
| CN117236421B (en) | Large model training method based on federal knowledge distillation | |
| CN109389424A (en) | Flow allocation method, device, electronic equipment and storage medium | |
| CN114760308B (en) | Edge calculation unloading method and device | |
| CN116781343A (en) | Terminal credibility assessment method, device, system, equipment and medium | |
| CN117950879B (en) | Self-adaptive cloud server distribution method and device and computer equipment | |
| CN116235529A (en) | Method for implementing an ad hoc network of a plurality of access network devices and electronic device for implementing the method | |
| CN113543160A (en) | 5G slice resource allocation method and device, computing equipment and computer storage medium | |
| CN115345306A (en) | Deep neural network scheduling method and scheduler | |
| CN114912627A (en) | Recommendation model training method, system, computer device and storage medium | |
| CN111092755B (en) | Edge service migration simulation method based on resource occupation | |
| CN110928683B (en) | Edge computing resource allocation method based on two types of intensive virtual machines | |
| CN113762972A (en) | Data storage control method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |