CN112433815A - Cloud data center energy saving method and system based on container control - Google Patents

Cloud data center energy saving method and system based on container control Download PDF

Info

Publication number
CN112433815A
CN112433815A CN202011332616.0A CN202011332616A CN112433815A CN 112433815 A CN112433815 A CN 112433815A CN 202011332616 A CN202011332616 A CN 202011332616A CN 112433815 A CN112433815 A CN 112433815A
Authority
CN
China
Prior art keywords
server
container
containers
optional
utilization
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
Application number
CN202011332616.0A
Other languages
Chinese (zh)
Inventor
徐敏贤
宋承浩
须成忠
叶可江
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Institute of Advanced Technology of CAS
Original Assignee
Shenzhen Institute of Advanced Technology of CAS
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 Shenzhen Institute of Advanced Technology of CAS filed Critical Shenzhen Institute of Advanced Technology of CAS
Priority to CN202011332616.0A priority Critical patent/CN112433815A/en
Priority to PCT/CN2020/139676 priority patent/WO2022110443A1/en
Publication of CN112433815A publication Critical patent/CN112433815A/en
Pending legal-status Critical Current

Links

Images

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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • 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
    • G06F1/3206—Monitoring of events, devices or parameters that trigger a change in power modality
    • 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
    • G06F1/329—Power saving characterised by the action undertaken by task scheduling
    • 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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G06F2009/4557—Distribution of virtual machine instances; Migration and load balancing
    • 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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G06F2009/45575—Starting, stopping, suspending or resuming virtual machine instances
    • 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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G06F2009/45595—Network integration; Enabling network access in virtual machine instances
    • 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)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)
  • Power Sources (AREA)

Abstract

The invention relates to the field of cloud computing energy conservation, in particular to a cloud data center energy-saving method and system based on container control, which comprises the following steps: monitoring whether the server is in an overload state or not in real time; calculating a numerical value of expected reduction in utilization of the server in the overload state if there is the server in the overload state; based on the value of the expected utilization reduction, selecting an optional container deactivation strategy to deactivate the corresponding optional container; the idle server switches to a low power mode to save power consumption. The invention selects the strategy of the optional container deactivation according to the utilization ratio value expected to be reduced, so that the corresponding optional container is deactivated, and the standby server is switched to a low power consumption mode by the deactivation of the optional container so as to save power consumption; according to the application, the server can dynamically close the optional container according to the running state, so that the pressure of the server is relieved under the condition of overload, the energy-saving effect is achieved, and the purpose of reducing the total energy consumption of the cloud data center is achieved.

Description

Cloud data center energy saving method and system based on container control
Technical Field
The invention relates to the field of cloud computing energy conservation, in particular to a cloud data center energy conservation method and system based on container control.
Background
With the development of the information technology industry, the cloud computing technology is used as an important component of the information technology, and is widely applied to various fields in life due to the characteristics of a pay-as-you-go pricing mode, low operation cost, high expansibility, easy accessibility and the like. However, the enormous energy consumption and the amount of carbon emissions generated by cloud data centers have attracted extensive attention from researchers. The high energy consumption of the cloud computing data center also becomes a constraint restricting the cloud computing development, and the existing energy-saving technologies such as virtual machine migration and dynamic voltage frequency adjustment are difficult to generate effective effects when the data center is in an overall overload state.
Finding an effective method for reducing energy consumed by a cloud computing center server has become a main target of research in the field of cloud computing energy conservation at present. The energy consumption of the cloud server is reduced, so that the operation cost of the server can be reduced, and the reliability of the whole system can be improved. Currently, the mainstream methods for reducing energy consumption of the existing cloud data center include virtual machine migration (VM coherence) and Dynamic Voltage Frequency Scaling (DVFS).
(1) Virtual machine migration (VM consistency): to an energy saving method for minimizing energy consumption by distributing tasks among fewer machines while shutting down unused machines. By using the strategy, the cloud computing server can cause the virtual machines working on the underutilized servers to be migrated to other servers, and the spare servers enter a low-energy consumption mode or are shut down.
(2) Dynamic voltage frequency adjustment (DVFS): the energy-saving method is a method for balancing the computing performance and the energy consumption of the server according to the load of the current state. For example, DVFS techniques reduce frequency and voltage during light processor load to achieve reduced power consumption. While increasing frequency and voltage when the machine is heavily loaded.
The existing method is mainly aimed at the problem of optimizing the energy consumption of the cloud server with the virtual machine as the granularity; moreover, when the entire cloud server is in an overload state, the existing methods such as virtual machine migration (VM coherence) and Dynamic Voltage Frequency Scaling (DVFS) cannot effectively reduce the energy consumption of the cloud server.
Disclosure of Invention
The embodiment of the invention provides a cloud data center energy-saving method and system based on container control, which can relieve server pressure under the condition of overload and play a role in energy saving according to an operation state, and achieve the purpose of reducing the total energy consumption of a cloud data center.
According to an embodiment of the invention, a cloud data center energy saving method based on container control is provided, which includes the following steps:
monitoring whether the server is in an overload state or not in real time;
calculating a numerical value of expected reduction in utilization of the server in the overload state if there is the server in the overload state;
based on the value of the expected utilization reduction, selecting an optional container deactivation strategy to deactivate the corresponding optional container;
the idle server switches to a low power mode to save power consumption.
Further, before monitoring whether the server is in an overload state, the method further comprises:
predicting the workload of the server based on the historical data;
and adjusting the number of the servers performing work according to the predicted work load.
Further, monitoring whether the server is in an overload state comprises:
presetting an overload threshold value for judging whether the server is overloaded or not;
detecting the utilization rate of a server;
comparing the utilization rate with an overload threshold;
if the utilization is above the overload threshold, the server is deemed to be in an overloaded state.
Further, the number of servers in the overload state is calculated according to an overload server calculation formula:
the overload server has the calculation formula as follows:
Figure BDA0002796248850000031
Figure BDA0002796248850000032
wherein the content of the first and second substances,
Figure BDA0002796248850000033
i is the server number, uiFor the utilization of server i, TuIs an overload threshold; if u isiNot less than TuThen, then
Figure BDA0002796248850000034
Is 1, otherwise is 0; n is0The total number of servers in the server cluster that are in an overloaded state.
Further, in calculating the value of the expected reduction in utilization of the server in the overload state includes:
firstly, the value of the adjustor of the server in the overload state is calculated based on an adjustor calculation formula, wherein the adjustor calculation formula is as follows:
Figure BDA0002796248850000035
where t represents the current time, θtIs the value of the adjuster when the time is t, and n is the number of servers;
then calculating a value of expected utilization reduction according to a utilization calculation formula based on the value of the regulator;
the utilization calculation formula is as follows:
Figure BDA0002796248850000036
wherein the content of the first and second substances,
Figure BDA0002796248850000037
is a value where a reduction in utilization is expected.
Further, the selectable container deactivation strategy comprises a container priority strategy with the lowest utilization rate, and the container priority strategy with the lowest utilization rate is to reduce the utilization rate of the overload host by selecting a group of selectable containers to deactivate based on a deactivation calculation formula, so that the reduced utilization rate is lower than an overload threshold;
the deactivation calculation formula is:
Figure BDA0002796248850000041
wherein u'iIs the utilization factor of the server, u 'after optional Container deactivation'iIs equal to
Figure BDA0002796248850000042
Figure BDA0002796248850000043
Is shown as
Figure BDA0002796248850000044
Minimum absolute value of (d), dcliA list formed for optional containers that are deactivated;
the optional container deactivation strategy further comprises the following steps:
a minimum number of containers prioritization policy that selects the least number of containers to disable, such that fewer containers can be disabled while achieving power savings, so as to provide more optional functionality;
and a random container selection strategy, namely randomly selecting a plurality of containers to be deactivated to achieve the aim of reducing energy consumption based on a random minimum number of container priority strategies.
According to another embodiment of the present invention, there is provided a cloud data center energy saving system based on container policing, including: the system comprises a cloud service storage library, an execution environment and a regulation center; the cloud service storage library comprises optional containers, and the optional containers can be deactivated or activated according to the running state of the servers; the execution environment provides an environment for the operation of the cloud service storage library, and the regulation and control center comprises a Brown out controller, a system monitor and a scheduling policy manager;
the system monitor is used for monitoring the operation condition of the server and collecting the state of the server workload;
the scheduling policy manager is used for providing a policy for controlling the deactivation or opening of the optional container for the Brown out controller;
the Brownout controller controls the deactivation or opening of the optional container according to the operation state of the server.
Furthermore, the Brown out controller comprises a regulator, and the value of the regulator is calculated according to the number of overloaded servers; the regulator controls the deactivation or opening of an optional container of the overload server;
further, the scheduling policy manager includes:
the container priority strategy with the lowest utilization rate is used for reducing the utilization rate of the overload host by selecting a group of containers and deactivating the containers based on a deactivation calculation formula, so that the reduced utilization rate is lower than an overload threshold; the optional containers which are stopped are brought into a stopping list through a stopping calculation formula, and the stopping list sorts the optional containers based on ascending utilization rate;
a minimum number of containers prioritization policy that selects the least number of containers to disable, such that fewer containers can be disabled while achieving power savings, so as to provide more optional functionality;
and a random container selection strategy, namely randomly selecting a plurality of containers to be deactivated to achieve the aim of reducing energy consumption based on a random minimum number of container priority strategies.
Further, the cloud service storage library also comprises necessary containers, and the necessary containers are always kept in a running state when being started and cannot be stopped.
The invention has the beneficial effects that: whether a server in work is in a load condition or not is monitored in real time, if the server in the overload condition exists, the expected reduction quantity of the utilization rate of the server in the overload condition is calculated, a strategy for deactivating the optional containers is selected according to the expected reduction utilization rate value, the corresponding optional containers are deactivated, and the overloaded server is switched to a low power consumption mode by deactivating the optional containers so as to save power consumption; according to the application, the server can dynamically close the optional container according to the running state, so that the pressure of the server is relieved under the condition of overload, the energy-saving effect is achieved, and the purpose of reducing the total energy consumption of the cloud data center is achieved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
FIG. 1 is a flowchart of a cloud data center energy saving method based on container policing according to the present invention;
FIG. 2 is a flow chart of the present invention for monitoring whether a server is in an overloaded state;
fig. 3 is a schematic block diagram of a cloud data center energy saving system based on container policing according to the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Example 1
According to an embodiment of the present invention, a cloud data center energy saving method based on container policing is provided, referring to fig. 1 and fig. 2, including the following steps:
s101: and monitoring whether the server is in an overload state or not in real time.
S102: if there is a server in an overload state, a value is calculated that the utilization of the server expected to be in the overload state is reduced.
S103: based on the value of the expected utilization reduction, an optional container deactivation strategy is selected to deactivate the corresponding optional container.
S104: the overloaded server switches to a low power mode to save power.
In the embodiment, whether a server in work is in a load condition or not is monitored in real time, if the server in the overload condition exists, the number of expected reductions of the utilization rate of the server in the overload condition is calculated, a strategy for stopping the selectable containers is selected according to the utilization rate value expected to be reduced, the corresponding selectable containers are stopped, and the overload server is switched to a low power consumption mode to save power consumption due to the stop of the selectable containers; according to the application, the server can dynamically close the optional container according to the running state, so that the pressure of the server is relieved under the condition of overload, the energy-saving effect is achieved, and the purpose of reducing the total energy consumption of the cloud data center is achieved.
In the embodiment, the containers are divided into the necessary containers and the optional containers, and the necessary containers are always kept in a running state when being started and cannot be stopped; for example, the container associated with the database will be marked as a necessary service.
The optional containers may temporarily deactivate the containers based on the state of the server. If the containers are communicated with each other, the connection relation exists between the containers. By definition, if an optional service is disabled, other services connected to this optional service (optional container) will also be placed in a disabled state. If the service or content provided by the container provider is defined as optional by its creator, it may be identified as an optional service. For example, an online recommendation engine in an online store system and a spell checker in an online editor system may be set as optional services in the event of limited or overloaded resources.
In a preferred technical solution, before monitoring whether the server is in an overload state, the method further includes:
the workload of the server is predicted based on the historical data.
And adjusting the number of the servers performing work according to the predicted work load.
In this embodiment, a slipping windows algorithm is used to predict the future workload based on the existing workload. The sliding window algorithm is as follows:
setting the size L of the sliding windowwAt a time interval of LwThe number of requests in, t is the time interval to be predicted (t ≧ L)w)；
Figure BDA0002796248850000071
A predicted value output for the algorithm requesting the number of times within the time interval t; k represents time when k is at t-LwAnd between t-1
Figure BDA0002796248850000072
Up to
Figure BDA0002796248850000073
Time return
Figure BDA0002796248850000074
The value of (c).
Sliding motionThe window gives more weight to the request rate separated by the most recent time interval; let the sliding window size LwAre constant integer values. num (k) is the actual number of requests at time interval k, with the number of requests within time interval t as the sliding window LwThe average value of the request in (1) is shown in the following formula:
Figure BDA0002796248850000081
to ensure that enough historical data is available for prediction, the time interval for which prediction is requested should not be less than the sliding window size LwThe time within the time interval, i.e. t ≧ Lw. Depending on the predicted workload, the number of servers in operation may be dynamically adjusted.
In this embodiment, the monitoring whether the server is in the overload state includes:
s201: and presetting an overload threshold value for judging whether the server is overloaded or not.
S202: and detecting the utilization rate of the server.
S203: the utilization is compared to an overload threshold.
S204: if the utilization is above the overload threshold, the server is deemed to be in an overloaded state.
In this embodiment, the utilization rate of the server is measured by a Central Processing Unit (CPU); for example, the overload threshold is set to 85%, and when the CPU utilization of the server is detected to be above 85%, the server will be considered to be in an overloaded state.
In this embodiment, the number of servers in an overload state is calculated according to an overload server calculation formula, where the overload server calculation formula is:
Figure BDA0002796248850000082
Figure BDA0002796248850000083
wherein the content of the first and second substances,
Figure BDA0002796248850000084
i is the server number, uiFor utilization of servers, TuIs an overload threshold; if u isiNot less than TuThen, then
Figure BDA0002796248850000085
Is 1, otherwise is 0; n is0Is the number of total servers in the server cluster that are in an overloaded state.
In this embodiment, the calculating the value of the utilization reduction utilization expected of the server in the overload state includes:
firstly, the value of the adjustor of the server in the overload state is calculated based on an adjustor calculation formula, wherein the adjustor calculation formula is as follows:
Figure BDA0002796248850000091
where t represents the current time, θtIs the value of the adjuster when the time is t, and n is the number of servers;
then calculating a value of expected utilization reduction according to a utilization calculation formula based on the value of the regulator;
the utilization calculation formula is as follows:
Figure BDA0002796248850000092
wherein the content of the first and second substances,
Figure BDA0002796248850000093
is a value where a reduction in utilization is expected.
In this embodiment, the optional Container deactivation policy includes a least-utilized Container priority policy (LUCF), where the least-utilized Container priority policy is to reduce the Utilization rate of the overloaded host by selecting a group of optional containers to deactivate based on a deactivation calculation formula, and make the reduced Utilization rate lower than an overload threshold;
the deactivation calculation formula is:
Figure BDA0002796248850000094
wherein u'iIs the utilization factor of the server, u 'after optional Container deactivation'iIs equal to
Figure BDA0002796248850000095
Figure BDA0002796248850000096
Is shown as
Figure BDA0002796248850000097
Minimum absolute value of (d), dcliIs an optional container for deactivation.
After the selectable containers are deactivated, the container priority strategy with the lowest utilization rate carries out an ordered list according to the calculated selectable containers in ascending order, so that the selectable containers with the lowest utilization rate are positioned at the beginning of the list; the specific sorting process is as follows:
dcliis dcl, the size of the list ofisize (); the algorithm checks the work servers one by one if the first optional container c in server i0Ratio of utilization
Figure BDA0002796248850000101
If so, c is0Placed in an optional container deactivation list. Since c also needs to be considered0Containers connected, so the policy will also hold all with c0The optional containers connected are all marked as being in the data set
Figure BDA0002796248850000102
(
Figure BDA0002796248850000103
A set of data) to describe how it is connected to other optional containers. However, if the first optional container c0Is less than the expected reduction in utilization, the LUCF policy may find a sub-list of optional containers to retire more optional containers. The sub-list is the list of the sum of the other utilization rates that are closest to the expected utilization reduction. These alternative containers are placed in the list of deactivated alternative containers and their connection tags are placed in the collection as previously described. Other connected optional containers can then be found according to the algorithm and put in the list of disabled optional containers.
In this embodiment, the optional container deactivation policy further includes:
a minimum number of containers prioritization policy that selects the least number of containers to disable, such that fewer containers can be disabled while achieving power savings, so as to provide more optional functionality;
and a random container selection strategy, namely randomly selecting a plurality of containers to be deactivated to achieve the aim of reducing energy consumption based on a random minimum number of container priority strategies.
In particular, in order to deactivate fewer Containers in order to provide more optional functionality, a Minimum Number of container priority policies (MNCF) are implemented that select a Minimum Number of Containers while saving power consumption. Very similar to LUCF;
as shown in the formula:
Figure BDA0002796248850000104
wherein min (dcl)isize ()) represents the size of the smallest list of disabled containers.
In particular, randomly selected dcliUniformly distributed discrete Random variables of the subset, Random container selection strategy (RCS) uses a uniformly distributed function U (0, dcl)isize () -1) randomly selecting a plurality of selectable containers to achieve the purpose of reducing energy consumption;
as shown in the formula:
Figure BDA0002796248850000111
example 2
According to another embodiment of the present invention, there is provided a cloud data center energy saving system based on container policing, referring to fig. 3, including: the system comprises a cloud service storage library, an execution environment and a regulation center; the cloud service storage library comprises optional containers, and the optional containers can be deactivated or activated according to the running state of the servers; the execution environment provides an environment for the operation of the cloud service storage library, and the regulation and control center comprises a Brown out controller, a system monitor and a scheduling policy manager;
the system monitor is used for monitoring the operation condition of the server and collecting the state of the server workload;
the scheduling policy manager is used for providing a policy for controlling the deactivation or opening of the optional container for the Brown out controller;
the Brownout controller controls the deactivation or opening of the optional container according to the operation state of the server.
In the embodiment, the running state of the server is monitored in real time through a system monitor, for example, whether the server has a working load or not is judged; if a server with load exists, the scheduling policy manager provides a policy for controlling the deactivation of the optional container for the Brown out controller; the Brown out controller controls the deactivation of the optional container according to the strategy; deactivation of the optional container switches the overloaded server to a low power mode to save power consumption; according to the application, the server can dynamically close the optional container according to the running state, so that the pressure of the server is relieved under the condition of overload, the energy-saving effect is achieved, and the purpose of reducing the total energy consumption of the cloud data center is achieved.
In this embodiment, the system monitor is a component that monitors the operating conditions of the server and collects the resource consumption states of the server. It uses third party toolkits to support its functionality, such as public APIs (public application programming interfaces) in Grid' 5000, which provide real-time data about the infrastructure, including server running status, CPU (central processing unit) utilization and power consumption.
In this embodiment, the scheduling policy manager provides and manages policies for the Brownout controller to schedule the optional containers. To ensure energy budget and quality of service constraints, different strategies need to be designed for different preferences. For example, if a cloud service provider wants to balance the trade-off between energy consumption and quality of service, a trade-off policy is preferred.
In this embodiment, the Brownout controller includes a regulator, and the value of the regulator is calculated according to the number of overloaded servers; the Brownout controller controls the disabling or enabling of the optional containers of the overloaded server depending on the value of the regulator.
The value of the regulator is calculated based on a regulator calculation formula, which is:
Figure BDA0002796248850000121
where t represents the current time, θtIs the value of the adjuster when the time is t, and n is the number of servers;
calculating a value of expected utilization reduction according to a utilization calculation formula based on the value of the governor; the utilization calculation formula is as follows:
Figure BDA0002796248850000122
wherein the content of the first and second substances,
Figure BDA0002796248850000123
is a value where a reduction in utilization is expected.
In this embodiment, an optional container deactivation strategy is selected based on the value of expected utilization reduction; the scheduling policy manager includes: the method comprises three optional container deactivation strategies, namely a container priority strategy with the lowest utilization rate, a container priority strategy with the least quantity and a random container selection strategy.
The first method comprises the following steps: a container priority strategy with the lowest utilization rate, wherein the utilization rate of the overloaded host is reduced by selecting a group of containers and stopping the containers based on a stopping calculation formula, and the reduced utilization rate is lower than an overload threshold value; calculating by using a deactivation calculation formula and summarizing deactivated optional containers into a deactivation list, wherein the deactivation list sorts the optional containers in an ascending order based on utilization rate;
the deactivation calculation formula is:
Figure BDA0002796248850000131
wherein u'iIs the utilization factor of the server, u 'after optional Container deactivation'iIs equal to
Figure BDA0002796248850000132
Figure BDA0002796248850000133
Is shown as
Figure BDA0002796248850000134
Minimum absolute value of (d), dcliA list formed for optional containers that are deactivated.
After the optional container column is deactivated, the least utilized container prioritization policy sorts the list according to the alternative containers to be calculated in ascending order so that the least utilized optional container is at the beginning of the list.
And the second method comprises the following steps: a minimum number of containers prioritization policy that selects the least number of containers to disable, such that fewer containers can be disabled while achieving power savings, so as to provide more optional functionality;
and the third is that: and a random container selection strategy, namely randomly selecting a plurality of containers to be deactivated to achieve the aim of reducing energy consumption based on a random minimum number of container priority strategies.
In this embodiment, the cloud service storage further includes a necessary container, and the necessary container is always kept in an operating state when being started and cannot be stopped.
Optional containers may temporarily deactivate these containers based on server status. If the containers are communicated with each other, the connection relation exists between the containers. By definition, if an optional service is disabled, other services connected to the optional service will also be placed in a disabled state. If the service or content provided by the container provider is defined as optional by its creator, it may be identified as an optional service. For example, an online recommendation engine in an online store system and a spell checker in an online editor system may be set as optional services in the event of limited or overloaded resources.
In this embodiment, the control center further includes a model manager, which is a model for maintaining energy consumption and service quality in the server;
in this embodiment, the execution environment provides an operating environment for the cloud service repository; the main execution environments for the common case containers are Docker, Kubernets and Mesos. In the present application, Docker is chosen as the execution environment for the optional container.
In this embodiment, the system is built on GRID' 5000 as a cloud infrastructure.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims (10)

1. A cloud data center energy-saving method based on container control is characterized by comprising the following steps:
monitoring whether the server is in an overload state or not in real time;
calculating a numerical value of expected reduction in utilization of a server in an overload state if the server in the overload state exists;
based on the value of the expected utilization reduction, selecting an optional container deactivation strategy to deactivate the corresponding optional container;
the overloaded server switches to a low power mode to save power.
2. The cloud data center energy saving method based on container policing according to claim 1, wherein before the monitoring whether the server is in an overload state in real time further comprises:
predicting a workload of the server based on historical data;
and adjusting the number of the servers performing work according to the predicted work load.
3. The cloud data center energy saving method based on container policing according to claim 1, wherein the monitoring whether the server is in an overload state comprises:
presetting an overload threshold value for judging whether the server is overloaded or not;
detecting the utilization rate of the server;
comparing the utilization rate to the overload threshold;
if the utilization is above the overload threshold, the server is deemed to be in an overload state.
4. The cloud data center energy saving method based on container policing of claim 3, wherein the number of servers in an overload state is calculated according to an overload server calculation formula:
the overload server has the calculation formula as follows:
Figure FDA0002796248840000021
Figure FDA0002796248840000022
wherein the content of the first and second substances,
Figure FDA0002796248840000023
for watchIndicating whether the server is in an overload state, i is a server number, uiFor the utilization of server i, TuIs the overload threshold; if u isiNot less than TuThen, then
Figure FDA0002796248840000024
Is 1, otherwise is 0; n is0Is the total number of the server clusters that are in an overload state.
5. The container policing-based cloud data center energy-saving method of claim 4, wherein the calculating the value of the expected reduction in utilization of the server in the overload state comprises:
first, a value of a governor of the server in an overload state is calculated based on a governor calculation formula, which is:
Figure FDA0002796248840000025
where t represents the current time, θtIs the value of the adjuster when the time is t, and n is the number of servers;
then calculating a value for which the utilization reduction is expected according to a utilization calculation formula based on the value of the regulator;
the utilization calculation formula is as follows:
Figure FDA0002796248840000026
wherein the content of the first and second substances,
Figure FDA0002796248840000027
is a value where a reduction in utilization is expected.
6. The cloud data center energy saving method based on container policing according to claim 5, wherein the selectable container deactivation policies include a container priority policy with lowest utilization rate, and the container priority policy with lowest utilization rate is based on a deactivation calculation formula, and is used for reducing utilization rate of an overloaded host by selecting a group of the selectable container deactivations, so that the reduced utilization rate is lower than the overload threshold;
the deactivation calculation formula is:
Figure FDA0002796248840000031
wherein u'iIs optional Container post-deactivation Server utilization, u'iIs equal to
Figure FDA0002796248840000032
Is shown as
Figure FDA0002796248840000033
Minimum absolute value of (d), dcliA list formed for optional containers that are deactivated;
the optional container deactivation strategy further comprises:
the minimum number of containers is a priority strategy, and the strategy selects the containers with the minimum deactivation, so that less containers can be deactivated to provide more optional functions while achieving power saving;
and a random container selection strategy, namely randomly selecting a plurality of containers to be deactivated to achieve the aim of reducing energy consumption based on a random minimum number of container priority strategies.
7. A cloud data center energy-saving system based on container control is characterized by comprising: the system comprises a cloud service storage library, an execution environment and a regulation center; the cloud service storage library comprises optional containers which can be deactivated or activated according to the running state of the server; the execution environment provides an environment for the operation of the cloud service storage library, and the regulation and control center comprises a Brown out controller, a system monitor and a scheduling policy manager;
the system monitor is used for monitoring the operation condition of the server and collecting the state of the server workload;
the scheduling policy manager is used for providing a policy for controlling the deactivation or opening of the optional container for the Brown out controller;
the Brownout controller controls the deactivation or opening of the optional container according to the operation state of the server.
8. The cloud data center energy-saving system based on container policing of claim 7, wherein the Brownout controller comprises a regulator, and the value of the regulator is calculated according to the number of overloaded servers; the regulator controls the deactivation or opening of the optional container of the overloaded server.
9. The cloud data center energy saving system based on container policing of claim 8, wherein the scheduling policy manager comprises:
a least utilized container priority policy that reduces utilization of overloaded hosts by selecting a set of containers and disabling the selected containers based on a disabling calculation formula such that the reduced utilization is below the overload threshold; including the optional containers that are disabled into a disable list by a disable calculation formula, the disable list ordering the optional containers based on the utilization ascending;
the minimum number of containers is a priority strategy, and the strategy selects the containers with the minimum deactivation, so that less containers can be deactivated to provide more optional functions while achieving power saving;
the random container selection strategy randomly selects a plurality of containers to be deactivated to achieve the purpose of reducing energy consumption based on a random minimum container priority strategy.
10. The cloud data center energy saving system based on container policing of claim 7, wherein the cloud service repository further comprises necessary containers, and the necessary containers will always keep running and cannot be stopped when being started.
CN202011332616.0A 2020-11-24 2020-11-24 Cloud data center energy saving method and system based on container control Pending CN112433815A (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202011332616.0A CN112433815A (en) 2020-11-24 2020-11-24 Cloud data center energy saving method and system based on container control
PCT/CN2020/139676 WO2022110443A1 (en) 2020-11-24 2020-12-25 Cloud data center energy saving method and system based on container control

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011332616.0A CN112433815A (en) 2020-11-24 2020-11-24 Cloud data center energy saving method and system based on container control

Publications (1)

Publication Number Publication Date
CN112433815A true CN112433815A (en) 2021-03-02

Family

ID=74694043

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011332616.0A Pending CN112433815A (en) 2020-11-24 2020-11-24 Cloud data center energy saving method and system based on container control

Country Status (2)

Country Link
CN (1) CN112433815A (en)
WO (1) WO2022110443A1 (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110099403A1 (en) * 2009-10-26 2011-04-28 Hitachi, Ltd. Server management apparatus and server management method
CN108170522A (en) * 2017-12-06 2018-06-15 南京邮电大学 A kind of cloud computing virtual machine (vm) migration control method based on dynamic threshold
CN109901932A (en) * 2019-03-12 2019-06-18 东北大学 A kind of Server Consolidation method based on virtual machine
CN109981396A (en) * 2019-01-22 2019-07-05 平安普惠企业管理有限公司 The monitoring method and device, medium and electronic equipment of docker service container cluster

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103605418B (en) * 2013-10-23 2017-01-04 曙光信息产业(北京)有限公司 The regulating power consumption method and apparatus of cluster server
CN110347498B (en) * 2019-06-10 2022-12-16 华南理工大学 Load dynamic migration method in container and virtual machine mixed cloud environment

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110099403A1 (en) * 2009-10-26 2011-04-28 Hitachi, Ltd. Server management apparatus and server management method
CN108170522A (en) * 2017-12-06 2018-06-15 南京邮电大学 A kind of cloud computing virtual machine (vm) migration control method based on dynamic threshold
CN109981396A (en) * 2019-01-22 2019-07-05 平安普惠企业管理有限公司 The monitoring method and device, medium and electronic equipment of docker service container cluster
CN109901932A (en) * 2019-03-12 2019-06-18 东北大学 A kind of Server Consolidation method based on virtual machine

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
MIN XIAN XU: "iBrownout :An Integrated Approach for Managing Energy and Brownout in Container-Based Clouds", 《IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING》 *

Also Published As

Publication number Publication date
WO2022110443A1 (en) 2022-06-02

Similar Documents

Publication Publication Date Title
US10212220B2 (en) Self-adaptive control system for dynamic capacity management of latency-sensitive application servers
US8104041B2 (en) Computer workload redistribution based on prediction from analysis of local resource utilization chronology data
US8225119B2 (en) Energy-aware server management
CN103473139B (en) Cluster virtual machine resource allocation scheduling method
US8001403B2 (en) Data center power management utilizing a power policy and a load factor
US8381215B2 (en) Method and system for power-management aware dispatcher
US8185893B2 (en) Starting up at least one virtual machine in a physical machine by a load balancer
US7870241B2 (en) Automated power control policies based on application-specific redundancy characteristics
Wang et al. Power optimization with performance assurance for multi-tier applications in virtualized data centers
US20050060590A1 (en) Power-aware workload balancing usig virtual machines
CN113672383A (en) Cloud computing resource scheduling method, system, terminal and storage medium
CN102662750A (en) Virtual machine resource optimal control method and control system based on elastic virtual machine pool
US20110161707A1 (en) Power management of computers
CN111090505B (en) Task scheduling method and system in multiprocessor system
CN102208986B (en) Cluster power consumption distribution and control method
CN112380005A (en) Data center energy consumption management method and system
Jiang et al. Adaptive thresholds determination for saving cloud energy using three-way decisions
Jiang et al. An energy-aware virtual machine migration strategy based on three-way decisions
CN114880129A (en) Resource utilization rate prediction perception cloud data center virtual machine resource allocation method
CN112433815A (en) Cloud data center energy saving method and system based on container control
Kant et al. Enhancing data center sustainability through energy-adaptive computing
Wu et al. Energy-Efficient Virtual Machine Migration Algorithm in Data Centers
Huixi et al. A combination of host overloading detection and virtual machine selection in cloud server consolidation based on learning method
Naeen Cost reduction using SLA-aware genetic algorithm for consolidation of virtual machines in cloud data centers
Zhu et al. Multi-Objective Scheduling of Cloud Data Centers Prone to Failures.

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