CN105739929B - The selection method of data center when big data is migrated to cloud - Google Patents
The selection method of data center when big data is migrated to cloud Download PDFInfo
- Publication number
- CN105739929B CN105739929B CN201610067866.3A CN201610067866A CN105739929B CN 105739929 B CN105739929 B CN 105739929B CN 201610067866 A CN201610067866 A CN 201610067866A CN 105739929 B CN105739929 B CN 105739929B
- Authority
- CN
- China
- Prior art keywords
- data
- user
- criterion
- cost
- cmdp
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0628—Interfaces specially adapted for storage systems making use of a particular technique
- G06F3/0655—Vertical data movement, i.e. input-output transfer; data movement between one or more hosts and one or more storage devices
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The selection method of data center when being migrated the invention proposes from a kind of big data to cloud, firstly, having carried out noncomplete graph modeling in view of leading to the unavailable situation of DC because of factors such as user preference and legal restrictions；The data yield of user is described otherwise using activation grade；It defines fair data and places four kinds of criterion such as FDP, preference data placement PDP, transmission cost minimum data placement TCMDP and cost minimization data placement CMDP；The selection of DC is carried out based on above-mentioned criterion.Demand of the method proposed by the present invention for BD to cloud when mobile, has studied mobile mechanism from user perspective, can shorten data access time delay, reduce data cost.Method of the invention can reflect the availability of DC and the preference of user.Method of the invention can use network and carry out low cost automatically, and the Data Migration of low latency avoids being conducive to the implementation of automatic management using hardware mode.
Description
Technical field
A kind of choosing of data center when being migrated the present invention relates to field of cloud computer technology more particularly to from big data to cloud
Selection method.
Background technique
Cloud computing has become the preferred platform of big data (BD) analysis.Particularly when data from multiple cross-regions point
The place of cloth generates, and local user needs often to use local data, and sometimes data need again further integration with
It is especially true when being further analyzed.For example, having the transnational sales company of many subsidiaries all over the world for one
For, the subsidiary of each country needs the data generated in time to native country user to analyze for commercial object.All
Data are aggregated analysis again to offer general headquarters, or support Foreign Transactions.In general, a large-scale cloud in a distributed manner into
Row networking and with the distribution of multiple cross-regions data center (DC, for example Amazon is at least throughout 11 DC in 4 continents,
Google is at least throughout 13 DC in 4 continents).Each DC pay-for-use mode configured with calculate and storage money
Source.This infrastructure is capable of providing nearest service, is distributed particularly suitable for cross-region.
In order to handle BD in cloud, precondition is migrated BD and stores on suitable DC.Direct mobile hardware is
A kind of optional way of mobile large-scale data.For example, Amazon Import/Export service recommendation is set with removable Storage
It is standby to transport data.Sometimes, it could even be possible to mobile entire machine.But this is suitable only for intermittent or disposable big
Batch data is mobile.This mode has very big delay, is not able to satisfy ever-increasing data and analyzes demand in real time.And it
It is contradicted with automatic management philosophy, and needs more to become more and more expensive labour and participate in.Data are passed on the net in Inter
It is very expensive and impracticable because of too big delay.According to Amazon data, the number of transmission 1TB is netted by 10MB Inter
According to substantially needing 13 day time.Real time data, which is usually proposed, connects transmission (such as AWS direct with high-speed dedicated
connect).This mode can accelerate transmission speed.Even if but connected dependent on high-speed dedicated, carry out data transmission across continent
Still very difficult.For example, AWS direct connect does not provide the service across continent.And International Line is too expensive.This is just
It limits and the large-scale data for usually extending over the whole world is moved on a DC.Moreover, will lead to a DC come storing data
More frequent local data analysis delay is bigger.
Especially in some regions, some data of data safety legal requirement must be stored in local (some states of such as European Union
Family).To sum up, it is necessary to follow some rules to select suitable storage location for their data by user.Just as
As Amazon suggests: it is closer to the user to meet specific law rule to reduce data using delay and require, or reduce at
This etc..
Currently, some frames based on MapReduce, such as G-Hadoop and G-MR, can be realized across cluster and DC
Data analysis.It is only compared with the mechanism of a DC, the demand of comprehensive analysis is not only able to satisfy using the mechanism of multiple DC, and
And it can guarantee that faster data use and have lower cost.
The select permeability and facility select permeability (facility location of multiple DC when BD is moved to cloud
Problem, FLP) and k- intermediate point problem correlation.FLP is intended to select facility come services client based on different criterion.DC can be with
It is seen as facility, and local data user is client.K- intermediate point problem attempts to find not more than k point, remaining does not have
The point selected will be assigned to the point selected, so that these side lengths and minimum between.
In the mutation of FLP problem, k- supplier problem needs select at most k supplier (corresponding DC) from given set
So that the maximum distance between each client and the nearest supplier is minimum.In general, supplier and customer network are built
K- supplier problem mutation of the mould at a complete graph for a broad sense, each supplier are endowed a weight, it is desirable that
All supplier's weights selected are not more than k.But it is limited to regulation, some DC may not be able to be used to service certain data,
So figure not always complete graph.Moreover, data are relevant with user, rather than related to DC.
Facility select permeability (Uncapacitated facility location, UFL) without capacity limit is FLP
Another mutation, wherein facility does not have a capacity limit, and each facility has one fixed to open up cost weight.Problem mesh
Mark is the total fixed cost of minimization and total cost of serving.Current obtained algorithm has the promise breaking factor an of constant.
I.e. this means that the facility number that algorithm needs is no less than some multiple of k.And k- intermediate point problem does not consider the weight of facility,
But the movement of data is relevant to the size of data.
Summary of the invention
In order to solve the problems in the prior art, data center when being migrated the invention proposes a kind of big data to cloud
Selection method, realize the data access object of BD in distributed cloud computing to low cost of cloud when mobile, high-speed.
The invention is realized by the following technical scheme:
The selection method of data center when a kind of big data is migrated to cloud, which comprises building bottom is non-complete
Full figure describes the data yield of user otherwise using activation grade, defines that fair data place FDP, preference data is put
Set PDP, transmission cost minimizes data and places four kinds of criterion such as TCMDP and cost minimization data placement CMDP, and is based on
The selection of one of above-mentioned criterion progress DC；Wherein, the noncomplete graph G=(U, V, E), U represent user, and V represents DC, side length eij
∈ E (i ∈ U, j ∈ V) meets triangle inequality, positive integer k (k≤| U |, k≤| V |), for any i ∈ U and j ∈ V, if
The data of user i can be moved to DCj, then there are a lines between them；The method is intended to from available DC set V
Find the data that DC subset D (| D |≤k) to store all users in U according to different criterion；The FDP criterion are as follows: maximum
User and the distance between the DC that is assigned to minimization, allow each local user to access number with the smallest time delay
According to:The PDP criterion are as follows: the Weighted distance between maximum user and the DC being assigned to
Minimization, allowing, there is more multidata local user to access data with the smallest time delay:The TCMDP criterion are as follows: the Weighted distance between all users and its DC being assigned to
And minimization:The CMDP) criterion are as follows: the sum of cost of all users minimization:
The beneficial effects of the present invention are: demand of the method proposed by the present invention for BD to cloud when mobile, from user angle
Degree has studied mobile mechanism, can shorten data access time delay, reduces data cost.The present invention has studied four kinds of criterion: fair
Data place FDP, and preference data places PDP, and transmission cost minimizes data and places TCMDP and the placement of cost minimization data
CMDP, method of the invention can reflect the availability of DC and the preference of user, and for first two criterion, algorithm can guarantee
The solution found is at least no worse than 3 times of optimal solution.Method of the invention can use network and carry out low cost automatically, low latency
Data Migration avoids being conducive to the implementation of automatic management using hardware mode.
Detailed description of the invention
Fig. 1 is the non-fully bipartite graph of distributed user data and data center of the invention；
Fig. 2 is the schematic diagram of thresholding figure.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that described herein, specific examples are only used to explain the present invention, not
For limiting the present invention.
The present invention considers the feasibility that the big data that cross-region is distributed is moved to cloud, has studied multiple target DC selections
Problem indicates bottom facilities with non-fully bipartite graph, thus overcome it is problematic be all Complete Bipartite Graph limitation.More
Due to not being the case where each DC can be used caused by user preference or legal restriction in tallying with the actual situation.
Attached drawing 1 simulates the non-fully bipartite graph that the DC in a distributed user data and distributed cloud computing is constituted,
Wherein user has preferred or is limited by safety legislation, is not that each DC can be selected by each user.There is side connection
Indicate that this DC is available or user does not repel.
The present invention is that purpose is to seek faster data access and lower cost.This problem has been promoted traditional k- and has been supplied
Answer quotient's problem, UFL and k- intermediate point problem.
Consider bottom noncomplete graph G=(U, V, E), wherein U represents user, and V represents DC, side length eij∈E(i∈U,j∈V)
Meet triangle inequality, positive integer k (k≤| U |, k≤| V |), the present invention is directed to find DC from available DC set V
Collection D (| D |≤k) to store the data of all users in U according to different criterion.For any i ∈ U and j ∈ V, if user i
Data can be moved to DCj(do not limited at least or do not excluded by user by regulation), then there are a lines between them.
It is assumed that all i close at least one j, otherwise problem is without solution.It is assumed that | E |=m, wherein m≤| U | * | V |.
User's weight definition: each user is endowed a weight wi, indicate the data in current or visible future
The activation rank of generation or the significance level of local user.wiIncrease with the increase of data volume or significance level.With
Activation rank rather than data volume can tolerate data dynamic change simultaneously provide to data volume appropriateness approximation.Activate rank
It can be depending on the daily amount of uploading data.Such as one is typically uploaded the company of 200GB daily, 10GB can be with
It is used as the judgement thresholding of activation rank.If the data that a subsidiary generates daily are less than 10GB, power can be assigned
Value 1.For the subsidiary between 20-30GB, then weight 3 is assigned, and so on.There is activation rank w for useriUse
Family i, moves data over into DCjIt will need to pay expense wi eij。
DC weight definition: each DC has different calculating and storage resource price.For more economical storage and
Data are handled, certainly preferred lower price.Given DCj, it is assumed that the price that a VM example handles data per hour is aj.It is flat
For, this example can analyze b per hourjGB data.The price for then handling 10GB data is p'j=10/bj*aj.If
The storage expense of 10GB data is p "j, then, it is p for having total cost of the user of activation rank 1 in the side DCj=p'j+
p”j.For having activation rank wiUser i, if it wants in DCjData are stored and processed, it needs to pay wi pj。
With activation rank wiThe total cost of user be wi(pj+eij.In view of e in actual environmentij(such as thousands of kilometers)
And pjDifference in the number of levels of (U.S. dollars several per hour of Amazon), we are using the form after standardization: cij=wi(pj'
+eij'), wherein pj'=pj/maxh∈V(ph), eij'=eij/maxl∈U,h∈V(elh).Side length e after paying attention to standardizationij' still full
Sufficient triangle inequality.
Then target can specifically be stated are as follows:
A) fair data place (FDP).The distance between maximum user and the DC being assigned to minimization, so that each
Local user can access data with the smallest time delay:
B) preference data places (PDP).Weighted distance minimization between maximum user and the DC being assigned to, so that
Data can be accessed with more multidata local user with the smallest time delay:If needed
It wants, we also use w (i, j)=wieijTo indicate Weighted distance.
C) transmission cost minimizes data and places (TCMDP).Transmission cost, is defined as all users and it is assigned to
DC between Weighted distance and minimization:
D) cost minimization (CMDP).Totle drilling cost is defined as the sum of the cost of all users minimization:
Because a) and c) being respectively special shape b) and d), the subsequent algorithm provided b) and d).
Preference data places the algorithm basic thought of (PDP)
Several concepts are introduced first, for describing algorithm.
1) bottleneck figure: notice that the optimal solution one of PDP problem is scheduled on some user weighting side and reaches, so we should
Weighting side is checked one by one from small to large, until all constraints all meet.The building of bottleneck figure is based on this with thought.M item
User weights side and is sorted according to non-decreasing sequence, and be denoted as w (1, j)≤w (2, g)≤...≤w (m, h), wherein j, g, h ∈ V
And it may be identical.Bottleneck figure G1,G2,…,GmIt is the edge subgraph of G, and Gr=(U, V, Er) (r=1,2 ..., m), wherein Er={ eij
|wieij≤w(r,g)}.That is, GrBy the vertex of all G and no more than r-th, the side of most short weighting side w (r, g) is formed.
2) thresholding figure: for each Gr, corresponding thresholding figure TrIt is constructed as follows on user's set U.To every two point
U, v ∈ U, side (u, v) is in TrIn, and if only if there are a DC j ∈ V and (u, j) and (v, j) is all in GrIn.For example, in Fig. 2
In, side (5,6) are not present, because in GmIn, user 5,6 not adjacent public DC.
3) Maximum Clique: given non-directed graph H, the group of H is a complete subgraph.If a group is not included in other groups,
Then it is referred to as Maximum Clique, is denoted as C (H).It is easy to find C (H) in polynomial time.The simple method used herein is such as
Under, C (H) first is added in any one point of H, its neighbours' point is then added, these neighbours' points and the institute in C (H)
A complete graph can a little be constituted.Neighbours' point of all the points is all checked one by one in C (H), it is known that not putting can be added.This
Sample just has found a Maximum Clique.Now, all the points of this Maximum Clique are deleted, and repeat this process in left point, directly
Become sky to H.In this way, just having found a series of Maximum Clique, and each Maximum Clique is without friendship.
It is noted that corresponding thresholding figure is a group if user can be serviced by the same DC.For example, in attached drawing 1
In, user 6-9 can be serviced by common DC, therefore it is in TmIn correspondence thresholding figure be exactly in attached drawing 2 include 6-9 group.
Maximum Clique is found in thresholding figure to be meaned for user to be grouped according to the availability of DC, wherein the user in each group may be used
To be represented with one of member.
4) weight limit priority algorithm (BWF)
Input G=(U, V, E): user weights non-fully bipartite graph；K: positive integer
Export D: selected DC set
It is as follows by the sequence of non-decreasing sequence that user is weighted into side: w (1, j)≤w (2, g)≤...≤w (m,
h)
To each GrConstruct corresponding thresholding figure Tr.To each TrLook for its Maximum Clique set.It suppose there is H group.Note pair
It should be in TrMaximum Clique set be Cr=C (Tr)h, wherein r=1,2 ..., m, h=1 ..., H
Enable t=min r | | Cr|≤k}
For h=1 ..., H }
{
Enabling l is with weight limit wl=max { wj|j∈C(Tt)h, j is not assigned a little }.Enable C (Tt)hNeighbours DC be
E=minj∈N{elj|elj∈Gt}.Enabling v is the point in N, and elv=e.D=D ∪ v
}
It can be proved that BWF algorithm gives a 3- approximate solution of PDP problem, (value of the solution obtained is no more than optimal
3 times of the value of solution), wherein the bipartite graph of bottom is non-complete.And this solution is that tight (can not have more smaller than 3
Approximation ratio).
When bottom figure is Complete Bipartite Graph, Maximum Clique can be replaced with independent sets, algorithm faster, solves identical in quality.It calculates
Method BWF attempts for the data of maximum-norm to be placed into nearest DC.It weights the sequence of side from small to large according to user and examines one by one
Corresponding bottleneck figure is looked into, and constructs corresponding thresholding figure.Then the agglomeration for finding each thresholding figure closes.Each group represents
One group of user that can be at least serviced by an identical DC, and the weighting side between user and common DC is most no more than bottleneck figure
Big weighting side.It is selected that agglomeration of the number no more than k rolled into a ball in all agglomerations conjunctions closes subscript, so that maximum user weights side
It is small as far as possible.Speed can of course be further speeded up with binary search.
C is closed for each agglomeration in thresholding figuretIn user, in accordance with the following methods find target DC.At each
In group, since the user with weight limit, it is assigned to G by wetIn nearest neighbours DC.In all same groups
Remaining user also assigns to this DC by implicit.This process repeats, until all CtIn user be all assigned that (for is followed
Ring).In this way, algorithm establishes a mapping between user and DC.Bipartite graph is divided into multiple clusters, each cluster by mapping
Center be all a DC.
Cost minimization data place (CMDP) algorithm basic thought
To CMDP problem, The present invention gives NPA-CMDP algorithm, it attempts to assign to each user (uses recently
Cost minimization between family and DC) DC.If the number of DC has been more than k, some DC will be removed, and corresponding user understands quilt
Again it assigns.This process is repeated until constraint satisfaction.
Algorithm 2.CMDP best-first algorithm (NPA-CMDP).
Input G=(U, V, E): user weights non-fully bipartite graph；K: positive integer
Export D: selected DC set
Each user is assigned to nearest DC, and (Weighted distance (cost) between user and DC is defined as wieij+wipj)。
The DC set of selection is denoted as D
While{|D|>k}
{
DC is found in D to compare with remaining DC in D, between this DC and the user for assigning to it away from weighting from being most
Small.This DC is deleted from D.
And the user for assigning to this DC originally is assigned into remaining nearest DC in D again
}
The present invention considers the feasibility that the big data that cross-region is distributed is moved to cloud, has studied multiple target DC selections
Problem indicates bottom facilities with non-fully bipartite graph, thus overcome it is problematic be all Complete Bipartite Graph limitation.More
Due to not being the case where each DC can be used caused by user preference or legal restriction in tallying with the actual situation.It grinds
Study carefully four kinds of criterion, including.Fair data place (FDP), and preference data places (PDP), and transmission cost minimizes data and places
(TCMDP) and cost minimization data place (CMDP), in practical application, can be selected according to the demand of operator and user
It selects.
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be said that
Specific implementation of the invention is only limited to these instructions.For those of ordinary skill in the art to which the present invention belongs, exist
Under the premise of not departing from present inventive concept, a number of simple deductions or replacements can also be made, all shall be regarded as belonging to of the invention
Protection scope.
Claims (7)
1. the selection method of data center DC when big data is migrated to cloud a kind of, which is characterized in that the described method includes:
Bottom noncomplete graph is constructed, using activation rank wiMode the data yield of user is described, define fair data and put
Set FDP, preference data places PDP, transmission cost minimizes data and places TCMDP and cost minimization data placement CMDP etc. four
Kind criterion, and the selection based on one of above-mentioned criterion progress DC；
Wherein, the noncomplete graph G=(U, V, E), U represent user, and V represents DC set, side length eij∈ E meets triangle etc.
Formula, i ∈ U, j ∈ V, for any i ∈ U and j ∈ V, if the data of user i can be moved to DCj, then deposited between them
In a line；The method be intended to from found in available DC set V a DC subset D come according to different criterion store U in institute
Have the data of user, wherein | D |≤k, positive integer k meet k≤| U |, k≤| V |；
The FDP criterion are as follows: the distance between maximum user and the DC being assigned to minimization, so that each local user can
To access data with the smallest time delay:
The PDP criterion are as follows: the Weighted distance minimization between maximum user and the DC being assigned to, so that having most
According to local user can with the smallest time delay access data:
The TCMDP criterion are as follows: Weighted distance between all users and its DC being assigned to and minimization:
The CMDP criterion are as follows: the sum of cost of all users minimization:cijFor with activation
Rank wiUser total cost.
2. according to the method described in claim 1, weighting is non-complete it is characterized by: the noncomplete graph is weighting noncomplete graph
The side length of full figure is wi eij, wherein wiThere is activation rank, w for useri eijTo move data over into DCjThe expense that needs are paid
With.
3. according to the method described in claim 1, it is characterized by: the activation rank wiIt is determined according to the data volume uploaded daily
It is fixed.
4. according to the method described in claim 1, it is characterized by: for having activation rank wiUser i, if it thinks
DCjData are stored and processed, it needs to pay wi pj, wherein pj=p'j+p”j, p'jThe valence of unit capacity data is handled for DC
Lattice, p "jFor the storage expense of DC storage cell capacity data.
5. according to the method described in claim 1, it is characterized by: using weight limit based on the selection that PDP criterion carries out DC
Preferential BWF algorithm, the BWF algorithm include the following steps:
User is weighted into side by the sequence of non-decreasing sequence, be denoted as w (1, j)≤w (2, g)≤...≤w (m, h), wherein j, g, h ∈ V,
Bottleneck figure G1,G2,…,GmIt is the edge subgraph of G, and Gr=(U, V, Er), r=1,2 ..., m, Er={ eij|wieij≤w(r,g)}；
To each GrConstruct corresponding thresholding figure Tr, to every two point u, v ∈ U, and if only if there are a DCj ∈ V and (u,
J) and (v, j) is all in GrWhen middle, side (u, v) is in TrIn；
To each TrIts Maximum Clique set is looked for, if a group is not included in other groups, referred to as Maximum Clique；It suppose there is H
A group, note correspond to TrMaximum Clique set be Cr=C (Tr)h, wherein r=1,2 ..., m, h=1 ..., H, enable t=min r | |
Cr|≤k}
For (h=1 ..., H)
{ enabling l is with weight limit wl=max { wj|j∈C(Tt)h, j is not assigned a little }；
Enable C (Tt)hNeighbours DC beE=minj∈N{elj|elj∈Gt}；
Enabling v is the point in N, and elv=e, D=D ∪ v }.
6. according to the method described in claim 5, it is characterized by: being replaced when bottom figure is Complete Bipartite Graph with independent sets
Maximum Clique accelerates BWF algorithm.
7. according to the method described in claim 4, it is characterized by: nearest using CMDP based on the selection that CMDP criterion carries out DC
Preferential NPA-CMDP algorithm, the NPA-CMDP algorithm include the following steps:
Each user is assigned to nearest DC, the Weighted distance between user and DC is defined as wieij+wipj, by the DC of selection
Set is denoted as D；
While{|D|>k}
Find DC in set D, wherein compared with remaining DC in D, this DC for finding meet and assign to it user it
Between Weighted distance be the smallest；
This DC is deleted from D；
And the user for assigning to this DC originally is assigned into remaining nearest DC in D again }.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610067866.3A CN105739929B (en) | 2016-01-29 | 2016-01-29 | The selection method of data center when big data is migrated to cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610067866.3A CN105739929B (en) | 2016-01-29 | 2016-01-29 | The selection method of data center when big data is migrated to cloud |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105739929A CN105739929A (en) | 2016-07-06 |
| CN105739929B true CN105739929B (en) | 2019-01-11 |
Family
ID=56247304
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610067866.3A Expired - Fee Related CN105739929B (en) | 2016-01-29 | 2016-01-29 | The selection method of data center when big data is migrated to cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105739929B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106776876A (en) * | 2016-11-29 | 2017-05-31 | 用友网络科技股份有限公司 | Data migration method and data mover system |
| CN107103381A (en) * | 2017-03-17 | 2017-08-29 | 华为技术有限公司 | A kind of method and system for planning of data center |
| CN109388486B (en) * | 2018-10-09 | 2021-08-24 | 北京航空航天大学 | Data placement and migration method for heterogeneous memory and multi-type application mixed deployment scene |
| CN111427926B (en) * | 2020-03-23 | 2023-02-03 | 平安医疗健康管理股份有限公司 | Abnormal medical insurance group identification method and device, computer equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103441942A (en) * | 2013-08-26 | 2013-12-11 | 重庆大学 | Data center network system and data communication method based on software definition |
| CN104022928A (en) * | 2014-05-21 | 2014-09-03 | 中国科学院计算技术研究所 | Topology construction method of high-density server and system thereof |
| CN104809539A (en) * | 2014-01-29 | 2015-07-29 | 宏碁股份有限公司 | Dynamic planning method of data center server resource |
| CN105264457A (en) * | 2014-02-28 | 2016-01-20 | 华为技术有限公司 | Energy consumption monitoring method and device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120155468A1 (en) * | 2010-12-21 | 2012-06-21 | Microsoft Corporation | Multi-path communications in a data center environment |
-
2016
- 2016-01-29 CN CN201610067866.3A patent/CN105739929B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103441942A (en) * | 2013-08-26 | 2013-12-11 | 重庆大学 | Data center network system and data communication method based on software definition |
| CN104809539A (en) * | 2014-01-29 | 2015-07-29 | 宏碁股份有限公司 | Dynamic planning method of data center server resource |
| CN105264457A (en) * | 2014-02-28 | 2016-01-20 | 华为技术有限公司 | Energy consumption monitoring method and device |
| CN104022928A (en) * | 2014-05-21 | 2014-09-03 | 中国科学院计算技术研究所 | Topology construction method of high-density server and system thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105739929A (en) | 2016-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Du et al. | Managing noncooperative behaviors in large-scale group decision-making: Integration of independent and supervised consensus-reaching models | |
| CN105739929B (en) | The selection method of data center when big data is migrated to cloud | |
| Kuo et al. | Application of metaheuristics-based clustering algorithm to item assignment in a synchronized zone order picking system | |
| WO2018196525A1 (en) | Goods handling method and device | |
| CN107330715B (en) | Method and device for selecting picture advertisement material | |
| CN107203789B (en) | Distribution model establishing method, distribution method and related device | |
| George et al. | Genetic algorithm based airlines booking terminal open/close decision system | |
| CN104008428B (en) | Service of goods requirement forecasting and resource preferred disposition method | |
| CN109840730B (en) | Method and device for data prediction | |
| Chen et al. | Courier dispatch in on-demand delivery | |
| CN108921472A (en) | A kind of two stages vehicle and goods matching method of multi-vehicle-type | |
| CN116501711A (en) | Computing power network task scheduling method based on 'memory computing separation' architecture | |
| CN112685138B (en) | Multi-workflow scheduling method based on multi-population hybrid intelligent optimization in cloud environment | |
| CN107633358B (en) | Facility site selection and distribution method and device | |
| Roh et al. | A block transportation scheduling system considering a minimisation of travel distance without loading of and interference between multiple transporters | |
| CN114048924A (en) | Multi-distribution center site selection-distribution path planning method based on hybrid genetic algorithm | |
| CN110516985A (en) | Warehouse selection method, system, computer system and computer readable storage medium storing program for executing | |
| CN110489556A (en) | Quality evaluating method, device, server and storage medium about follow-up record | |
| Gitinavard et al. | A bi-objective multi-echelon supply chain model with Pareto optimal points evaluation for perishable products under uncertainty | |
| CN111260288B (en) | Order management method, device, medium and electronic equipment | |
| CN110232487A (en) | A kind of task allocating method and device | |
| CN109934427B (en) | Method and device for generating item distribution scheme | |
| Chirici et al. | Tackling the storage problem through genetic algorithms | |
| Fontana et al. | Multi-criteria assignment model to solve the storage location assignment problem | |
| CN112231299B (en) | Method and device for dynamically adjusting feature library |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190111 Termination date: 20220129 |