CN115878260A - Low-carbon self-adaptive cloud host task scheduling system - Google Patents
Low-carbon self-adaptive cloud host task scheduling system Download PDFInfo
- Publication number
- CN115878260A CN115878260A CN202211260300.4A CN202211260300A CN115878260A CN 115878260 A CN115878260 A CN 115878260A CN 202211260300 A CN202211260300 A CN 202211260300A CN 115878260 A CN115878260 A CN 115878260A
- Authority
- CN
- China
- Prior art keywords
- host
- task scheduling
- module
- task
- low
- 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.)
- Pending
Links
Images
Classifications
-
- 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
Abstract
The invention discloses a low-carbon self-adaptive cloud host task scheduling system which comprises a state detection module, a task scheduling module and a task scheduling module, wherein the state detection module is used for detecting the running state of a virtual machine in a cloud host service area; the execution configuration module is used for storing historical data of the virtual machine state, periodically predicting the configuration amount of computing resources to be allocated, allocating the starting number of the virtual machines in the next period according to the running state of the virtual machines in the host service area in the current period, and transmitting the execution result to the task scheduling module; the task scheduling module is used for selecting a host and selecting the virtual machines to be started according to the starting number of the virtual machines transmitted by the execution module, and performing cloud host computing resource allocation; according to the cloud host computing resource scheduling method, computing resources are rapidly configured after the state of the virtual machine is detected, and then cloud host computing resources are scheduled, so that the cloud host computing resources are rapidly distributed, meanwhile, in the physical host selection process, the problem of energy consumption is fully considered, and the purposes of reducing energy consumption, and being low-carbon and energy-saving are achieved.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a low-carbon self-adaptive cloud host task scheduling system.
Background
The high-speed development of the cloud data center server always aims at improving efficiency, reducing cost and reducing energy consumption, and along with the gradual expansion of the scale of the cloud data center server, the number of tasks borne by the cloud data center is more and more, and the types of the tasks are more and more complicated. On the other hand, due to the continuous development of cloud computing and the continuous increase of the number of users, massive data is a severe test for background operation. Cloud computing systems often have large-scale servers, and the computing power and the operation quality of various resources are different. In this context, how to schedule tasks to corresponding virtual machines with a reasonable policy and efficiently allocate computing resources in the cloud system at the same time makes it an important research topic that tasks submitted by users can be processed at a faster speed and at a lower cost.
For example, chinese patent CN201510422559.8 discloses a fault-tolerant task scheduling method based on task backward movement in cloud, which replaces the traditional PB model by establishing a real-time fault-tolerant model in virtualized cloud, and establishes a fault-tolerant task scheduling method that makes full use of idle resources by adopting a task backward movement strategy, so as to improve resource utilization under fault-tolerant guarantee and schedulability of fault-tolerant tasks; however, the scheme only considers the relevance of the tasks, and does not fully consider the state of the host, so that the resources of the cloud host cannot be rapidly allocated.
Disclosure of Invention
The invention mainly solves the problem that the resources of the cloud host can not be rapidly distributed in the prior art; the low-carbon self-adaptive cloud host task scheduling system is provided, the physical host is selected after the load state of the physical host is fully considered, the virtual host is started in combination with the execution difficulty of the tasks, and the cloud host computing resources are rapidly distributed.
The technical problem of the invention is mainly solved by the following technical scheme: a low-carbon self-adaptive cloud host task scheduling system comprises a state detection module, a task scheduling module and a task scheduling module, wherein the state detection module is used for detecting the running state of a virtual machine in a cloud host service area; the execution module is used for storing historical data of the state of the virtual machine, periodically predicting the allocation amount of computing resources to be allocated, allocating the starting number of the virtual machine in the next period according to the running state of the virtual machine in the host service area in the current period, and transmitting an execution result to the task scheduling module; and the task scheduling module is used for selecting a host and selecting the virtual machines to be started according to the starting number of the virtual machines transmitted by the execution module, and performing cloud host computing resource allocation.
Preferably, the task scheduling module runs a task scheduling algorithm, and the specific method for performing task scheduling by the task scheduling algorithm includes:
s1: carrying out dynamic load adjustment on the host;
s2: selecting a host based on host applicability and host matching distance;
s3: and selecting the virtual machine to be started on the selected host machine based on the improved MAX-MIN algorithm.
Preferably, the dynamic load adjustment includes:
s11: selecting load parameters to divide the load state of the host;
s12: recording the change times of the load state of the host;
s13: and adjusting the maximum number of executed tasks based on the change times of the load state of the host.
Preferably, the load parameters include one or more of an average number of run queues, CPU utilization, GPU utilization, and memory utilization.
Preferably, the specific method for performing host selection in step S2 is as follows:
s21: evaluating the applicability of available physical hosts, selecting the total number of the currently started virtual machines of each physical host, and defining an applicability division threshold;
s22: dividing the applicability of the host into a high-applicability host set, a low-applicability host set and a dormant host set;
s23: the virtual machines are preferentially distributed to the high-applicability host machine, and the virtual machines on the low-applicability host machine are preferentially closed to enter a dormant state.
Preferably, step S2 further comprises the steps of:
s24: abstracting available resources of a physical host into a three-dimensional vector;
s25: calculating the performance matching distance from the virtual resource to the physical resource based on the performance vectors of the virtual machine and the server;
s26: and selecting a host with a small matching distance for task allocation.
Preferably, the specific method of step S3 is:
s31: dividing the tasks into n levels, wherein each virtual machine corresponds to the highest task level capable of being executed;
s32: and arranging the tasks in the task flow from high to low according to the task level, and scheduling and executing the tasks with higher difficulty in priority.
The invention has the beneficial effects that: the method comprises the steps of carrying out rapid configuration on computing resources after detecting the state of a virtual machine, further carrying out scheduling on computing resources of a cloud host, carrying out physical host selection after fully considering the load state of the physical host in the scheduling process, and starting the virtual host by combining the execution difficulty of tasks, so that rapid distribution of computing resources of the cloud host is realized, rapid configuration and distribution of the computing resources are achieved, meanwhile, in the physical host selection process, the energy consumption problem is fully considered, and the purposes of reducing energy consumption, low carbon and saving energy are achieved.
Drawings
Fig. 1 is a block diagram of a task scheduling system according to an embodiment of the present invention.
FIG. 2 is a flow chart of a task scheduling algorithm of an embodiment of the present invention.
In the figure, the system comprises a user interaction module 1, a cloud resource configuration module 2, a cloud resource configuration module 3, an execution module 4, a resource configuration module 5, a task scheduling module 6, a system evaluation module 7 and a node state detection module.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions in the embodiments of the present invention are further described in detail by the following embodiments in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
The main computing resources in the cloud data center server include a CPU and a GPU.
The energy consumption of the CPU consists of two parts, static energy consumption and dynamic energy consumption, respectively. The dynamic energy consumption is caused by the flow of electric energy caused by the charge and discharge process of a transistor in the CPU, the size of the dynamic energy consumption is related to the running state of the CPU, the main frequency of the CPU refers to the switching frequency of the transistor per second, and the higher the main frequency is, the more the charge and discharge frequency is, the power consumption of the transistor is increased. The higher the voltage of the CPU and the higher the main frequency, the greater the dynamic energy consumption. The static energy consumption mainly comes from the leakage and conduction (short circuit) of the transistor, and is not considered as a constant in the optimization process. Under the environment of the cloud platform, the dynamic energy consumption of the CPU is closely related to the utilization rate of the virtual machine, so that the following power consumption model can be established:
P(u)＝P s +P d ＝P min +(P max -P min )u
wherein P is s And P d Static and dynamic power consumption, P, respectively, of the CPU min For power consumption when the CPU is idle, P max And u is the utilization rate of the CPU, and is the maximum power consumption when the CPU is completely utilized.
Under the actual working environment, the CPU utilization rate changes in real time, and u (t) represents the CPU utilization rate at the time t, so that the CPU utilization rate at the time t 0 To t 1 During a time interval, the power consumption of the CPU can be expressed as:
the energy consumption of the GPU can be divided into runtime energy consumption and idle energy consumption, which are expressed as:
E＝E run +E idle
wherein E run And E idle Respectively the energy consumption of the GPU during operation and the energy consumption of the GPU when idling, and the energy consumption of the GPU when idling can pass through the power P of the GPU when idling idle The product with time yields:
E idle ＝P idle ·t idle
the energy consumption during the operation comprises two parts of energy consumption of a stream processor and energy consumption of a memory, the energy consumption of the stream processor is the sum of the energy consumption of all components in a stream processor unit multiplied by the number of the stream processors in the GPU, and the energy consumption of the memory is equal to the energy consumption of the memory unit multiplied by the number of the memory units:
according to the above analysis, the energy consumption caused by the CPU and the GPU is generally reduced by optimizing the dynamic process thereof, and the energy consumption in this process is mainly generated by the stable energy consumption generated in the running process of the virtual machine or the host and the additional control energy consumption caused by the startup and shutdown process, so that the energy consumption of the system can be effectively reduced by reducing the number of the virtual machines or the hosts which run stably and the startup and shutdown times of the virtual machine and the host.
Example (b): a low-carbon self-adaptive cloud host task scheduling system is shown in figure 1 and comprises a user interaction module 1 and a cloud resource configuration module 2, wherein the user interaction module is connected with the cloud resource configuration module and used for a user to perform task interaction with the cloud resource configuration module, the user interaction module comprises resource configuration checking, resource allocation is performed through user instruction input, and the virtual machine is remotely opened and closed, and serves as an interface for interaction of the user and the cloud resource, so that the user is helped to simplify the task submission process, and the task response is efficiently acquired in real time.
The cloud resource configuration module comprises an execution module 3, a resource configuration module 4, a task scheduling module 5, a system evaluation module 6 and a node state detection module 7, the execution module and the resource configuration module form an execution configuration module, the system evaluation module and the node state detection module form a state detection module, the execution module is connected with the user interaction module, the execution module is used for storing cloud computing resources, receiving feedback information of a user and calculating task allocation quantity based on the predicted resource allocation task quantity of the next period and the virtual machine usage quantity of the current period, the resource configuration module is provided with a task quantity prediction algorithm, historical data of the virtual machine state are stored, the calculation resource allocation quantity needing to be allocated is predicted periodically, and the virtual machine starting quantity of the next period is allocated to the execution module according to the running state of the virtual machine in the host service area of the current period; the task scheduling module is used for selecting a physical host and a virtual machine, distributing the task scheduling amount to the selected virtual machine, realizing the distribution of cloud host computing resources and is connected with the execution module; the system evaluation module is used for recording the execution quality of each batch processing task queue, carrying out accuracy evaluation on the task scheduling amount of the execution module and the predicted task amount of the resource allocation module, calculating the energy consumption of the task amount of the current period, and is connected with the task scheduling module and the resource allocation module; the node state detection module is used for detecting the running state of the host and the running state of the virtual machine in the cloud host service area, including various information such as the execution state of tasks and the resource utilization rate, summarizing the information and reporting the information to the corresponding module in a classified mode, and assisting the whole system in reducing energy consumption.
The execution module comprises a processing module and a storage module, the storage module is used for storing cloud computing resources, the storage module is connected with the processing module, and the processing module calls the cloud computing resources of the storage module according to the calculated task scheduling amount and distributes the cloud computing resources to the task scheduling module.
As shown in fig. 2, the task scheduling module runs a task scheduling algorithm, and the specific method for performing task scheduling by the task scheduling algorithm includes:
s1: carrying out dynamic load adjustment on the host; the dynamic load adjustment includes:
s11: selecting load parameters to divide the load state of the host; the process of scheduling the tasks to the virtual machines is based on a dynamic load adjustment strategy, a reasonable load evaluation method is provided for task scheduling of the system, each host is adaptive to the change of the load, the tasks are obtained according to the computing capacity to realize self-adjustment, and meanwhile, huge system overhead caused by adopting a complex scheduling algorithm is avoided, and the system becomes a system performance bottleneck. The dynamic change of the load of each host, the difference of different host performances and the difference of different task load requirements are fully considered, the average number of running queues, the CPU utilization rate, the GPU utilization rate and the memory utilization rate are selected as load parameters required by evaluation, different priorities are given, and the task scheduling method for multi-level load evaluation is provided. In order to overcome the defects of a load evaluation mechanism of a cloud platform, the load evaluation precision is improved as much as possible, and meanwhile, the system overhead brought by a load evaluation method is reduced.
Firstly, in order to effectively evaluate the load condition of a host, various load parameters are required to be selected as evaluation standards, wherein the number of the running queue processes counts the number of tasks which are being executed and are waiting to be executed, and the state of the system is indicated; the utilization rate of the CPU and the GPU reflects the size of system resources occupied by the executing task, and whether the current host has enough resources to execute the new task is judged. The host is divided into three states according to the above parameters: starvation (HUNGER), OPTIMAL (OPTIMAL), and SATURATION (saturition).
Recording the change times of the load state of the host; in the process of host operation, corresponding adjustment is made according to the change of own resources and load, so that the operation performance of the host can achieve the optimal self-adaptive adjustment. Each host balances the load condition by collecting load parameters when the processing of each batch processing task queue is finished, and dynamically adjusts the size of the maximum execution task number MaxTasks (task slots), so that the host can only obtain partial tasks at each judgment point instead of filling the capacity at one time, and the dynamic controllability of each node in the system is improved. And re-evaluating the load condition of the system in the next period and then deciding whether to accept a new task. The system is unstable due to frequent adjustment of the maximum number of executed tasks, so that the self-adaptive adjustment is performed only when a saturated state or a starvation state continuously occurs k times, the granularity of the self-adaptive adjustment of the system is determined by the size of k, and the value of k can be adjusted according to the actual running condition in the cloud system during the actual running of the system. The num _ LightLoad and num _ OverLoad parameters record the number of times starvation and saturation occur consecutively at the host.
Adjusting the maximum number of executed tasks based on the change times of the load state of the host; when the processing of each batch processing task queue is completed, the num _ LightLoad and num _ OverLoad parameters of the host are updated, and then whether the host asks for a new task is determined by judging whether the number of the current queued tasks exceeds the maximum execution task number MaxTasks. And finally, updating whether the maximum execution task number MaxTasks is increased or decreased according to the parameter states of the num _ LightLoad and the num _ OverLoad of the host, and circulating the same steps at the next decision point.
On the basis of a dynamic load adjustment strategy, a host is selected for each task in a task flow in a host set with a task slot and a vacancy, and a virtual machine is preferentially selected from the hosts with high scores to allocate the tasks according to scores of two scales of applicability and matching distance.
S2: selecting a host based on host applicability and host matching distance; the method comprises the steps of firstly evaluating the applicability of available physical hosts, selecting the total number of virtual machines currently started by each physical host, defining an applicability division threshold, and dividing the applicability of the hosts into a high-applicability host set, a low-applicability host set and a dormant host set. The number of the starting virtual machines on the high-applicability host set exceeds the applicability division threshold value, and the virtual machines are preferentially distributed to the high-applicability hosts so as to reduce the starting number of the hosts and improve the resource utilization rate of the system; the number of the starting virtual machines on the low-applicability host set is lower than the applicability division threshold value, and the virtual machines on the low-applicability host set are preferentially closed to enter a dormant state, so that the energy consumption of the system is reduced.
And searching an optimized feasible solution through the division of the applicability host set and the matching between supply and demand resources. The available resources of the physical host are abstracted into a three-dimensional vector. In consideration of load balancing and resource maximum utilization among high-applicability hosts, the usage of various resources of the hosts needs to be balanced to prevent the occurrence of a barrel effect, and therefore the matching degree of the virtual machine to be placed and the target host needs to be considered. The performance requirement of the virtual machine is matched with the performance of the physical server, and a proper physical server is selected to deploy the virtual machine, so that the performance matching distance from the virtual resource to the physical resource is established based on the performance vectors of the virtual machine and the server, and the smaller the matching distance is, the more the performance of the physical server resource is matched with the performance required by the virtual machine. The virtual machines with different specifications and the host machine have different resource performances, so that the resources need to be normalized, and the host machine with a small matching distance is selected for task allocation.
S3: selecting a virtual machine to start on the selected host based on an improved MAX-MIN algorithm; dividing the tasks into n levels, wherein each virtual machine corresponds to the highest task level capable of being executed; and arranging the tasks in the task flow from high to low according to the task level, and scheduling and executing the tasks with higher difficulty in priority.
The conventional MIN-MIN algorithm allocates the task with the minimum length to the virtual machine with the minimum sum of the ready time and the task execution time from the unallocated tasks in the current task list in consideration of the expected completion time of the tasks, that is, the unallocated task with the lowest complexity is placed on the virtual machine with the shortest expected completion time to run. And selecting a virtual machine with stronger processing performance in a biased manner, wherein the virtual machine with poorer performance is gradually used only under the condition that the waiting time is too long due to too many tasks on the high-performance virtual machine.
The improved MAX-MIN algorithm is adopted in the scheme. The tasks are divided into n levels, each virtual machine corresponds to the highest task level capable of being executed, the tasks of the n levels are arranged according to the execution difficulty, the task of the level 1 is the lowest in execution difficulty, and the task of the level n is the highest in execution difficulty. And according to the MAX-MIN algorithm idea, arranging the tasks in the task flow from high to low according to the task grade, and preferentially scheduling and executing the tasks with higher difficulty.
The specific method for predicting the task quantity of the resource allocation of the next period by the task quantity prediction algorithm comprises the following steps: summarizing task quantity values of tasks of various types reaching a system in each period into a vector D; performing nonlinear prediction on the task quantity by adopting a cubic exponential smoothing algorithm; performing task quantity linear prediction by adopting an ARIMA prediction model; distributing weight coefficients by adopting a composite prediction model; and combining the prediction result of the cubic exponential smoothing algorithm and the prediction result of the ARIMA prediction model with a weight coefficient, and outputting a comprehensive task quantity prediction result.
By adopting the task scheduling method, the rapid allocation of the cloud host computing resources is realized, and the purpose of reducing energy consumption is realized.
The above-described embodiment is a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, and other variations and modifications may be made without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (7)
1. A low-carbon self-adaptive cloud host task scheduling system is characterized by comprising:
the state detection module is used for detecting the running state of the virtual machine in the cloud host service area;
the execution configuration module is used for storing historical data of the virtual machine state, periodically predicting the configuration amount of computing resources to be allocated, allocating the starting number of the virtual machines in the next period according to the running state of the virtual machines in the host service area in the current period, and transmitting the execution result to the task scheduling module;
and the task scheduling module is used for selecting a host and selecting the virtual machines to be started according to the starting number of the virtual machines transmitted by the execution module, and performing cloud host computing resource allocation.
2. The low-carbon adaptive cloud host task scheduling system of claim 1,
the task scheduling module runs a task scheduling algorithm, and the specific method for performing task scheduling by the task scheduling algorithm comprises the following steps:
s1: carrying out dynamic load adjustment on the host;
s2: selecting a host based on host applicability and host matching distance;
s3: and selecting the virtual machine to be started on the selected host machine based on the improved MAX-MIN algorithm.
3. The low-carbon adaptive cloud host task scheduling system of claim 2,
the dynamic load adjustment includes:
s11: selecting load parameters to divide the load state of the host;
s12: recording the change times of the load state of the host;
s13: and adjusting the maximum number of executed tasks based on the change times of the load state of the host.
4. The low-carbon adaptive cloud host task scheduling system of claim 3,
the load parameters include one or more of run queue averages, CPU utilization, GPU utilization, and memory utilization.
5. The low-carbon adaptive cloud host task scheduling system according to claim 2 or 3,
the specific method for selecting the host in step S2 is as follows:
s21: evaluating the applicability of available physical hosts, selecting the total number of the virtual machines currently started by each physical host, and defining an applicability division threshold;
s22: dividing the applicability of the host into a high-applicability host set, a low-applicability host set and a dormant host set;
s23: the virtual machines are preferentially distributed to the high-applicability host machine, and the virtual machines on the low-applicability host machine are preferentially closed to enter a dormant state.
6. The low-carbon adaptive cloud host task scheduling system of claim 5,
step S2 further includes the steps of:
s24: abstracting available resources of a physical host into a three-dimensional vector;
s25: calculating the performance matching distance from the virtual resource to the physical resource based on the performance vectors of the virtual machine and the server;
s26: and selecting the host with small matching distance to distribute the tasks.
7. The low-carbon adaptive cloud host task scheduling system according to claim 2 or 3,
the specific method of the step S3 is as follows:
s31: dividing the tasks into n levels, wherein each virtual machine corresponds to the highest task level capable of being executed;
s32: and arranging the tasks in the task flow from high to low according to the task level, and scheduling and executing the tasks with higher difficulty in priority.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211260300.4A CN115878260A (en) | 2022-10-14 | 2022-10-14 | Low-carbon self-adaptive cloud host task scheduling system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211260300.4A CN115878260A (en) | 2022-10-14 | 2022-10-14 | Low-carbon self-adaptive cloud host task scheduling system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115878260A true CN115878260A (en) | 2023-03-31 |
Family
ID=85770417
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211260300.4A Pending CN115878260A (en) | 2022-10-14 | 2022-10-14 | Low-carbon self-adaptive cloud host task scheduling system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115878260A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117407178A (en) * | 2023-12-14 | 2024-01-16 | 成都凯迪飞研科技有限责任公司 | Acceleration sub-card management method and system for self-adaptive load distribution |
| CN117632380A (en) * | 2024-01-25 | 2024-03-01 | 泰德网聚(北京)科技股份有限公司 | Low-code workflow system for automatically generating script based on user demand |
-
2022
- 2022-10-14 CN CN202211260300.4A patent/CN115878260A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117407178A (en) * | 2023-12-14 | 2024-01-16 | 成都凯迪飞研科技有限责任公司 | Acceleration sub-card management method and system for self-adaptive load distribution |
| CN117407178B (en) * | 2023-12-14 | 2024-04-02 | 成都凯迪飞研科技有限责任公司 | Acceleration sub-card management method and system for self-adaptive load distribution |
| CN117632380A (en) * | 2024-01-25 | 2024-03-01 | 泰德网聚(北京)科技股份有限公司 | Low-code workflow system for automatically generating script based on user demand |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Praveenchandar et al. | RETRACTED ARTICLE: Dynamic resource allocation with optimized task scheduling and improved power management in cloud computing | |
| CN109324875B (en) | Data center server power consumption management and optimization method based on reinforcement learning | |
| Hui et al. | Improved strategies for dynamic load balancing | |
| CN115878260A (en) | Low-carbon self-adaptive cloud host task scheduling system | |
| CN101488098B (en) | Multi-core computing resource management system based on virtual computing technology | |
| US9104498B2 (en) | Maximizing server utilization within a datacenter | |
| CN106528266B (en) | Method and device for dynamically adjusting resources in cloud computing system | |
| Arshad et al. | Utilizing power consumption and SLA violations using dynamic VM consolidation in cloud data centers | |
| WO2011031459A2 (en) | A method and apparatus for data center automation | |
| CN103927225A (en) | Multi-core framework Internet information processing and optimizing method | |
| Li et al. | Energy-efficient and QoS-aware model based resource consolidation in cloud data centers | |
| CN107370799B (en) | A kind of online computation migration method of multi-user mixing high energy efficiency in mobile cloud environment | |
| CN103179048A (en) | Method and system for changing main machine quality of service (QoS) strategies of cloud data center | |
| CN110795323A (en) | Load statistical method, device, storage medium and electronic equipment | |
| CN109144664B (en) | Dynamic migration method of virtual machine based on user service quality demand difference | |
| Zhao et al. | Insight and reduction of MapReduce stragglers in heterogeneous environment | |
| Ma et al. | SE-PSO: resource scheduling strategy for multimedia cloud platform based on security enhanced virtual migration | |
| CN104731662B (en) | A kind of resource allocation methods of variable concurrent job | |
| Fan et al. | GreenSleep: a multi-sleep modes based scheduling of servers for cloud data center | |
| Sun et al. | An energy efficient and runtime-aware framework for distributed stream computing systems | |
| EP3982258A1 (en) | Method and apparatus for reducing power consumption of virtual machine cluster | |
| Cao et al. | A QoS-guaranteed energy-efficient VM dynamic migration strategy in cloud data centers | |
| Khojasteh et al. | Characterizing energy consumption of IaaS clouds in non-saturated operation | |
| Das et al. | Mapreduce scheduler: A 360-degree view | |
| Chakroun et al. | Reducing energy consumption for reconfiguration in cloud data centers |
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 |