CN108182105B - Local dynamic migration method and control system based on Docker container technology - Google Patents
Local dynamic migration method and control system based on Docker container technology Download PDFInfo
- Publication number
- CN108182105B CN108182105B CN201711322296.9A CN201711322296A CN108182105B CN 108182105 B CN108182105 B CN 108182105B CN 201711322296 A CN201711322296 A CN 201711322296A CN 108182105 B CN108182105 B CN 108182105B
- Authority
- CN
- China
- Prior art keywords
- container
- migration
- server
- utilization rate
- resource
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
- G06F9/505—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the load
-
- 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
-
- 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)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Hardware Redundancy (AREA)
- Debugging And Monitoring (AREA)
Abstract
A local dynamic migration method and a control system based on a Docker container technology belong to the technical field of cloud computing. The problem of the wasting of resources that exists among the current container migration technique is solved. The method is characterized in that occurrence of an overload server is judged according to node resource utilization rate and predicted resource utilization rate in the last n sampling periods, container migration is triggered, and the selection of a container to be migrated in the overload server is the container with the higher total utilization rate but smaller memory, so that the effect of reducing migration overhead while balancing loads is achieved. The target server is chosen to first ensure that there are enough resources available in the target server. For the case that a new container is generated in the migration process, and overload is caused to the target server in parallel, it means that there is no server capable of accommodating the container in all servers of the cluster, so that it is not necessary to migrate the container, and replace the container to be migrated. The application improves the resource utilization rate and reduces the task execution time.
Description
Technical Field
The application relates to a Docker container migration technology, and belongs to the technical field of cloud computing.
Background
With the wide application of cloud computing technology, the scale of cloud services is gradually expanded, and data center networks are continuously expanded. In order to realize load balancing in the cloud service platform, how to effectively allocate computing resources, a resource migration technology becomes a research hotspot. By load balancing is meant that requests/data are evenly spread out over multiple units of operation.
In the traditional mode, the core of cloud computing is virtualization, and the utilization rate of resources is improved mainly through a virtual machine. The container (Docker) is an open source engine that can easily create a lightweight, portable, self-sufficient container for any application, is an emerging virtualization tool in recent years that can achieve isolation of resources and system environments as virtual machines, and has the advantages of light weight, high performance and convenience. Different from the traditional virtual machine, the container elastically distributes computing resources in a mode of sharing the host kernel, so that the resource utilization rate is improved. With respect to the operating system hierarchy, the container corresponds to the set of processes running therein. Because of the close association of the container and the kernel, the computing resources and state information in the migration process are dispersed among the operating system and the kernel, which provides challenges for ensuring load balancing in the migration process.
The container provides process-based isolation relative to conventional virtual machines. The creation, start, stop and destruction of the container can be achieved at a millisecond rate. Since the lifecycle of the container is only one-fourth of that of the virtual machine, it is an NP-hard problem to implement load balancing globally, the present application considers the dynamic migration of the container from a locally balanced state. The migration of the container is classified into static migration and dynamic migration. Static migration is to immediately close the container of the source server, then create a new container in the target server, and migrate the container state of the source server to the new container of the target server. In static migration, the entire cloud service will be suspended, thereby causing a large economic loss. Live migration, also called live migration, refers to the process of moving applications between different physical computers or cloud platforms with no interruption in client access ensured, and the entire migration process is imperceptible to the user. The dynamic migration is to perform information collaboration between the source server and the target server after determining the two servers, thereby determining the state information of the container to be migrated. Currently, the dynamic migration of a container refers to the mode of dynamic migration of a virtual machine, and is mainly divided into two types, namely, a pre-copy memory (pre-copy memory) and a post-copy memory (post-copy memory).
The research of Virtuozzo on container hot migration technology by the Virtuozzo team can provide a container engine for production environment and realizing container hot migration, and the disadvantage is that the technology can only be oriented to a Linux kernel customized by the Virtuozzo. The container thermomigration function can be completed through modification or patching of the kernel, and the thermomigration can not be realized by facing the mainstream Linux kernel.
The main disadvantages of the existing container migration technology are:
(1) Container internal state information acquisition problem: the container memory is distributed among the processes, corresponding memory information needs to be sequentially acquired according to the process information, the state information and the resource objects which are related to the container and need to be stored in the freezing step are quite complex, about 1000 object information needs to be stored, the cores of the object information do not provide corresponding interfaces for direct acquisition, and the virtual machine only needs to store about 70 object information and can directly acquire the object information.
(2) Uncertainty of data volume: the programs contained in the containers may vary widely in the amount of resources required due to the different computing tasks. After a large number of calculations are performed inside the container, a large amount of random data is generated quickly, thereby greatly affecting the storage space of the container.
(3) Relationship between processes: because nested parent-child relationships exist between groups of processes between containers, the hot migration process needs to maintain relationships between processes, so the migration problem of process subtrees needs to be solved,
(4) Waste of resources: the lifecycle management of the container is not considered, which not only causes the waste of computing resources, but also causes the loss of user data after the container is consumed.
Disclosure of Invention
The application solves the problems existing in the prior container migration technology.
In order to solve the above problems, the present application provides a local dynamic migration method and a control system based on a Docker container technology, wherein:
the local dynamic migration method based on the Docker container technology comprises the following steps of
Step 1, according to the resource utilization rate of each node in the cluster nodes, monitoring the resource utilization condition of each node in the cluster nodes, determining whether hot spots exist, and triggering and starting the following steps to perform container migration when the hot spots exist;
step 2, each server in the statistical cluster integrates the utilization parameters volume of different resources,
in the formula, mem represents the memory utilization rate of a server or a container, CPU represents the utilization rate of a CPU, and net represents the bandwidth utilization rate; the N servers in the cluster are ordered in descending order according to the volume value;
step 3, aiming at the overload server with the maximum volume value, the following processing is carried out:
ordering all containers inside it in descending order of VSR value, the vsr=volume/size, wherein size represents the memory size of the container;
sequentially performing container migration according to the sequence of the VSR values from large to small until the utilization rate of all resources in the overload server with the maximum volume value is lower than a corresponding threshold value, and completing container migration;
and returning to the execution of the step 1.
The method for determining whether the hot spot exists in the step 1 is that the following processing is performed for the resource utilization rate of each node:
predicting the utilization rate at the next moment according to the utilization rates obtained in the last n acquisition periods, if the utilization rate exceeds a given threshold value of the node, further judging whether the number exceeding the given threshold value in the n utilization rates exceeds k, and if so, determining that the corresponding node is a hot spot, namely: there are hot spots.
The utilization rate v of the next moment is predicted according to the utilization rate obtained in the last n acquisition periods t The method is realized by adopting an autoregressive model AR (n), and the specific method is as follows:
v t ＝β 0 +β 1 v t-1 +β 2 v t-2 +...+β n v t-n +ω (1)
prediction, in the formula, v t-n ，...，v t-2 ，v t-1 The usage rates, beta, obtained for the last n acquisition cycles, respectively 0 、β 1 、β 2 …β n For the autoregressive coefficients, ω is white noise.
The process of sequentially performing container migration according to the order from the high VSR value to the low VSR value in the step 3 is as follows:
step 31, selecting a container with the largest VSR value as a container to be migrated, selecting a light-load server with the smallest volume as a target server, and migrating the container to be migrated to the target server;
and step 32, judging that the utilization rate of all resources in the overload server with the maximum volume value is lower than a corresponding threshold value, if so, completing container migration, otherwise, selecting a container with the VSR value of times as a container to be migrated, and returning to the step 31.
In the step 31, if a new container is generated during the process of migrating the container to be migrated to the target server, thereby causing overload of the target server, the second smallest volume of the light-load server is used as the target server to continue migration, and the steps are repeated until the target server capable of being migrated is found to complete the migration; if no target server capable of being migrated is found finally, the migration is canceled, a container with the inferior VSR value is selected as a container to be migrated, and the step 31 is repeatedly executed.
The control system for realizing the local dynamic migration based on the Docker container technology comprises a cluster node and a migration control unit; each server node comprises a container monitoring unit and a plurality of Docker containers;
the container monitoring unit is used for monitoring the resource use state of all the Docker containers in the server node;
the migration control unit comprises a migration manager, a hot spot detector and a resource collector, wherein:
the resource collector is used for collecting resource use states of the Docker containers obtained by the container monitoring units in all the server nodes; and is further configured to generate a resource usage profile for each container; and generating an integrated resource usage profile for each server;
the hot spot detector is used for accessing the resource use configuration file generated by the resource collector to obtain the resource use conditions of all the Docker containers in all the server nodes, so as to realize the detection of hot spots;
the migration manager is used for controlling and realizing the migration of the Docker container in the hot spot determined by the hot spot detector.
The method for realizing the hot spot detection by the hot spot detector comprises the following steps:
and predicting the utilization rate at the next moment according to the utilization rates obtained in the last n acquisition periods, if the utilization rate exceeds a given threshold value, further judging whether the number exceeding the given threshold value in the n utilization rates exceeds k, and if so, determining that the corresponding node is a hot spot.
The method for predicting the utilization rate at the next moment according to the utilization rate obtained in the last n acquisition periods is realized by adopting an autoregressive model AR (n), and the specific method is as follows:
v t ＝β 0 +β 1 v t-1+ β 2 v t-2 +...+β n v t-n +ω (1)
realizing prediction, v in formula t Indicating the utilization rate at the next time, v t-n ，...，v t-2 ，v t-1 The usage rates, beta, obtained for the last n acquisition cycles, respectively 0 、β 1 、β 2 …β n For the autoregressive coefficients, ω is white noise.
The local dynamic migration method based on the Docker container technology considers the dynamic migration among the containers from the local load balancing, takes the local load balancing as a principle, realizes the dynamic migration among the containers, further achieves the effects of improving the utilization rate of resources and reducing the execution time of tasks, and has the specific advantages that:
(1) Timeliness: and monitoring the load condition of the cluster in real time, wherein containers run in a process relative to the virtual machine, and the resource consumption of each container is at the thread level. The resource state information of the container is acquired, so that finer granularity of resource monitoring can be realized.
(2) Local optimization: compared with the traditional virtual machine, the container has the characteristic of short life cycle, and the operation and stop of the container are realized at the millisecond level, so that the load balance of the whole container network can not be realized. And the dynamic migration of the container is carried out by monitoring the running state of the cluster, so that the optimization of local load is realized.
(3) Load balancing: after local dynamic migration, a container with large resource occupation amount and rapid growth trend is migrated to a server with low resource occupation rate. And finally, the load balancing rate of the server cluster is smaller than a threshold value, so that final load balancing can be realized.
The application provides a local dynamic migration method based on a container technology aiming at the problem of load balancing in cloud computing, wherein the method adopts a lighter virtualization scheme-a container, realizes resource interaction and cooperation between containers, considers the life cycle characteristics of the containers, adopts the concept of local balancing, realizes partial optimization inside a single node, and finally realizes the scheme of dynamic migration between the containers.
Drawings
Fig. 1 is a schematic block diagram of container resource monitoring in a cluster, and fig. 2 is a framework of a Docker container dynamic migration system according to the present application.
Description of the embodiments
An embodiment one, a local dynamic migration method based on a Docker container technology described in this embodiment includes the following steps:
step 1, according to the resource utilization rate of each node in the cluster nodes, monitoring the resource utilization condition of each node in the cluster nodes, determining whether hot spots exist, and triggering and starting the following steps to perform container migration when the hot spots exist;
step 2, each server in the statistical cluster integrates the utilization parameters volume of different resources,
in the formula, mem represents the memory utilization rate of a server or a container, CPU represents the utilization rate of a CPU, and net represents the bandwidth utilization rate; the N servers in the cluster are ordered in descending order according to the volume value;
according to the formula of the parameter volume, the higher the utilization rate of the resource is, the larger the volume value is. If the usage of multiple resources is large, that is, if multiple resources are overloaded at the same time, the volume value is large.
And 3, marking the overload server with the maximum volume value as a hot spot server, and performing the following processing:
ordering all containers inside it in descending order of VSR value, the vsr=volume/size, wherein size represents the memory size of the container;
sequentially performing container migration according to the sequence of the VSR values from large to small until the utilization rate of all resources in the overload server with the maximum volume value is lower than a corresponding threshold value, and completing container migration;
and returning to the execution of the step 1.
In step 3 of the above method, which container is selected for migration and the appropriate migration target server is selected. The relocation of the container belongs to the NP-hard problem, so that the embodiment adopts a heuristic algorithm to carry out task migration from the server with the heaviest load to the server with the lightest load, and the data volume is ensured not to be excessively large in the migration process so as to reduce the migration consumption. Often, a shortage of one or more resources can cause the server to be overloaded.
In practical situations, after an overloaded node (hot spot) is detected, the additional resources required for eliminating the hot spot container are firstly evaluated, if the hot spot can be eliminated through the allocation of resources in the node, a hot spot relieving algorithm can be called to relieve the hot spot, and container migration operation is not needed, so that migration cost is saved. But if the detected usage of the hotspot server exceeds a given threshold, it can only be addressed by container migration, as is the case in this embodiment.
In this embodiment, the method for determining whether the hot spot exists in step 1 may be implemented by the following method: the resource usage rate for each node is processed as follows:
predicting the utilization rate at the next moment according to the utilization rate obtained in the last n acquisition periods, if the utilization rate exceeds a given threshold, further judging whether the number exceeding the given threshold in the n utilization rates exceeds k, and if so, determining that the corresponding node is a hot spot, namely: there are hot spots.
Above-mentionedPredicting the utilization v at the next moment according to the utilization obtained in the last n acquisition periods t The method is realized by adopting an autoregressive model AR (n), and the specific method is as follows:
v t ＝β 0 +β 1 v t-1 +β 2 v t-2 +...+β n v t-n +ω (1)
prediction, in the formula, v t-n ，...，v t-2 ，v t-1 The usage rates, beta, obtained for the last n acquisition cycles, respectively 0 、β 1 、β 2 …β n For the autoregressive coefficients, ω is white noise.
The method of step 1 can ensure that overload condition can be found in time, unnecessary migration caused by instantaneous resource request is avoided, a regression model AR (n) is adopted in the prediction of container resources, n past values are utilized to predict the state value of the next moment, and the use rate in the last n sampling periods is combined to trigger the starting of container migration.
When one server node in the cluster is overloaded, a suitable container is selected from the overloaded node for migration, and the step 3 is a process of selecting a suitable container for migration.
In the step 3, the process of sequentially performing container migration according to the order from the larger VSR value to the smaller VSR value is as follows:
step 31, selecting a container with the largest VSR value as a container to be migrated, selecting a light-load server with the smallest volume as a target server, and migrating the container to be migrated to the target server;
and step 32, judging that the utilization rate of all resources in the overload server with the maximum volume value is lower than a corresponding threshold value, if so, completing container migration, otherwise, selecting a container with the VSR value of times as a container to be migrated, and returning to the step 31.
In the step 31, if a new container is generated during the process of migrating the container to be migrated to the target server, thereby causing overload of the target server, the second smallest volume of the light-load server is used as the target server to continue migration, and the steps are repeated until the target server capable of being migrated is found to complete the migration; if no target server capable of being migrated is found finally, the migration is canceled, a container with the inferior VSR value is selected as a container to be migrated, and the step 31 is repeatedly executed.
If a new container is generated and the target server is overloaded, the migration of the container is invalid, and the container must be retried to be migrated to the light-load server of volume times.
If in the above process, no target server capable of migration is found finally, it means that there is no server capable of accommodating the container in all servers of the cluster, so that it is not necessary to migrate the container, but rather a container with a secondary VSR value is selected for migration.
Step 3 is to first migrate the container with larger aggregate usage but smaller memory, i.e. reduce migration overhead while balancing load.
In the step 3, when the target server is selected, it is first ensured that there are enough available resources in the target server, so that the light-load server with the smallest volume value is preferred as the target server, and when the target server cannot realize migration, the light-load server with the volume value is selected.
The local dynamic migration method in this embodiment restores the overload nodes in the cluster to a light load state, i.e. maintains a local load balance state.
The embodiment provides a local dynamic migration method based on a container technology aiming at the problem of load balancing in cloud computing, the method adopts a lighter virtualization scheme-a container, realizes resource interaction and cooperation between containers, adopts a local balancing idea in consideration of life cycle characteristics of the containers, realizes partial optimization inside a single node, and finally realizes the scheme of dynamic migration between the containers.
Embodiment two, this embodiment will be described with reference to fig. 1 and 2. The embodiment describes a local dynamic migration control system based on a Docker container technology, the control system comprises a cluster node, and the control system further comprises a migration control unit; each server node comprises a container monitoring unit and a plurality of Docker containers;
the container monitoring unit is used for monitoring the resource use state of all the Docker containers in the server node;
the migration control unit comprises a migration manager, a hot spot detector and a resource collector, wherein:
the resource collector is used for collecting resource use states of the Docker containers obtained by the container monitoring units in all the server nodes; and is further configured to generate a resource usage profile for each container; and generating an integrated resource usage profile for each server;
the hot spot detector is used for accessing the resource use configuration file generated by the resource collector to obtain the resource use conditions of all the Docker containers in all the server nodes, and detecting hot spots according to the latest resource use conditions;
the migration manager is used for controlling and realizing the migration of the Docker container in the hot spot determined by the hot spot detector. The migration manager effectively mitigates the hot spot in the manner of container migration.
The hot spot refers to a server node with the resource utilization rate exceeding a rated threshold.
The control system described in this embodiment is used to implement container migration of an overloaded server.
The method for realizing the hot spot detection by the hot spot detector comprises the following steps:
and predicting the utilization rate at the next moment according to the utilization rates obtained in the last n acquisition periods, if the utilization rate exceeds a given threshold value, further judging whether the number exceeding the given threshold value in the n utilization rates exceeds k, and if so, determining that the corresponding node is a hot spot.
The method for predicting the utilization rate at the next moment according to the utilization rate obtained in the last n acquisition periods is realized by adopting an autoregressive model AR (n), and the specific method is as follows:
v t ＝β 0 +β 1 v t-1 +β 2 v t-2 +...+β n v t-n +ω (1)
realizing prediction, v in formula t Indicating the utilization rate at the next time, v t-n ，...，v t-2 ，v t-1 The usage rates, beta, obtained for the last n acquisition cycles, respectively 0 、β 1 、β 2 …β n For the autoregressive coefficients, ω is white noise.
FIG. 1 is a schematic diagram of a cluster of n nodes according to the present embodiment, S 1 ，S 2 ...S n Representing the 1 st, 2 nd and … n server nodes in the cluster respectively, then virtualizing the inside of each server node to generate m Docker containers, and configuring a resource monitoring unit inside each server node.
Claims (9)
1. A local dynamic migration method based on a Docker container technology, which is characterized by comprising the following steps:
step 1, according to the resource utilization rate of each node in the cluster nodes, monitoring the resource utilization condition of each node in the cluster nodes, determining whether hot spots exist, and triggering and starting the following steps to perform container migration when the hot spots exist;
step 2, each server in the statistical cluster integrates the utilization parameters volume of different resources,
in the formula, mem represents the memory utilization rate of a server or a container, CPU represents the utilization rate of a CPU, and net represents the bandwidth utilization rate; the N servers in the cluster are ordered in descending order according to the volume value;
step 3, aiming at the overload server with the maximum volume value, the following processing is carried out:
ordering all containers inside it in descending order of VSR value, the vsr=volume/size, wherein size represents the memory size of the container;
sequentially performing container migration according to the sequence of the VSR values from large to small until the utilization rate of all resources in the overload server with the maximum volume value is lower than a corresponding threshold value, and completing container migration;
and returning to the execution of the step 1.
2. The method for determining whether a hotspot exists in step 1 according to claim 1, wherein the method for determining whether a hotspot exists in step 1 is that, for each node, the following process is performed for the resource usage rate of each node:
predicting the utilization rate at the next moment according to the utilization rates obtained in the last n acquisition periods, if the utilization rate exceeds a given threshold value of the node, further judging whether the number exceeding the given threshold value in the n utilization rates exceeds k, and if so, determining that the corresponding node is a hot spot, namely: there are hot spots.
3. The method for local dynamic migration based on the Docker container technology according to claim 2, wherein the usage v at the next moment is predicted according to the usage obtained in the last n acquisition cycles t The method is realized by adopting an autoregressive model AR (n), and the specific method is as follows:
v t ＝β 0 +β 1 v t-1 +β 2 v t-2 +...+β n v t-n +ω (1)
prediction, in the formula, v t-n ，...，v t-2 ，v t-1 The usage rates, beta, obtained for the last n acquisition cycles, respectively 0 、β 1 、β 2 …β n For the autoregressive coefficients, ω is white noise.
4. The local dynamic migration method based on the Docker container technology according to claim 1, wherein the process of sequentially performing the container migration in the order from the higher VSR value to the lower VSR value in the step 3 is as follows:
step 31, selecting a container with the largest VSR value as a container to be migrated, selecting a light-load server with the smallest volume as a target server, and migrating the container to be migrated to the target server;
and step 32, judging that the utilization rate of all resources in the overload server with the maximum volume value is lower than a corresponding threshold value, if so, completing container migration, otherwise, selecting a container with the VSR value of times as a container to be migrated, and returning to the step 31.
5. The method for local dynamic migration based on dock container technology according to claim 4, wherein in step 31, if a new container is generated during migration of a container to be migrated to a target server, thereby causing overload of the target server, the migration is continued with the light-load server with the volume value of the second smallest as the target server, and the above steps are repeated until a target server capable of being migrated is found to complete the present migration; if no target server capable of being migrated is found finally, the migration is canceled, a container with the inferior VSR value is selected as a container to be migrated, and the step 31 is repeatedly executed.
6. A local dynamic migration control system based on a Docker container technology, which comprises cluster nodes, and is characterized by further comprising a migration control unit; each server node comprises a container monitoring unit and a plurality of Docker containers;
the container monitoring unit is used for monitoring the resource use state of all the Docker containers in the server node;
the migration control unit comprises a migration manager, a hot spot detector and a resource collector, wherein:
the resource collector is used for collecting resource use states of the Docker containers obtained by the container monitoring units in all the server nodes; and is further configured to generate a resource usage profile for each container; and generating an integrated resource usage profile for each server;
the hot spot detector is used for accessing the resource use configuration file generated by the resource collector to obtain the resource use conditions of all the Docker containers in all the server nodes, so as to realize the detection of hot spots;
the migration manager is used for controlling and realizing the migration of the Docker container in the hot spot determined by the hot spot detector.
7. The system of claim 6, wherein the hot spot is a server node whose resource usage exceeds a nominal threshold.
8. The local dynamic migration control system of claim 6, wherein the hot spot detector implements a method for detecting hot spots by:
and predicting the utilization rate at the next moment according to the utilization rates obtained in the last n acquisition periods, if the utilization rate exceeds a given threshold value, further judging whether the number exceeding the given threshold value in the n utilization rates exceeds k, and if so, determining that the corresponding node is a hot spot.
9. The local dynamic migration control system of claim 8, wherein the method for predicting the usage rate at the next moment according to the usage rate obtained in the last n acquisition cycles is implemented by adopting an autoregressive model AR (n), and the specific method is as follows:
v t ＝β 0 +β 1 v t-1 +β 2 v t-2 +...+β n v t-n +ω (1)
realizing prediction, v in formula t Indicating the utilization rate at the next time, v t-n ，...，v t-2 ，v t-1 The usage rates, beta, obtained for the last n acquisition cycles, respectively 0 、β 1 、β 2 …β n For the autoregressive coefficients, ω is white noise.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711322296.9A CN108182105B (en) | 2017-12-12 | 2017-12-12 | Local dynamic migration method and control system based on Docker container technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711322296.9A CN108182105B (en) | 2017-12-12 | 2017-12-12 | Local dynamic migration method and control system based on Docker container technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108182105A CN108182105A (en) | 2018-06-19 |
| CN108182105B true CN108182105B (en) | 2023-08-15 |
Family
ID=62546178
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711322296.9A Active CN108182105B (en) | 2017-12-12 | 2017-12-12 | Local dynamic migration method and control system based on Docker container technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108182105B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108762912A (en) * | 2018-07-04 | 2018-11-06 | 郑州云海信息技术有限公司 | A kind of method and apparatus of container cluster elastic telescopic |
| CN109271249B (en) * | 2018-07-27 | 2022-02-08 | 暨南大学 | Cloud container pre-copy online migration method based on P.haul framework |
| CN109710376A (en) * | 2018-12-12 | 2019-05-03 | 中国联合网络通信集团有限公司 | The dynamic dispatching method and device of container cluster management system |
| CN110119377B (en) * | 2019-04-24 | 2021-01-15 | 华中科技大学 | Docker container-oriented online migration system implementation and optimization method |
| CN110134495B (en) * | 2019-05-21 | 2022-08-12 | 山东大学 | Container cross-host online migration method, storage medium and terminal equipment |
| 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 |
| CN110413436B (en) * | 2019-07-18 | 2021-11-23 | 腾讯科技（深圳）有限公司 | Fault processing method and device |
| CN111190688B (en) * | 2019-12-19 | 2022-12-09 | 西安交通大学 | Cloud data center-oriented Docker migration method and system |
| CN110990160B (en) * | 2019-12-27 | 2023-06-27 | 广西电网有限责任公司 | Static security analysis container cloud elastic telescoping method based on load prediction |
| CN111367632B (en) * | 2020-02-14 | 2023-04-18 | 重庆邮电大学 | Container cloud scheduling method based on periodic characteristics |
| CN112181605A (en) * | 2020-10-27 | 2021-01-05 | 北京城市网邻信息技术有限公司 | Load balancing method and device, electronic equipment and computer readable medium |
| CN113821340A (en) * | 2021-08-27 | 2021-12-21 | 济南浪潮数据技术有限公司 | Dynamic balancing method, system, terminal and storage medium of distributed system |
| CN118055023B (en) * | 2024-02-20 | 2024-10-11 | 中国科学院计算机网络信息中心 | Method for migrating cross-node parallel computing task local processes |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102244685A (en) * | 2011-08-11 | 2011-11-16 | 中国科学院软件研究所 | Distributed type dynamic cache expanding method and system supporting load balancing |
| CN105338118A (en) * | 2015-11-30 | 2016-02-17 | 上海斐讯数据通信技术有限公司 | Distributed storage system |
| WO2016134542A1 (en) * | 2015-02-28 | 2016-09-01 | 华为技术有限公司 | Virtual machine migration method, apparatus and device |
| CN106528270A (en) * | 2016-11-16 | 2017-03-22 | 航天信息股份有限公司 | Automatic migration method and system of virtual machine based on OpenStack cloud platform |
| CN106790726A (en) * | 2017-03-30 | 2017-05-31 | 电子科技大学 | A kind of priority query's dynamic feedback of load equilibrium resource regulating method based on Docker cloud platforms |
-
2017
- 2017-12-12 CN CN201711322296.9A patent/CN108182105B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102244685A (en) * | 2011-08-11 | 2011-11-16 | 中国科学院软件研究所 | Distributed type dynamic cache expanding method and system supporting load balancing |
| WO2016134542A1 (en) * | 2015-02-28 | 2016-09-01 | 华为技术有限公司 | Virtual machine migration method, apparatus and device |
| CN105338118A (en) * | 2015-11-30 | 2016-02-17 | 上海斐讯数据通信技术有限公司 | Distributed storage system |
| CN106528270A (en) * | 2016-11-16 | 2017-03-22 | 航天信息股份有限公司 | Automatic migration method and system of virtual machine based on OpenStack cloud platform |
| CN106790726A (en) * | 2017-03-30 | 2017-05-31 | 电子科技大学 | A kind of priority query's dynamic feedback of load equilibrium resource regulating method based on Docker cloud platforms |
Non-Patent Citations (1)
| Title |
|---|
| 基于Docker Swarm集群的容器迁移策略的实现;毛祺，卢胜林;《信息技术》(第09期);156-160 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108182105A (en) | 2018-06-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108182105B (en) | Local dynamic migration method and control system based on Docker container technology | |
| Farahnakian et al. | Utilization prediction aware VM consolidation approach for green cloud computing | |
| CN107273185B (en) | Load balancing control method based on virtual machine | |
| Kanakala et al. | Performance analysis of load balancing techniques in cloud computing environment | |
| Vignesh et al. | Resource management and scheduling in cloud environment | |
| Al-Dulaimy et al. | Type-aware virtual machine management for energy efficient cloud data centers | |
| Tarafdar et al. | Energy and quality of service-aware virtual machine consolidation in a cloud data center | |
| CN105491138B (en) | Distributed load scheduling method based on load rate graded triggering | |
| Deboosere et al. | Efficient resource management for virtual desktop cloud computing | |
| Fu et al. | Layered virtual machine migration algorithm for network resource balancing in cloud computing | |
| Li | An adaptive overload threshold selection process using Markov decision processes of virtual machine in cloud data center | |
| CN104679594B (en) | A kind of middleware distributed computing method | |
| Abdelsamea et al. | Virtual machine consolidation challenges: A review | |
| Wu et al. | An energy efficient VM migration algorithm in data centers | |
| Nie et al. | Energy-aware multi-dimensional resource allocation algorithm in cloud data center | |
| CN109960579B (en) | Method and device for adjusting service container | |
| Zhang et al. | Dynamic energy-efficient virtual machine placement optimization for virtualized clouds | |
| Pandya et al. | Dynamic resource allocation techniques in cloud computing | |
| Garg et al. | Optimal virtual machine scheduling in virtualized cloud environment using VIKOR method | |
| Jaiswal et al. | An approach towards the dynamic load management techniques in cloud computing environment | |
| Guo | Ant colony optimization computing resource allocation algorithm based on cloud computing environment | |
| CN106775942B (en) | Cloud application-oriented solid-state disk cache management system and method | |
| CN104683480A (en) | Distribution type calculation method based on applications | |
| Jayamala et al. | An Enhanced Decentralized Virtual Machine Migration Approach for Energy-Aware Cloud Data Centers. | |
| Theja et al. | An evolutionary computing based energy efficient VM consolidation scheme for optimal resource utilization and QoS assurance |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |