WO2015139374A1 - 一种云计算平台下的虚拟机分布式任务调度方法 - Google Patents

一种云计算平台下的虚拟机分布式任务调度方法 Download PDF

Info

Publication number
WO2015139374A1
WO2015139374A1 PCT/CN2014/079039 CN2014079039W WO2015139374A1 WO 2015139374 A1 WO2015139374 A1 WO 2015139374A1 CN 2014079039 W CN2014079039 W CN 2014079039W WO 2015139374 A1 WO2015139374 A1 WO 2015139374A1
Authority
WO
WIPO (PCT)
Prior art keywords
physical server
load
server
physical
scheduling
Prior art date
Application number
PCT/CN2014/079039
Other languages
English (en)
French (fr)
Inventor
李凡
柳岸
王流一
王铁军
Original Assignee
成都盛思睿信息技术有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 成都盛思睿信息技术有限公司 filed Critical 成都盛思睿信息技术有限公司
Publication of WO2015139374A1 publication Critical patent/WO2015139374A1/zh

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08Configuration management of networks or network elements
    • H04L41/0803Configuration setting
    • H04L41/0813Configuration setting characterised by the conditions triggering a change of settings
    • H04L41/0816Configuration setting characterised by the conditions triggering a change of settings the condition being an adaptation, e.g. in response to network events
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08Configuration management of networks or network elements
    • H04L41/0896Bandwidth or capacity management, i.e. automatically increasing or decreasing capacities
    • H04L41/0897Bandwidth or capacity management, i.e. automatically increasing or decreasing capacities by horizontal or vertical scaling of resources, or by migrating entities, e.g. virtual resources or entities
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00Arrangements for monitoring or testing data switching networks
    • H04L43/08Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
    • H04L43/0805Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters by checking availability
    • H04L43/0817Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters by checking availability by checking functioning
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00Arrangements for monitoring or testing data switching networks
    • H04L43/10Active monitoring, e.g. heartbeat, ping or trace-route
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/01Protocols
    • H04L67/10Protocols in which an application is distributed across nodes in the network
    • H04L67/1001Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
    • H04L67/1004Server selection for load balancing
    • H04L67/1008Server selection for load balancing based on parameters of servers, e.g. available memory or workload
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/01Protocols
    • H04L67/10Protocols in which an application is distributed across nodes in the network
    • H04L67/1001Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
    • H04L67/1029Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers using data related to the state of servers by a load balancer
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/50Network services
    • H04L67/60Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
    • H04L67/62Establishing a time schedule for servicing the requests
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • G06F2009/4557Distribution of virtual machine instances; Migration and load balancing
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/01Protocols
    • H04L67/10Protocols in which an application is distributed across nodes in the network
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L69/00Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
    • H04L69/28Timers or timing mechanisms used in protocols
    • YGENERAL 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
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00Reducing energy consumption in communication networks

Definitions

  • the present invention relates to the field of cloud computing, and in particular, to a virtual machine distributed task scheduling method under a cloud computing platform.
  • Cloud computing As an emerging business computing model, has become a research hotspot in academia and major IT vendors.
  • Cloud computing is currently the most widely used distributed application system. It distributes computing power across resource pools composed of a large number of servers, enabling users to access computing, storage, and communication services on demand.
  • the singularity of the task model and the performance of the task scheduling algorithm are low, and it is not possible to balance the QoS (Quality of Service) requirements. How to make it reasonable? Allocating cloud resources and efficiently performing virtual machine task scheduling are key issues in the field of cloud computing research.
  • the traditional scheduling mechanism for resource balancing is based on the scheduling of a single virtual machine CPU load to migrate a virtual machine with a higher CPU load to another physical server.
  • This traditional scheduling mechanism has many drawbacks: 1 Because the virtual machine VCPU will be mapped to a CPU core of the physical server, when the virtual machine load is high, it can only affect several CPU cores mapped by the physical server. Affects the operation of a virtual machine mapped to other cores. Even if it is migrated to another physical server, if the computing power of the physical server is the same as that of the original physical server, there will be no improvement after the migration. 2 The virtual machine load may be too high. Instantaneous peak, the CPU will calculate the loss and a large amount of 10 load during the migration process. 3 The traditional mode of scheduling mechanism is not considered as a whole, and the method of “removing the east wall to complement the western wall” does not achieve the effect of resource balance.
  • the traditional power-saving scheduling mechanism is based on virtual machine CPU load and physical server CPU load scheduling, and is not considered from the overall perspective, and from the overall CPU load development trend.
  • This kind of power-saving scheduling mechanism has many drawbacks: Because the instantaneous peak and valley value are easy to cause misjudgment, and the host machine starts and stops for a long time, usually more than ten minutes, frequent start and stop need to consume ten times the usual time. The energy consumption, not only does not save energy, but affects business operations.
  • the object of the present invention is to overcome the deficiencies of the prior art, and provide a distributed task scheduling method for a virtual machine under a cloud computing platform, which has various scheduling modes and strong flexibility; and comprehensively considers resource balancing scheduling from a virtualized VCPU mapping mechanism. Online migration, high availability, and load balancing of virtual machines; Considering power-saving scheduling from the virtualization VCPU mapping mechanism as a whole, realizing the economic operation of virtual machines and achieving energy-saving effects.
  • a virtual machine distributed task scheduling method under a cloud computing platform which includes the following situations: 1 When a physical server crashes, the physical event of the down event occurs. The load on the server is migrated to the normal physical server. 2 When the load of a physical server is high, the high load is migrated to the physical server with low load when the application is not interrupted. When the load of a physical server is small, the load on the physical server is migrated to other physical servers without interruption, and will be idle. The physical server sleeps or shuts down, reducing power consumption.
  • the migration of the physical server on the downtime event to the normal physical server includes the following specific steps:
  • the primary management server establishes a connection with a plurality of physical servers located in the resource pool to maintain communication;
  • Each physical server continuously sends a heartbeat line to the primary management server at intervals tl;
  • the primary management server checks the heartbeat request of each physical server. If the heartbeat request of a physical server is not received after the preset time t2, and the heartbeat of the physical server is abnormal, the network of the physical server is automatically detected. If a network interruption is detected and does not recover within a certain period of time, a physical server down event is triggered, thereby starting the HA mechanism, and automatically transferring the load on the physical server to another physical server for re-running.
  • Migrating a high load to a low-load physical server involves the following specific steps:
  • the primary management server establishes a connection with a plurality of physical servers located in the resource pool to maintain communication;
  • S22 The primary management server monitors the running status of the load on each physical server in real time, and calculates the load LC of each physical server in the resource pool.
  • the calculation formula is:
  • Load LC SUM (virtual machine VCPU count) / physical server CPU core count;
  • S26 Migrate the selected virtual machine from the MAX (LC) physical server to the MIN (LC) physical server, and calculate the load LC of the migrated MAX (LC) physical server and the MIN (LC) physical server:
  • MIN (LC) (MIN (LC) * physical server CPU core +MINV) / physical server CPU core number;
  • MAX (LC) (MIN (LC) * physical server CPU core -MINV) / physical server CPU core number;
  • step S22 the process proceeds to step S22 to start the next round of calculation.
  • the primary management server establishes a connection with a plurality of physical servers located in the resource pool to maintain communication;
  • S32 The primary management server monitors the running status of the load on each physical server in real time, and calculates the load LC of each physical server in the resource pool.
  • the calculation formula is:
  • Load LC SUM (number of virtual machine VCPUs) / number of physical server CPU cores; S33: Determine whether SUM (LC) is greater than "scheduling threshold SF* physical server number", if "SUM (LC) scheduling threshold SF* physical server number", proceed to the next step, if "SUM (LC) > scheduling threshold SF* physical server number, perform the following operations: Select the physical server in the resource pool from the resource pool, enter the resource balance scheduling, and migrate the virtual machine from the higher-load physical server to the lower-load physical server until the calculation SUM (LC) scheduling threshold SF* physical server CPU core, or the resource pool has no resources;
  • MIN (LC) (MIN (LC) * physical server CPU core -VC) / physical server CPU core number;
  • CHOICE (LC) (CHOICE (LC) * physical server CPU core + VC) / physical server CPU core; and migrate the virtual machine;
  • step S33 the process proceeds to step S33 to start the next round of calculation.
  • the main management server and each physical server form a distributed scheduling mechanism of a server-agent architecture system, and the main management server and each physical server are embedded with a script parsing language, and implement a task-tasklet framework distributed computing model.
  • the main management server executes the task part code, and forwards the context parameter and the tasklet part code to each physical server for execution.
  • the whole process adopts an asynchronous manner, and sends a different stage event trigger execution code through the distributed event mechanism to complete the information and process synchronization. .
  • the invention has the beneficial effects that: the invention performs task scheduling from a plurality of situations, realizes diversity and flexibility of the scheduling manner; comprehensively considers resource balancing scheduling from the virtualized VCPU mapping mechanism, and realizes online migration of the virtual machine, High availability and load balancing; Considering power saving scheduling from the virtualized VCPU mapping mechanism as a whole, realizing the economic operation of the virtual machine and achieving energy saving effect.
  • FIG. 1 is a flow chart of a method for migrating a load on a physical server in which a down event occurs to a normal physical server according to the present invention
  • 2 is a flow chart of a method for migrating a high load to a physical server with a low load according to the present invention
  • 3 is a flow chart of a method for migrating a load on a physical server with a small load to another physical server and sleeping or shutting down the idle physical server according to the present invention.
  • a virtual machine distributed task scheduling method under a cloud computing platform which includes the following situations: 1 when a physical server down event occurs, the load on the physical server in which the down event occurs is migrated to a normal physical server; When the load of a physical server is high, the application load is not interrupted, and the high load is migrated to the physical server with low load; 3 when the load of a physical server is small, the application is not interrupted. In this case, the load on the physical server is migrated to other physical servers, and the idle physical server is hibernated or shut down to reduce energy consumption.
  • the main management server and each physical server form a distributed scheduling mechanism of a server-agent architecture system, and the main management server and each physical server are embedded with a script parsing language, and implement a task-tasklet framework distributed computing model.
  • the main management server executes the task part code, and forwards the context parameter and the tasklet part code to each physical server for execution.
  • the whole process adopts an asynchronous manner, and sends a different stage event trigger execution code through the distributed event mechanism to complete the information and process synchronization. .
  • the load on the physical server where the downtime event occurs is migrated to the normal physical server, including the following specific steps:
  • the primary management server establishes a connection with a plurality of physical servers located in the resource pool to maintain communication;
  • S12 Each physical server sends a heartbeat line to the primary management server through the management network at intervals of time;
  • the server-agent architecture consisting of the main management server and each physical server separates the resource scheduling from the heartbeat check.
  • the structure is clear, the configuration is easier to manage, and a large cluster of more than 100 nodes can be built. The traditional maximum size cannot exceed 32 nodes.
  • the primary management server is only responsible for managing resource scheduling. Even if the primary management server is down, the physical server can operate normally, but it is out of management, and there is no physical server crash.
  • the migration of the high load to the physical server with low load includes the following specific steps:
  • the primary management server establishes a connection with a plurality of physical servers located in the resource pool to maintain communication;
  • S22 The primary management server monitors the running status of the load on each physical server in real time, that is, monitors each physical server. Run the VM state and calculate the load LC of each physical server in the resource pool. The calculation formula is:
  • Load LC SUM (virtual machine VCPU count) / physical server CPU core count;
  • S24 determining whether MAX(LC) is greater than a preset scheduling threshold SF, if greater than a preset scheduling threshold SF, proceeding to the next step, if less than or equal to a preset scheduling threshold SF, the scheduling ends; Said that the number of virtual machine VCPU / physical server CPU core is 3 for the physical server CPU to use the best parameters, less than 1, it means that some physical server CPU core may be idle;
  • S26 Migrate the selected virtual machine from the MAX (LC) physical server to the MIN (LC) physical server, and calculate the load LC of the migrated MAX (LC) physical server and the MIN (LC) physical server:
  • MIN (LC) (MIN (LC) * physical server CPU core +MINV) / physical server CPU core number;
  • MAX (LC) (MIN (LC) * physical server CPU core -MINV) / physical server CPU core number;
  • step S22 the process proceeds to step S22 to start the next round of calculation.
  • the load on the physically loaded physical server is migrated to another physical server, and the idle physical server is hibernated or shut down, including the following specific steps:
  • the primary management server establishes a connection with a plurality of physical servers located in the resource pool to maintain communication;
  • S32 The primary management server monitors the running status of the load on each physical server in real time, and calculates the load LC of each physical server in the resource pool.
  • the calculation formula is:
  • Load LC SUM (virtual machine VCPU count) / physical server CPU core count;
  • S33 Determine whether SUM(LC) is greater than "scheduling threshold SF* physical server number", and if "SUM(LC) scheduling threshold SF* physical server number", proceed to the next step, if "SUM(LC)> scheduling threshold SF* physical server number, first perform the following operations: select the physical server in the stop state from the resource pool, enter the resource balance scheduling, and migrate the virtual machine from the higher-load physical server to the lower-load physical server until Calculate the SUM(LC) scheduling threshold SF* physical server CPU core, or the resource pool has no resources;
  • S34 It is judged whether "5 0 ⁇ scheduling threshold 5?* (physical server number - 2)" is established. If “SUM(LC) ⁇ scheduling threshold SF* (physical server number - 2)", the next step is entered. If the "SUM (LC) scheduling threshold SF* (physical server number - 2)", the scheduling ends; generally, a certain capacity is reserved to meet the business burst access demand, where the reserved capacity is 1;
  • MIN (LC) (MIN (LC) * physical server CPU core -VC) / physical server CPU core number;
  • CHOICE (LC) (CHOICE (LC) * physical server CPU core + VC) / physical server CPU core; and migrate the virtual machine;
  • step S33 the process proceeds to step S33 to start the next round of calculation.

Abstract

本发明公开了一种云计算平台下的虚拟机分布式任务调度方法,包括以下情况:①当物理服务器宕机事件时,将该物理服务器上的负载迁移到正常的物理服务器上;②当某个物理服务器的负载较高时,在应用不中断的情况下,将偏高的负载迁移到负载偏低的物理服务器上;③当某个物理服务器的负载较小时,在应用不中断的情况下,将该物理服务器上的负载迁移到其他物理服务器上,并将空闲出来的物理服务器进行休眠或关机。本发明实现了调度方式的多样性及灵活性;从虚拟化VCPU映射机制上整体考虑资源均衡调度,实现了虚拟机的在线迁移、高可用以及负载均衡;从虚拟化VCPU映射机制上整体考虑省电调度,实现了虚拟机的经济运行,达到了节能的效果。

Description

一种云计算平台下的虚拟机分布式任务调度方法
技术领域
本发明涉及云计算领域, 特别是涉及一种云计算平台下的虚拟机分布式任务调度方法。
背景技术
当前, 云计算作为一种新兴的商业计算模式, 已经成为学术界和各大 IT厂商的研究热 点。 云计算是目前研究应用最广泛的分布式应用系统, 它将计算能力分布在大量服务器构成 的资源池上, 用户能够按需获取计算、 存储和通信服务。 针对当前云计算虚拟机任务调度问 题的研究中所存在的任务模型的单一性以及任务调度算法性能较低,且不能够兼顾云计算 QoS (Qual ity of Service , 服务质量)需求的情况, 如何合理分配云资源, 高效的进行虚拟机任 务调度是云计算研究领域的关键问题。
传统的针对资源均衡的调度机制是基于单个虚拟机 CPU负载进行调度, 将某个 CPU负载 较高的虚拟机迁移到其他物理服务器上。 这种传统的调度机制存在很大弊端: ①由于虚拟机 VCPU将映射到物理服务器的某个 CPU核上, 虚拟机负载较高时, 也只能影响物理服务器映射 的几个 CPU核, 不会影响映射到其他核的虚拟机运行, 即使迁移到其他物理服务器上, 如果 该物理服务器的运算能力与原物理服务器的运算能力相同, 迁移后也不会有改善; ②虚拟机 负载过高可能是瞬时高峰, 迁移过程中本身会到来 CPU计算损耗以及大量 10负载; ③传统方 式的调度机制没有从整体上进行考虑, 采用 "拆东墙补西墙"的方式达不到资源均衡的效果。
传统的针对省电的调度机制是基于虚拟机 CPU负载以及物理服务器 CPU负载进行调度, 并没有从整体考虑, 以及从 CPU整体负载发展趋势考虑。 这种针对省电的调度机制存在很大 弊端: 因为瞬间的峰值以及谷值容易引起错误判断, 而宿主机启停需要很长时间, 一般在十 分钟以上, 频繁启动和停止需要消耗平时十倍的能耗, 不但不节能反而影响业务运行。
发明内容
本发明的目的在于克服现有技术的不足, 提供一种云计算平台下的虚拟机分布式任务调 度方法, 调度方式多样, 灵活性强; 从虚拟化 VCPU映射机制上整体考虑资源均衡调度, 实现 了虚拟机的在线迁移、 高可用以及负载均衡; 从虚拟化 VCPU映射机制上整体考虑省电调度, 实现了虚拟机的经济运行, 达到了节能的效果。
本发明的目的是通过以下技术方案来实现的: 一种云计算平台下的虚拟机分布式任务调 度方法, 它包括以下情况: ①当物理服务器宕机事件时, 将该出现宕机事件的物理服务器上 的负载迁移到正常的物理服务器上; ②当某个物理服务器的负载较高时, 在应用不中断的情 况下, 将偏高的负载迁移到负载偏低的物理服务器上; ③当某个物理服务器的负载较小时, 在应用不中断的情况下, 将该物理服务器上的负载迁移到其他物理服务器上, 并将空闲出来 的物理服务器进行休眠或关机, 降低能耗。
其中, 将出现宕机事件的物理服务器上的负载迁移到正常的物理服务器上, 包括以下具 体步骤:
S11 : 主管理服务器和多个位于资源池的物理服务器建立连接保持通讯;
S12: 各物理服务器每隔一段时间 tl持续的向主管理服务器发送心跳线;
S13: 主管理服务器检査各物理服务器的心跳线请求, 若在预设时间 t2后没有收到某物 理服务器的心跳线请求, 发现某物理服务器的心跳异常, 则自动探测该物理服务器网络是否 连通, 若检测到网络中断, 并在一段时间内没有恢复, 则触发物理服务器宕机事件, 从而启 动 HA机制, 自动的将该物理服务器上的负载转移到其他物理服务器上重新运行。
将偏高的负载迁移到负载偏低的物理服务器上, 包括以下具体步骤:
S21 : 主管理服务器和多个位于资源池的物理服务器建立连接保持通讯;
S22: 主管理服务器实时监控各物理服务器上负载的运行情况, 计算资源池中各物理服务 器的负载 LC, 计算公式为:
负载 LC= SUM (虚拟机 VCPU数) /物理服务器 CPU核数;
S23: 对上述步骤计算所得的负载 LC进行排序, 即 SORT (LC) ;
S24: 判断 MAX (LC)是否大于预设的调度阀值 SF, 若大于预设的调度阀值 SF, 则进入下 一步, 若小于或等于预设的调度阀值 SF, 则调度结束;
S25: 从 MAX (LC)物理服务器中选取 MINV=MIN (虚拟机 VCPU) , 判断 "MIN (LC) - MIN (虚拟 机 VCPU) "是否小于或等于调度阀值 SF, 若 "MIN (LC) -MIN (虚拟机 VCPU) 调度阀值 SF", 则 进入下一步, 否则调度结束;
S26: 将选取的虚拟机从 MAX (LC)物理服务器迁移到 MIN (LC)物理服务器上, 并计算迁移 后 MAX (LC)物理服务器及 MIN (LC)物理服务器的负载 LC:
MIN (LC) = (MIN (LC) * 物理服务器 CPU核数 +MINV) /物理服务器 CPU核数;
MAX (LC) = (MIN (LC) * 物理服务器 CPU核数 -MINV) /物理服务器 CPU核数;
然后, 进入步骤 S22, 开始下一轮计算。
将负载较小的物理服务器上的负载迁移到其他物理服务器上, 并将空闲出来的物理服务 器进行休眠或关机, 包括以下具体步骤:
S31 : 主管理服务器和多个位于资源池的物理服务器建立连接保持通讯;
S32: 主管理服务器实时监控各物理服务器上负载的运行情况, 计算资源池中各物理服务 器的负载 LC, 计算公式为:
负载 LC=SUM (虚拟机 VCPU数) /物理服务器 CPU核数; S33: 判断 SUM (LC)是否大于 "调度阀值 SF*物理服务器数", 若 " SUM (LC) 调度阀值 SF* 物理服务器数"则进入下一步, 若 " SUM (LC)〉调度阀值 SF*物理服务器数", 则进行如下操 作- 从资源池中选择停止状态的物理服务器, 进入资源均衡调度, 将虚拟机从负载较高的物 理服务器迁移到负载较低的物理服务器上, 直到计算 SUM (LC) 调度阀值 SF*物理服务器 CPU 核数, 或者资源池已无资源为止;
S34: 判断 " 5 0 <调度阀值5?* (物理服务器数-2) "是否成立, 若 " SUM (LC) <调度 阀值 SF* (物理服务器数 -2) ",则进入下一步,若" SUM (LC) 调度阀值 SF* (物理服务器数 -2) ", 则调度结束;
S35: 依次判断负载 LC==0是否成立, 若负载 LC==0, 从集合 (LC) 中移出, 并调度该物 理服务器, 将该物理服务器进行停止操作, 对集合 (LC) 按照计算的负载 LC进行排序, 即 SORT (LC);
S36: 依次从 MIN (LC)物理服务器中选取一个虚拟机 VC=H0ST OF (虚拟机 VCPU), 依次从 集合 (LC)中, 由大到小顺序选取物理服务器计算:
MIN (LC) = (MIN (LC) * 物理服务器 CPU核数 -VC) / 物理服务器 CPU核数;
CHOICE (LC) = (CHOICE (LC) * 物理服务器 CPU核数 +VC) / 物理服务器 CPU核数; 并迁移该虚拟机;
然后进入步骤 S33, 开始下一轮计算。
所述主管理服务器和各物理服务器构成一种 server-agent架构体系的分布式调度机制, 主管理服务器和各物理服务器均内嵌脚本解析语言,并实现一种 task-tasklet框架分布式计 算模型, 主管理服务器执行 task部分代码, 将上下文参数、 tasklet部分代码转发到各物理 服务器上进行执行, 整个过程采用异步方式, 通过分布式事件机制发送不同阶段的事件触发 执行代码从而完成信息以及过程的同步。
本发明的有益效果是: 本发明从多种情况下进行任务调度, 实现了调度方式的多样性及 灵活性; 从虚拟化 VCPU映射机制上整体考虑资源均衡调度, 实现了虚拟机的在线迁移、 高可 用以及负载均衡; 从虚拟化 VCPU映射机制上整体考虑省电调度, 实现了虚拟机的经济运行, 达到了节能的效果。
附图说明
图 1为本发明将出现宕机事件的物理服务器上的负载迁移到正常的物理服务器情况下的 方法流程图;
图 2为本发明将偏高的负载迁移到负载偏低的物理服务器情况下的方法流程图; 图 3为本发明将负载较小的物理服务器上的负载迁移到其他物理服务器并将空闲出来的 物理服务器进行休眠或关机情况下的方法流程图。
具体实施方式
下面结合附图进一步详细描述本发明的技术方案, 但本发明的保护范围不局限于以下所 述。
一种云计算平台下的虚拟机分布式任务调度方法, 它包括以下情况: ①当物理服务器宕 机事件时, 将该出现宕机事件的物理服务器上的负载迁移到正常的物理服务器上; ②当某个 物理服务器的负载较高时, 在应用不中断的情况下, 将偏高的负载迁移到负载偏低的物理服 务器上; ③当某个物理服务器的负载较小时, 在应用不中断的情况下, 将该物理服务器上的 负载迁移到其他物理服务器上, 并将空闲出来的物理服务器进行休眠或关机, 降低能耗。
所述主管理服务器和各物理服务器构成一种 server-agent架构体系的分布式调度机制, 主管理服务器和各物理服务器均内嵌脚本解析语言,并实现一种 task-tasklet框架分布式计 算模型, 主管理服务器执行 task部分代码, 将上下文参数、 tasklet部分代码转发到各物理 服务器上进行执行, 整个过程采用异步方式, 通过分布式事件机制发送不同阶段的事件触发 执行代码从而完成信息以及过程的同步。
如图 1所示, 将出现宕机事件的物理服务器上的负载迁移到正常的物理服务器上, 包括 以下具体步骤:
S11 : 主管理服务器和多个位于资源池的物理服务器建立连接保持通讯;
S12: 各物理服务器每隔一段时间 tl, 通过管理网持续的向主管理服务器发送心跳线; S13: 主管理服务器检査各物理服务器的心跳线请求, 若在预设时间 t2后没有收到某物 理服务器的心跳线请求, 发现某物理服务器的心跳异常, 则自动探测该物理服务器所在网络 (应用网) 是否连通, 此时物理服务器会自动的检査应用网是否中断, 若网络中断, 并在一 段时间内没有恢复, 则触发物理服务器宕机事件, 从而启动 HA机制, 自动的将该物理服务器 上的负载转移到其他物理服务器上重新运行, 出现宕机事件的物理服务器被强行停止。
主管理服务器和各物理服务器构成的 server-agent架构, 将资源调度与心跳检査分离, 结构清晰, 更易于管理配置, 并且可以构建超过 100个节点的大型集群, 而传统的一般最大 规模不能超过 32个节点。 另外, 主管理服务器只负责管理资源调度, 即使主管理服务器出现 宕机, 物理服务器也能正常运行, 只是脱离了管理, 不会出现物理服务器崩溃的现象。
如图 2所示, 将偏高的负载迁移到负载偏低的物理服务器上, 包括以下具体步骤:
S21 : 主管理服务器和多个位于资源池的物理服务器建立连接保持通讯;
S22: 主管理服务器实时监控各物理服务器上负载的运行情况, 即监控每个物理服务器上 运行虚拟机状态, 并计算资源池中各物理服务器的负载 LC, 计算公式为:
负载 LC= SUM (虚拟机 VCPU数) /物理服务器 CPU核数;
S23: 对上述步骤计算所得的负载 LC进行排序, 即 SORT(LC);
S24: 判断 MAX(LC)是否大于预设的调度阀值 SF, 若大于预设的调度阀值 SF, 则进入下 一步, 若小于或等于预设的调度阀值 SF, 则调度结束; 一般来说, 虚拟机 VCPU数 /物理服务 器 CPU核数为 3时为物理服务器 CPU利用最佳参数, 小于 1, 则表示有部分物理服务器 CPU 核可能空闲;
S25: 从 MAX(LC)物理服务器中选取 MINV=MIN (虚拟机 VCPU), 判断 "MIN(LC) - MIN (虚拟 机 VCPU) "是否小于或等于调度阀值 SF, 若 "MIN(LC)-MIN (虚拟机 VCPU) 调度阀值 SF", 则 进入下一步, 否则调度结束;
S26: 将选取的虚拟机从 MAX (LC)物理服务器迁移到 MIN(LC)物理服务器上, 并计算迁移 后 MAX (LC)物理服务器及 MIN(LC)物理服务器的负载 LC:
MIN (LC) = (MIN (LC) * 物理服务器 CPU核数 +MINV) /物理服务器 CPU核数;
MAX (LC) = (MIN (LC) * 物理服务器 CPU核数 -MINV) /物理服务器 CPU核数;
然后, 进入步骤 S22, 开始下一轮计算。
如图 3所示, 将负载较小的物理服务器上的负载迁移到其他物理服务器上, 并将空闲出 来的物理服务器进行休眠或关机, 包括以下具体步骤:
S31: 主管理服务器和多个位于资源池的物理服务器建立连接保持通讯;
S32: 主管理服务器实时监控各物理服务器上负载的运行情况, 计算资源池中各物理服务 器的负载 LC, 计算公式为:
负载 LC=SUM (虚拟机 VCPU数) /物理服务器 CPU核数;
S33: 判断 SUM(LC)是否大于 "调度阀值 SF*物理服务器数",若 "SUM(LC) 调度阀值 SF* 物理服务器数"则进入下一步, 若 "SUM(LC)〉调度阀值 SF*物理服务器数", 则先进行如下 操作- 从资源池中选择停止状态的物理服务器, 进入资源均衡调度, 将虚拟机从负载较高的物 理服务器迁移到负载较低的物理服务器上, 直到计算 SUM(LC) 调度阀值 SF*物理服务器 CPU 核数, 或者资源池已无资源为止;
S34: 判断 "5 0<调度阀值5?*(物理服务器数-2)"是否成立, 若 "SUM(LC)<调度 阀值 SF*(物理服务器数 -2) ",则进入下一步,若" SUM(LC) 调度阀值 SF*(物理服务器数 -2) ", 则调度结束; 一般会预留一定容量, 来应对业务突发访问需求, 此处预留容量为 1;
S35: 依次判断负载 LC==0是否成立, 若负载 LC==0, 从集合 (LC) 中移出, 并调度该物 理服务器, 将该物理服务器进行停止操作, 对集合 (LC) 按照计算的负载 LC进行排序, 即 SORT (LC);
S36: 依次从 MIN (LC)物理服务器中选取一个虚拟机 VC=H0ST OF (虚拟机 VCPU), 依次从 集合 (LC)中, 由大到小顺序选取物理服务器计算:
MIN (LC) = (MIN (LC) * 物理服务器 CPU核数 -VC) / 物理服务器 CPU核数;
CHOICE (LC) = (CHOICE (LC) * 物理服务器 CPU核数 +VC) / 物理服务器 CPU核数; 并迁移该虚拟机;
然后进入步骤 S33, 开始下一轮计算。

Claims

权 利 要 求 书
1. 一种云计算平台下的虚拟机分布式任务调度方法, 其特征在于: 它包括以下情况: ① 当物理服务器宕机事件时, 将该出现宕机事件的物理服务器上的负载迁移到正常的物理服务 器上; ②当某个物理服务器的负载较高时, 在应用不中断的情况下, 将偏高的负载迁移到负 载偏低的物理服务器上; ③当某个物理服务器的负载较小时, 在应用不中断的情况下, 将该 物理服务器上的负载迀移到其他物理服务器上,并将空闲出来的物理服务器进行休眠或关机, 降低能耗;
其中, 将出现宕机事件的物理服务器上的负载迁移到正常的物理服务器上, 包括以下具 体步骤:
S11 : 主管理服务器和多个位于资源池的物理服务器建立连接保持通讯;
S12: 各物理服务器每隔一段时间 tl持续的向主管理服务器发送心跳线;
S13: 主管理服务器检査各物理服务器的心跳线请求, 若在预设时间 t2后没有收到某物 理服务器的心跳线请求, 发现某物理服务器的心跳异常, 则自动探测该物理服务器网络是否 连通, 若检测到网络中断, 并在一段时间内没有恢复, 则触发物理服务器宕机事件, 从而启 动 HA机制, 自动的将该物理服务器上的负载转移到其他物理服务器上重新运行;
将偏高的负载迁移到负载偏低的物理服务器上, 包括以下具体步骤:
S21 : 主管理服务器和多个位于资源池的物理服务器建立连接保持通讯;
S22: 主管理服务器实时监控各物理服务器上负载的运行情况, 计算资源池中各物理服务 器的负载 LC, 计算公式为- 负载 LC= SUM (虚拟机 VCPU数) /物理服务器 CPU核数;
S23: 对上述步骤计算所得的负载 LC进行排序, 即 SORT (LC) ;
S24: 判断 MAX (LC)是否大于预设的调度阔值 SF, 若大于预设的调度阀值 SF, 则进入下 一步, 若小于或等于预设的调度阅值 SF, 则调度结束;
S25: 从 MAX (LC)物理服务器中选取 MINV=MIN (虚拟机 VCPU) , 判断 "MIN (LC) - MIN (虚拟 机 VCPU) "是否小于或等于调度阀值 SF, 若 "MIN (LC) -MIN (虛拟机 VCPU) 调度阀值 SF", 则 进入下一步, 否则调度结束;
S26: 将选取的虚拟机从 MAX (IX)物理服务器迁移到 MIN (LC)物理服务器上, 并计算迁移 后 MAX (LC)物理服务器及 MIN (LC)物理服务器的负载 LC:
MIN (LC) = (MIN (LC) *物理服务器 CPU核数 +MINV) /物理服务器 CPU核数;
MAX (LC) = (MIN (LC) *物理服务器 CPU核数 -MINV) /物理服务器 CPU核数;
然后, 进入步骤 S22, 开始下一轮计算:
将负载较小的物理服务器上的负载迁移到其他物理服务器上, 并将空闲出来的物理服务 器进行休眠或关机, 包括以下具体步骤:
S31 : 主管理服务器和多个位于资源池的物理服务器建立连接保持通讯;
S32: 主管理服务器实时监控各物理服务器上负载的运行情况, 计算资源池中各物理服务 器的负载 LC, 计算公式为:
负载 LC=SUM (虚拟机 VCPU数) /物理服务器 CPU核数;
S33: 判断 SUM (LC)是否大于 "调度阀值 SF*物理服务器数", 若 " SUM (LC) 调度阀值 SF* 物理服务器数"则进入下一步, 若 " SUM (LC)〉调度阀值 SF*物理服务器数", 则进行如下操 作- 从资源池中选择停止状态的物理服务器, 进入资源均衡调度, 将虚拟机从负载较高的物 理服务器迁移到负载较低的物理服务器上, 直到计算 SUM (LC) 调度阀值 SF*物理服务器 CPU 核数, 或者资源池已无资源为止;
S34: 判断 " 5 0 <调度阀值5?* (物理服务器数-2) "是否成立, 若 " SUM (LC) <调度 阀值 SF* (物理服务器数 -2) ",则进入下一步,若" SUM (LC) 调度阀值 SF* (物理服务器数 -2) ", 则调度结束;
S35: 依次判断负载 LC==0是否成立, 若负载 LC==0, 从集合 (LC) 中移出, 并调度该物 理服务器, 将该物理服务器进行停止操作, 对集合 (LC) 按照计算的负载 LC进行排序, 即 SORT (LC);
S36: 依次从 MIN (LC)物理服务器中选取一个虚拟机 VC=H0ST OF (虚拟机 VCPU), 依次从 集合 (LC)中, 由大到小顺序选取物理服务器计算:
MIN (LC) = (MIN (LC) * 物理服务器 CPU核数 -VC) / 物理服务器 CPU核数;
CHOICE (LC) = (CHOICE (LC) * 物理服务器 CPU核数 +VC) / 物理服务器 CPU核数; 并迁移该虚拟机;
然后进入步骤 S33, 开始下一轮计算。
2.根据权利要求 1所述的一种云计算平台下的虚拟机分布式任务调度方法,其特征在于: 所述主管理服务器和各物理服务器构成一种 server-agent架构体系的分布式调度机制,主管 理服务器和各物理服务器均内嵌脚本解析语言,并实现一种 task-tasklet框架分布式计算模 型, 主管理服务器执行 task部分代码, 将上下文参数、 tasklet部分代码转发到各物理服务 器上进行执行, 整个过程采用异步方式, 通过分布式事件机制发送不同阶段的事件触发执行 代码从而完成信息以及过程的同步。
PCT/CN2014/079039 2014-03-18 2014-05-31 一种云计算平台下的虚拟机分布式任务调度方法 WO2015139374A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201410099845.0A CN103957231B (zh) 2014-03-18 2014-03-18 一种云计算平台下的虚拟机分布式任务调度方法
CN201410099845.0 2014-03-18

Publications (1)

Publication Number Publication Date
WO2015139374A1 true WO2015139374A1 (zh) 2015-09-24

Family

ID=51334454

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2014/079039 WO2015139374A1 (zh) 2014-03-18 2014-05-31 一种云计算平台下的虚拟机分布式任务调度方法

Country Status (2)

Country Link
CN (1) CN103957231B (zh)
WO (1) WO2015139374A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111078363A (zh) * 2019-12-18 2020-04-28 深信服科技股份有限公司 一种虚拟机的numa节点调度方法、装置、设备及介质
CN111930469A (zh) * 2020-07-20 2020-11-13 湖北美和易思教育科技有限公司 一种基于云计算的高校大数据竞赛管理系统及方法
CN112148462A (zh) * 2020-10-20 2020-12-29 上海新炬网络信息技术股份有限公司 基于Jenkins的CICD流程的处理方法
CN112269632A (zh) * 2020-09-25 2021-01-26 北京航空航天大学杭州创新研究院 一种优化云数据中心的调度方法和系统
CN113434262A (zh) * 2021-08-30 2021-09-24 云宏信息科技股份有限公司 dom0调度方法、虚拟机运行方法、存储介质和计算机
CN114296868A (zh) * 2021-12-17 2022-04-08 中电信数智科技有限公司 一种多云环境下基于用户体验的虚拟机自动迁移决策方法

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105700951B (zh) * 2014-11-25 2021-01-26 中兴通讯股份有限公司 一种实现cpu业务迁移的方法及装置
CN105704107A (zh) * 2014-11-28 2016-06-22 中兴通讯股份有限公司 实现心跳通信的方法、注册中心、服务端和客户端
CN105760240A (zh) * 2014-12-16 2016-07-13 航天信息股份有限公司 分布式任务处理方法及装置
CN105868004B (zh) * 2015-01-23 2020-10-16 南京中兴新软件有限责任公司 一种基于云计算的业务系统的调度方法及调度装置
CN104866375B (zh) * 2015-05-22 2018-05-11 中国联合网络通信集团有限公司 一种迁移虚拟机的方法及装置
CN106331036B (zh) * 2015-06-30 2020-05-26 联想(北京)有限公司 一种服务器控制方法及装置
CN105025104A (zh) * 2015-07-23 2015-11-04 浪潮(北京)电子信息产业有限公司 一种云桌面调度方法和系统
CN105141675B (zh) * 2015-08-10 2019-02-22 浪潮(北京)电子信息产业有限公司 一种多路径访问远程逻辑设备的方法、发送端及系统
CN105262799B (zh) * 2015-09-21 2019-04-23 重庆大学 一种面向性能的云数据中心轮休方法、装置和系统
CN105159752B (zh) * 2015-09-22 2018-03-30 中国人民解放军国防科学技术大学 虚拟化云中机器启动时间感知的实时任务与资源调度方法
CN106900012A (zh) * 2015-12-17 2017-06-27 中兴通讯股份有限公司 基站与虚拟机的绑定方法及装置
CN105912406B (zh) * 2016-05-05 2018-01-12 中国人民解放军国防科学技术大学 一种低能耗的独立任务调度与资源配置方法
CN107835130A (zh) * 2017-12-12 2018-03-23 中国联合网络通信集团有限公司 一种流量分配方法及装置
TWI689823B (zh) * 2018-01-04 2020-04-01 財團法人工業技術研究院 動態工作移轉方法及伺服器
CN109101336A (zh) * 2018-07-20 2018-12-28 深圳市瑞云科技有限公司 一种均匀调度节点机的方法
CN110297692B (zh) * 2019-07-04 2020-12-15 北京伟杰东博信息科技有限公司 一种分布式软件任务动态管理的方法及其系统
CN110347495B (zh) * 2019-07-24 2023-04-28 张�成 一种使用深度强化学习进行移动边缘计算的任务迁移方法
CN115118535B (zh) * 2022-05-25 2023-08-25 成都吉胜科技有限责任公司 一种基于循环责任链的网吧分布式并行计费方法及系统

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102427475A (zh) * 2011-12-08 2012-04-25 曙光信息产业(北京)有限公司 一种云计算环境中负载均衡调度的系统
CN102637138A (zh) * 2012-03-20 2012-08-15 浪潮电子信息产业股份有限公司 一种计算调度虚拟机的方法
CN102981910A (zh) * 2012-11-02 2013-03-20 曙光云计算技术有限公司 虚拟机调度的实现方法和装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9032397B2 (en) * 2008-05-28 2015-05-12 Hewlett-Packard Development Company, L.P. Virtual machine migration with direct physical access control
CN102681895B (zh) * 2011-03-11 2015-03-11 北京市国路安信息技术有限公司 一种动态自迁移云服务方法
CN102932418B (zh) * 2012-09-27 2015-04-15 东软集团股份有限公司 一种云应用部署方法和装置

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102427475A (zh) * 2011-12-08 2012-04-25 曙光信息产业(北京)有限公司 一种云计算环境中负载均衡调度的系统
CN102637138A (zh) * 2012-03-20 2012-08-15 浪潮电子信息产业股份有限公司 一种计算调度虚拟机的方法
CN102981910A (zh) * 2012-11-02 2013-03-20 曙光云计算技术有限公司 虚拟机调度的实现方法和装置

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111078363A (zh) * 2019-12-18 2020-04-28 深信服科技股份有限公司 一种虚拟机的numa节点调度方法、装置、设备及介质
CN111078363B (zh) * 2019-12-18 2024-02-23 深信服科技股份有限公司 一种虚拟机的numa节点调度方法、装置、设备及介质
CN111930469A (zh) * 2020-07-20 2020-11-13 湖北美和易思教育科技有限公司 一种基于云计算的高校大数据竞赛管理系统及方法
CN111930469B (zh) * 2020-07-20 2024-04-09 武汉美和易思数字科技有限公司 一种基于云计算的高校大数据竞赛管理系统及方法
CN112269632A (zh) * 2020-09-25 2021-01-26 北京航空航天大学杭州创新研究院 一种优化云数据中心的调度方法和系统
CN112269632B (zh) * 2020-09-25 2024-02-23 北京航空航天大学杭州创新研究院 一种优化云数据中心的调度方法和系统
CN112148462A (zh) * 2020-10-20 2020-12-29 上海新炬网络信息技术股份有限公司 基于Jenkins的CICD流程的处理方法
CN112148462B (zh) * 2020-10-20 2023-03-28 上海新炬网络信息技术股份有限公司 基于Jenkins的CICD流程的处理方法
CN113434262A (zh) * 2021-08-30 2021-09-24 云宏信息科技股份有限公司 dom0调度方法、虚拟机运行方法、存储介质和计算机
CN114296868A (zh) * 2021-12-17 2022-04-08 中电信数智科技有限公司 一种多云环境下基于用户体验的虚拟机自动迁移决策方法

Also Published As

Publication number Publication date
CN103957231B (zh) 2015-08-26
CN103957231A (zh) 2014-07-30

Similar Documents

Publication Publication Date Title
WO2015139374A1 (zh) 一种云计算平台下的虚拟机分布式任务调度方法
Jin et al. Joint host-network optimization for energy-efficient data center networking
Lin et al. Energy-efficient virtual machine provision algorithms for cloud systems
CN103179048B (zh) 云数据中心的主机QoS策略变换方法及系统
CN103793278B (zh) 一种基于虚拟器件运维规则的资源自动调整方法
CN106059835B (zh) 一种低能耗计算机集群节点的高可靠性控制方法
TW201205441A (en) Multi-CPU domain mobile electronic device and operation method thereof
CN108023958B (zh) 一种基于云平台资源监视的资源调度系统
CN106020934A (zh) 一种基于虚拟集群在线迁移的优化部署方法
CN102662750A (zh) 基于弹性虚拟机池的虚拟机资源优化控制方法及其系统
CN109151045A (zh) 一种分布式云系统及监控方法
CN106603618A (zh) 一种基于云平台的应用弹性伸缩方法
WO2013097396A1 (zh) 一种虚拟化集群整合方法、装置及系统
CN103049332A (zh) 一种虚拟cpu调度方法
CN105975049B (zh) 一种任务同步偶发任务低能耗调度方法
WO2014008867A1 (zh) 虚拟机迁移方法、装置及系统
CN103108039B (zh) 一种低能耗集群环境下的服务质量保证方法
Ranjana et al. A survey on power aware virtual machine placement strategies in a cloud data center
Binder et al. Green computing: Energy consumption optimized service hosting
CN104391736B (zh) 虚拟机的休眠模式设置方法和装置
WO2017148253A1 (zh) 一种节能管理实现方法、装置及网络设备
CN105183563B (zh) 一种面向关键任务计算机的cpu资源动态自配置方法
CN106959885A (zh) 一种虚拟机高可用实现系统及其实现方法
CN103034539A (zh) 伺服器系统及其电源管理方法
Ghoreyshi Energy-efficient resource management of cloud datacenters under fault tolerance constraints

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14886445

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 14/03/2017)

122 Ep: pct application non-entry in european phase

Ref document number: 14886445

Country of ref document: EP

Kind code of ref document: A1