CN107273184A - A kind of optimized algorithm migrated based on high in the clouds big data with processing cost - Google Patents
A kind of optimized algorithm migrated based on high in the clouds big data with processing cost Download PDFInfo
- Publication number
- CN107273184A CN107273184A CN201710445796.5A CN201710445796A CN107273184A CN 107273184 A CN107273184 A CN 107273184A CN 201710445796 A CN201710445796 A CN 201710445796A CN 107273184 A CN107273184 A CN 107273184A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- msubsup
- munder
- data
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/4881—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/56—Provisioning of proxy services
- H04L67/563—Data redirection of data network streams
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
- G06F2009/4557—Distribution of virtual machine instances; Migration and load balancing
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
Abstract
The present invention relates to the algorithmic issue of high in the clouds big data migration optimization, belong to cloud computing applied technical field, the data and Resources Management that the present invention is handled multi-source big data high in the clouds are studied, and to optimize the cost of big data high in the clouds processing, improve its service quality;Based on this, the Data Migration and resource provision problem for first handling big data high in the clouds are converted into joint stochastic optimization problems, and then model is solved and corresponding on-line decision algorithm is designed using Liapunov optimisation technique;The algorithm does not need the to-be of forecasting system, and the current state for being based only on system is made decision.
Description
Technical field:
The present invention relates to the algorithmic issue of high in the clouds big data migration optimization, belong to cloud computing applied technical field.
Background technology:
With the fast development of network technology and mobile communication technology, the growth of data volume shows the rule of exponential increase
Rule.Figure spirit prize winner Jim Gray propose a new empirical law within 1998:
Present every 18 months newly-increased amount of storage are equal to amount of storage sum since the dawn of human civilization, up to the present, the increasing of data volume
Length substantially meets this rule, root it was predicted that the year two thousand twenty data universe is up to 35.2ZB (PB of 1Z,B=1 million), than 2009
0.8ZB increases by 44 times.In the case where so powerful actual demand is promoted, people constantly pursue mass memory capacity, high-performance, Gao An
The characteristics such as Quan Xing, high availability, scalability, manageability, the demand to storage is improved constantly.Information content is presented explosion type and increased
Long trend so that storage performance has become the bottleneck for being badly in need of improving.The data of enterprise will be stored using high in the clouds to be greatly improved
Free space, but on condition that Data Migration will be carried out, it is most important that to greatest extent safely, effectively, least cost by this
A little magnanimity big datas move to high in the clouds, and this is a system engineering, and not the work being easily accomplished, and is migrated for big data
Demand increases into geometric now, is counted according to IDC, Data Migration cost accounts for more than the 40% of whole big data application charges.
The mode operation that cloud computing is according to payable at sight so that user dynamic can adjust the money rented according to needed for itself
Source, and with high-performance and high failure tolerance, provide a kind of efficient and economy solution for big data processing and exist
Under cloud computing mode, effectively management how is carried out to data and cloud resource reduces data processing cost to closing weight for data manager
Will.Wherein, mostly important the problem of, will count:1) how dynamically by the large-scale data of the real-time generation of diverse location distribute to
The data center of geographical distribution2) need to provide how many computing resource in these data centers to ensure service quality while again
Run minimized expenseThe dynamic of the dynamic, polyphyly and the resource price that are produced due to data causes above mentioned problem to become
Obtain great challenge.
Currently, the research to big data is concentrated mainly on the high-speed parallel processing of different types of data (such as batch
The MapReduce frameworks of data processing, for the Spark systems of interactive data, for the Dreme systems of stream data processing
System, and for the Prege systems of diagram data), big data analysis apply (such as personalized recommendation, software classification, gene selects)
And in terms of big data processing basic technology, but large-scale data transmission is managed to high in the clouds and to its data and resource
The seldom of research at present, in order to solve data migration problems, usually using some simple poorly efficient method for example, data are copied
Shellfish carries out physical transportation into the hard disk of Large Copacity, entire machine directly even is transported into these methods of the such as data center again
Intolerable data processing delay can be not only produced, and can be damaged in view of hard disk in transportation, with great peace
Full hidden danger also has actual items to realize automatic duplication and the transmission data as needed between data center, but main focusing
The business demand of data, the resource required for data processing is not considered.
The content of the invention
The data and Resources Management that the present invention is handled multi-source big data high in the clouds are studied, to optimize big data cloud
The cost of processing is held, its service quality is improved;Based on this, the Data Migration and resource provision for first handling big data high in the clouds are asked
Topic is converted into joint stochastic optimization problems, and then model is solved using Liapunov optimisation technique and designs corresponding
On-line decision algorithm;The algorithm does not need the to-be of forecasting system, and the current state for being based only on system is made decision.Hair
Bright content is as follows:
The symbol of table one represents implication
(1) a kind of across data center complex optimization Data Migration and resource provision unified model are proposed, it is contemplated that many
Dynamic and the dynamic of high in the clouds different virtual machine type and its price that data source data is produced
(2) joint stochastic optimization problems is solved based on the analytic solutions derived by using Liapunov optimisation technique
Corresponding efficiently on-line decision algorithm is devised, the algorithm can make Data Migration and resource provision decision-making simultaneously and can divide
Cloth is realized.
Specifically include following steps:
(1) generally, because different VPN typically belong to different ISPs, its bandwidth price is each not
It is identical, orderTo transmit 1GB data to data center d ∈ D price from data source r ∈ R, then the bandwidth total cost of t can
It is defined as:
(2) because big data analyzes the huge of the data scale of application, the carrying cost of data is also influence data center
One key factor of selection makes sdTo store the cost required for 1GB data in single time slot on data center d ∈ D, then during t
The totle drilling cost of storage that etching system is produced is:
(3) because each cloud service provider generally uses dynamic pricing mechanism, thus the virtual machine rented from data center
Quantity the totle drilling cost and service quality of system are had a major impact orderThe k classes rented for t from data center d
Type virtual machine quantity makesFor the price of the virtual machine of k types in t data center d, then the meter required for data processing
Be counted as originally be:
(4) consider that data source is distributed in diverse geographic location with data center, needed herein using delay as data processing
The important performance indexes to be considered, will reduce the influence that delay is caused as far as possible during Data Migration,For data source r ∈ R
Transmit data to data center d delay.α is that the weight coefficient for postponing to be converted to financial cost is then postponed into the calculating changed
Cost is:
Wherein α is that delay is converted into cost formulas of the weight coefficient of financial cost based on more than, can be with guiding system
The totle drilling cost of middle generation is:
C (t)=Cp(t)+Cs(t)+Cb(t)+Cl(t) (5)
If ar(t) it is movable to for the data volume that t data source r is generated due to the data generated from arbitrary source
Arbitrary data center is handled, Wo MensheFor data center d data volume is moved to from data source r in t,
Then have for the data source r maximum amount of data produced:
The definition of each expense cost based on more than, the target studied according to this paper minimizes period [0, T] interior data
The time average unit cost of migration and processing can be turned in form:
Wherein, constraint (10) is when being equal in order to ensure distributing to the summation of each data center's data in single time slot at this
Carve the total amount of data constraints (11) produced and ensure that required virtual machine quantity is no more than the scope that data center can provide
From the point of view of problem P1 expression, because data generation is unknown and dynamic, resource variableIt is integer type, therefore problem above
It is that it is an object of the present invention under longtime running state, distributed to by optimization each by a constraint random integers optimization problem
The quantity for the virtual machine that the data of data center and data center rent is so that long term data processing average unit cost minimum is
This problem is handled, employs the optimisation technique developed a recently-Liapunov Optimization Framework to solve problem,
The details of solution will be described in detail.
Make Hd(t) in time series in data center d untreated data volume first, we define Hd(0)=0, then team
Arrange Hd(t) evolution can be described as follows:
The renewal rule of above-mentioned queue means that handled data volume isNewly arrived data volume isIn order to ensure queue Hd(t),Delay in the worst cases is in maximum functional load delay l, I
Slightly devise a respective fictional queue Zd(t) that (can be considered delay-tolerant queue) wherein, virtual queue Zd(t) at the beginning of load
Beginning turns to Zd=0, (0) and following rule is updated:
Wherein indicator functionH is worked as in expressiond(t) it is equal to 1 during > 0, otherwise equal to 0. similarly,H is worked as in expressiond
(t) it is 1 when=0, is otherwise 0. εdFor preset constant, for controlling the scope of queue delay it is possible thereby to prove, if proposing calculation
Method ensure that queue HdAnd Z (t)d(t) stablizing for a long time, then all data can be obtained at most l time-slot delay
To processing, also, l may be configured asWhereinWithIt is queue H respectivelydAnd Z (t)d(t) upper
Limit;
Make Z (t)=(Zd(t)), H (t)=(Hd(t)),Virtual queue and the matrix of actual arrays are represented respectively,
Then the confederate matrix of actual arrays and virtual queue can be represented with θ (t)=[H (t), Z (t)].According to Liapunov framework,
It is as follows that we define liapunov function:
Wherein L (θ (t)) is that the Liapunov drift function that overstocked measurement then list time slots are loaded in system can then be determined
Justice is:
Δ (θ (t))=E L (θ (t+1)) L (θ (t)) | θ (t) } (16)
For the cost produced by also minimizing system while ensureing that system queue is stable, then Liapunov drift-
Penalty term can be in above formula (16)
Increase system synthesis this function acquisition in drift function, i.e.,:
Δ(θ(t))+V·E{C(t)|θ(t)} (17)
Wherein V is non-negative parameter, and it can be traded off between the stability of a system and cost, and .V is bigger, what system was produced
Cost is just smaller, otherwise cost bigger is therefore, originally the problem of P1 P2 the problem of reformed into following:
P2.min(17) (18)
s.t.:(9)(10)(11)(12) (19)
In order to solve P2, drift-penalty is not minimized directly herein, and the upper bound for being directed to minimize it is right
And, therefore this mode of theoretical proof does not destroy the optimality and performance of algorithm, and the key for solving P2 is to find its upper bound to lead to
Cross that theory deduction is provable, the boundary of formula (17) is:
Wherein
By carefully studying on the right of inequality (20), it is found that the optimization problem can equivalently resolve into two subproblems:
I.e. data distribution problem and resource provisioning problem solve the details as described below of two above subproblem
A, Data Migration:To minimize on the right of formula (20), by the relation between observation variable, wherein with data distribution phase
The part of pass can be extracted as:
Additionally, because the data of each data source are independently generated, the multi-data source global optimization side described in formula (21)
Formula can perform in each Data Source Independent respectively and consider data distribution on t data source r, then to be converted into solution as follows for problem
Problem:
In fact, above mentioned problem is the minimal weight problem of a broad sense, and data center d power is moved to from data source r
Weight isIt overstocks H with datad(t), bandwidth costStorage cost sd, delay costRelevant is by using linear
Planning theory, we can be in the hope of solution below:
WhereinObviously, t algorithm tends to produce data source r
Data Migration to the moment there is most short task queue and the data center of minimum operating cost to be handled.
B, resource distribution:Such as remove the constant term B on the right of formula (20), then variableRelated part can be recognized
To be resource provisioning problem.Therefore, we can obtain the optimal provisioning policy of virtual machine by solving following problem:
Similarly, because the resource provision in each data center is independent, similar to data distribution problem, formula (23) can
To be solved with building distribution in each data center.Thus, for individual data center d, resource provisioning problem can further change
It is written as:
The solution of above-mentioned linear problem of being easy to get is:
Above-mentioned solution shows, when the price of t k class virtual machinesIt is smaller, and its virtual machine capacity vkIt is bigger
When, the virtual machine of k types has bigger possibility to be rented so far, and former problem is turned by using Liapunov framework
Change, for a long time in the cost minimization problem of Data Migration and resource provision obtain effectively solving that solution has briefly above
Help dispose the details of carried algorithm its on-line Algorithm as shown in algorithm 1 online in real system
Formally, this problem consider geographical distribution data center set D, its sum be D=| D |, value for d (1≤
D≤D) each data center configurations of have different types of virtual machine K (have size K=| K |), have different CPU per class virtual machine
And memory configurations, and the ability of k class virtual machines is set as vk, represent the speed of such virtual machine processing data itself and MapReduce
Concrete application is relevant, and there are different speed data managers to manage R=for different data processings | R | individual data source (is represented
For set R), and each data source (value is 1≤r≤R) dynamically produces that to need data to be processed be this, the number of any data source
Carried out according to data center that it is rented can be moved to by Virtual Private Network (Virtual Private Network, VPN)
Analysis is simulation of real scenes, we assume that from the bandwidth of the VPN connections (r, d) on data source r to data center dIt is to have
Limit, and be one of system bottleneck in addition, the data volume of each geographical position generation be it is independent, each data center
For example, resource price (virtual machine, storage) is different, and changes over time
The system is run according to time series, is divided into t=0,1 ..., T. is in each time series, data manager
Need to determine to move how many data to data center d and each data center rental how many resource to support number from data source r
The totle drilling cost that the target optimized according to processing is analyzed big data for minimum high in the clouds,
Embodiment:
On-line Algorithm step:
1st, H is inputtedd(t),Zd(t),ar(t),vk,
2nd, export
3rd, resource distribution
For each data center d ∈ D, type of virtual machine k ∈ Kdo
Pass through application
Solve the problems, such as (24)
Obtain virtual machine supply strategy
4th, Data Migration
For each data source r ∈ R, data center d ∈ D do
By applying (23)
Solve the problems, such as (21)
Obtain data distribution strategy (Xd(t))
Queue H is updated according to queue Dynamic Equation (13) (14) respectivelyd(t), Zd(t).
Claims (1)
1. a kind of optimized algorithm migrated based on high in the clouds big data with processing cost, it is characterised in that comprise the following steps:
(1) defineIt is the price that 1GB data to data center d ∈ D are transmitted from data source r ∈ R, then the bandwidth of t is total
Expense may be defined as:
<mrow>
<msub>
<mi>C</mi>
<mi>b</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>r</mi>
<mo>&Element;</mo>
<mi>R</mi>
</mrow>
</munder>
<msubsup>
<mi>&lambda;</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<msubsup>
<mi>b</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
(2) s is assumeddTo store the cost required for 1GB data in single time slot on data center d ∈ D, then t system is produced
The totle drilling cost of data storage is:
<mrow>
<msub>
<mi>C</mi>
<mi>s</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>r</mi>
<mo>&Element;</mo>
<mi>R</mi>
</mrow>
</munder>
<msubsup>
<mi>&lambda;</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<msub>
<mi>s</mi>
<mi>d</mi>
</msub>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
(3) defineThe k type Virtual machine quantity rented for t from data center d, orderFor in t data
The price of the virtual machine of k types in heart d, then the calculating cost required for data processing be:
<mrow>
<msub>
<mi>C</mi>
<mi>p</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<msubsup>
<mi>n</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<msubsup>
<mi>p</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
(4) assumeData center d ∈ D delay is transmitted data to for data source r ∈ R, α is that delay is converted into financial cost
Weight coefficient, then postpone conversion calculating cost be:
<mrow>
<msub>
<mi>C</mi>
<mi>l</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>r</mi>
<mo>&Element;</mo>
<mi>R</mi>
</mrow>
</munder>
<mi>&alpha;</mi>
<mo>&CenterDot;</mo>
<msubsup>
<mi>&lambda;</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<msubsup>
<mi>L</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
Based on cost formula above, the totle drilling cost that can obtain producing in system is:
C (t)=Cp(t)+Cs(t)+Cb(t)+Cl(t) (5)
Assuming that ar(t) data volume generated for t data source r, because the data generated from arbitrary source are movable to arbitrarily
Data center is handled, ifFor data center d data volume is moved to from data source r in t,For data source
The maximum amount of data that r is produced;Then have:
<mrow>
<msub>
<mi>a</mi>
<mi>r</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&le;</mo>
<msubsup>
<mi>A</mi>
<mi>r</mi>
<mi>max</mi>
</msubsup>
<mo>,</mo>
<mo>&ForAll;</mo>
<mi>r</mi>
<mo>,</mo>
<mi>t</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>,</mo>
<mi>T</mi>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>6</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>a</mi>
<mi>r</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&le;</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<msubsup>
<mi>&lambda;</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mo>&ForAll;</mo>
<mi>r</mi>
<mo>,</mo>
<mi>t</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>,</mo>
<mi>T</mi>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>7</mn>
<mo>)</mo>
</mrow>
</mrow>
According to definition above and it is assumed that the time average unit cost for minimizing period [0, T] interior Data Migration and processing can be with
Form is turned to:
<mrow>
<mi>P</mi>
<mn>1.</mn>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
<mo>:</mo>
<munder>
<mi>lim</mi>
<mrow>
<mi>T</mi>
<mo>&RightArrow;</mo>
<mi>&infin;</mi>
</mrow>
</munder>
<mfrac>
<mn>1</mn>
<mi>T</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>t</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mrow>
<mi>T</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<mi>E</mi>
<mo>{</mo>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>}</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>8</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>a</mi>
<mi>r</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&le;</mo>
<msubsup>
<mi>A</mi>
<mi>r</mi>
<mi>max</mi>
</msubsup>
<mo>,</mo>
<mo>&ForAll;</mo>
<mi>r</mi>
<mo>,</mo>
<mi>t</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>,</mo>
<mi>T</mi>
<mo>&rsqb;</mo>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>9</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>a</mi>
<mi>r</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<msubsup>
<mi>&lambda;</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mo>&ForAll;</mo>
<mi>r</mi>
<mo>,</mo>
<mi>t</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>,</mo>
<mi>T</mi>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>10</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mn>0</mn>
<mo>&le;</mo>
<msup>
<msub>
<mi>n</mi>
<mi>d</mi>
</msub>
<mi>k</mi>
</msup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&le;</mo>
<msubsup>
<mi>N</mi>
<mi>d</mi>
<mrow>
<mi>k</mi>
<mo>,</mo>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msubsup>
<mo>,</mo>
<mo>&ForAll;</mo>
<mi>d</mi>
<mo>,</mo>
<mo>&ForAll;</mo>
<mi>k</mi>
<mo>,</mo>
<mi>t</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>,</mo>
<mi>T</mi>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>11</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msup>
<msub>
<mi>n</mi>
<mi>d</mi>
</msub>
<mi>k</mi>
</msup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&Element;</mo>
<msup>
<mi>Z</mi>
<mo>+</mo>
</msup>
<mo>&cup;</mo>
<mn>0</mn>
<mo>,</mo>
<mo>&ForAll;</mo>
<mi>d</mi>
<mo>,</mo>
<mo>&ForAll;</mo>
<mi>k</mi>
<mo>,</mo>
<mi>t</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>,</mo>
<mi>T</mi>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>12</mn>
<mo>)</mo>
</mrow>
</mrow>
From the point of view of problem P1 expression, because data generation is unknown and dynamic, resource variableIt is integer type, therefore more than
Problem is a constraint random integers optimization problem;Wherein constraint (10) is in order to ensure being distributed in single time slot in each data
The summation of calculation evidence is equal to the total amount of data constraints (11) produced at the moment and ensure that required virtual machine quantity no more than number
Is enclosed according to what center can be provided
(5) Liapunov Optimization Framework Theoretical Design On-line Control algorithm is utilized
Make Hd(t) it is untreated data volume in data center d in time series, first, defines Hd(0)=0, then queue Hd(t)
Evolution can be described as follows:
<mrow>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
<mo>&lsqb;</mo>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<msubsup>
<mi>n</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<mn>0</mn>
<mo>&rsqb;</mo>
<mo>+</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>r</mi>
<mo>&Element;</mo>
<mi>R</mi>
</mrow>
</munder>
<msubsup>
<mi>&lambda;</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>13</mn>
<mo>)</mo>
</mrow>
</mrow>
The renewal rule of above-mentioned queue means that handled data volume isNewly arrived data volume isIn order to ensure queue Hd(t),Delay in the worst cases is in maximum functional load delay l, if
A respective fictional queue Z is countedd(t), wherein, virtual queue Zd(t) load is initialized as Zd=0, (0) and update it is as follows
Rule:
<mrow>
<msub>
<mi>Z</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
<mo>&lsqb;</mo>
<msub>
<mi>Z</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msub>
<mn>1</mn>
<mrow>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>></mo>
<mn>0</mn>
</mrow>
</msub>
<mo>(</mo>
<mrow>
<mi>&epsiv;</mi>
<mo>-</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<msubsup>
<mi>n</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>-</mo>
<msub>
<mn>1</mn>
<mrow>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mn>0</mn>
</mrow>
</msub>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<msubsup>
<mi>N</mi>
<mi>d</mi>
<mrow>
<mi>k</mi>
<mo>,</mo>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msubsup>
<mo>&CenterDot;</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<mn>0</mn>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>14</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein indicator functionH is worked as in expressiond(t) it is equal to 1 during > 0, otherwise equal to 0;Similarly,H is worked as in expressiond(t)
It is 1 when=0, is otherwise 0;εdFor preset constant, for controlling the scope of queue delay.It is possible thereby to prove, if carried algorithm
It ensure that queue HdAnd Z (t)d(t) stablizing for a long time, then all data can be obtained at most l time-slot delay
Handle and, l may be configured asWhereinWithIt is queue H respectivelydAnd Z (t)d(t) upper
Limit;
Make Z (t)=(Zd(t)), H (t)=(Hd(t)),Virtual queue and the matrix of actual arrays are represented respectively, then may be used
To represent the confederate matrix of actual arrays and virtual queue with θ (t)=[H (t), Z (t)];According to Liapunov framework, definition
Liapunov function is as follows:
<mrow>
<mi>L</mi>
<mrow>
<mo>(</mo>
<mi>&theta;</mi>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<mo>{</mo>
<msub>
<mi>Z</mi>
<mi>d</mi>
</msub>
<msup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<msup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>}</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>15</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein L (θ (t)) be system in load overstocked measurement.Then the Liapunov drift function of single time slot then may be defined as:
Δ (θ (t))=E L (θ (t+1)) L (θ (t)) | θ (t) } (16)
For the cost produced by also minimizing system while ensureing that system queue is stable, then Liapunov drift-punishment
Item can increase system synthesis this function acquisition in above formula (16) drift function, i.e.,:
Δ(θ(t))+V·E{C(t)|θ(t)} (17)
Wherein V is non-negative parameter, and it can be traded off between the stability of a system and cost, and .V is bigger, the cost that system is produced
It is just smaller, otherwise cost bigger is therefore, originally the problem of P1 P2 the problem of reformed into following:
P2.min(17) (18)
s.t.:(9)(10)(11)(12) (19)
The key for solving P2 is to find its upper bound, and provable by deriving, the boundary of formula (17) is:
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mi>&Delta;</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>&theta;</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mi>V</mi>
<mo>&CenterDot;</mo>
<mi>E</mi>
<mrow>
<mo>{</mo>
<mrow>
<munder>
<mi>lim</mi>
<mrow>
<mi>T</mi>
<mo>&RightArrow;</mo>
<mi>&infin;</mi>
</mrow>
</munder>
<mfrac>
<mn>1</mn>
<mi>T</mi>
</mfrac>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>t</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mrow>
<mi>T</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>|</mo>
<mi>&theta;</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
<mo>}</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>&le;</mo>
<mi>B</mi>
<mo>+</mo>
<mi>E</mi>
<mrow>
<mo>{</mo>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<msubsup>
<mi>n</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>Vp</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msubsup>
<mi>H</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>-</mo>
<msubsup>
<mi>Z</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mo>|</mo>
<mi>&theta;</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
<mo>}</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>+</mo>
<mi>E</mi>
<mrow>
<mo>{</mo>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>r</mi>
<mo>&Element;</mo>
<mi>R</mi>
</mrow>
</munder>
<msubsup>
<mi>&lambda;</mi>
<mi>d</mi>
<mi>r</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>Vs</mi>
<mi>d</mi>
</msub>
<mo>+</mo>
<msubsup>
<mi>Vb</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mo>+</mo>
<msubsup>
<mi>VL</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mo>+</mo>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
<mo>|</mo>
<mi>&theta;</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
<mo>}</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>20</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein
By carefully studying on the right of inequality (20), optimization problem is equivalently resolved into two subproblems:I.e. data distribution is asked
Topic and resource provisioning problem;
(6) details for solving two above subproblem is as described below:
A, Data Migration:To minimize on the right of formula (20), wherein the part related to Data Migration can be extracted as:
<mrow>
<mi>min</mi>
<mi> </mi>
<mi>E</mi>
<mo>{</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<msubsup>
<mi>&lambda;</mi>
<mi>d</mi>
<mi>r</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>Vs</mi>
<mi>d</mi>
</msub>
<mo>+</mo>
<msubsup>
<mi>Vb</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mo>+</mo>
<msubsup>
<mi>V&alpha;L</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mo>+</mo>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>|</mo>
<mi>&theta;</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>}</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>21</mn>
<mo>)</mo>
</mrow>
</mrow>
Additionally, because the data of each data source are independently generated, the multi-data source global optimization mode described in formula (21) can
To be performed respectively in each Data Source Independent, it is considered to data distribution on t data source r, then the data migration problems are converted into
Solve following problem:
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<msubsup>
<mi>&lambda;</mi>
<mi>d</mi>
<mi>r</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&lsqb;</mo>
<msub>
<mi>Vs</mi>
<mi>d</mi>
</msub>
<mo>+</mo>
<msubsup>
<mi>Vb</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mo>+</mo>
<msubsup>
<mi>V&alpha;L</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mo>+</mo>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&rsqb;</mo>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
<mo>.</mo>
<mrow>
<mo>(</mo>
<mn>9</mn>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mn>10</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>22</mn>
<mo>)</mo>
</mrow>
</mrow>
The problem is the minimal weight problem of a broad sense, is from the data source r weights for moving to data center dIt with
Data overstock Hd(t), bandwidth costStorage cost sd, delay costIt is relevant, by using programming theory, we
Can be in the hope of solution below:
<mrow>
<msubsup>
<mi>&lambda;</mi>
<mi>r</mi>
<mi>d</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msub>
<mi>&alpha;</mi>
<mi>r</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>d</mi>
<mo>=</mo>
<msup>
<mi>d</mi>
<mo>*</mo>
</msup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mrow>
<mi>e</mi>
<mi>l</mi>
<mi>s</mi>
<mi>e</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>23</mn>
<mo>)</mo>
</mrow>
</mrow>
WhereinObviously, t algorithm tends to the number for producing data source r
According to migrating to the moment there is most short task queue and the data center of minimum operating cost to be handled.
B, resource distribution:Such as remove the constant term B on the right of formula (20), then variableRelated part is considered
Resource provisioning problem, the optimal provisioning policy of virtual machine is obtained by solving following problem:
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
<mi> </mi>
<mi>E</mi>
<mo>{</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>d</mi>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<msubsup>
<mi>n</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<mrow>
<mo>(</mo>
<msubsup>
<mi>Vp</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
<mo>-</mo>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>-</mo>
<msub>
<mi>Z</mi>
<mi>d</mi>
</msub>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>|</mo>
<mi>&theta;</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
<mo>.</mo>
<mrow>
<mo>(</mo>
<mn>11</mn>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mn>12</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>24</mn>
<mo>)</mo>
</mrow>
</mrow>
Similarly, because the resource provision in each data center is independent, similar to data distribution problem, formula (23) can be
Each data center solves with building distribution, thus, for individual data center d, resource provisioning problem can be further rewritten as:
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mi>min</mi>
<mi> </mi>
<mi>E</mi>
<mrow>
<mo>{</mo>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<msubsup>
<mi>n</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>Vp</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>-</mo>
<msub>
<mi>Z</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mo>|</mo>
<mi>&theta;</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
<mo>}</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
<mo>.</mo>
<mrow>
<mo>(</mo>
<mn>11</mn>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mn>12</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>25</mn>
<mo>)</mo>
</mrow>
</mrow>
The solution of above-mentioned linear problem of being easy to get is:
<mrow>
<msubsup>
<mi>n</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msubsup>
<mi>N</mi>
<mi>d</mi>
<mrow>
<mi>k</mi>
<mo>,</mo>
<mi>max</mi>
</mrow>
</msubsup>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mi>f</mi>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msub>
<mi>Z</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>></mo>
<mfrac>
<mrow>
<msubsup>
<mi>Vp</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
</mfrac>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>0</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mi>f</mi>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>H</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msub>
<mi>Z</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>&le;</mo>
<mfrac>
<mrow>
<msubsup>
<mi>Vp</mi>
<mi>d</mi>
<mi>k</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
</mfrac>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>26</mn>
<mo>)</mo>
</mrow>
</mrow>
4
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710445796.5A CN107273184A (en) | 2017-06-14 | 2017-06-14 | A kind of optimized algorithm migrated based on high in the clouds big data with processing cost |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710445796.5A CN107273184A (en) | 2017-06-14 | 2017-06-14 | A kind of optimized algorithm migrated based on high in the clouds big data with processing cost |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107273184A true CN107273184A (en) | 2017-10-20 |
Family
ID=60066572
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710445796.5A Pending CN107273184A (en) | 2017-06-14 | 2017-06-14 | A kind of optimized algorithm migrated based on high in the clouds big data with processing cost |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107273184A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108446383A (en) * | 2018-03-21 | 2018-08-24 | 吉林大学 | A kind of data task redistribution method based on geographically distributed data query |
CN108664328A (en) * | 2018-04-02 | 2018-10-16 | 上海交通大学 | A kind of effective computing architecture optimization method of accelerating type perception |
CN109089271A (en) * | 2018-09-17 | 2018-12-25 | 中南大学 | A kind of wireless cooperation Internet resources random distribution approach of mixed tensor energy supply |
CN109819032A (en) * | 2019-01-24 | 2019-05-28 | 中山大学 | A kind of base station selected cloud robot task distribution method with computation migration of joint consideration |
CN110413231A (en) * | 2019-07-16 | 2019-11-05 | 东华大学 | The data-optimized storage method of inexpensive high availability under a kind of cloudy environment |
CN110704504A (en) * | 2019-09-20 | 2020-01-17 | 天翼征信有限公司 | Data source acquisition interface distribution method, system, storage medium and terminal |
CN112540727A (en) * | 2020-12-04 | 2021-03-23 | 北京百度网讯科技有限公司 | Data management method and device, computing equipment, storage medium and cloud platform |
CN117610970A (en) * | 2024-01-04 | 2024-02-27 | 成都开元精创信息技术有限公司 | Intelligent evaluation method and system for data migration work |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106126340A (en) * | 2016-06-23 | 2016-11-16 | 中国人民解放军国防科学技术大学 | A kind of reducer system of selection across data center's cloud computing system |
CN106155785A (en) * | 2016-06-23 | 2016-11-23 | 中国人民解放军国防科学技术大学 | A kind of data migration method across data center's cloud computing system |
US9524243B1 (en) * | 2011-09-27 | 2016-12-20 | Emc Ip Holdng Company Llc | Scalable monolithic data storage system for cloud environment |
-
2017
- 2017-06-14 CN CN201710445796.5A patent/CN107273184A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9524243B1 (en) * | 2011-09-27 | 2016-12-20 | Emc Ip Holdng Company Llc | Scalable monolithic data storage system for cloud environment |
CN106126340A (en) * | 2016-06-23 | 2016-11-16 | 中国人民解放军国防科学技术大学 | A kind of reducer system of selection across data center's cloud computing system |
CN106155785A (en) * | 2016-06-23 | 2016-11-23 | 中国人民解放军国防科学技术大学 | A kind of data migration method across data center's cloud computing system |
Non-Patent Citations (1)
Title |
---|
肖文华,包卫东等: ""面向多源大数据云端处理的成本最小化方法"", 《软件学报》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108446383A (en) * | 2018-03-21 | 2018-08-24 | 吉林大学 | A kind of data task redistribution method based on geographically distributed data query |
CN108446383B (en) * | 2018-03-21 | 2021-12-10 | 吉林大学 | Data task redistribution method based on geographic distributed data query |
CN108664328B (en) * | 2018-04-02 | 2021-08-17 | 上海交通大学 | Accelerated sensing high-utility computing architecture optimization method |
CN108664328A (en) * | 2018-04-02 | 2018-10-16 | 上海交通大学 | A kind of effective computing architecture optimization method of accelerating type perception |
CN109089271A (en) * | 2018-09-17 | 2018-12-25 | 中南大学 | A kind of wireless cooperation Internet resources random distribution approach of mixed tensor energy supply |
CN109089271B (en) * | 2018-09-17 | 2021-12-14 | 中南大学 | Hybrid energy powered wireless cooperative network resource random allocation method |
CN109819032A (en) * | 2019-01-24 | 2019-05-28 | 中山大学 | A kind of base station selected cloud robot task distribution method with computation migration of joint consideration |
CN109819032B (en) * | 2019-01-24 | 2021-06-29 | 中山大学 | Cloud robot task allocation method considering base station selection and computing migration in combined manner |
CN110413231A (en) * | 2019-07-16 | 2019-11-05 | 东华大学 | The data-optimized storage method of inexpensive high availability under a kind of cloudy environment |
CN110704504A (en) * | 2019-09-20 | 2020-01-17 | 天翼征信有限公司 | Data source acquisition interface distribution method, system, storage medium and terminal |
CN112540727A (en) * | 2020-12-04 | 2021-03-23 | 北京百度网讯科技有限公司 | Data management method and device, computing equipment, storage medium and cloud platform |
CN117610970A (en) * | 2024-01-04 | 2024-02-27 | 成都开元精创信息技术有限公司 | Intelligent evaluation method and system for data migration work |
CN117610970B (en) * | 2024-01-04 | 2024-04-02 | 成都开元精创信息技术有限公司 | Intelligent evaluation method and system for data migration work |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107273184A (en) | A kind of optimized algorithm migrated based on high in the clouds big data with processing cost | |
Senthil Kumar et al. | Multi-objective task scheduling using hybrid genetic-ant colony optimization algorithm in cloud environment | |
Zhang et al. | Virtual network embedding based on modified genetic algorithm | |
Yoshizaki | Scatter search for a real-life heterogeneous fleet vehicle routing problem with time windows and split deliveries in Brazil | |
CN103226762B (en) | A kind of Logistics Distribution Method based on cloud computing platform | |
Li et al. | A two-level self-adaptive variable neighborhood search algorithm for the prize-collecting vehicle routing problem | |
CN108170530B (en) | Hadoop load balancing task scheduling method based on mixed element heuristic algorithm | |
CN106126340B (en) | A kind of reducer selection method across data center's cloud computing system | |
Jangiti et al. | Scalable and direct vector bin-packing heuristic based on residual resource ratios for virtual machine placement in cloud data centers | |
Goldbarg et al. | A transgenetic algorithm applied to the traveling car renter problem | |
CN106155785B (en) | A kind of data migration method across data center's cloud computing system | |
Naik et al. | Multiobjective virtual machine selection for task scheduling in cloud computing | |
Blythe et al. | Farm: Architecture for distributed agent-based social simulations | |
Adoni et al. | The MapReduce-based approach to improve the shortest path computation in large-scale road networks: the case of A* algorithm | |
Akhmetov et al. | Mobile platform for decision support system during mutual continuous investment in technology for smart city | |
CN103970612A (en) | Load balancing method and device based on pre-division of virtual machine | |
Shao et al. | Production system performance improvement by assembly line-seru conversion | |
Khanzadi et al. | Application of CBO and CSS for resource allocation and resource leveling problem | |
CN106209975B (en) | A kind of resource provision method across data center's cloud computing system | |
CN115917509A (en) | Reduction server for fast distributed training | |
Ma et al. | Cloud-based multidimensional parallel dynamic programming algorithm for a cascade hydropower system | |
Jing et al. | Quantum-inspired immune clonal algorithm for railway empty cars optimization based on revenue management and time efficiency | |
CN105589752A (en) | Cross-data center big data processing based on key value distribution | |
Siemiński | Verifying usefulness of ant colony community for solving dynamic tsp | |
Qian et al. | A Cost-driven multi-objective optimization algorithm for SaaS applications placement |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171020 |