CN103473368A - Virtual machine real-time migration method and system based on counting rank ordering - Google Patents
Virtual machine real-time migration method and system based on counting rank ordering Download PDFInfo
- Publication number
- CN103473368A CN103473368A CN2013104515341A CN201310451534A CN103473368A CN 103473368 A CN103473368 A CN 103473368A CN 2013104515341 A CN2013104515341 A CN 2013104515341A CN 201310451534 A CN201310451534 A CN 201310451534A CN 103473368 A CN103473368 A CN 103473368A
- Authority
- CN
- China
- Prior art keywords
- page
- dirty
- virtual machine
- count
- sort
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Memory System Of A Hierarchy Structure (AREA)
Abstract
The invention relates to a virtual machine real-time migration method based on counting rank ordering. The method specifically comprises the following steps of a step 1 that the times that each memory page is dirtied are stored in a storage array; a step 2 that the memory pages are ranked from low to high according to the dirtying times; a step 3 that the ranked memory pages are transmitted from a source host to a target host according to the sequence. The method has the beneficial effects that the memory pages with more dirtying times are placed on the last period of each round as far as possible for carrying out transmission judgment, and the condition that the same page is transmitted for many times due to the frequent dirtying can be reduced as far as possible, so that the time that the page with high dirtying probability is transmitted is delayed as far as possible, and the probability that the page is retransmitted frequently is reduced; the transition time is reduced; the migratory data size is reduced; the migration efficiency is thus improved.
Description
Technical field
The present invention relates to a kind of virtual machine real-time migration method and system based on count sort, relate to virtual and cloud computing field.
Background technology
In virtual field, the virtual machine real-time migration is as one of gordian technique, is widely used in the load balancing of modern data center, cluster, cloud computing aspect, fault-tolerant and system support etc.Virtual machine is to have the computer system of whole hardware characteristics by the mode of software simulation.In the situation that guarantee the normal operation of virtual in-fight service, virtual machine can be moved between different physical hosts, and this is called as the virtual machine real-time migration.
Real-time migration is a key feature of virtual machine technique.In the process of virtual machine real-time migration, when the physical store of virtual machine is shared, (for example SAN or NFS service), only need to transmit memory mirror and CPU relevant information.Most of the time is spent in the transmission of memory mirror data in this case.For the memory mirror data, the pre-copy algorithm is the most frequently used virtual machine real-time migration method.
For the virtual machine real-time migration that uses the pre-copy algorithm, the gross migration time comprises four different time periods usually.The page that the transmission that at first will circulate is dirty.Then virtual machine is suspended on the physical host of source.Next transmit last take turns in dirty page.Virtual machine restarts on destination host after this.Just be comprised of three time periods of back the stop time in transition process, sees Fig. 1.Used the virtual machine real-time migration iterative process of pre-copy algorithm as shown in Figure 2.The pre-copy algorithm attempts to be minimized in the page that last takes turns middle transmission, short as far as possible to guarantee stop time.Physical memory mirror image in this process, be transferred to new destination host by network, simultaneously source physical host continuous service.So, in order to keep the virutal machine memory state consistent between the source and target main frame, some send to the page of destination host, after again dirty, destination host should be sent to again.Therefore the pre-copy algorithm checks by the mechanism of an iteration every now and then that dirty page is laid equal stress on and newly transmits these pages.In the first round of iterative process, all pages are marked as dirty, and the page more next remodified also is marked as dirty, and the page be not modified is marked as clean.Internal memory number of pages dirty in a certain wheel the in iterative process is less than predefined threshold values, or iterative process is when reaching bull wheel and counting, and iterative process will be terminated, virtual machine will be suspended and enter iteration last take turns.In the end one take turns in dirty page be unfixed, dirty page quantity has directly determined stop time.
Concerning a page, if we mean dirty state with 1, with 0, mean clean state.The state of a page in adjacent two-wheeled will be as shown in Figure 3 so.As seen from Figure 3, for a page, in the adjacent two-wheeled of iterative process, having four kinds of combinations of states, is respectively 11,10,00, and 01.The meaning of combination 10 is that a page is dirty at previous round, and is clean in rear one takes turns.Only, for 10 this combinations, we should send this page.Because this page was modified, in visible future, it can not be revised again.For other three kinds of combinations, we should not send this page.For 11 combinations, in visible future, this page may be changed again, there is no need to transmit at once this page.For 00 combination, this page was not modified, and we do not need to transmit it.For combination 01, this page was not modified now, and in visible future, it likely is modified, and we should, in the process of next round, do further judgement.
The Pre-copy algorithm will repeat this process, until one take turns in dirty internal memory number of pages be less than a default threshold values, or iterative process reaches bull wheel number, this takes turns iteration becomes critical the wheel usually.After transferring critical dirty page of taking turns, virtual machine will be suspended and enter last and take turns.We see now: in the process iterated at the pre-copy algorithm, what have many pages to be repeated sends to destination host.Cause the page volume of transmitted data to increase, gross migration time and stop time extend.
Summary of the invention
Technical matters to be solved by this invention is, exist and repeat to send dirty page problem for the pre-copy algorithm, a kind of page that reduces re-transmission is provided, save transit time and reduce the data volume of moving, thereby improve the virtual machine real-time migration method based on count sort of transport efficiency.
The technical scheme that the present invention solves the problems of the technologies described above is as follows: a kind of virtual machine real-time migration method based on count sort specifically comprises the following steps:
Step 1: store dirty inferior the counting in a storage array of each page;
Step 2: page is sorted from low to high according to dirty number of times;
Step 3: the page after sequence is transferred to destination host from source host in order.
The invention has the beneficial effects as follows: the method for the invention can be placed in dirty page often as far as possible every final stage of taking turns and transmit judgement, just can reduce as much as possible the same page due to frequent dirty and situation about repeatedly being transmitted, thereby postpone as far as possible the time that the page that dirty probability is high is transmitted, may with what reduce that the page frequently retransmitted; The data volume that realizes saving transit time and reduce migration, thus transport efficiency improved.
On the basis of technique scheme, the present invention can also do following improvement.
Further, the sequence in described step 2 adopts the count sort method.
Further, the number of described disposable transmission page may be defined as batch size, and when the method for usage count sequence, batch size is also an influence factor; Reduce batch size, realize reducing the page that repeats transmission.
Further, described step 2 specifically comprises the following steps:
Step 2.1: record the original position of page by a position array, record the quantity of the page that different dirty number of times are corresponding by another auxiliary storage array;
Step 2.2: with 0 initialization, store each in array;
Step 2.3: page is sorted, make the page that dirty number of times is little come front, the order that the page that dirty number of times is large comes in the array of ，Shi position, back rearranges;
Step 2.4: complete page and sort from low to high according to dirty number of times.
Another technical matters to be solved by this invention is, exist and repeat to send dirty page problem for the pre-copy algorithm, a kind of page that reduces re-transmission is provided, save transit time and reduce the data volume of moving, thereby improve the virtual machine real-time migration system based on count sort of transport efficiency.
The technical scheme that the present invention solves the problems of the technologies described above is as follows: a kind of virtual machine real-time migration system based on count sort comprises: memory module, order module and transport module;
Described memory module for store each page dirty time count to a storage array;
Described order module is for being sorted page from low to high according to dirty number of times;
Described transport module transfers to destination host from source host in order for the page by after sequence.
The invention has the beneficial effects as follows: system of the present invention can be placed in dirty page often as far as possible every final stage of taking turns and transmit judgement, just can reduce as much as possible the same page due to frequent dirty and situation about repeatedly being transmitted, thereby postpone as far as possible the time that the page that dirty probability is high is transmitted, may with what reduce that the page frequently retransmitted; The data volume that realizes saving transit time and reduce migration, thus transport efficiency improved.
On the basis of technique scheme, the present invention can also do following improvement.
Further, the sort method in described order module adopts the count sort method.
Further, the number of described disposable transmission page may be defined as batch size, and when the method for usage count sequence, batch size is also an influence factor; Reduce batch size, realize reducing the page that repeats transmission.
Further, described order module comprises: logging modle, initialization module and reordering module;
Described logging modle, for by a position array, recording the original position of page, records the quantity of the page that different dirty number of times are corresponding by another auxiliary storage array;
Described initialization module is used for each by 0 initialization storage array;
Described reordering module, for page is sorted, makes the page that dirty number of times is little come front, and the order that the page that dirty number of times is large comes in the array of ，Shi position, back rearranges, and completes page and sorts from low to high according to dirty number of times.
Exist and repeat to send dirty page problem for the pre-copy algorithm.If reduce the page retransmitted, just can save transit time and reduce the data volume of moving, thereby improve transport efficiency.Take turns at first when one, we can't know that page is at several definite states of taking turns next.But, during one of iterative process is taken turns, if we postpone the transmission time of a certain page, the combinations of states of this page likely becomes 11 by 10.This will cause the page that should be transmitted in epicycle according to normal pre-copy algorithm needn't be transmitted due to state variation, thereby reduces the page retransmitted.
During one of pre-copy algorithm iteration is taken turns, the page that comes the back location transmission is all longer than the time of the page delay that comes the anterior locations transmission, thereby the possibility of dirty (page is carried out to " writing " operation) is also just larger.The dirty probability of each page is also different simultaneously, and some part of internal memory is more frequent than the other parts change.We think, the dirty number of times of page is more, and dirty probability is also just larger again for it.Yet generally the dirty probability of page is different during the different periods.If during the T1 period, the dirty probability of page k is A
k1, the dirty probability of page w is A
w1, during the T2 period, the dirty probability of page k is A
k2, the dirty probability of page w is A
w2, until during the Tn period, the dirty probability of page k is A
kn, the dirty probability of page w is A
wn.The average dirty probability of page k is A
kavg=(A
k1+ A
k2+ ... + A
kn)/n, the average dirty probability of page w is A
wavg=(A
w1+ A
w2+ ... + A
wn)/n, at (T1+T2+ ... + Tn), during the whole period, the dirty number of times of page k is C
k, the dirty number of times of page w is C
w.The dirty probability of page during although we can't add up each period, if we replace the occurrence during each period with mean value, and hypothesis (A
kavg>=A
wavg) and (C
k<C
w), by such hypothesis, calculate, if dirty in taking turns in iterative process of page w, page k also will be dirty because one take turns during, the time of two page experience is identical, and (A
kavg>=A
wavg), so during the whole period, (C will be arranged
k>=C
w) one of this and assumed condition (C
k<C
w) be conflicting.So as (A
kavg>=A
wavg) time, (C can only be arranged
k>=C
w).So the height of dirty number of times has reflected the size of average dirty probability from the side.For a page, its dirty probability during the different periods may fluctuate near average dirty probability.But, from whole structure, average dirty probability can be taken fully, so we think that dirty number of times is more, average dirty probability larger (be dirty probability larger).Therefore, if can as far as possible dirty page often be placed in to every final stage of taking turns, transmit judgement, just can reduce as much as possible the same page due to frequent dirty and situation about repeatedly being transmitted.Based on this thinking, the virtual machine real-time migration technology based on count sort has been proposed, at first, we transmit them according to the dirty number of times of page order from low to high.Do like this, can reduce the page that repeats transmission.Secondly, we adopt the count sort method that time complexity is O (n).Than other sort method, we can save sequence and transit time.Finally, the number of disposable transmission page is called batch size.When the method for usage count sequence, batch size is also an influence factor.We reduce batch size, further reduce the page that repeats transmission.In each of pre-copy algorithm iteration process, take turns, break the fixed form according to the original sequential delivery of virutal machine memory page, but when taking turns beginning, each at first according to the dirty number of times of page order from low to high, is sorted, thereby postpone as far as possible the time that the page that dirty probability is high is transmitted, may with what reduce that the page frequently retransmitted.
For the page of a 4KB size, the time of the whole internal memory that sorts will be a no small duration, and we should select optimal sort algorithm, and count sort has catered to this needs.In computer science, count sort be one stable, effective and simple sort algorithm.The object of a set can carry out count sort according to its key assignments, as long as these key assignments are little integers.So count sort is based on again the sort method of integer.Count sort is added up the number of key assignments, and the number of key assignments is carried out to arithmetical operation to determine the position of each key assignments in output sequence.The quantity of the project in sorting time and set and the difference of largest key value and minimum key value are linear scaling.Because count sort be take key assignments as array indexing, so it is not the sort algorithm of a comparison, the minimum event horizon O (nlogn) of sort algorithm relatively is not suitable for count sort.Use the real-time migration of pre-copy algorithm that the maximum wheel number iterated restriction is arranged, we can utilize this point to realize count sort.This point is pointed out us: the dirty number of times of each page can not surpass the bull wheel number iterate, and the dirty number of times of each page should be { 0,1,2 ... between maxround}.So we can utilize the method for count sort to transmit it by the dirty number of times of page order from low to high.
The accompanying drawing explanation
The different period schematic diagram that Fig. 1 is real-time migration in prior art;
The virtual machine real-time migration iterative process schematic diagram that Fig. 2 is Pre-copy algorithm in prior art;
Fig. 3 is the constitutional diagram of a page in adjacent two-wheeled in prior art;
Fig. 4 is for being the described a kind of virtual machine real-time migration method process flow diagram based on count sort of the specific embodiment of the invention 1;
Fig. 5 is the described a kind of virtual machine real-time migration system architecture diagram based on count sort of the specific embodiment of the invention 2;
Fig. 6 adopt the method for the invention compared with prior art stop time comparison diagram;
Fig. 7 adopts compared with prior art gross migration time comparison diagram of the method for the invention;
Fig. 8 adopts the compared with prior art internal memory number of pages comparison diagram of gross migration of the method for the invention;
Fig. 9 Tpcc loads on the page transmitted in different internal memory situations.
In accompanying drawing, the list of parts of each label representative is as follows:
1, memory module, 2, order module, 3, transport module, 21, logging modle, 22, initialization module, 23, reordering module.
Embodiment
Below in conjunction with accompanying drawing, principle of the present invention and feature are described, example, only for explaining the present invention, is not intended to limit scope of the present invention.
As shown in Figure 4, be the described a kind of virtual machine real-time migration method based on count sort of the specific embodiment of the invention 1, specifically comprise the following steps:
Step 1: store dirty inferior the counting in a storage array of each page;
Step 2: record the original position of page by a position array, record the quantity of the page that different dirty number of times are corresponding by another auxiliary storage array;
Step 3: with 0 initialization, store each in array;
Step 4: page is sorted, make the page that dirty number of times is little come front, the order that the page that dirty number of times is large comes in the array of ，Shi position, back rearranges;
Step 5: complete page and sort from low to high according to dirty number of times;
Step 6: the page after sequence is transferred to destination host from source host in order.
Sequence in described step 4 adopts the count sort method.
The number of described disposable transmission page may be defined as batch size, and when the method for usage count sequence, batch size is also an influence factor; Reduce batch size, realize reducing the page that repeats transmission.
As shown in Figure 5, for the described a kind of virtual machine real-time migration system based on count sort of the specific embodiment of the invention 2, comprising: memory module 1, order module 2 and transport module 3;
Described memory module 1 for store each page dirty time count to a storage array;
Described order module 2 is for being sorted page from low to high according to dirty number of times;
Described transport module 3 transmits in order for the page by after sequence.
Sort method in described order module 2 adopts the count sort method.
The number of described disposable transmission page may be defined as batch size, and when the method for usage count sequence, batch size is also an influence factor; Reduce batch size, realize reducing the page that repeats transmission.
Described order module 2 comprises: logging modle 21, initialization module 22 and reordering module 23;
Described logging modle 21, for by a position array, recording the original position of page, records the quantity of the page that different dirty number of times are corresponding by another auxiliary storage array;
Described initialization module 22 is for by each of 0 initialization storage array;
Described reordering module 23, for page is sorted, makes the page that dirty number of times is little come front, and the order that the page that dirty number of times is large comes in the array of ，Shi position, back rearranges, and completes page and sorts from low to high according to dirty number of times.
In pre-copy algorithm iteration process, at first we use an array A[p2m_size]: store the dirty number of times of each page.The population size that wherein p2m_size is page is used another array D[p2m_size simultaneously]: record the original position of page.And with another auxiliary storage array C[tmax+1]: record the page quantity that different dirty number of times are corresponding.During beginning with 0 initialization array A[p2m_size] each.Then according to following algorithm, processed:
If not last iterating, take turns
To the page, i is checked, if memory pages i is dirty being write, just makes A[i] add 1;
If not the first round iterated
Its code is as follows:
After being processed according to above arthmetic statement, we no longer from position 0 to position the order of (p2m_size-1) transmit page.On the contrary, we will be according to from position D[0] to position D[p2m_size-1] order transmit them.
The number of the page of disposable transmission is called batch size.When the method for usage count sequence, batch size is also an influence factor.In the actual process realized, we should select suitable batch size, and in the test based on the Xen-4.1.3 platform, our reality is reduced to 64 to batch size by 1024.
In order to verify the validity of count sort algorithm, we have realized this algorithm on Ubuntu-12.04 operating system and Xen-4.1.3 virtual platform, and compare experiment with the pre-copy algorithm.The cpu type of our destination host is Intel (R) Core (TM) i3-2120@3.30GZ, memory size 2GB.The cpu type of our source host is Intel (R) Core (TM) 2DuoE8200@2.66GZ, and memory size is also 2GB.Two main frames are connected in the LAN (Local Area Network) of 100Mbps by a FR400 router.In source host, we utilize NFS service shared disk to video to destination host, so that the destination host disk image can be loaded.In experiment, adopt following scene as operating load.
(1) Idle: the Ubuntu-12.04 operating system of a free time, do not have special load to move in the above, this scene is used as comparison and reference.
(2) Tpcc:Tpcc is an online issued transaction benchmark, this benchmark simulation a group terminal operation person database is carried out to the whole environment that affairs behaviour does.We have used this software of tpcc-mysql, and have configured ten connections.
(3) Dbench: this is a benchmark of increasing income and instrument, is used for producing the input and output load to a file system.Dbench can be used to exert pressure to a file system or service, and in experiment, we have configured a connection and have produced suitable disk load.
(4) Linpack: the Linpack that our type of service is HPL produces huge vector sum matrix manipulation, and this will produce CPU and memory pressure.HPL can be regarded as the high performance calculating benchmark that can be performed freely.We use 3 rank matrixes and are provided with 4 processes.
(5) Webbench: this is a simple instrument, is WWW or the acting server of benchmark, and it can be simulated a plurality of clients and use the http request.We have configured the webpage of 2063 byte-sized of 5 connections and a static state.
Virutal machine memory is 512MB simultaneously, and has configured two VCPU, and the greatest iteration wheel number of algorithm is made as 30 and takes turns.Experimental result shows; under different loads; in the situation that fixedly batch size is 64; the stop time of real-time migration, the gross migration time, and total page volume of transmitted data isometry index has produced reduction in various degree simultaneously; on average; reduced 18% stop time and seen Fig. 6, the gross migration time decreased 23% see Fig. 7, the page volume of transmitted data of gross migration has reduced 16% and has seen Fig. 8.The method of count sort all can work in the internal memory of different sizes, for the same load in different internal memories (256MB, 512MM, 1024MB) situation, the page volume of transmitted data reduced in breast wheel about equally, with regard to Tpcc load decreased average 47% see Fig. 9.
The foregoing is only preferred embodiment of the present invention, in order to limit the present invention, within the spirit and principles in the present invention not all, any modification of doing, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.
Claims (8)
1. the virtual machine real-time migration method based on count sort, is characterized in that, specifically comprises the following steps:
Step 1: store dirty inferior the counting in a storage array of each page;
Step 2: page is sorted from low to high according to dirty number of times;
Step 3: the page after sequence is transferred to destination host from source host in order.
2. a kind of virtual machine real-time migration method based on count sort according to claim 1, is characterized in that, the sequence in described step 2 adopts the count sort method.
3. a kind of virtual machine real-time migration method based on count sort according to claim 2, it is characterized in that, the number of described disposable transmission page may be defined as batch size, and when the method for usage count sequence, batch size is also an influence factor; Reduce batch size, realize reducing the page that repeats transmission.
4. according to the described a kind of virtual machine real-time migration method based on count sort of claim 1-3 any one, it is characterized in that, described step 2 specifically comprises the following steps:
Step 2.1: record the original position of page by a position array, record the quantity of the page that different dirty number of times are corresponding by another auxiliary storage array;
Step 2.2: with 0 initialization, store each in array;
Step 2.3: page is sorted, make the page that dirty number of times is little come front, the order that the page that dirty number of times is large comes in the array of ，Shi position, back rearranges;
Step 2.4: complete page and sort from low to high according to dirty number of times.
5. the virtual machine real-time migration system based on count sort, comprising: memory module, order module and transport module;
Described memory module for store each page dirty time count to a storage array;
Described order module is for being sorted page from low to high according to dirty number of times;
Described transport module transfers to destination host from source host in order for the page by after sequence.
6. a kind of virtual machine real-time migration system based on count sort according to claim 5, is characterized in that, the sort method in described order module adopts the count sort method.
7. a kind of virtual machine real-time migration system based on count sort according to claim 6, it is characterized in that, the number of described disposable transmission page may be defined as batch size, and when the method for usage count sequence, batch size is also an influence factor; Reduce batch size, realize reducing the page that repeats transmission.
8. according to the described a kind of virtual machine real-time migration system based on count sort of claim 5-7 any one, it is characterized in that, described order module comprises: logging modle, initialization module and reordering module;
Described logging modle, for by a position array, recording the original position of page, records the quantity of the page that different dirty number of times are corresponding by another auxiliary storage array;
Described initialization module is used for each by 0 initialization storage array;
Described reordering module, for page is sorted, makes the page that dirty number of times is little come front, and the order that the page that dirty number of times is large comes in the array of ，Shi position, back rearranges, and completes page and sorts from low to high according to dirty number of times.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013104515341A CN103473368A (en) | 2013-09-27 | 2013-09-27 | Virtual machine real-time migration method and system based on counting rank ordering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013104515341A CN103473368A (en) | 2013-09-27 | 2013-09-27 | Virtual machine real-time migration method and system based on counting rank ordering |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103473368A true CN103473368A (en) | 2013-12-25 |
Family
ID=49798216
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2013104515341A Pending CN103473368A (en) | 2013-09-27 | 2013-09-27 | Virtual machine real-time migration method and system based on counting rank ordering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103473368A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN107094160A (en) * | 2016-02-18 | 2017-08-25 | 北大方正集团有限公司 | The method and system that a kind of data are downloaded |
| CN108228634A (en) * | 2016-12-15 | 2018-06-29 | 杭州华为数字技术有限公司 | A kind of data processing method and device |
| CN109660576A (en) * | 2017-10-10 | 2019-04-19 | 武汉斗鱼网络科技有限公司 | User data real-time migration method, storage medium, electronic equipment and system |
| CN114625474A (en) * | 2020-12-08 | 2022-06-14 | 中移(苏州)软件技术有限公司 | Container migration method and device, electronic equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101493797A (en) * | 2009-03-12 | 2009-07-29 | 华为技术有限公司 | Data dynamic migration method and device |
| US20100175063A1 (en) * | 2009-01-05 | 2010-07-08 | International Business Machines Corporation | Detection and Management of Dynamic Migration of Virtual Environments |
| CN102136993A (en) * | 2010-07-29 | 2011-07-27 | 华为技术有限公司 | Data transfer method, device and system |
-
2013
- 2013-09-27 CN CN2013104515341A patent/CN103473368A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100175063A1 (en) * | 2009-01-05 | 2010-07-08 | International Business Machines Corporation | Detection and Management of Dynamic Migration of Virtual Environments |
| CN101493797A (en) * | 2009-03-12 | 2009-07-29 | 华为技术有限公司 | Data dynamic migration method and device |
| CN102136993A (en) * | 2010-07-29 | 2011-07-27 | 华为技术有限公司 | Data transfer method, device and system |
Non-Patent Citations (2)
| Title |
|---|
| QINGXIN ZOU ET AL.,: "Counting sort for the live migration of virtual machines", 《CLUSTER COMPUTING (CLUSTER), 2013 IEEE INTERNATIONAL CONFERENCE ON》 * |
| 张井昆: "基于Xen的虚拟机动态迁移技术研究及应用", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN107094160A (en) * | 2016-02-18 | 2017-08-25 | 北大方正集团有限公司 | The method and system that a kind of data are downloaded |
| CN107094160B (en) * | 2016-02-18 | 2020-07-14 | 北大方正集团有限公司 | Method and system for downloading data file |
| CN108228634A (en) * | 2016-12-15 | 2018-06-29 | 杭州华为数字技术有限公司 | A kind of data processing method and device |
| CN109660576A (en) * | 2017-10-10 | 2019-04-19 | 武汉斗鱼网络科技有限公司 | User data real-time migration method, storage medium, electronic equipment and system |
| CN109660576B (en) * | 2017-10-10 | 2021-09-07 | 武汉斗鱼网络科技有限公司 | User data real-time migration method, storage medium, electronic device and system |
| CN114625474A (en) * | 2020-12-08 | 2022-06-14 | 中移(苏州)软件技术有限公司 | Container migration method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11281404B2 (en) | Storage volume migration scheduling based on storage volume priorities and specified constraints | |
| US11977734B2 (en) | Storage block balancing using volume part migration | |
| CN103810237B (en) | Data managing method and system | |
| US20160132541A1 (en) | Efficient implementations for mapreduce systems | |
| Rasmussen et al. | {TritonSort}: A Balanced {Large-Scale} Sorting System | |
| CN103873559A (en) | Database all-in-one machine capable of realizing high-speed storage | |
| US11914894B2 (en) | Using scheduling tags in host compute commands to manage host compute task execution by a storage device in a storage system | |
| CN106569896B (en) | A kind of data distribution and method for parallel processing and system | |
| CN103473368A (en) | Virtual machine real-time migration method and system based on counting rank ordering | |
| CN105320773A (en) | Distributed duplicated data deleting system and method based on Hadoop platform | |
| CN102521419A (en) | Hierarchical storage realization method and system | |
| CN103077197A (en) | Data storing method and device | |
| US20150193526A1 (en) | Schemaless data access management | |
| CN109767274B (en) | Method and system for carrying out associated storage on massive invoice data | |
| Nachman et al. | GoSeed: Optimal seeding plan for deduplicated storage | |
| JP2022547673A (en) | DATA PROCESSING METHOD AND RELATED DEVICE, AND COMPUTER PROGRAM | |
| CN113723443A (en) | Distributed training method and system for large visual model | |
| CN113568908A (en) | Parallel scheduling method and system for key value requests | |
| CN105426119A (en) | Storage apparatus and data processing method | |
| CN108595251A (en) | Dynamic Graph update method, device, storage engines interface and program medium | |
| CN117093538A (en) | Sparse Cholesky decomposition hardware acceleration system and solving method thereof | |
| Wang et al. | Improved intermediate data management for mapreduce frameworks | |
| CN103559204A (en) | Database operation request processing method, unit and system | |
| Catania et al. | Design and performance analysis of a disk array system | |
| US20220075530A1 (en) | Method, device, and computer program product for rebuilding storage 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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20131225 |