CN104991825A - Hypervisor resource hyper-allocation and dynamic adjusting method and system based on load awareness - Google Patents
Hypervisor resource hyper-allocation and dynamic adjusting method and system based on load awareness Download PDFInfo
- Publication number
- CN104991825A CN104991825A CN201510140721.7A CN201510140721A CN104991825A CN 104991825 A CN104991825 A CN 104991825A CN 201510140721 A CN201510140721 A CN 201510140721A CN 104991825 A CN104991825 A CN 104991825A
- Authority
- CN
- China
- Prior art keywords
- hypervisor
- excessive distribution
- resource
- main frame
- load
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention relates to a Hypervisor resource hyper-allocation and dynamic adjusting method and a system based on load awareness. According to the method, the load and hyper-allocation conditions of Hypervisor resources are monitored, the resource hyper-allocation coefficient of each Hypervisor host is set, recorded, and adjusted, the set resource hyper-allocation coefficient of each Hypervisor resource in a cluster is adjusted in a dynamic manner, the resource reusability of Hypervisor is quantified rather than being an empirical value, the allocation of virtual machines in the Hypervisor cluster is migrated so that normal operation of the virtual machines can be guaranteed, the resource utilization rate of Hypervisor is increased, the overall utilization rate of the cloud is increased, and the energy consumption and the maintenance cost are reduced.
Description
Technical field
The present invention relates to cloud computing management domain, particularly relate to a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic adjusting method and system.
Background technology
Hypervisor, a kind of intermediate software layer operated between basic physics server and operating system, can allow multiple operating system and Application share hardware.Also VMM (virtual machine monitor) can be called, i.e. virtual machine monitor.
Hypervisors be a kind of in virtual environment " unit " operating system.They access services device can comprise all physical equipments in disk and interior existence.
Hypervisors not only coordinates the access of these hardware resources, also between each virtual machine, applies protection simultaneously.When startup of server and when performing Hypervisor, the operating system that it can load all virtual machine client ends can distribute to the appropriate internal memory of each virtual machine, CPU, network and disk simultaneously.
The resource excessive distribution ability of Hypervisor makes physical machine have higher resource consolidation rate, and more effectively utilize computational resource, it can cut operating costs simultaneously.Use resource excessive distribution, Hypervisor makes physical resource provide virtual machine of good performance in the best way.Such as, when virtual desktop framework is disposed, user can operate multiple Windows virtual machine, eachly comprises a process application program.Likely use Hypervisor and the excessive virtual machine that identical virtual desktop framework is provided.Because virtual machine comprises similar operating system and application program, their many memory pages and comprise similar content.
Hypervisor finds and consolidates the content memory pages identical with these virtual machines, thus saves CPU/ internal memory.This just utilizes CPU/ internal memory better and realizes higher integration rate.
At present, the resource excessive distribution technology of Hypervisor in each large virtual manufacturer and product very ripe (Hyper-v, Xenserver, KVM, Esxi etc.).
But when Hypervisor excessive distribution resource is too much, real resource use amount is prescribed a time limit more than the upper of Hypervisor resource for a long time, and partial virtual machine cannot continue to obtain enough CPU/ memory sources.Hypervisor inside occurs that the contradiction of contention for resources can grow in intensity, and finally causes virtual-machine fail, even Hypervisor fault.
Summary of the invention
Technical matters to be solved by this invention is for the deficiencies in the prior art, a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic adjusting method and system are provided, solve the islanding problem of Hypervisor resource multiplex technology, dynamically, the adjustment virtual machine of intelligence is in the deployment of Hypervisor cluster, distribution situation, reduce all kinds of faults that resource multiplex technology causes, improve the overall resource utilization in high in the clouds.
The technical scheme that the present invention solves the problems of the technologies described above is as follows: a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic adjusting method, comprise the steps:
Step 1, arranges Hypervisor resource load threshold, Hypervisor resource excessive distribution threshold value, the excessive distribution coefficient of each Hypervisor main frame and excessive distribution regulation coefficient;
Step 2, arranges the scheduling strategy of Hypervisor cluster virtual machine;
Step 3, monitors the loading condition of each Hypervisor main frame;
Step 4, judges whether the load of each Hypervisor main frame is more than or equal to Hypervisor resource load threshold successively, if it is reduces the excessive distribution coefficient of this Hypervisor main frame according to excessive distribution regulation coefficient, performs step 5, otherwise performs step 6;
Step 5, according to the excessive distribution coefficient of this Hypervisor main frame after adjustment and the adjustment of the scheduling strategy described in step 2 virtual machine in the distribution of Hypervisor cluster, returns step 3;
Step 6, detects the resource excessive distribution situation of this Hypervisor main frame;
Step 7, judge whether the resource excessive distribution ratio of this Hypervisor main frame is more than or equal to Hypervisor resource excessive distribution threshold value, if it is increase the excessive distribution coefficient of this Hypervisor main frame according to excessive distribution regulation coefficient, return step 6, otherwise process ends.
The invention has the beneficial effects as follows: the load by monitoring Hypervisor resource of the present invention and excessive distribution situation, arrange, record, and adjust the resource excessive distribution coefficient of each Hypervisor main frame, resource excessive distribution coefficient is arranged to each Hypervisor in cluster and carries out dynamic conditioning, the resource multiplex ability of Hypervisor is quantized, instead of an empirical value, by the distribution of migration virtual machine at Hypervisor cluster, under ensureing the prerequisite that virtual machine normally runs, promote the resource utilization of Hypervisor, thus improve the overall utilization rate of cloud, reduce energy consumption, reduce maintenance cost.
On the basis of technique scheme, the present invention can also do following improvement.
Further, the cluster of Hypervisor described in step 2 virtual machine scheduling policy adopts any one in first-fit, best-fit, packing, loadbalance, Round-Robin or random.
Further, technique scheme also comprises the excessive distribution coefficient default value pre-setting Hypervisor main frame, the excessive distribution coefficient of each Hypervisor main frame is according in excessive distribution regulation coefficient dynamic adjustment process, and the excessive distribution coefficient after adjustment can not lower than the excessive distribution coefficient default value pre-setting Hypervisor main frame.
Further, being implemented as of step 5:
Step 5.1, detects each Hypervisor main frame one by one;
Step 5.2, judges whether the load of Hypervisor main frame to be detected is less than Hypervisor resource load threshold, if perform step 5.3, otherwise performs step 5.5;
Step 5.3, judges whether the resource excessive distribution ratio of this Hypervisor main frame is less than Hypervisor resource excessive distribution threshold value, if perform step 5.4, otherwise performs step 5.5;
Step 5.4, is more than or equal to virtual machine (vm) migration on the Hypervisor main frame of Hypervisor resource load threshold to the Hypervisor main frame described in step 5.3 by load;
Step 5.5, end process flow process.
Further, technique scheme also comprises according to adjustment result adjustment Hypervisor resource load threshold and/or Hypervisor resource excessive distribution threshold value.
Another technical scheme that the present invention solves the problems of the technologies described above is as follows: a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic debugging system, comprises parameter setting module, scheduling strategy arranges module, load monitoring module, load dispatch module and excessive distribution monitoring module;
Described parameter setting module, it is for arranging Hypervisor resource load threshold, Hypervisor resource excessive distribution threshold value, the excessive distribution coefficient of each Hypervisor main frame and excessive distribution regulation coefficient;
Described scheduling strategy arranges module, and it is for arranging the scheduling strategy of Hypervisor cluster virtual machine;
Described load monitoring module, it is for judging whether the load of each Hypervisor main frame is more than or equal to Hypervisor resource load threshold successively, if it is the excessive distribution coefficient of this Hypervisor main frame is reduced according to excessive distribution regulation coefficient, call load dispatch module, otherwise call excessive distribution monitoring module;
Described load dispatch module, it is for according to the distribution at Hypervisor cluster of the excessive distribution coefficient of this Hypervisor main frame after adjustment and scheduling strategy adjustment virtual machine, until the load of this Hypervisor main frame is less than Hypervisor resource load threshold;
Described excessive distribution monitoring module, it is for detecting the resource excessive distribution situation of this Hypervisor main frame, judge whether the resource excessive distribution ratio of this Hypervisor main frame is more than or equal to Hypervisor resource excessive distribution threshold value, if it is the excessive distribution coefficient of this Hypervisor main frame is increased according to excessive distribution regulation coefficient, continue the resource excessive distribution situation detecting this Hypervisor main frame, until the excessive distribution ratio of this Hypervisor main frame is less than Hypervisor resource load threshold, otherwise process ends.
Further, described scheduling strategy arranges any one in the employing of Hypervisor cluster virtual machine scheduling policy first-fit, best-fit, packing, loadbalance, Round-Robin or random of module installation.
Further, described parameter setting module also pre-sets the excessive distribution coefficient default value of Hypervisor main frame, the excessive distribution coefficient of each Hypervisor main frame is according in excessive distribution regulation coefficient dynamic adjustment process, and the excessive distribution coefficient after adjustment can not lower than the excessive distribution coefficient default value pre-setting Hypervisor main frame.
Further, described load dispatch module comprises resources measurement unit, the first judging unit, the second judging unit and virtual machine (vm) migration unit;
Described resources measurement unit, it is for detecting each Hypervisor main frame one by one;
Described first judging unit, it is for judging whether the load of Hypervisor main frame to be detected is less than Hypervisor resource load threshold, if call the second judging unit, otherwise the flow process that ends process;
Described second judging unit, it is for judging whether the resource excessive distribution ratio of this Hypervisor main frame is less than Hypervisor resource excessive distribution threshold value, if call virtual machine (vm) migration unit, end process flow process;
Described virtual machine (vm) migration unit, it is for being more than or equal to virtual machine (vm) migration on the Hypervisor main frame of Hypervisor resource load threshold to described Hypervisor main frame by load.
Further, technique scheme also comprises according to adjustment result adjustment Hypervisor resource load threshold and/or Hypervisor resource excessive distribution threshold value.
Accompanying drawing explanation
Fig. 1 is a kind of Hypervisor resource excessive distribution based on Load-aware of the present invention and dynamic adjusting method process flow diagram;
Fig. 2 is a kind of Hypervisor resource excessive distribution based on Load-aware of the present invention and dynamic debugging system block diagram;
Fig. 3 is load dispatch modular structure schematic diagram of the present invention.
In accompanying drawing, the list of parts representated by each label is as follows:
1, parameter setting module, 2, scheduling strategy arranges module, 3, load monitoring module, 4, load dispatch module, 5, excessive distribution monitoring module, 41, monitoring resource unit, the 42, first judging unit, 4,3, the second judging unit, 44, virtual machine (vm) migration unit.
Embodiment
Be described principle of the present invention and feature below in conjunction with accompanying drawing, example, only for explaining the present invention, is not intended to limit scope of the present invention.
The present invention relates to a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic adjusting method and system.What relate generally to is CPU, memory source excessive distribution management and dynamic conditioning.
As shown in Figure 1, a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic adjusting method, comprise the steps:
Step 1, arranges Hypervisor resource load threshold, Hypervisor resource excessive distribution threshold value (the excessive distribution coefficient of the virtual machine aggregate resource namely on Hypervisor/(Hypervisor resource * excessive distribution coefficient), each Hypervisor main frame and excessive distribution regulation coefficient;
Step 2, arranges the scheduling strategy of Hypervisor cluster virtual machine;
Step 3, monitors the loading condition of each Hypervisor main frame;
Step 4, judges whether the load of each Hypervisor main frame is more than or equal to Hypervisor resource load threshold successively, if it is reduces the excessive distribution coefficient of this Hypervisor main frame according to excessive distribution regulation coefficient, performs step 5, otherwise performs step 6;
Step 5, according to the excessive distribution coefficient of this Hypervisor main frame after adjustment and the adjustment of the scheduling strategy described in step 2 virtual machine in the distribution of Hypervisor cluster, returns step 3;
Step 6, detects the resource excessive distribution situation of this Hypervisor main frame;
Step 7, judge whether the resource excessive distribution ratio of this Hypervisor main frame is more than or equal to Hypervisor resource excessive distribution threshold value, if it is increase the excessive distribution coefficient of this Hypervisor main frame according to excessive distribution regulation coefficient, return step 6, otherwise process ends.
The cluster of Hypervisor described in step 2 virtual machine scheduling policy adopts any one in first-fit, best-fit, packing, loadbalance, Round-Robin or random.
Technique scheme also comprises the excessive distribution coefficient default value pre-setting Hypervisor main frame, the excessive distribution coefficient of each Hypervisor main frame is according in excessive distribution regulation coefficient dynamic adjustment process, and the excessive distribution coefficient after adjustment can not lower than the excessive distribution coefficient default value pre-setting Hypervisor main frame.
The excessive distribution coefficient default value of such as Hypervisor main frame is set to 1, and the excessive distribution coefficient of each Hypervisor main frame is set to 2,3,4,5,6 respectively, and the coefficient after adjustment can be 2,3,4,3,2, namely all can not be less than default value 1; For another example, the excessive distribution coefficient default value of Hypervisor main frame is set to 2, and the excessive distribution coefficient of each Hypervisor main frame is set to 2,4,6,8,10 respectively, and the coefficient after adjustment can be 2,4,6,4,2, namely all can not be less than default value 2.
Being implemented as of step 5:
Step 5.1, detects each Hypervisor main frame one by one;
Step 5.2, judges whether the load of Hypervisor main frame to be detected is less than Hypervisor resource load threshold, if perform step 5.3, otherwise performs step 5.5;
Step 5.3, judges whether the resource excessive distribution ratio of this Hypervisor main frame is less than Hypervisor resource excessive distribution threshold value, if perform step 5.4, otherwise performs step 5.5;
Step 5.4, is more than or equal to virtual machine (vm) migration on the Hypervisor main frame of Hypervisor resource load threshold to the Hypervisor main frame described in step 5.3 by load;
Step 5.5, end process flow process.
Technique scheme also comprises according to adjustment result adjustment Hypervisor resource load threshold and/or Hypervisor resource excessive distribution threshold value.
Technique scheme also comprises and pre-sets sense cycle, in each sense cycle to cluster in virtual machine distribution situation adjust.
As shown in Figure 2, a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic debugging system, comprise parameter setting module 1, scheduling strategy arrange module 2, load monitoring module 3, load dispatch module 4 and excessive distribution monitoring module 5; Described parameter setting module 1, it is for arranging Hypervisor resource load threshold, Hypervisor resource excessive distribution threshold value, the excessive distribution coefficient of each Hypervisor main frame and excessive distribution regulation coefficient; Described scheduling strategy arranges module 2, and it is for arranging the scheduling strategy of Hypervisor cluster virtual machine; Described load monitoring module 3, it is for judging whether the load of each Hypervisor main frame is more than or equal to Hypervisor resource load threshold successively, if it is the excessive distribution coefficient of this Hypervisor main frame is reduced according to excessive distribution regulation coefficient, call load dispatch module 4, otherwise call excessive distribution monitoring module 5; Described load dispatch module 4, it is for according to the distribution at Hypervisor cluster of the excessive distribution coefficient of this Hypervisor main frame after adjustment and scheduling strategy adjustment virtual machine, until the load of this Hypervisor main frame is less than Hypervisor resource load threshold; Described excessive distribution monitoring module 5, it is for detecting the resource excessive distribution situation of this Hypervisor main frame, judge whether the resource excessive distribution ratio of this Hypervisor main frame is more than or equal to Hypervisor resource excessive distribution threshold value, if it is the excessive distribution coefficient of this Hypervisor main frame is increased according to excessive distribution regulation coefficient, continue the resource excessive distribution situation detecting this Hypervisor main frame, until the excessive distribution ratio of this Hypervisor main frame is less than Hypervisor resource load threshold, otherwise process ends.
Described scheduling strategy arranges any one in the employing of Hypervisor cluster virtual machine scheduling policy first-fit, best-fit, packing, loadbalance, Round-Robin or random of module 2 setting.
Described parameter setting module 1 also pre-sets the excessive distribution coefficient default value of Hypervisor main frame, the excessive distribution coefficient of each Hypervisor main frame is according in excessive distribution regulation coefficient dynamic adjustment process, and the excessive distribution coefficient after adjustment can not lower than the excessive distribution coefficient default value pre-setting Hypervisor main frame.
As shown in Figure 3, described load dispatch module 4 comprises resources measurement unit 41, first judging unit 42, second judging unit 43 and virtual machine (vm) migration unit 44; Described resources measurement unit 41, it is for detecting each Hypervisor main frame one by one; Described first judging unit 42, it is for judging whether the load of Hypervisor main frame to be detected is less than Hypervisor resource load threshold, if call the second judging unit 43, otherwise the flow process that ends process; Described second judging unit 43, it is for judging whether the resource excessive distribution ratio of this Hypervisor main frame is less than Hypervisor resource excessive distribution threshold value, if call virtual machine (vm) migration unit 44, end process flow process; Described virtual machine (vm) migration unit 44, it is for being more than or equal to virtual machine (vm) migration on the Hypervisor main frame of Hypervisor resource load threshold to described Hypervisor main frame by load.
Suppose the cluster be made up of 3 Hypervisor main frames, configuration is: 16CPU*2.0Ghz, 32G internal memory, and Hypervisor is designated H1 respectively, H2, H3.
Create 100 1CPU*1.0Ghz, 2G memory virtual machines, virtual machine is divided into 4 groups, creates successively, represents below with T1, T2, T3, T4:
T1:30 platform, the load of virtual machine average resource is 10%.
T2:30 platform, the load of virtual machine average resource is 30%
T3:30 platform, the load of virtual machine average resource is 70%
T4:10 platform, the load of virtual machine average resource is between 30 ~ 70%
In order to easy understand, save as example within the present invention and be described in detail:
Total memory source number that virtual machine needs is 200G, is far longer than the physical memory (96G) of Hypervisor.The resource multiplex technology of Hypervisor must be used just to meet the demands.
Traditional resource multiplex technology merely provides resource excessive distribution ability, after virtual machine creating completes, and the following form of distribution tend of virtual machine:
Form 1
The internal memory having reached more than 96%, H3 as can be seen from the memory usage of form 1, H1, H2 mostly is 56.8% most.When the internal memory fluctuation of virtual machine, H1, H2 just there will be serious contention for resources, and cannot meet actual user demand at all, often cause virtual machine to respond slowly, notably the problem such as Hypervisor fault.And H1 deploys 65 virtual machines, and H3 only has 13 virtual machines, distributes also unreasonable.
By the present invention, the method for Resourse Distribute can be quantized, the distribution of dynamic conditioning virtual machine, the stability of elevator system.
Precondition:
1. the load of pair each Hypervisor main frame and resource allocation conditions are monitored;
2. arranging sense cycle is 5 minutes;
3. the CPU/ internal memory acquiescence excessive distribution coefficient arranging each Hypervisor main frame is 1;
4. arranging excessive distribution regulation coefficient is 1;
5. arranging Hypervisor resource load threshold is 90%;
6. arranging Hypervisor resource excessive distribution threshold value is 85%;
7. arranging Hypervisor cluster virtual machine scheduling policy is best free, the Hypervisor that namely prioritizing selection resources occupation rate is minimum.
In order to the present invention can be allowed more convincing, adopt and the Hypervisor of form 1 same size and virtual machine, save as example also and be described.
Application scenarios of the present invention, mainly carries out from the constructive process of virtual machine and adjustment two aspects after having created are described, and elaborates respectively below.
First how the constructive process describing virtual machine applies method of the present invention, in order to easy understand, below illustrates that the internal memory ignoring Hypervisor and virtual machine itself is reserved and consumes.Original state is as shown in table 2.
Form 2
| Hypervisor | Can storage allocation | Internal memory average load | Virtual machine |
| H1 | 32G*1＝32G | 0％ | Nothing |
| H2 | 32G*1＝32G | 0％ | Nothing |
| H3 | 32G*1＝32G | 0％ | Nothing |
Create 100 virtual machines successively:
T1:30 platform virtual machine, average resource load is 10%;
T2:30 platform virtual machine, average resource load is 30%;
T3:30 platform virtual machine, average resource load is 70%;
T4:10 platform virtual machine, average resource load is between 30 ~ 70%.
Create the 1st virtual machine: internal memory 2G, virtual machine load 10%, according to arranging cluster virtual machine scheduling policy, the H1 that prioritizing selection memory usage is minimum, owing to also not having load (virtual machine), internal memory load occupancy is less than threshold value 90%, continue to judge Memory Allocation situation, 2G/32G=6.2% after distributing, is less than resource excessive distribution threshold value 85%, process ends judges, creates virtual machine.
Because each Hypervisor resource is more idle, according to the scheduling strategy arranged at present, the deployment of virtual machine is successively with H1, H2, H3 is uniformly distributed, until resource load or excessive distribution reach threshold value, when the first time of this example adjusts, virtual machine is as shown in table 3 in the state of Hypervisor.
Form 3
Because acquiescence excessive distribution threshold value is 85%, before adjustment excessive distribution coefficient, cannot allow to distribute new virtual machine.Excessive distribution proportion threshold value is that 85%, H1 distributes 10 T1, and during 4 T2, taking storage allocation is 28G, reach virtual machine can storage allocation 87.5%.Because internal memory average load only has 13.7%, lower than load threshold, the excessive distribution coefficient of Hypervisor need be increased according to excessive distribution regulation coefficient, in conjunction with precondition, after adjustment as shown in Table 4.
Form 4
The Hypervisor cluster virtual machine scheduling policy arranged due to this example is the establishment of bestfree, next virtual machine, can select H2, instead of can the larger H1 of storage allocation.So the excessive distribution coefficient of H2 also can be increased.The like, until created 100 virtual machines, result is as form 5.
Form 5
More than description can understand the present invention following advantage:
The excessive distribution ability of 1.Hypervisor can quantitative statistics, and automatically adjusts;
2. make the distribution of virtual machine more balanced in conjunction with scheduling strategy;
3. load threshold controls, and makes Hypervisor can not distribute excess resource causing trouble.
Above-described embodiment, the application scenarios that just description one is fairly simple, have employed the virtual machine of more single specification, and create in order, virtual machine load all within the specific limits.In actual applications, the specification of virtual machine is varied, and especially load is unpredictalbe.List uses method of the present invention to be inadequate from virtual machine creating angle, also need after establishment completes, and the load of each Hypervisor of periodic detection, excessive distribution coefficient, carry out dynamic conditioning.
Next, we have a look how to use flow process of the present invention, adjustment form 1 Problems existing.Initial situation is as form 6.
Form 6
In form 6, the excessive distribution coefficient of each Hypervisor main frame be 1, H1, H2 reality storage allocation far exceeded can apportioning cost, but be greater than the setting of load threshold, so the net result of flow process of the present invention can not be affected due to the actual loading of internal memory.The virtual machine of new establishment also can not continue to be assigned to H1, on H2, increases the burden of Hypervisor.
One by one Hypervisor main frame is detected according to method of the present invention:
Step 1: H1 is detected with the cycle of pre-conditioned setting;
Step 2: detect that the resource load of H1 is greater than threshold value 90%, the excessive distribution coefficient due to H1 has been 1, does not reduce excessive distribution coefficient;
Step 3: according to the scheduling strategy arranged and load threshold migration virtual machine, load is minimum the most in the cluster in H3 load, after the T1 virtual machine of a migration H1, load, still lower than load threshold 90%, selects H3 for migration target Hypervisor;
Step 4: move a T1 virtual machine to H3 from H1;
Step 5: the internal memory loading condition again detecting H1, load is 96.2%, is greater than load threshold 90%, repeat step 2-4, until the load of H1 lower than in threshold value 90% or cluster without suitable Hypervisor;
Step 6: when the load of H1 is lower than 90%, when being 89%, virtual machine is distributed as: T1,18; T2,30; T3,5;
Step 7:H1's can storage allocation be 32G, and it is 106G that real resource distribution takies situation, and be greater than to surpass and join threshold value 85%, increase the super distribution coefficient of H1, after increasing, the super distribution coefficient of H1 is 2, i.e. 32*2=64G;
Step 8: again detect distribution, resource allocation conditions is 165.5%, is greater than to surpass to join threshold value 85%, continues the super distribution coefficient increasing H1, until resource allocation conditions is less than to surpass join threshold value.
By the adjustment in 1 cycle, Hypervisor and virtual machine distribution are as form 7.
Form 7
As can be seen from form 7, the load of each Hypervisor main frame is lower than the Hypervisor load threshold arranged, and the distribution of virtual machine is relatively balanced with form 1.Also by amendment Hypervisor load threshold, can continue dynamically to reduce H1, the load of H2, improve the load of H3.
Form 5 is compared with form 7, and excessive distribution coefficient and the virtual distribution of each Hypervisor are not identical.But all meet pre-conditioned, visible adaptability of the present invention and extensibility.
The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (10)
1., based on the excessive distribution of Hypervisor resource and the dynamic adjusting method of Load-aware, it is characterized in that, comprise the steps:
Step 1, arranges Hypervisor resource load threshold, Hypervisor resource excessive distribution threshold value, the excessive distribution coefficient of each Hypervisor main frame and excessive distribution regulation coefficient;
Step 2, arranges the scheduling strategy of Hypervisor cluster virtual machine;
Step 3, monitors the loading condition of each Hypervisor main frame;
Step 4, judges whether the load of each Hypervisor main frame is more than or equal to Hypervisor resource load threshold successively, if it is reduces the excessive distribution coefficient of this Hypervisor main frame according to excessive distribution regulation coefficient, performs step 5, otherwise performs step 6;
Step 5, according to the excessive distribution coefficient of this Hypervisor main frame after adjustment and the adjustment of the scheduling strategy described in step 2 virtual machine in the distribution of Hypervisor cluster, returns step 3;
Step 6, detects the resource excessive distribution situation of this Hypervisor main frame;
Step 7, judge whether the resource excessive distribution ratio of this Hypervisor main frame is more than or equal to Hypervisor resource excessive distribution threshold value, if it is increase the excessive distribution coefficient of this Hypervisor main frame according to excessive distribution regulation coefficient, return step 6, otherwise process ends.
2. a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic adjusting method according to claim 1, it is characterized in that, the cluster of Hypervisor described in step 2 virtual machine scheduling policy adopts any one in first-fit, best-fit, packing, loadbalance, Round-Robin or random.
3. a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic adjusting method according to claim 1, it is characterized in that, also comprise the excessive distribution coefficient default value pre-setting Hypervisor main frame, the excessive distribution coefficient of each Hypervisor main frame is according in excessive distribution regulation coefficient dynamic adjustment process, and the excessive distribution coefficient after adjustment can not lower than the excessive distribution coefficient default value pre-setting Hypervisor main frame.
4. a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic adjusting method according to claim 1, is characterized in that, being implemented as of step 5:
Step 5.1, detects each Hypervisor main frame one by one;
Step 5.2, judges whether the load of Hypervisor main frame to be detected is less than Hypervisor resource load threshold, if perform step 5.3, otherwise performs step 5.5;
Step 5.3, judges whether the resource excessive distribution ratio of this Hypervisor main frame is less than Hypervisor resource excessive distribution threshold value, if perform step 5.4, otherwise performs step 5.5;
Step 5.4, is more than or equal to virtual machine (vm) migration on the Hypervisor main frame of Hypervisor resource load threshold to the Hypervisor main frame described in step 5.3 by load;
Step 5.5, end process flow process.
5. a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic adjusting method according to claim 1, is characterized in that, also comprises according to adjustment result adjustment Hypervisor resource load threshold and/or Hypervisor resource excessive distribution threshold value.
6., based on the excessive distribution of Hypervisor resource and the dynamic debugging system of Load-aware, it is characterized in that, comprise parameter setting module, scheduling strategy arranges module, load monitoring module, load dispatch module and excessive distribution monitoring module;
Described parameter setting module, it is for arranging Hypervisor resource load threshold, Hypervisor resource excessive distribution threshold value, the excessive distribution coefficient of each Hypervisor main frame and excessive distribution regulation coefficient;
Described scheduling strategy arranges module, and it is for arranging the scheduling strategy of Hypervisor cluster virtual machine;
Described load monitoring module, it is for judging whether the load of each Hypervisor main frame is more than or equal to Hypervisor resource load threshold successively, if it is the excessive distribution coefficient of this Hypervisor main frame is reduced according to excessive distribution regulation coefficient, call load dispatch module, otherwise call excessive distribution monitoring module;
Described load dispatch module, it is for according to the distribution at Hypervisor cluster of the excessive distribution coefficient of this Hypervisor main frame after adjustment and scheduling strategy adjustment virtual machine, until the load of this Hypervisor main frame is less than Hypervisor resource load threshold;
Described excessive distribution monitoring module, it is for detecting the resource excessive distribution situation of this Hypervisor main frame, judge whether the resource excessive distribution ratio of this Hypervisor main frame is more than or equal to Hypervisor resource excessive distribution threshold value, if it is the excessive distribution coefficient of this Hypervisor main frame is increased according to excessive distribution regulation coefficient, continue the resource excessive distribution situation detecting this Hypervisor main frame, until the excessive distribution ratio of this Hypervisor main frame is less than Hypervisor resource load threshold, otherwise process ends.
7. a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic debugging system according to claim 6, it is characterized in that, described scheduling strategy arranges any one in the employing of Hypervisor cluster virtual machine scheduling policy first-fit, best-fit, packing, loadbalance, Round-Robin or random of module installation.
8. a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic debugging system according to claim 6, it is characterized in that, described parameter setting module also pre-sets the excessive distribution coefficient default value of Hypervisor main frame, the excessive distribution coefficient of each Hypervisor main frame is according in excessive distribution regulation coefficient dynamic adjustment process, and the excessive distribution coefficient after adjustment can not lower than the excessive distribution coefficient default value pre-setting Hypervisor main frame.
9. a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic debugging system according to claim 6, it is characterized in that, described load dispatch module comprises resources measurement unit, the first judging unit, the second judging unit and virtual machine (vm) migration unit;
Described resources measurement unit, it is for detecting each Hypervisor main frame one by one;
Described first judging unit, it is for judging whether the load of Hypervisor main frame to be detected is less than Hypervisor resource load threshold, if call the second judging unit, otherwise the flow process that ends process;
Described second judging unit, it is for judging whether the resource excessive distribution ratio of this Hypervisor main frame is less than Hypervisor resource excessive distribution threshold value, if call virtual machine (vm) migration unit, end process flow process;
Described virtual machine (vm) migration unit, it is for being more than or equal to virtual machine (vm) migration on the Hypervisor main frame of Hypervisor resource load threshold to described Hypervisor main frame by load.
10. a kind of Hypervisor resource excessive distribution based on Load-aware and dynamic debugging system according to claim 6, is characterized in that, also comprises according to adjustment result adjustment Hypervisor resource load threshold and/or Hypervisor resource excessive distribution threshold value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510140721.7A CN104991825B (en) | 2015-03-27 | 2015-03-27 | A kind of excessive distribution of Hypervisor resource and dynamic adjusting method and system based on Load-aware |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510140721.7A CN104991825B (en) | 2015-03-27 | 2015-03-27 | A kind of excessive distribution of Hypervisor resource and dynamic adjusting method and system based on Load-aware |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104991825A true CN104991825A (en) | 2015-10-21 |
| CN104991825B CN104991825B (en) | 2019-07-05 |
Family
ID=54303641
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510140721.7A Active CN104991825B (en) | 2015-03-27 | 2015-03-27 | A kind of excessive distribution of Hypervisor resource and dynamic adjusting method and system based on Load-aware |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104991825B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106059940A (en) * | 2016-05-25 | 2016-10-26 | 杭州昆海信息技术有限公司 | Flow control method and device |
| CN107707424A (en) * | 2017-09-11 | 2018-02-16 | 厦门集微科技有限公司 | The control method and system of load condition |
| CN110930291A (en) * | 2019-11-15 | 2020-03-27 | 山东英信计算机技术有限公司 | GPU video memory management control method and related device |
| CN113722100A (en) * | 2021-09-02 | 2021-11-30 | 上海仪电（集团）有限公司中央研究院 | Dynamic super-allocation method and device for cloud platform memory resources based on trend prediction |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080034370A1 (en) * | 2006-08-07 | 2008-02-07 | International Business Machines Corporation | Method For Balancing Resource Sharing And Application Latency Within A Data Processing System |
| US20080307188A1 (en) * | 2007-06-06 | 2008-12-11 | International Business Machines Corporation | Management of Guest OS Memory Compression In Virtualized Systems |
| CN101968749A (en) * | 2010-09-26 | 2011-02-09 | 华中科技大学 | Method for receiving message passing interface (MPI) information under circumstance of over-allocation of virtual machine |
| CN102508718A (en) * | 2011-11-22 | 2012-06-20 | 杭州华三通信技术有限公司 | Method and device for balancing load of virtual machine |
| US20120272234A1 (en) * | 2011-04-20 | 2012-10-25 | Cisco Technology, Inc. | Scoring of Computing Equipment for Virtualized Computing Environments |
| US20130047159A1 (en) * | 2011-08-18 | 2013-02-21 | International Business Machines Corporation | Preserving, From Resource Management Adjustment, Portions Of An Overcommitted Resource Managed By A Hypervisor |
| CN103502944A (en) * | 2011-11-28 | 2014-01-08 | 华为技术有限公司 | Method and device for adjusting memories of virtual machines |
| CN104050008A (en) * | 2013-03-15 | 2014-09-17 | 中兴通讯股份有限公司 | Excessive distribution management system and method for memory |
-
2015
- 2015-03-27 CN CN201510140721.7A patent/CN104991825B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080034370A1 (en) * | 2006-08-07 | 2008-02-07 | International Business Machines Corporation | Method For Balancing Resource Sharing And Application Latency Within A Data Processing System |
| US20080307188A1 (en) * | 2007-06-06 | 2008-12-11 | International Business Machines Corporation | Management of Guest OS Memory Compression In Virtualized Systems |
| CN101968749A (en) * | 2010-09-26 | 2011-02-09 | 华中科技大学 | Method for receiving message passing interface (MPI) information under circumstance of over-allocation of virtual machine |
| US20120272234A1 (en) * | 2011-04-20 | 2012-10-25 | Cisco Technology, Inc. | Scoring of Computing Equipment for Virtualized Computing Environments |
| US20130047159A1 (en) * | 2011-08-18 | 2013-02-21 | International Business Machines Corporation | Preserving, From Resource Management Adjustment, Portions Of An Overcommitted Resource Managed By A Hypervisor |
| CN102508718A (en) * | 2011-11-22 | 2012-06-20 | 杭州华三通信技术有限公司 | Method and device for balancing load of virtual machine |
| CN103502944A (en) * | 2011-11-28 | 2014-01-08 | 华为技术有限公司 | Method and device for adjusting memories of virtual machines |
| CN104050008A (en) * | 2013-03-15 | 2014-09-17 | 中兴通讯股份有限公司 | Excessive distribution management system and method for memory |
Non-Patent Citations (1)
| Title |
|---|
| KAPIL ARYA，YURY BASKAKOV，ALEX GARTHWAITE: "Tesseract：reconciling guest IO and hypervisor swapping in a VM", 《PROCEEDINGS OF THE 10TH ACM SIGPLAN/SIGOPS INTERNATIONAL CONFERENCE ON VIRTUAL EXECUTION ENVIRONMENTS》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106059940A (en) * | 2016-05-25 | 2016-10-26 | 杭州昆海信息技术有限公司 | Flow control method and device |
| CN106059940B (en) * | 2016-05-25 | 2019-07-09 | 新华三信息技术有限公司 | A kind of flow control methods and device |
| CN107707424A (en) * | 2017-09-11 | 2018-02-16 | 厦门集微科技有限公司 | The control method and system of load condition |
| CN110930291A (en) * | 2019-11-15 | 2020-03-27 | 山东英信计算机技术有限公司 | GPU video memory management control method and related device |
| CN110930291B (en) * | 2019-11-15 | 2022-06-17 | 山东英信计算机技术有限公司 | GPU video memory management control method and related device |
| CN113722100A (en) * | 2021-09-02 | 2021-11-30 | 上海仪电（集团）有限公司中央研究院 | Dynamic super-allocation method and device for cloud platform memory resources based on trend prediction |
| CN113722100B (en) * | 2021-09-02 | 2023-12-15 | 上海仪电（集团）有限公司中央研究院 | Dynamic super-allocation method and equipment for cloud platform memory resources based on trend prediction |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104991825B (en) | 2019-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10824215B2 (en) | Managing power budget of multiple computing node clusters in a computing rack system | |
| CN103761139B (en) | General purpose computation virtualization implementation method based on dynamic library interception | |
| CN104834569B (en) | A kind of cluster resource dispatching method and system based on application type | |
| CN103605613A (en) | Method and system for dynamically adjusting virtual-machine memory in cloud computing environment | |
| CN104991825A (en) | Hypervisor resource hyper-allocation and dynamic adjusting method and system based on load awareness | |
| CN102567072B (en) | Resource allocation method, resource allocation device and resource allocation system | |
| CN103336722A (en) | Virtual machine CPU source monitoring and dynamic distributing method | |
| CN106293871A (en) | A kind of resource regulating method of cluster virtual machine | |
| CN106020934A (en) | Optimized deploying method based on virtual cluster online migration | |
| CN103502944A (en) | Method and device for adjusting memories of virtual machines | |
| CN103856512A (en) | Cloud computing management server, working host, idle host and resource scheduling method | |
| GB2493812A (en) | Scaling resources for virtual machines, using comparison with established threshold values | |
| CN104133727A (en) | Load distribution method based on real-time resources | |
| CN103051564A (en) | Dynamic resource allocation method and device | |
| CN103455880A (en) | Electric network dispatching automation system based on virtualization technology | |
| CN102339233A (en) | Cloud computing centralized management platform | |
| CN104461731A (en) | Virtual machine resource recycling method and device in dynamic resource expanding | |
| CN103577306A (en) | Active-detecting high-availability software operating method in cloud platform | |
| CN108279967A (en) | A kind of virtual machine and container mixed scheduling method | |
| CN106775925B (en) | Quota processing method and device for CPU of virtual machine | |
| CN105183563A (en) | CPU resource dynamic self-configuration method facing mission critical computer | |
| CN105528054A (en) | Cluster system comprehensive scheduling energy saving method and device | |
| CN105025104A (en) | Cloud desktop scheduling method and system | |
| CN103955397A (en) | Virtual machine scheduling multi-strategy selection method based on micro-architecture perception | |
| CN104111892A (en) | Internal memory dynamic distribution method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A method and system of hypervisor resource over allocation and dynamic adjustment based on load perception Effective date of registration: 20210510 Granted publication date: 20190705 Pledgee: Beijing ustron Tongsheng financing Company limited by guarantee Pledgor: BEIJING SKYCLOUD RONGCHUANG SOFTWARE TECHNOLOGY Co.,Ltd. Registration number: Y2021990000410 |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right |