CN103365704A - Memory pre-copying method in virtual machine migration, device executing memory pre-copying method and system - Google Patents
Memory pre-copying method in virtual machine migration, device executing memory pre-copying method and system Download PDFInfo
- Publication number
- CN103365704A CN103365704A CN2012100824002A CN201210082400A CN103365704A CN 103365704 A CN103365704 A CN 103365704A CN 2012100824002 A CN2012100824002 A CN 2012100824002A CN 201210082400 A CN201210082400 A CN 201210082400A CN 103365704 A CN103365704 A CN 103365704A
- Authority
- CN
- China
- Prior art keywords
- dirty page
- migration
- leaf
- virtual machine
- copy
- 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
Images
Abstract
Disclosed are a memory pre-copying method in virtual machine migration, a virtual machine migration monitor executing the memory pre-copying method and a system comprising the virtual machine migration monitor. The memory pre-copying method includes: calculating a current average dirty page factor after every turn of iterative copying is finished, wherein the current average dirty page factor is an average value of dirty page factors of all turns of iterative copying already finished, and a dirty page factor of every turn of iterative copying is a ratio between a dirty page generating speed of the current turn and a migration bandwidth of the same turn; comparing the calculated average dirty page factor with a preset value; adaptively adjusting and the migration bandwidth for executing the next turn of iterative copying according to the comparison result.
Description
Technical field
The application relates to the virtual machine (vm) migration technology, is specifically related to the internal memory pre-copy method in the virtual machine (vm) migration and carries out device and the system of the method.
Background technology
In recent years, along with the continuous rise of the technology such as cloud computing, data center, distributed environment, the server virtualization technology more and more becomes the focus of IT field.The virtual machine (vm) migration technology provides easily method for server virtualization.In the situations such as the resource that virtual machine (vm) migration need between the host in the data center to refer to adjust virtual machine distributes, certain host needs repairing, a dummy machine system is moved to the process of another physical host from a physical host.The essence of the migration of virtual machine is the migration of complete state, such as the content in CPU state, internal memory and the disk.
The virtual machine migration method of at present industry proposition has three kinds: Cold Static Migration (calm attitude migration), Warm Static Migration (hot static migrating), Live Migration (dynamic migration).Two kinds of fronts all are the methods of shutting down migration, and are larger on the impact of the application service on the VME operating system because it is longer to shut down the stop time that the method for migration brings, therefore general less employing.At present research is more, use wider is the method for dynamic migration, and the stop time that this method is brought is extremely short, does not affect or very littlely has influence on application service, therefore has very much practical value.
Because in modern data center, generally between the virtualization system adopt the shared network storage, so disc content do not need to move veritably, virtual machine (vm) migration is mainly reflected on the internal memory migration.For dynamic migration, its migration to internal memory is main with two kinds of methods at present: copy method behind the internal memory (Post-copy), internal memory pre-copy method (Pre-copy).Copy refers in the virtual machine (vm) migration process behind the internal memory, at first stop source virtual machine, the cpu status information is copied on the purpose physical machine, then start this virtual machine in the purpose physical machine, virtual machine also constantly wins over internal memory in the operation of purpose physical machine from the source physical machine afterwards.On the contrary, the internal memory pre-copy is that at first the mode by iteration constantly copies the internal memory of source virtual machine on the purpose physical machine, arrival to a certain degree stops afterwards source virtual machine and the cpu status information is copied to the purpose physical machine, then starts this virtual machine in the purpose physical machine.
Because high-quality and the high efficiency of internal memory pre-copy method, commercial virtual instrument now all are to adopt the method mostly.
In order to realize the dynamic migration of virtual machine, the process of a complete internal memory migration is divided into following 3 stages usually: Push stage, Stop-and-Copy stage and Pull stage.Push stage and Stop-and-Copy stage have only been adopted based on the migration of internal memory pre-copy, its main thought is as follows: when the virtual machine on a certain physical machine of needs migrations is to another physical machine, at first in the Push stage page is copied on the purpose machine so that the mode of iteration is rounds of, the first round copies all pages, second takes turns the page that copy was only revised in first round iterative process, by that analogy, n wheel copy is the page of revising in the n-1 wheel iterative process.When the number of dirty page or leaf was reduced to certain predefined constant or iterations and reaches certain predefined maximum times, the pre-copy stage finished, and enters into the Stop-and-Copy stage.At this moment, the information copy such as shutdown and a remaining dirty page or leaf and running status, CPU state are gone over.
For example, in the system architecture of existing Xen Intel Virtualization Technology, be provided with the migration management module as the main modular of carrying out virtual machine (vm) migration.Suppose that the virtual machine on the host A need to move on host B, concrete virtual machine (vm) migration flow process is as follows:
Step 0: on the host A operation virtual machine is arranged after, for to future possible virtual machine (vm) migration ready in advance, resource management system RMS will select in advance what a main frame as the migration target, such as host B.
Step 1: resource management system RMS gives an order, and determines virtual machine on the migration host A such as VM1 to host B, and host B feedback information (whether can hold another virtual machine such as oneself) is given RMS, and RMS determines whether prepare migration according to feedback information.
Step 2: the result who supposes step 1 can move, and the migration management module is carried out virtual machine (vm) migration on the host A, enters into the iteration pre-copy stage.VM still moves at source host A, host A copies on the host B with the mode of the iteration page with VM, the first round copies all pages, later each is taken turns and is only copied the page or leaf that was modified in the previous round iterative process, by that analogy, until iterations reaches maximum iteration time or remaining dirty page or leaf is reduced to a predetermined value, then enter into the Stop-and-Copy stage.
Step 3: enter into the Stop-and-Copy stage, the VMM on the host A hangs up source virtual machine, and network redirection is arrived the target virtual machine, and then the migration management module moves to host B to remaining dirty page or leaf and CPU state at this one-phase.
Step 4: host B sends message to host A and represents successful sink virtual machine, and host A confirms and destroy source virtual machine, and host B is enabled the virtual machine of migration.
Step 5: move to the device drives on the virtual machine use B on the host B, and issue new ARP message.
In said method, the enormous bandwidth that memory copying takies when limiting virtual machine (vm) migration is supported a kind of method of dynamic adjustment migration occupied bandwidth in the migration management module of Xen.In the method, the keeper predesignates the bandwidth upper and lower bound that migration can take; First round pre-copy is moved internal memory with lower band; The every wheel calculated the dirty speed (dirtying rate) of page when finishing, this is to take turns the lasting time with this size of taking turns dirty page or leaf divided by this to obtain; The bandwidth value of rear one setting of taking turns is that the dirty speed of previous round adds 50Mbit/sec; When reaching the default bandwidth upper limit or reaching predetermined iterations or be lower than default minimum amount of ram, stop pre-copy; And transmit to guarantee that with maximum rate stop time is the shortest in the stop-and-copy stage.
In existing dynamic migration scheme, only try one's best short and shorten the internal memory pre-copy time as target as far as possible take the downtime (downtime) of guaranteeing the stop-and-copy stage, can reduce to minimum degree downtime by iteration pre-copy internal memory, and the gross migration time is controlled in the acceptable scope, yet, it does not consider the pressure of internal memory pre-copy local area network switch, may cause LAN switch congested, transmission or other virtual machine (vm) migration of the LAN (Local Area Network) that coexists or effective transmission of administrative messag of the important business data of a LAN (Local Area Network) thereby impact coexists.
Generally, the amount of ram that virtual machine need to move is very large, if in transition process, do not carry out limit bandwidth, internal memory might be within the extremely short time at a high speed or accelerate to migrate to target machine, local switch is caused huge pressure, affect other network application of local network.In addition, carry out at the same time then more likely so that local switch pressure is excessive, causing the congested of network under the scene that the multi-section virtual machine moves simultaneously, also affect all virtual machine (vm) migration processes.More especially for the very high application service of reliability, time delay and qos requirement, very responsive for network congestion, as therefore to need especially the control virtual machine (vm) migration that network is caused impact.
In addition; although the method for existing Xen restricted migration bandwidth can be controlled the migration bandwidth; but because its every bandwidth of taking turns all is to add 50Mbit/sec on the basis of last round of dirty page or leaf speed; taking of bandwidth can not be adjusted as required, stop time, gross migration time can not be well realized, on compromise between this three of affecting of network side.For some application services very high to quality of service requirement, need the method for new restricted migration bandwidth to control migration bandwidth and stop time, make both and keep within the acceptable range.
Summary of the invention
Device and system that the application is intended to propose the internal memory pre-copy method in a kind of virtual machine (vm) migration and carries out the method are so that realize reaching balance between downtime, gross migration time, these three factors of pressure to network in the migration of virtual machine.
First aspect according to the application, internal memory pre-copy in a kind of virtual machine (vm) migration method is provided, comprise: after each takes turns the end of iteration copy, calculate the current average dirty page or leaf factor, the described current average dirty page or leaf factor is the mean value that current all that have finished are taken turns the dirty page or leaf factor in the iteration copy, and the described dirty page or leaf factor that each is taken turns is that this dirty page or leaf of taking turns produces the ratio between speed and this migration bandwidth of taking turns; The average dirty page or leaf factor and the predetermined value that calculate are compared; And according to the result of described comparison, adjust adaptively the migration bandwidth of carrying out next round iteration copy.
Second aspect according to the application, a kind of virtual machine (vm) migration watch-dog is provided, comprise: dirty page or leaf factor computing module, after each takes turns the end of iteration copy, calculate the current average dirty page or leaf factor, the described current average dirty page or leaf factor is the mean value that current all that have finished are taken turns the dirty page or leaf factor in the iteration copy, and the described dirty page or leaf factor that each is taken turns is that this dirty page or leaf of taking turns produces the ratio between speed and this migration bandwidth of taking turns; The migration management module comprises comparer, and described comparer compares the average dirty page or leaf factor and the predetermined value that described dirty page or leaf factor computing module calculates; And the bandwidth control module, adjust adaptively the migration bandwidth of carrying out next round iteration copy according to the comparative result of described comparer.
The third aspect according to the application, a kind of virtual machine (vm) migration system is provided, comprise a plurality of above-mentioned virtual machine (vm) migration watch-dogs and the resource management system of communicating by letter with described a plurality of virtual machine (vm) migration watch-dogs, wherein, each virtual machine (vm) migration watch-dog corresponds respectively to a main frame setting that can move virtual machine.
According to the application, in the virtual machine (vm) migration process, added the restriction to memory copying institute dedicated bandwidth.Therefore; the target of migration no longer only is downtime and the minimum gross migration time that reaches minimum; but realize downtime, gross migration time, to a balance between these three factors of pressure of network; stop time and the bandwidth that takies all are controlled in certain scope, thereby can adapt to better some application services very high to quality of service requirement.
Description of drawings
Fig. 1 shows the Organization Chart according to the virtual machine (vm) migration system of an embodiment of the application.
Fig. 2 shows the exemplary process diagram according to the internal memory pre-copy method in the application's the virtual machine (vm) migration.
Embodiment
The application is described in conjunction with exemplary embodiment hereinafter with reference to accompanying drawing.
As indicated above, in the migration of virtual machine, downtime, gross migration time and be several very important factors to the pressure of network side.The gross migration time refers to that virtual machine intactly moves on the target physical machine and starts the needed time from the source physical machine.In internal memory pre-copy mechanism, downtime mainly is to stop the source physical machine and copy remaining page and the needed time of CPU state.Thereby the transmission of a large amount of internal memories took huge bandwidth the service of network side is impacted when the pressure of network side then was the migration internal memory.
Fig. 1 shows the Organization Chart according to the virtual machine (vm) migration system of an embodiment of the application.The virtual machine (vm) migration system comprises resource management system (RMS) 110 and corresponding to a plurality of monitor of virtual machine of a plurality of main frames.For for simplicity, only show two the monitor of virtual machine 120a and the 120b that correspond respectively to host A and host B among Fig. 1.But should be appreciated that the virtual machine (vm) migration system can comprise the monitor of virtual machine of any right quantity, and be not limited in two.Can move one or more virtual machines on each main frame.Exemplarily, Fig. 1 shows two virtual machine VM1 of operation and VM2 on the host A, two virtual machine VM3 of operation and VM4 on the host B.Move respectively two virtual machines on each main frame although show, should be appreciated that the virtual machine that can move any right quantity on each main frame.Monitor of virtual machine corresponding to each main frame manages the virtual machine that moves on the respective host.For example, in example shown in Figure 1, monitor of virtual machine 120a manages virtual machine VM1 and the VM2 that moves on the host A, and monitor of virtual machine 120b manages virtual machine VM3 and the VM4 that moves on the main frame.
Hereinafter, in the description of not distinguishing monitor of virtual machine 120a and 120b, monitor of virtual machine 120a and 120b are referred to as monitor of virtual machine 120.According to the application, monitor of virtual machine 120 comprises dirty page or leaf factor computing module 1201, migration management module 1202 and bandwidth control module 1203.
The processing of carrying out in virtual machine (vm) migration below in conjunction with monitor of virtual machine 120 is described monitor of virtual machine 120.
So that the virtual machine VM1 on the host A is moved on the host B as example.
At first, send the inquiry message that whether the virtual machine VM1 on the host A can be migrated to host B to host B by resource management system (RMS) 110, then host B feedback information (whether can hold another virtual machine such as oneself) is given RMS 110, and RMS 110 determines whether prepare migration according to the feedback information of host B.If feedback result is to move, RMS110 selects another main frame to carry out above-mentioned inquiry again.If feedback result is to move, 1202 of the migration management modules among the monitor of virtual machine 120a on the host A are carried out virtual machine (vm) migration, enter into the iteration pre-copy stage.At this moment, VM1 still moves at host A, and host A copies on the host B with the mode of the iteration page with VM1.
Although should be appreciated that dirty page or leaf factor computing module 1201, migration management module 1202 and bandwidth control module 1203 will be shown among Fig. 1 to be illustrated as three parts, one or more functions of these three modules can be integrated in the module and realize.
Describe according to the internal memory pre-copy method 200 in the application's the virtual machine (vm) migration referring to the exemplary process diagram shown in Fig. 2.
At first, at step S201, set a transmission bandwidth (or migration bandwidth), and carry out first round copy with the bandwidth of setting.In first round copy, all pages that copy virtual machine VM1 need to copy.After the first round, copy finished, at step S202, calculate the current average dirty page or leaf factor by dirty page or leaf factor computing module 1201, namely, current all that have finished are taken turns the mean value of the dirty page or leaf factor in the iteration copy, wherein, the described dirty page or leaf factor that each is taken turns is that this dirty page or leaf of taking turns produces the ratio between speed (dirtying rate) and this migration bandwidth of taking turns, and dirty page or leaf generation speed is to obtain divided by this duration of taking turns with this dirty page or leaf of taking turns (page or leaf that was modified in the copy procedure that this is taken turns) size.The average dirty page or leaf factor of calculating after the first round, the iteration copy finished is exactly the dirty page or leaf factor of the first round.
Then, at step S203, migration management module 1202 compares dirty page or leaf factor computing module 1201 in the average dirty page or leaf factor and the predefined reference value (predetermined value) that step S202 calculates.The comparative example here is as realizing by the comparer (not shown) is set in migration management module 1202.For example, the predetermined value of use can rule of thumb be set by the keeper, also can be that the full memory amount according to virtual machine needs copy to be migrated, described predetermined iterations threshold value and described predetermined dirty page or leaf surplus calculate.Hereinafter will introduce a kind of method of determining this predetermined value.
Then, bandwidth control module 1203 based on the comparison the result adjust adaptively the next round iteration and copy employed migration bandwidth.For example, the average dirty page or leaf factor that calculates when dirty page or leaf factor computing module 1201 is during greater than this predetermined value, and bandwidth control module 1203 determines that at step S204 the migration bandwidth that increases carries out next round iteration copy; When the average dirty page or leaf factor that calculates when dirty page or leaf factor computing module 1201 was less than or equal to this predetermined value, bandwidth control module 1203 was determined to reduce or the migration bandwidth that remains unchanged is carried out next round iteration copy at step S205.Then, at step S206, carry out the next round copy with the migration bandwidth after adjusting.
According to an embodiment, the constraint that bandwidth control module 1203 can be less than or equal to based on the average dirty page or leaf factor after next round iteration copy is finished above-mentioned predetermined value comes definite (or adjustment) to carry out the migration bandwidth of next round iteration copy.For example, the average dirty page or leaf factor that calculates when dirty page or leaf factor computing module 1201 is during greater than this predetermined value, bandwidth control module 1203 can be less than or equal to based on the average dirty page or leaf factor after described next round iteration copy is finished the constraint of described predetermined value, determines the migration bandwidth of the increase of execution next round iteration copy.When the average dirty page or leaf factor that calculates when dirty page or leaf factor computing module 1201 is less than or equal to this predetermined value, bandwidth control module 1203 also can be less than or equal to based on the average dirty page or leaf factor after described next round iteration copy is finished the constraint of described predetermined value, determines to carry out the migration bandwidth that reduces of next round iteration copy.Hereinafter will introduce bandwidth control module 1203 according to a kind of method of average dirty page or leaf factor adjustment migration bandwidth.
Like this, when each took turns iteration, the dirty page or leaf of respectively taking turns in the iterative process according to the front produced speed and internal memory transfer rate (corresponding to the migration bandwidth), determined that this takes turns the transfer rate in the iterative process.Owing to having controlled the speed of internal memory transmission, therefore having reduced the impact on network.
According to an embodiment, after step S206, go to step S207, migration management module 1202 determines whether iterations reaches predetermined iterations threshold value and remain dirty page or leaf whether be less than or equal to predetermined dirty page or leaf surplus.
When determining that at step S207 iterations reaches predetermined iterations threshold value or remains dirty page or leaf when being less than or equal to predetermined dirty page or leaf surplus, process at step S208 finishing iteration copy, enter Stop-and-Copy, monitor of virtual machine 120a on the host A hangs up source virtual machine VM1, and network redirection to the target virtual machine, then the migration management module 1202 of monitor of virtual machine 120a moves to host B to remaining dirty page or leaf and CPU state at this one-phase, so far sets up virtual machine VM1 at host B.Send to host A after message represents successful sink virtual machine VM1 at host B, host A confirms and destroys virtual machine VM1, and host B is enabled the virtual machine VM1 of migration.Move to the device drives on the virtual machine VM1 use B on the host B, and issue new ARP message.
If when step S207 determines that iterations does not reach predetermined iterations threshold value and remains dirty page or leaf greater than predetermined dirty page or leaf surplus, then go to step S202, calculate the current average dirty page or leaf factor, and the next operation of repeating step S203-S206.
In above-mentioned iteration copy method, two important parameters have been considered, that is, the speed that dirty page or leaf produces and the copy speed (migration bandwidth) of internal memory reach balance thereby make between downtime in the virtual machine (vm) migration process, gross migration time, these three factors of pressure to network
Below by exemplary embodiment a kind of definite method of the reference point that compares with the average dirty page or leaf factor and a kind of method of adjustment of migration bandwidth are described.
At first, as follows to parameter and the variable-definition that relates in the internal memory transition process:
M: the full memory amount that virtual machine to be migrated need to copy;
P: dirty page or leaf produces speed;
B: copy bandwidth;
N: constant is the upper limit of predefined iteration copy number of times;
H: constant, the amount of ram that needs copy when iteration pre-copy residue stop the iteration pre-copy during less than or equal to h, enter into and shut down the copy stage;
B
i: the bandwidth of i wheel copy;
p
i: the speed that i wheel copy procedure apoplexy involving the solid organs page or leaf produces, the i.e. dirty number of pages that is modified in the unit interval;
t
i: the time of i wheel copy;
N: the wheel number of actual iteration copy.
According to an embodiment, the predetermined value that compares with the average dirty page or leaf factor is
Its derivation is as follows:
P wherein
I-1* t
I-1Being the dirty page or leaf that produces in the i-1 wheel process, also is that the i wheel needs copy.In actual iteration copy procedure, each is taken turns before the copy all and can put into directly that stop-and-copy transmitted in the stage to some frequent pages of revising and iteration copy no longer, so the epicycle page that need to copy generally is less than dirty page of last round of middle generation.In order to simplify mathematical model, ignored these frequent pages of revising in the above-mentioned formula, the page that copies in this hypothesis epicycle namely is the dirty page or leaf of last round of middle generation.
n＝min({i+1|pi×ti≤h}，N) ②
By formula 1. as seen, the ratio p/B of p and B is the important factor that affects transit time and stop time.At this, p/B is the dirty page or leaf factor defined above.
For internal memory pre-copy method is come into force, will remain on stop time in the acceptable scope, dirty page or leaf need to reduce to h before N wheel copy so, therefore has:
p
N-1* t
N-1≤ h, 1. the substitution formula then has
Suppose that each p/B that takes turns is a constant, i.e. p
i/ B
iBe certain value, and the formula distortion, then have:
4. formula illustrates, make the internal memory pre-copy reach preferably effect, each take turns in the iteration copy procedure dirty page or leaf factor p/B preferably can less than or equal to
Like this, before the i wheel iteration except the first round began, bandwidth control function obtained last round of iterative process apoplexy involving the solid organs page or leaf factor p from dirty page or leaf factor computing function
I-1/ B
I-1And the average dirty page or leaf factor that all are taken turns before calculating thus
If
Illustrate i-1 wheel and before all dirty page or leaf factors of taking turns in the iterative process too large, the i wheel needs to increase B so that the dirty page or leaf factor that this is taken turns reduces, thereby make that this takes turns iterative process toward good future development; If
The iteration copy that all are taken turns above illustrating maintains acceptable scope, and function is carried out in this Time Bandwidth control can be according to network load condition, the bandwidth of not adjusting bandwidth or suitably reducing to move.
After the bandwidth value of having determined next round is to increase or to reduce, according to an embodiment, can adopt following bandwidth estimation method to determine the size adjustment of next round bandwidth value.
Before i wheel iteration, the dirty page or leaf of i-1 wheel produces speed and can calculate, and is p
I-1, in order to draw the bandwidth predicted value of i wheel iteration, suppose that the dirty page or leaf of i wheel produces speed p
iBe all average dirty page or leaf generation rate, i.e. p that take turns before
i=(p
1+ p
2+ ... ..p
I-1)/(i-1).All average dirty page or leaf factors of taking turns reach acceptable scope in order to make the i wheel reach before, have so:
Can get:
The bandwidth value that 5. formula draws namely can be used as a reference value setting i wheel bandwidth.
Although above the application's illustrative embodiments and embodiment are illustrated; but be to be understood that; the application is not limited to the concrete restriction of the above-described embodiment and examples; the various suitable distortion that those skilled in the art can make under the prerequisite that does not depart from spirit and scope of the invention according to content disclosed herein, the application's protection domain is defined by the claims.
Claims (13)
1. the internal memory pre-copy method in the virtual machine (vm) migration comprises:
After each takes turns the end of iteration copy, calculate the current average dirty page or leaf factor, the described current average dirty page or leaf factor is the mean value that current all that have finished are taken turns the dirty page or leaf factor in the iteration copy, and the described dirty page or leaf factor that each is taken turns is that this dirty page or leaf of taking turns produces the ratio between speed and this migration bandwidth of taking turns;
The average dirty page or leaf factor and the predetermined value that calculate are compared; And
According to the result of described comparison, adjust adaptively the migration bandwidth of carrying out next round iteration copy.
2. the method for claim 1, wherein the described migration bandwidth of carrying out next round iteration copy of adjusting adaptively comprises:
When the average dirty page or leaf factor that calculates during greater than described predetermined value, carry out next round iteration copy with the migration bandwidth that increases; And
When the average dirty page or leaf factor that calculates is less than or equal to described predetermined value, carry out next round iteration copy with migration bandwidth constant or that reduce.
3. the method for claim 1 further comprises:
After each takes turns the end of iteration copy, determine whether iterations reaches predetermined iterations threshold value and remain dirty page or leaf whether be less than or equal to predetermined dirty page or leaf surplus; And
When determining that iterations reaches predetermined iterations threshold value or remains dirty page or leaf when being less than or equal to predetermined dirty page or leaf surplus, the process of finishing iteration copy.
4. method as claimed in claim 3, wherein, described predetermined value is according to the full memory amount of virtual machine needs copy to be migrated, described predetermined iterations threshold value and described predetermined dirty page or leaf surplus and definite.
5. method as claimed in claim 4, wherein, described predetermined value is
Wherein, M is the full memory amount that described virtual machine to be migrated need to copy, and N is described predetermined iterations, and h is described predetermined dirty page or leaf surplus.
6. the method for claim 1 further comprises:
Be less than or equal to the constraint of described predetermined value based on the average dirty page or leaf factor after described next round iteration copy is finished, determine to carry out the migration bandwidth of described next round iteration copy.
7. virtual machine (vm) migration watch-dog comprises:
Dirty page or leaf factor computing module, after each takes turns the end of iteration copy, calculate the current average dirty page or leaf factor, the described current average dirty page or leaf factor is the mean value that current all that have finished are taken turns the dirty page or leaf factor in the iteration copy, and the described dirty page or leaf factor that each is taken turns is that this dirty page or leaf of taking turns produces the ratio between speed and this migration bandwidth of taking turns;
The migration management module comprises comparer, and described comparer compares the average dirty page or leaf factor and the predetermined value that described dirty page or leaf factor computing module calculates; And
The bandwidth control module is adjusted the migration bandwidth of carrying out next round iteration copy adaptively according to the comparative result of described comparer.
8. virtual machine (vm) migration watch-dog as claimed in claim 7, wherein, when described comparer was determined the described average dirty page or leaf factor greater than described predetermined value, described bandwidth control module increased the migration bandwidth of carrying out next round iteration copy; And
When described comparer determined that the described average dirty page or leaf factor is less than or equal to described predetermined value, described bandwidth control module reduced to carry out the migration bandwidth of next round iteration copy or it is remained unchanged.
9. virtual machine (vm) migration watch-dog as claimed in claim 7, wherein, described migration management module further is configured to after each takes turns the end of iteration copy, determines whether iterations reaches predetermined iterations threshold value and remain dirty page or leaf whether be less than or equal to predetermined dirty page or leaf surplus; And when determining that iterations reaches predetermined iterations threshold value or remains dirty page or leaf when being less than or equal to predetermined dirty page or leaf surplus, the process of finishing iteration copy.
10. virtual machine (vm) migration watch-dog as claimed in claim 9, wherein, described predetermined value is according to the full memory amount of virtual machine needs copy to be migrated, described predetermined iterations threshold value and described predetermined dirty page or leaf surplus and definite.
11. virtual machine (vm) migration watch-dog as claimed in claim 10, wherein, described predetermined value is
Wherein,
M is the full memory amount that described virtual machine to be migrated need to copy, and N is described predetermined iterations, and h is described predetermined dirty page or leaf surplus.
12. virtual machine (vm) migration watch-dog as claimed in claim 7, wherein, described bandwidth control module is less than or equal to the constraint of described predetermined value based on the average dirty page or leaf factor after described next round iteration copy is finished, determine to carry out the migration bandwidth of described next round iteration copy.
13. a virtual machine (vm) migration system comprises:
A plurality of such as each described virtual machine (vm) migration watch-dog in the claim 7 to 12, each virtual machine (vm) migration watch-dog corresponds respectively to a main frame setting that can move virtual machine; And
The resource management system of communicating by letter with described virtual machine (vm) migration watch-dog.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210082400.2A CN103365704B (en) | 2012-03-26 | 2012-03-26 | Internal memory pre-copy method in virtual machine (vm) migration and perform device and the system of the method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210082400.2A CN103365704B (en) | 2012-03-26 | 2012-03-26 | Internal memory pre-copy method in virtual machine (vm) migration and perform device and the system of the method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103365704A true CN103365704A (en) | 2013-10-23 |
| CN103365704B CN103365704B (en) | 2016-12-14 |
Family
ID=49367133
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210082400.2A Active CN103365704B (en) | 2012-03-26 | 2012-03-26 | Internal memory pre-copy method in virtual machine (vm) migration and perform device and the system of the method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103365704B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103955399A (en) * | 2014-04-30 | 2014-07-30 | 华为技术有限公司 | Migrating method and device for virtual machine, as well as physical host |
| CN104268003A (en) * | 2014-09-30 | 2015-01-07 | 南京理工大学 | Memory state migration method applicable to dynamic migration of virtual machine |
| CN104767824A (en) * | 2015-04-22 | 2015-07-08 | 成都睿峰科技有限公司 | Cloud platform data processing method |
| CN105786585A (en) * | 2014-12-15 | 2016-07-20 | 中国电信股份有限公司 | System and method for online transfer of memory data |
| CN106101211A (en) * | 2016-06-08 | 2016-11-09 | 西安电子科技大学 | A kind of carrier wave emigration method rewriting probabilistic forecasting based on page |
| US9552233B1 (en) | 2015-09-10 | 2017-01-24 | Red Hat Israel, Ltd. | Virtual machine migration using free page hinting |
| CN107085534A (en) * | 2017-03-30 | 2017-08-22 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
| CN107368349A (en) * | 2017-07-17 | 2017-11-21 | 郑州云海信息技术有限公司 | A kind of method for managing resource and device |
| CN107430518A (en) * | 2015-03-27 | 2017-12-01 | 英特尔公司 | Technology for virtual machine (vm) migration |
| CN107463424A (en) * | 2016-06-02 | 2017-12-12 | 北京金山云网络技术有限公司 | A kind of virtual machine migration method and device |
| CN107832118A (en) * | 2017-11-18 | 2018-03-23 | 浙江网新恒天软件有限公司 | A kind of KVM live migration of virtual machine optimization methods of reduction VCPU temperatures |
| CN108121589A (en) * | 2016-11-30 | 2018-06-05 | 中标软件有限公司 | A kind of online migration automatic convergent method of virtual machine |
| US10013205B2 (en) | 2014-09-12 | 2018-07-03 | Huawei Technologies Co., Ltd. | Memory migration method and device |
| CN108369524A (en) * | 2015-12-21 | 2018-08-03 | 英特尔公司 | The technology of collaboration migration for virtual machine |
| CN108363611A (en) * | 2017-11-02 | 2018-08-03 | 北京紫光恒越网络科技有限公司 | Method for managing security, device and the omnidirectional system of virtual machine |
| CN110134490A (en) * | 2018-02-08 | 2019-08-16 | 中兴通讯股份有限公司 | Dynamic migration of virtual machine method, equipment and storage medium |
| CN110737509A (en) * | 2019-10-18 | 2020-01-31 | 杭州朗和科技有限公司 | Thermal migration processing method and device, storage medium and electronic equipment |
| WO2021169618A1 (en) * | 2020-02-29 | 2021-09-02 | 苏州浪潮智能科技有限公司 | Virtual machine migration method, apparatus and device, and medium |
| CN115827169A (en) * | 2023-02-07 | 2023-03-21 | 天翼云科技有限公司 | Virtual machine migration method and device, electronic equipment and medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101101562A (en) * | 2007-07-10 | 2008-01-09 | 北京大学 | Dummy machine external storage on-line migration method |
| US20100299666A1 (en) * | 2009-05-25 | 2010-11-25 | International Business Machines Corporation | Live Migration of Virtual Machines In a Computing environment |
| CN102185774A (en) * | 2011-05-10 | 2011-09-14 | 中兴通讯股份有限公司 | Method, monitor and system for seamless migration of virtual machine |
-
2012
- 2012-03-26 CN CN201210082400.2A patent/CN103365704B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101101562A (en) * | 2007-07-10 | 2008-01-09 | 北京大学 | Dummy machine external storage on-line migration method |
| US20100299666A1 (en) * | 2009-05-25 | 2010-11-25 | International Business Machines Corporation | Live Migration of Virtual Machines In a Computing environment |
| CN102185774A (en) * | 2011-05-10 | 2011-09-14 | 中兴通讯股份有限公司 | Method, monitor and system for seamless migration of virtual machine |
Non-Patent Citations (1)
| Title |
|---|
| 孙国飞等: "基于预拷贝的虚拟机动态内存迁移机制改进", 《计算机工程》, vol. 37, no. 13, 31 July 2011 (2011-07-31), pages 36 - 39 * |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103955399A (en) * | 2014-04-30 | 2014-07-30 | 华为技术有限公司 | Migrating method and device for virtual machine, as well as physical host |
| CN103955399B (en) * | 2014-04-30 | 2017-06-20 | 华为技术有限公司 | A kind of virtual machine migration method and device and physical host |
| US10013205B2 (en) | 2014-09-12 | 2018-07-03 | Huawei Technologies Co., Ltd. | Memory migration method and device |
| CN104268003A (en) * | 2014-09-30 | 2015-01-07 | 南京理工大学 | Memory state migration method applicable to dynamic migration of virtual machine |
| CN104268003B (en) * | 2014-09-30 | 2017-11-14 | 南京理工大学 | A kind of internal storage state moving method suitable for dynamic migration of virtual machine |
| CN105786585A (en) * | 2014-12-15 | 2016-07-20 | 中国电信股份有限公司 | System and method for online transfer of memory data |
| CN105786585B (en) * | 2014-12-15 | 2019-01-29 | 中国电信股份有限公司 | The system and method for online migration internal storage data |
| CN107430518A (en) * | 2015-03-27 | 2017-12-01 | 英特尔公司 | Technology for virtual machine (vm) migration |
| EP3274822A4 (en) * | 2015-03-27 | 2018-11-07 | Intel Corporation | Technologies for virtual machine migration |
| CN104767824B (en) * | 2015-04-22 | 2018-03-02 | 成都睿峰科技有限公司 | A kind of cloud platform data processing method |
| CN104767824A (en) * | 2015-04-22 | 2015-07-08 | 成都睿峰科技有限公司 | Cloud platform data processing method |
| US9552233B1 (en) | 2015-09-10 | 2017-01-24 | Red Hat Israel, Ltd. | Virtual machine migration using free page hinting |
| CN108369524A (en) * | 2015-12-21 | 2018-08-03 | 英特尔公司 | The technology of collaboration migration for virtual machine |
| CN107463424A (en) * | 2016-06-02 | 2017-12-12 | 北京金山云网络技术有限公司 | A kind of virtual machine migration method and device |
| CN107463424B (en) * | 2016-06-02 | 2020-01-24 | 北京金山云网络技术有限公司 | Virtual machine migration method and device |
| CN106101211B (en) * | 2016-06-08 | 2019-09-06 | 西安电子科技大学 | A kind of page based on memory rewrites the carrier wave emigration method of probabilistic forecasting |
| CN106101211A (en) * | 2016-06-08 | 2016-11-09 | 西安电子科技大学 | A kind of carrier wave emigration method rewriting probabilistic forecasting based on page |
| CN108121589A (en) * | 2016-11-30 | 2018-06-05 | 中标软件有限公司 | A kind of online migration automatic convergent method of virtual machine |
| CN107085534A (en) * | 2017-03-30 | 2017-08-22 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
| CN107085534B (en) * | 2017-03-30 | 2020-06-23 | 联想(北京)有限公司 | Information processing method and electronic equipment |
| CN107368349A (en) * | 2017-07-17 | 2017-11-21 | 郑州云海信息技术有限公司 | A kind of method for managing resource and device |
| CN107368349B (en) * | 2017-07-17 | 2020-09-08 | 郑州云海信息技术有限公司 | Resource management method and device |
| CN108363611A (en) * | 2017-11-02 | 2018-08-03 | 北京紫光恒越网络科技有限公司 | Method for managing security, device and the omnidirectional system of virtual machine |
| CN107832118A (en) * | 2017-11-18 | 2018-03-23 | 浙江网新恒天软件有限公司 | A kind of KVM live migration of virtual machine optimization methods of reduction VCPU temperatures |
| CN110134490A (en) * | 2018-02-08 | 2019-08-16 | 中兴通讯股份有限公司 | Dynamic migration of virtual machine method, equipment and storage medium |
| CN110134490B (en) * | 2018-02-08 | 2023-12-29 | 中兴通讯股份有限公司 | Virtual machine dynamic migration method, device and storage medium |
| CN110737509A (en) * | 2019-10-18 | 2020-01-31 | 杭州朗和科技有限公司 | Thermal migration processing method and device, storage medium and electronic equipment |
| CN110737509B (en) * | 2019-10-18 | 2022-04-19 | 杭州朗和科技有限公司 | Thermal migration processing method and device, storage medium and electronic equipment |
| WO2021169618A1 (en) * | 2020-02-29 | 2021-09-02 | 苏州浪潮智能科技有限公司 | Virtual machine migration method, apparatus and device, and medium |
| CN115827169A (en) * | 2023-02-07 | 2023-03-21 | 天翼云科技有限公司 | Virtual machine migration method and device, electronic equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103365704B (en) | 2016-12-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103365704A (en) | Memory pre-copying method in virtual machine migration, device executing memory pre-copying method and system | |
| US10289451B2 (en) | Method, apparatus, and system for adjusting deployment location of virtual machine | |
| Zheng et al. | Joint power optimization of data center network and servers with correlation analysis | |
| US10630737B2 (en) | Data processing method in stream computing system, control node, and stream computing system | |
| CN103955399B (en) | A kind of virtual machine migration method and device and physical host | |
| US9223616B2 (en) | Virtual machine resource reduction for live migration optimization | |
| US8732717B2 (en) | Resource allocation | |
| CN104503838A (en) | Method for scheduling virtual CPU (Central Processing Unit) | |
| RU2573733C1 (en) | Method and apparatus for adjusting i/o channel on virtual platform | |
| WO2017106997A1 (en) | Techniques for co-migration of virtual machines | |
| Saxena et al. | Energy aware resource efficient-(eare) server consolidation framework for cloud datacenter | |
| CN104123171A (en) | Virtual machine migrating method and system based on NUMA architecture | |
| CN104951352A (en) | Method and equipment for migrating data flow | |
| KR101264895B1 (en) | Virtual machine migration processing apparatus and method thereof | |
| CN103763403B (en) | Message flux control method and relevant apparatus and calculate node | |
| CN103838520A (en) | Input/output traffic control method and equipment | |
| JP6300969B2 (en) | Optimization and automatic fan control mechanism in rack system | |
| Kaushi et al. | A computation offloading framework to optimize energy utilisation in mobile cloud computing environment | |
| CN105430083B (en) | A kind of cloud platform dispositions method | |
| Hu et al. | Research of scheduling strategy on OpenStack | |
| CN108123893A (en) | A kind of multiple target bandwidth allocation methods of real time virtual machine migration | |
| US9871732B2 (en) | Dynamic flow control in multicast systems | |
| CN111694635A (en) | Service quality control method and device | |
| CN108073443B (en) | Virtual machine selection and placement method based on shared memory page in cloud data center | |
| Fang et al. | Improvement of live migration mechanism for virtual machine based on pre-copy |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |