CN108845886A - Cloud computing energy consumption optimization method and system based on phase space - Google Patents
Cloud computing energy consumption optimization method and system based on phase space Download PDFInfo
- Publication number
- CN108845886A CN108845886A CN201810845198.1A CN201810845198A CN108845886A CN 108845886 A CN108845886 A CN 108845886A CN 201810845198 A CN201810845198 A CN 201810845198A CN 108845886 A CN108845886 A CN 108845886A
- Authority
- CN
- China
- Prior art keywords
- energy consumption
- phase space
- cloud computing
- task
- node
- 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.)
- Granted
Links
- 238000005265 energy consumption Methods 0.000 title claims abstract description 337
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000005457 optimization Methods 0.000 title claims abstract description 27
- 230000003068 static effect Effects 0.000 claims abstract description 53
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 39
- 239000013598 vector Substances 0.000 claims description 40
- 230000005484 gravity Effects 0.000 claims description 28
- 238000013507 mapping Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 abstract description 7
- 238000004088 simulation Methods 0.000 description 11
- 230000008859 change Effects 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000007613 environmental effect Effects 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 238000010168 coupling process Methods 0.000 description 4
- 238000005859 coupling reaction Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000013508 migration Methods 0.000 description 2
- 230000005012 migration Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000013079 data visualisation Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 239000005431 greenhouse gas Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000000302 molecular modelling Methods 0.000 description 1
- 230000037361 pathway Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
- G06F9/5044—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering hardware capabilities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/26—Power supply means, e.g. regulation thereof
- G06F1/32—Means for saving power
- G06F1/3203—Power management, i.e. event-based initiation of a power-saving mode
- G06F1/3234—Power saving characterised by the action undertaken
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/36—Circuit design at the analogue level
- G06F30/367—Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to field of cloud calculation, disclose a kind of cloud computing energy consumption optimization method and system based on phase space, to optimize the energy consumption in cloud computing, reduce task and distribute the time, improve task allocative efficiency；The method of the present invention includes the context informations for obtaining node and node all in cloud computing system, and establish energy consumption model；The total energy consumption of all being carrying out for tasks of measuring node establishes static energy consumption phase space according to the total energy consumption；The energy consumption of newly-increased required by task is predicted using energy consumption model, and establishes dynamic energy consumption phase space；Optimal phase subspace is calculated according to static energy consumption phase space；The energy consumption for predicting newly-increased task each node in optimal phase subspace according to static energy consumption phase space and dynamic energy consumption phase space design objective dispatching algorithm, and determines the optimal scheduling scheme of newly-increased task.
Description
Technical Field
The invention relates to the field of cloud computing, in particular to a cloud computing energy consumption optimization method and system based on a phase space.
Background
With the adoption of the cloud computing, enterprises can acquire a large amount of computing and storage resources on line, payment is only needed according to the time length of using the resources, and the investment on expensive IT infrastructure is avoided. With the rapid development of cloud computing technology, the number of global cloud computing data centers is increasing continuously, and the proportion of energy consumed by the data centers each year accounts for the energy consumption of the whole world, which increases year by year. Cloud computing data centers consume a great amount of energy every year and also exert a certain pressure on the environment (such as greenhouse gases generated by the data centers), and the problem of cloud computing energy consumption has gradually become a focus of attention of researchers. The cloud computing data center is a whole body of coordination work, and a large number of high-coupling operations such as computing migration, storage migration and the like exist in the cloud computing system, so that the cloud computing system is a high-coupling system with a large number of nodes. To study the cloud computing energy consumption problem on a system level, all nodes in the cloud computing system need to be treated and analyzed as a whole.
Currently, energy consumption optimization management of cloud computing is one of important contents of energy consumption related research, and research is mainly carried out around several main flow directions of a dynamic voltage frequency adjustment technology, a sleep shutdown technology and a virtualization technology. The accurate energy consumption measurement and prediction model is the basis of energy consumption optimization management, and the existing energy consumption evaluation model comprises an energy consumption model of the utilization rate of main components of the system and an energy consumption prediction model based on a performance counter. For an energy consumption evaluation method based on system utilization rate, early research is mostly based on an energy consumption model of processor utilization rate, a CPU is considered to be a component with the highest energy consumption proportion in a working node of a cloud server (cloud computing system), and the utilization rate of the CPU is increased due to the fact that the occupancy rates of other components are increased. Therefore, it is feasible to evaluate the system load by the CPU utilization and further estimate the system energy consumption. Some studies estimate server node energy consumption by focusing on the frequency and usage of the CPU and setting the energy consumption state of other components of the server to a constant. An energy consumption model based on a performance counter (PMC) is mainly characterized in that through a performance monitoring system provided by each large hardware manufacturer, such as monitoring of instructions, cache and page table cache in a processor, important influence factors are screened out, and the method for establishing the model is extremely low in measurement overhead and relatively accurate in result, can acquire real data of a bottom layer and can also evaluate energy consumption on the whole, but the method is relatively high in deployment difficulty. The methods have respective application scenes and can provide more accurate energy consumption prediction and measurement. However, the energy consumption model is more concerned about the energy consumption state of a single server node, and the energy consumption estimation of the high-performance computing center is only to simply accumulate the energy consumption of the server nodes. Therefore, the method is more suitable for the condition that the energy consumption state coupling between the servers is not obvious, and is not suitable for a high-coupling cloud computing system.
Disclosure of Invention
The invention aims to provide a cloud computing energy consumption optimization method and system based on a phase space, so as to optimize energy consumption in cloud computing, reduce task allocation time and improve task allocation efficiency.
In order to achieve the above object, the present invention provides a phase space-based cloud computing energy consumption optimization method, which includes the following steps:
s1: acquiring all nodes in a cloud computing system and context environment information of the nodes, and establishing an energy consumption model;
s2: measuring the total energy consumption of all tasks being executed by the nodes, and establishing a static energy consumption phase space according to the total energy consumption;
s3: predicting energy consumption required by a newly added task by adopting the energy consumption model, and establishing a dynamic energy consumption phase space;
s4: calculating according to the static energy consumption phase space to obtain an optimal sub-phase space;
s5: and predicting the energy consumption of each node of the newly added task in the optimal sub-phase space, designing a task scheduling algorithm according to the static energy consumption phase space and the dynamic energy consumption phase space, and determining an optimal task allocation scheme of the newly added task.
Preferably, in S2, the establishing the static energy consumption phase space specifically includes the following steps:
s21: acquiring all vectors formed by energy consumption parameters of all current nodes executing tasks in the cloud computing system;
s22: mapping all the vectors into a phase space, regarding the number of the energy consumption parameters as the dimensionality of the phase space, regarding the modular sum of all the vectors as the current total load energy consumption of the cloud computing system, and establishing a static energy consumption phase space according to the total load energy consumption.
Preferably, in S3, the establishing the dynamic energy consumption phase space specifically includes the following steps:
s31: predicting all energy consumption increasing vectors formed by energy consumption parameters of newly added tasks in the cloud computing system by adopting an energy consumption model;
s32: and mapping the energy consumption increasing vectors into a phase space, regarding the parameter number of the energy consumption of the newly added task as the dimensionality of the phase space, regarding the modular sum of all the energy consumption increasing vectors as newly added total energy consumption, and establishing a dynamic energy consumption phase space according to the newly added total energy consumption.
Preferably, the S4 specifically includes the following steps:
s41: establishing a coordinate axis reflecting the energy consumption state of the node in a static energy consumption phase space, and determining a projection point set of the static energy consumption phase space;
s42: in the static energy consumption phase space, making a vertical line to the coordinate axis through the gravity center of the projection point set to obtain a corresponding number of sub-item spaces;
s43: and if the coordinate value of any point in a certain sub-item space on each coordinate axis is smaller than the coordinate value of the center, the sub-item space is regarded as the optimal sub-item space.
Preferably, in S5, the determining the optimal task allocation scheme for the newly added task specifically includes the following steps:
s51: predicting the energy consumption of the newly added task on each node in the optimal subentry space by adopting the energy consumption model, and obtaining all possible allocation schemes;
s52: and mapping all the possible allocation schemes to the dynamic energy consumption phase space by adopting a task scheduling algorithm, and regarding a scheme which enables the phase space gravity center to have the minimum moving distance compared with the original point as the optimal task allocation scheme of the newly added task.
Preferably, in S1, the context information includes a node type, a network condition, and a cluster location.
Preferably, the vector includes CPU power consumption, memory power consumption, and disk power consumption.
Preferably, the increment vector includes a CPU energy consumption increment value, a memory energy consumption increment value, and a disk energy consumption increment value.
As a general technical concept, the present invention also provides a phase space-based cloud computing energy consumption optimization system, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method when executing the computer program.
The invention has the following beneficial effects:
the invention provides a cloud computing energy consumption optimization method and system based on a phase space, which comprises the steps of firstly, obtaining all nodes in a cloud computing system and context environment information of the nodes, and establishing an energy consumption model; measuring the total energy consumption of all tasks being executed by the nodes, and establishing a static energy consumption phase space according to the total energy consumption; then, predicting energy consumption required by a newly added task by adopting an energy consumption model, and establishing a dynamic energy consumption phase space; calculating according to the static energy consumption phase space to obtain an optimal sub-phase space and predicting the energy consumption of each node of the newly added task in the optimal sub-phase space, designing a task scheduling algorithm according to the static energy consumption phase space and the dynamic energy consumption phase space, and determining an optimal task allocation scheme of the newly added task; context environmental factors among nodes in the cloud computing system are considered, so that the energy consumption can be predicted more accurately, the energy consumption in the cloud computing is optimized, the task allocation time is shortened, and the task allocation efficiency is improved.
The present invention will be described in further detail below with reference to the accompanying drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:
FIG. 1 is a flow chart of a phase space-based cloud computing energy consumption optimization method according to a preferred embodiment of the invention;
FIG. 2 is an effect diagram of a cloud computing static energy consumption phase space according to a preferred embodiment of the present invention;
FIG. 3 is an effect diagram of the cloud computing dynamic energy consumption phase space according to the preferred embodiment of the present invention;
FIG. 4 is a diagram of the threshold setting of the optimal subentry space for the preferred embodiment of the present invention;
FIG. 5 is a distribution characteristic of projection points in a dynamic energy consumption phase space of four scheduling algorithms according to a preferred embodiment of the present invention;
FIG. 6 is a diagram illustrating the performance comparison of four algorithms based on phase space according to the preferred embodiment of the present invention;
FIG. 7 is a diagram illustrating the comparison of the gravity centers of the projected points in the static phase space with the distances from the origin as a result of the four algorithms according to the preferred embodiment of the present invention.
Detailed Description
The embodiments of the invention will be described in detail below with reference to the drawings, but the invention can be implemented in many different ways as defined and covered by the claims.
Example 1
Referring to fig. 1, the embodiment provides a cloud computing energy consumption optimization method based on a phase space, including the following steps:
s1: acquiring all nodes in a cloud computing system and context environment information of the nodes, and establishing an energy consumption model;
s2: measuring the total energy consumption of all tasks being executed by the nodes, and establishing a static energy consumption phase space according to the total energy consumption;
s3: predicting energy consumption required by a newly added task by adopting an energy consumption model, and establishing a dynamic energy consumption phase space;
s4: calculating according to the static energy consumption phase space to obtain an optimal sub-phase space;
s5: and predicting the energy consumption of each node of the newly added task in the optimal sub-phase space, designing a task scheduling algorithm according to the static energy consumption phase space and the dynamic energy consumption phase space, and determining the optimal task allocation scheme of the newly added task.
In the steps, context environmental factors among nodes in the cloud computing system are considered, a static energy consumption phase space and a dynamic energy consumption phase space are established, so that the energy consumption can be predicted more accurately, the energy consumption in the cloud computing is optimized, the task allocation time is shortened, and the task allocation efficiency is improved.
In practical application, the embodiment of the present invention may further add the following steps to optimize:
s1: acquiring all nodes in the cloud computing system and context environment information of the nodes, and establishing an energy consumption model.
Specifically, energy consumption optimization of the cloud computing system can be realized by optimizing the task allocation scheme, and therefore, energy consumption of a newly added task executed on a node that is likely to receive the task needs to be predicted. The target cloud computing system comprises M nodes, each node comprises N components, CiDenotes the ith node, CijThe jth component of the ith node is represented. Recording the component actual time overhead as Respectively representing the actual time overhead of the current task on the processor, the memory and the disk. The historical average power consumption of the node components is adopted to calculate the energy consumption of each component of the node, and the calculation is as follows:
wherein, PijRepresenting a component CijCalendar ofThe average power consumption of the power is used,representing tasks in component CijThe actual time overhead.
The node energy consumption is determined by actual time overhead and power, and the calculation formula is as follows:
wherein,is node CiThe actual energy consumption of the node is obtained by accumulating the energy consumption values of the N components of the node,is node CiThe j component CijThe actual energy consumption.
Because the ideal time overhead of the task is obtained after the idle waiting condition is eliminated as much as possible, and the actual time overhead of the task is larger than the ideal time overhead, the energy consumption obtained by adopting the ideal time overhead calculation is smaller than the energy consumption obtained by adopting the actual execution time overhead. Recording the ideal time overhead of the current task on each component Respectively representing the ideal time overhead of the current task on the processor, the memory and the disk.
In a large-scale cloud cluster, the overall energy consumption of the system is closely related to the context environment in which the cluster nodes are located, and the context environment comprises cluster positions, network conditions, system architecture, node types and the like. For example, two large-scale clusters with identical component configurations may cause a certain difference in energy consumption when system task requests are consistent due to different context environments. The context environment has consistency influence on cluster energy consumption within a certain time, namely, for a certain node, the context environment parameter reflects the overall state of the system and keeps unchanged within a certain time range.
It should be noted that, when energy consumption in the cloud computing system is optimized, the context environment where the node is located is fully considered, so that the computing result is more accurate. For a node, its current context environment parameter is θ ═ Tm/Tr. Further, using a weighted average method in time series analysis to predict the context environment parameters of the current node according to the previous K batches of historical measured time spending of the node, the calculation is as follows:
wherein, thetaK+1Representing the current estimated value of a contextual environmental parameter, thetaiRepresenting environmental parameters of the ith set of historical measured data from far to near, wiThe weight of the ith group of historical measured data from far to near is represented, the historical data closer to the current time is heavier,
in the case where the ideal time overhead for the current task request is known, some means is needed to get the actual time overhead for the task. In combination with the context parameters described above, the actual time overhead of the computing task is as follows:
representing tasks in component CijOf the ideal time overhead, thetaiThe new incoming task is executed on the node and is subject to the environment parameter characteristics of the node context. Thus, the energy consumption model can be established as:
s2: and measuring the total energy consumption of all the tasks being executed by the nodes, and establishing a static energy consumption phase space according to the total energy consumption. The method specifically comprises the following steps:
s21: acquiring all vectors formed by energy consumption parameters of tasks being executed by all current nodes in the cloud computing system;
s22: mapping all vectors into a phase space, regarding the number of energy consumption parameters as the dimension of the phase space, regarding the modulo sum of all vectors as the current total load energy consumption of the cloud computing system, and establishing a static energy consumption phase space according to the total load energy consumption.
Specifically, the cloud computing system currently comprises N real-time energy consumptions, a vector formed by the current N real-time energy consumption parameters of the cloud computing system is mapped into a phase space, the vector at the position comprises CPU energy consumption, memory energy consumption, disk energy consumption and the like, the dimensionality of the phase space is regarded as the number of the real-time energy consumption parameters, and a model of the vector represents the current total load energy consumption of the node, so that the phase space is called as a cloud computing static energy consumption phase space.
In order to facilitate understanding of the relationship between the nodes and the vectors in the cloud computing system, it should be noted that, assuming that the cloud computing system includes n nodes, n vectors should be formed, a module of the ith vector represents the total energy consumption of the ith node, and a sum of the modules of all the vectors represents the current total load energy consumption of the cloud computing system. The following dynamic energy consumption phase space situation is similar, where the sum of the modes of all energy consumption increase vectors represents the total energy consumption increased by the system, and the mode of a single vector represents the energy consumption increase value of a single node only.
The establishment of the static energy consumption phase space of the cloud computing system takes into account the total energy consumption of all executing tasks on the nodes of the cloud computing system. Node CiPosition coordinate A in static energy consumption phase spaceiIs represented as follows:
Ai＝(Xi1,Xi2,…Xij…XiN),1≤j≤N；
wherein,
in the formula, XijRepresenting a component CijThe above normalized value of the current total energy consumption,represents node CiUpper current total task load at component CijThe resulting energy consumption value. Lambda [ alpha ]iRepresents node CiEnergy consumption as a percentage of the total energy consumption of the cloud cluster, i.e. Is represented by CiHistorical average energy consumption of;representing the historical average energy consumption of the cloud data center. w is aijRepresenting a component CijEnergy consumption accounts for node CiPercentage of total energy consumption, i.e. Is represented by CijHistorical average energy consumption.
Note that, the node C is connected toiAfter the energy consumption under the current load is normalized into a generalized coordinate in a phase space, a unique projection point A is determined in the phase spaceiTherefore, the current energy consumption state of the nodes in the cloud computing system can be projected into the energy consumption phase space, and the position of the midpoint in the static energy consumption phase space represents the overall energy consumption state of the nodes. The effect of the cloud computing static energy consumption phase space is shown in fig. 2, and it should be noted that in this embodiment, only the 3-dimensional phase space of the CPU, the memory, and the disk is considered, and the multidimensional situation is similar to this, and is not described herein again.
The weight factors of nodes and components are considered when mapping from the node energy consumption state to the energy consumption phase space in the cloud computing system is established, and the gravity center of a projection point set of the energy consumption state in the energy consumption phase space is G (X)1,X2,…Xj…XN) Its position in phase space represents the current overall energy consumption state of the cloud computing system, where XjThe calculation is as follows:
in the formula, j is more than or equal to 1 and less than or equal to N.
S3: and predicting energy consumption required by the newly added task by adopting an energy consumption model, and establishing a dynamic energy consumption phase space. The method specifically comprises the following steps:
s31: predicting all energy consumption increasing vectors formed by energy consumption parameters of newly added tasks in the cloud computing system by adopting an energy consumption model;
s32: and mapping the energy consumption increasing vectors into a phase space, regarding the parameter number of the energy consumption of the newly added task as the dimensionality of the phase space, regarding the modular sum of all the energy consumption increasing vectors as newly added total energy consumption, and establishing a dynamic energy consumption phase space according to the newly added total energy consumption.
Then, an energy consumption increase vector composed of N energy consumption parameters under a newly added task in the cloud computing system is obtained, and the increase vector is mapped into a phase space, where the increase vector includes a CPU energy consumption increase value, a memory energy consumption increase value, a disk energy consumption increase value, and the like, the phase space dimension is regarded as the number of parameters, and when the vector is mapped into the phase space, the dimension of the energy consumption increase vector is the dimension of the phase space, and a modulus of the vector in the phase space represents the energy consumption change degree of a node, so that the phase space is called a cloud computing dynamic energy consumption phase space, as shown in fig. 3.
It is worth pointing out that the establishment of the dynamic energy consumption phase space of the cloud computing system considers the situation of energy consumption increase generated by the node receiving task. Node CiThe normalized value of the energy consumption increase value due to the reception of the task is taken as CiGeneralized coordinates of projected points in dynamic energy consumption phase space, node CiThe generalized coordinate of the energy consumption change value in the dynamic energy consumption phase space is as follows:
A′i＝(X′i1,X′i2…X′ij…X′iN),1≤j≤N；
wherein, X'ijRepresenting a component CijThe normalized value of the current energy consumption change above,represents node CiComponent C after receiving task loadijEnergy consumption variation value of.
Similarly, the gravity center of the projection point set in the dynamic energy consumption phase space is G '(X'1,X′1,…X′j…X′N) Its position in phase space represents an increase in energy consumption after the cloud computing system receives the task, where X'jThe calculation is as follows:
in the formula, j is more than or equal to 1 and less than or equal to N.
In this embodiment, the static energy consumption phase space and the dynamic energy consumption phase space are respectively established by using the above method, and the overall energy consumption state in the cloud computing system is depicted by the phase space, so that the method is convenient to use, and can assist in realizing accurate analysis of the energy consumption state in the cloud computing system.
As a person skilled in the art will understand, in the static energy consumption phase space, when the task on the node is executed and the resource is released, the distance from the projection point to the origin point decreases when the projection point moves towards the origin point of the phase space, that is, the total energy consumption of the current task load on the cloud computing system decreases; when the nodes in the cloud computing system receive tasks, the energy consumption generated by the nodes is increased, the projection point moves towards the direction far away from the phase space origin, namely the distance between the projection point and the origin is increased, and the total energy consumption generated by the current load of the cloud cluster is increased. Similarly, the distance between the projection point and the origin point in the dynamic energy consumption phase space corresponds to the added value of energy consumption after the node receives the load, and the larger the distance between the projection point and the origin point is, the larger the energy consumption generated by the task load received by the node on the node is.
It should be noted that the state change at the midpoint in the static or dynamic energy consumption phase space is not related to the motion process, but only to the external load request. And in thermodynamic systems, the variables of the state properties are only related to the permanent state and not to the pathway. The energy consumption state of the cloud computing system is put into a thermodynamic system research model, the state change characteristics are consistent with those of the thermodynamic system, so that the state change of a midpoint in a phase space is irrelevant to the process and only relevant to an external request, so that when the energy consumption state of the cloud computing system changes, a receiving load and a releasing load always exist at the same time, if the energy consumption state of the load transmitted outside the system is increased to D, the energy consumption state of the system is changed from S1 to S2, and meanwhile, the energy consumption state of the load released by the system is reduced to R, D is S2-S1+ R.
When a node in the cloud computing system releases or increases load, the location of the projected point in the static energy consumption phase space changes as follows:
assuming that the original projection point coordinates of the nodes of the cloud computing system in the energy consumption phase space are C _ node (x, y, z), the load added by the nodes is represented by a three-dimensional vector T _ add ═ a, b, C, and the coordinates of the projection points in the static energy consumption phase space after receiving the load are:
in the formula,and respectively representing generalized coordinate values of each dimensionality after the node energy consumption change value receiving the task is mapped to the dynamic energy consumption phase space.
S4: and calculating to obtain the optimal sub-phase space according to the static energy consumption phase space. The method specifically comprises the following steps:
s41: establishing a coordinate axis of the energy consumption state of the reaction node in the static energy consumption phase space, and determining a projection point set of the static energy consumption phase space;
s42: in a static energy consumption phase space, making a vertical line from the gravity center of the over-projection point set to a coordinate axis to obtain a corresponding number of sub-item spaces;
s43: and if the coordinate value of any point in a certain sub-item space on each coordinate axis is smaller than the coordinate value of the center, the sub-item space is regarded as the optimal sub-item space.
In practical application, in the static energy consumption phase space, the gravity center G of the over-projection point set forms vertical lines towards N coordinate axes, and the vertical lines decompose the N-dimensional phase space into 2NA subphase space. When the coordinate value of any point in the sub-phase space on each coordinate axis is smaller than the coordinate value on the coordinate axis of the gravity center G, the sub-phase space is called as the optimal sub-phase space.
The nodes in the optimal sub-phase space are roughly divided into two types, one type is that the energy efficiency of the nodes is better, and the energy consumption of the nodes is lower than that of other nodes when the nodes execute tasks, so the nodes can be considered preferentially when the tasks are distributed; the second type of nodes are low in energy consumption due to low resource utilization rate, and research shows that the electric energy consumption of the cloud computing system sometimes exceeds the peak energy consumption of the nodes when the utilization rate is low, so that the utilization rate of the nodes needs to be improved on the premise of not closing the nodes, tasks are allocated to the nodes, the energy consumption can be obviously increased, and the energy consumption of the whole system when the tasks are executed is reduced. Under the purpose of energy consumption optimization, the overall energy efficiency of the nodes in the optimal sub-phase space after receiving the tasks is better than that of the nodes in other sub-phase spaces after receiving the tasks.
Node CiGeneralized coordinate A in static energy consumption phase spaceiAnd the following formula is satisfied to judge whether the projection point is in the optimal subphase space.
As shown in fig. 2, in the cloud computing three-dimensional static energy consumption phase space, the phase space can be divided into 8 sub-phase spaces P1 to P8 by making a perpendicular line to the coordinate axis through the center of gravity. Coordinate values of all dimensions of any point in the sub-phase space P1 are all smaller than coordinate values of the center of gravity, and according to the definition of the optimal sub-phase space, P1 in the cloud computing three-dimensional static energy consumption phase space of fig. 2 is the optimal sub-phase space. Preferably, to simplify the finding of the optimal subphase space, X may be made1….XNThe value of S is set according to the actual situation.
S5: and predicting the energy consumption of each node of the newly added task in the optimal sub-phase space, designing a task scheduling algorithm according to the static energy consumption phase space and the dynamic energy consumption phase space, and determining the optimal task allocation scheme of the newly added task. The method specifically comprises the following steps:
s51: predicting the energy consumption of the newly added task on each node in the optimal subentry space by adopting an energy consumption model, and obtaining all possible allocation schemes;
s52: and mapping all possible allocation schemes to a dynamic energy consumption phase space by adopting a task scheduling algorithm, and regarding a scheme which enables the phase space gravity center to have the minimum moving distance compared with the original point as an optimal task allocation scheme of the newly added task.
In practical application, the weight of the task scheduling algorithm is calculated firstly, most task allocation methods adopt the random task selection and then iteration to find the minimum total energy consumption, and therefore the complexity of the algorithm can be increased. In the embodiment, a scheduling priority rank value of the task is calculated, wherein the rank value is determined by an average value of parameters of the task in each projection point in a phase space. The task scheduling according to the non-decreasing order of the rank values can ensure that the task scheduling meets the task with lower energy consumption of priority scheduling, and then the optimal node is selected according to the scheduling order for scheduling. When the optimal sub-phase space is not adopted, the number of the nodes to be matched is M, so that the time complexity of the algorithm is O (N M), and when the nodes corresponding to the projection points in the optimal sub-phase space are preferentially adopted for matching, the number of the nodes to be matched is smaller than M, so that the time complexity is smaller than O (N M).
Therefore, when a newly added task arrives at the cloud computing system, the task scheduling algorithm rapidly allocates the task under the target condition based on the minimum energy consumption. The number of the newly added tasks may be single or batch, and it should be noted that the number of the newly added tasks does not affect the method of this embodiment, and therefore, the number of the newly added tasks is not specifically analyzed. And predicting the energy consumption of the task on each node in the optimal sub-phase space by the power consumption model, then mapping various possible allocation schemes to the dynamic energy consumption phase space by a task scheduling algorithm, and finding out the scheme which enables the phase space gravity center moving distance to be minimum to carry out the task allocation of the current round. Here, the phase space gravity center movement distance is the smallest, specifically, the gravity center movement distance is the smallest compared to the origin in the dynamic power consumption phase space.
And (3) experimental verification:
in an actual situation, the energy consumption executed at a node when a next task arrives needs to be predicted according to historical energy consumption data of the node, and parameters of the node need to be known, in the experiment, the parameters of the node are shown in table 1, and the average power of components is divided into 5 levels of 1-5 according to the size.
TABLE 1 node parameters
Since the present embodiment is directed to a mixed type of application, a variety of practical applications are selected to represent a variety of different types of applications, including weather forecasting, image processing, molecular modeling, data visualization, and mail services. All the applications used in this experiment are shown in table 2 below.
By executing the above five types of applications, the monitored data is used as a simulation data set of the system. The applications with mixed load types are generated by using the applications, and the requirement degree of the applications on the resources of each component is divided into 3 levels according to the characteristics of the applications: low, medium, high. In Table 2, the numbers 1, 2 and 3 are shown, respectively.
TABLE 2 multiple types of application parameters
In the simulation system, tasks and the cloud computing system are simulated according to the real tasks and the parameter conditions, when batch simulation tasks reach the simulation system, the simulation system generates a task allocation scheme with minimum total energy consumption after predicting the execution energy consumption of the tasks by using an energy consumption prediction model, and the task scheduling of the current round is performed according to the generated scheme. The task request is simulated by adopting a vector formed by time slices, wherein the time slices represent the unit time length of the node components occupied by the simulation task, and the cpu request time, the memory request time and the disk request time are a simulation task. The task is reduced by one unit every unit time slice after being distributed, and when the time slices are all 0, the task is released after being executed. The simulation parameters of the nodes are vectors formed by a plurality of parameters, such as (the environment parameter theta of the nodes, the rated power consumption P of the energy consumption components, the energy consumption weight lambda of the nodes and the weight omega of the components), and finally the performance of the phase space energy consumption optimization framework is evaluated according to the data obtained by statistics.
In the experiment, the physical platform for carrying out the simulation experiment is that the processor is Inter (R) Core (TM) i5-6500, 3.20 GHZ; 4GB RAM; windows 764 bit operating system. And each node distributes virtual CPU, memory and disk space parameters according to the real server parameters. Therefore, a heterogeneous simulation server cluster is formed, after the tasks are distributed to the cluster, the distributed energy consumption is calculated according to the data fed back by the virtual servers, and the energy consumption is compared with results of other distribution modes.
In order to analyze the effect of the optimal sub-phase space in the phase space scheduling algorithm, fig. 4 records the proportion of the number of nodes in the optimal sub-phase space selected in multiple rounds to the total number of nodes and the proportion of the distance from the center of gravity to the origin when the optimal sub-phase space is not used, which exceeds the proportion of using the optimal sub-phase space.
As can be seen from fig. 4, the ratio of the selected projection points in the optimal subphase space when the thresholds are set to 0.2, 0.4, 0.6 and 0.8 respectively to the ratio when the distance from the center of gravity to the origin exceeds the value when the optimal subphase space is not used. When the threshold value is 0.2, fewer nodes are selected, the distance difference is large, the difference between the exceeding distance ratio when the threshold value is selected to be between 0.4 and 0.6 and the difference when the threshold value is selected to be 0.8 is not large, so that the complexity can be reduced to the maximum extent by selecting the threshold value to be between 0.4 and 0.6, and excessive extra energy consumption is not increased.
Further, the task scheduling algorithm (PSA) provided by the invention is compared with the Min-Min, RR and CPSO algorithms commonly used at present to evaluate the scheduling result in a phase space. Fig. 5(a) - (d) show distribution characteristics of projection points of the four scheduling algorithms in the dynamic energy consumption phase space when the number of task requests per time slice is 10000 for 1000 nodes, and the coordinates of the projection points are the average value of the scheduling results of a plurality of time slices in a stable stage after the simulation starts. As can be seen from comparison of fig. 5(a) - (d), after RR algorithm scheduling, the node projection point is in a randomly divergent state in the dynamic energy consumption phase space, and the center of gravity is farthest from the origin. The reason is that the RR algorithm performs scheduling according to the arrival queue of the task and the preparation queue of the cloud computing system, has high randomness, cannot select the most suitable task execution node, and causes the projection points in the dynamic phase space to be in a relatively random distribution state during task allocation. The projection points of the nodes scheduled by the Min-Min algorithm in the dynamic energy consumption phase space are relatively gathered, and the gravity center of the projection points is closer to the origin than the gravity center scheduled by the RR algorithm, so that the scheduling performance of the node is superior to that of the RR algorithm. Compared with a Min-Min algorithm, the projection points after CPSO scheduling are more concentrated near the origin. However, compared with the PSA, the distances from the gravity centers to the origin after the scheduling of the three scheduling algorithms are all larger than the distances from the gravity centers to the origin after the scheduling of the PSA, and the performance of the PSA scheduling algorithm is superior to that of the other three scheduling algorithms according to the corresponding relation that the closer the gravity centers of the projection points in the dynamic phase space to the origin, the smaller the energy consumption increase of the cloud computing cluster. Although the projection points of a small number of nodes are scattered after the PSA scheduling, the density of the projection points gathered near the origin in the dynamic energy consumption phase space is the highest after the PSA scheduling, and the reason for generating the characteristic is that the PSA aims to perform energy consumption optimization scheduling on an overall level, and Min-Min and CPSO fall into a local optimal condition during scheduling.
By combining the analysis, the performance of the PSA is superior to that of other three scheduling algorithms under the task scheduling requirement of minimum overall energy consumption of cloud computing.
Fig. 6 shows the scheduling performance comparison of the scheduling results of each scheduling algorithm with respect to the distances from the center of gravity to the origin in the static energy consumption phase space, PSA, NPSA (the number of steps for selecting the optimal subphase space is less than PSA, the same applies otherwise), RR, Min-Min, and CPSO, when the number of unit time slice tasks is from 5000 to 30000 at 1000 nodes. The velocity of increase of the distance from the center of gravity to the origin is minimized near the NPSA in PSA and NPSA. Although the NPSA is slightly better than the PSA, the PSA algorithm takes into account the optimal subphase space, can reduce the number of nodes to be considered, can perform fast allocation of arriving tasks, and has negligible advantages over the PSA, so that the overall scheduling performance of the PSA is optimal.
FIG. 7 shows the comparison of the distances from the center of gravity to the origin of the projection points of the PSA, Min-Min, RR and CPSO scheduling algorithms when the number of task requests per unit time is 10000 and the number of nodes varies from 100 to 5000. With the increasing of the number of the nodes, the projection after the scheduling of each algorithm can be seen, and the distances from the gravity center to the origin of the shadow point in the phase space are all reduced. This is because as the number of nodes increases, more alternative nodes and thus more alternatives appear, resulting in an optimal overall power consumption state. Meanwhile, the scheduling effect of the CPSO algorithm is better than that of other algorithms when the number of the nodes is small, but the scheduling performance of the PSA is greatly improved along with the increase of the number of the nodes and gradually exceeds that of the other three scheduling algorithms. Therefore, it can be concluded that PSA has a general task scheduling effect for a small number of nodes, and has a more obvious advantage for the data center scheduling problem of a cloud computing system with a large number of nodes.
Example 2
Corresponding to the above method embodiments, the present embodiment provides a phase space-based cloud computing energy consumption optimization system, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above method when executing the computer program.
As described above, the invention provides a phase space-based cloud computing energy consumption optimization method and system, which includes the steps of firstly obtaining all nodes in a cloud computing system and context environment information of the nodes, and establishing an energy consumption model; measuring the total energy consumption of all tasks being executed by the nodes, and establishing a static energy consumption phase space according to the total energy consumption; then, predicting energy consumption required by a newly added task by adopting an energy consumption model, and establishing a dynamic energy consumption phase space; calculating according to the static energy consumption phase space to obtain an optimal sub-phase space and predicting the energy consumption of each node of the newly added task in the optimal sub-phase space, designing a task scheduling algorithm according to the static energy consumption phase space and the dynamic energy consumption phase space, and determining an optimal task allocation scheme of the newly added task; context environmental factors among nodes in the cloud computing system are considered, so that the energy consumption can be predicted more accurately, the energy consumption in the cloud computing is optimized, the task allocation time is shortened, and the task allocation efficiency is improved.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. A cloud computing energy consumption optimization method based on a phase space is characterized by comprising the following steps:
s1: acquiring all nodes in a cloud computing system and context environment information of the nodes, and establishing an energy consumption model;
s2: measuring the total energy consumption of all tasks being executed by the nodes, and establishing a static energy consumption phase space according to the total energy consumption;
s3: predicting energy consumption required by a newly added task by adopting the energy consumption model, and establishing a dynamic energy consumption phase space;
s4: calculating according to the static energy consumption phase space to obtain an optimal sub-phase space;
s5: and predicting the energy consumption of each node of the newly added task in the optimal sub-phase space, designing a task scheduling algorithm according to the static energy consumption phase space and the dynamic energy consumption phase space, and determining an optimal task allocation scheme of the newly added task.
2. The phase space-based cloud computing energy consumption optimization method according to claim 1, wherein in S2, the establishing a static energy consumption phase space specifically includes the following steps:
s21: acquiring all vectors formed by energy consumption parameters of all current nodes executing tasks in the cloud computing system;
s22: mapping all the vectors into a phase space, regarding the number of the energy consumption parameters as the dimensionality of the phase space, regarding the modular sum of all the vectors as the current total load energy consumption of the cloud computing system, and establishing a static energy consumption phase space according to the total load energy consumption.
3. The phase space-based cloud computing energy consumption optimization method according to claim 1, wherein in S3, the establishing a dynamic energy consumption phase space specifically includes the following steps:
s31: predicting all energy consumption increasing vectors formed by energy consumption parameters of newly added tasks in the cloud computing system by adopting an energy consumption model;
s32: and mapping the energy consumption increasing vectors into a phase space, regarding the parameter number of the energy consumption of the newly added task as the dimensionality of the phase space, regarding the modular sum of all the energy consumption increasing vectors as newly added total energy consumption, and establishing a dynamic energy consumption phase space according to the newly added total energy consumption.
4. The phase space-based cloud computing energy consumption optimization method according to claim 1, wherein the S4 specifically includes the following steps:
s41: establishing a coordinate axis reflecting the energy consumption state of the node in a static energy consumption phase space, and determining a projection point set of the static energy consumption phase space;
s42: in the static energy consumption phase space, making a vertical line to the coordinate axis through the gravity center of the projection point set to obtain a corresponding number of sub-item spaces;
s43: and if the coordinate value of any point in a certain sub-item space on each coordinate axis is smaller than the coordinate value of the center, the sub-item space is regarded as the optimal sub-item space.
5. The phase space-based cloud computing energy consumption optimization method according to claim 1, wherein in S5, the determining an optimal task allocation scheme for the newly added task specifically includes the following steps:
s51: predicting the energy consumption of the newly added task on each node in the optimal subentry space by adopting the energy consumption model, and obtaining all possible allocation schemes;
s52: and mapping all the possible allocation schemes to the dynamic energy consumption phase space by adopting a task scheduling algorithm, and regarding a scheme which enables the phase space gravity center to have the minimum moving distance compared with the original point as the optimal task allocation scheme of the newly added task.
6. The phase space-based cloud computing energy consumption optimization method according to claim 1, wherein in S1, the context information includes node type, network condition, and cluster location.
7. The phase space-based cloud computing energy consumption optimization method according to claim 2, wherein the vector includes CPU energy consumption, memory energy consumption and disk energy consumption.
8. The phase space-based cloud computing energy consumption optimization method according to claim 3, wherein the increase vector includes a CPU energy consumption increase value, a memory energy consumption increase value, and a disk energy consumption increase value.
9. A phase space based cloud computing energy consumption optimization system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method according to any one of claims 1 to 8 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810845198.1A CN108845886B (en) | 2018-07-27 | 2018-07-27 | Cloud computing energy consumption optimization method and system based on phase space |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810845198.1A CN108845886B (en) | 2018-07-27 | 2018-07-27 | Cloud computing energy consumption optimization method and system based on phase space |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108845886A true CN108845886A (en) | 2018-11-20 |
| CN108845886B CN108845886B (en) | 2022-03-08 |
Family
ID=64195686
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810845198.1A Active CN108845886B (en) | 2018-07-27 | 2018-07-27 | Cloud computing energy consumption optimization method and system based on phase space |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108845886B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109947551A (en) * | 2019-03-19 | 2019-06-28 | 中南大学 | A kind of more round method for allocating tasks, edge calculations system and its storage medium |
| CN110096349A (en) * | 2019-04-10 | 2019-08-06 | 山东科技大学 | A kind of job scheduling method based on the prediction of clustered node load condition |
| CN111338649A (en) * | 2020-02-14 | 2020-06-26 | 浪潮商用机器有限公司 | Heterogeneous system, acceleration method and device thereof and readable storage medium |
| CN113205695A (en) * | 2021-04-13 | 2021-08-03 | 东南大学 | Multi-period length bidirectional trunk line green wave control method |
| CN118094601A (en) * | 2024-03-25 | 2024-05-28 | 数盾信息科技股份有限公司 | High-speed data transmission and encryption method and device based on hardware acceleration |
| CN118094601B (en) * | 2024-03-25 | 2024-10-15 | 数盾信息科技股份有限公司 | High-speed data transmission and encryption method and device based on hardware acceleration |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103365727A (en) * | 2013-07-09 | 2013-10-23 | 南京大学 | Host load forecasting method in cloud computing environment |
| US20140032185A1 (en) * | 2010-09-17 | 2014-01-30 | The University Of Texas M.D. Anderson Cancer Center | Gpu-based fast dose calculator for cancer therapy |
| CN105389212A (en) * | 2015-10-21 | 2016-03-09 | 浪潮(北京)电子信息产业有限公司 | Job assigning method and apparatus |
| CN105706076A (en) * | 2013-07-31 | 2016-06-22 | 埃克萨公司 | Temperature coupling algorithm for hybrid thermal lattice boltzmann method |
| CN106648890A (en) * | 2016-12-06 | 2017-05-10 | 中国科学院重庆绿色智能技术研究院 | Cloud computing server resource on-line management method and system with energy consumption sensing function |
| CN106951059A (en) * | 2017-03-28 | 2017-07-14 | 中国石油大学(华东) | Based on DVS and the cloud data center power-economizing method for improving ant group algorithm |
| CN107292061A (en) * | 2017-07-28 | 2017-10-24 | 西安交通大学 | A kind of process industry complex electromechanical systems information modelling approach of data-driven |
-
2018
- 2018-07-27 CN CN201810845198.1A patent/CN108845886B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140032185A1 (en) * | 2010-09-17 | 2014-01-30 | The University Of Texas M.D. Anderson Cancer Center | Gpu-based fast dose calculator for cancer therapy |
| CN103365727A (en) * | 2013-07-09 | 2013-10-23 | 南京大学 | Host load forecasting method in cloud computing environment |
| CN105706076A (en) * | 2013-07-31 | 2016-06-22 | 埃克萨公司 | Temperature coupling algorithm for hybrid thermal lattice boltzmann method |
| CN105389212A (en) * | 2015-10-21 | 2016-03-09 | 浪潮(北京)电子信息产业有限公司 | Job assigning method and apparatus |
| CN106648890A (en) * | 2016-12-06 | 2017-05-10 | 中国科学院重庆绿色智能技术研究院 | Cloud computing server resource on-line management method and system with energy consumption sensing function |
| CN106951059A (en) * | 2017-03-28 | 2017-07-14 | 中国石油大学(华东) | Based on DVS and the cloud data center power-economizing method for improving ant group algorithm |
| CN107292061A (en) * | 2017-07-28 | 2017-10-24 | 西安交通大学 | A kind of process industry complex electromechanical systems information modelling approach of data-driven |
Non-Patent Citations (1)
| Title |
|---|
| 王鹏: ""云计算系统相空间分析模型及仿真研究"", 《计算机学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109947551A (en) * | 2019-03-19 | 2019-06-28 | 中南大学 | A kind of more round method for allocating tasks, edge calculations system and its storage medium |
| CN109947551B (en) * | 2019-03-19 | 2021-04-23 | 中南大学 | Multi-turn task allocation method, edge computing system and storage medium thereof |
| CN110096349A (en) * | 2019-04-10 | 2019-08-06 | 山东科技大学 | A kind of job scheduling method based on the prediction of clustered node load condition |
| CN111338649A (en) * | 2020-02-14 | 2020-06-26 | 浪潮商用机器有限公司 | Heterogeneous system, acceleration method and device thereof and readable storage medium |
| CN113205695A (en) * | 2021-04-13 | 2021-08-03 | 东南大学 | Multi-period length bidirectional trunk line green wave control method |
| CN118094601A (en) * | 2024-03-25 | 2024-05-28 | 数盾信息科技股份有限公司 | High-speed data transmission and encryption method and device based on hardware acceleration |
| CN118094601B (en) * | 2024-03-25 | 2024-10-15 | 数盾信息科技股份有限公司 | High-speed data transmission and encryption method and device based on hardware acceleration |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108845886B (en) | 2022-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108845886B (en) | Cloud computing energy consumption optimization method and system based on phase space | |
| Liu et al. | A hierarchical framework of cloud resource allocation and power management using deep reinforcement learning | |
| He et al. | AMTS: Adaptive multi-objective task scheduling strategy in cloud computing | |
| Zhu et al. | Scheduling stochastic multi-stage jobs to elastic hybrid cloud resources | |
| Cariño et al. | Dynamic load balancing with adaptive factoring methods in scientific applications | |
| Kumar T et al. | Hybrid approach for resource allocation in cloud infrastructure using random forest and genetic algorithm | |
| Awad et al. | A novel intelligent approach for dynamic data replication in cloud environment | |
| Ng et al. | Defragmentation for efficient runtime resource management in NoC-based many-core systems | |
| Ataie et al. | Modeling and evaluation of dispatching policies in IaaS cloud data centers using SANs | |
| Hussin et al. | Efficient energy management using adaptive reinforcement learning-based scheduling in large-scale distributed systems | |
| Singh et al. | Value and energy optimizing dynamic resource allocation in many-core HPC systems | |
| Ma et al. | Virtual machine migration techniques for optimizing energy consumption in cloud data centers | |
| Petrovska et al. | Sequential Series-Based Prediction Model in Adaptive Cloud Resource Allocation for Data Processing and Security | |
| Li et al. | Batch jobs load balancing scheduling in cloud computing using distributional reinforcement learning | |
| Parsa et al. | Task dispatching approach to reduce the number of waiting tasks in grid environments | |
| He et al. | Dynamic scheduling of parallel real-time jobs by modelling spare capabilities in heterogeneous clusters | |
| Uchechukwu et al. | Scalable analytic models for performance efficiency in the cloud | |
| Alrammah et al. | Tri-Objective workflow scheduling and optimization in heterogeneous cloud environments | |
| Skrinarova | Implementation and evaluation of scheduling algorithm based on PSO HC for elastic cluster criteria | |
| Zhang et al. | Resource and performance prediction at high utilization for n-tier cloud-based service systems | |
| Mostafavi Amjad et al. | Locality-aware virtual machine placement based on similarity properties in mobile edge computing | |
| Zhao et al. | Distance-aware virtual cluster performance optimization: A hadoop case study | |
| Kamran et al. | Efficient HPC and Energy-Aware Proactive Dynamic VM Consolidation in Cloud Computing | |
| Yuan et al. | Tasks scheduling based on neural networks in grid | |
| Oltikar et al. | Robust resource allocation in weather data processing systems |
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 | ||
| GR01 | Patent grant |