CN102289766A - Method for scheduling grid resources based on continuous two-way auction mechanism - Google Patents

Abstract

The invention provides a method for scheduling grid resources based on a continuous two-way auction mechanism and belongs to the technical field of network. The method comprises a preparation period, a screening period and an auction period. In the method, an auction mode serves as a pricing mode of grid resource capacity, and the grid resources are allocated in the auction mode, so that market information can be well reflected, lots of users are fully mobilized to add idle resources into grids, and resource scheduling efficiency in the grid environment is improved; the conventional continuous two-way auction mechanism is improved and applied to scheduling of the grid resources which divide blocky capacity; and a method for matching the grid resources beyond the requirements is designed and better applicable to practical situation of scheduling of short-term grid resources.

Description

A kind of grid resource scheduling method based on continuous two way auction mechanism

Technical field

The invention belongs to networking technology area, be specifically related to a kind of grid resource scheduling method based on continuous two way auction mechanism.

Background technology

Gridding technique can connect the various resources in the network and integrate, and the vacant share of resource is offered the user who needs them.Provide the slack resources of oneself in order to make resource provider add grid energetically, grid need be set up abundant driver resource supplier and with respect to the resource distribution mechanism of common interest.Traditional gridding resource configuration mechanism is based on the managerial personnel manual allotment that the user directly applies for resource and resource provider, can't embody the interests separately of grid user and resource provider, also can't realize the dynamic allotment of gridding resource.

Can reflect market information well and give full play to users as the pricing method of gridding resource ability with auction slack resources is joined grid.

Summary of the invention

Deficiency at existing grid resource scheduling method, the invention provides a kind of grid resource scheduling method based on continuous two way auction mechanism, with the right to use of pervasive resource in the short-term time range idle in the grid as commodity to carry out the resource rental transaction, and the combinatorial optimization problem in the mechanism is found the solution, and related resource is dispatched according to the result who obtains with genetic algorithm.

One, the basic principle of the grid resource scheduling method based on continuous two way auction mechanism of the present invention:

The classification of 1 gridding resource

According to application target, the ability kind that the present invention provides gridding resource according to resource is different and be divided into Mesh Processing resource and grid storage resources.The Mesh Processing resource comprises all can provide processing power for affairs or a certain class particular problem on grid resource, in the middle of these resources, computational resources such as computing machine provide the processing processing power for data, production equipment in the factory provides for product starting material or half-finished processing processing power, and the human processing power that provides the variety of issue that in actual life and work, runs into as the social man.And the grid storage resources comprises that all can provide certain space to hold the article of regulation or the resource of information on the grid in a period of time, in these resources, the external memory of server can be used for storing data information in the network, warehouse in the manufacturing district can be used for storing starting material, semifinished or finished goods, and the bandwidth resources in the Internet resources then are used for the packet that temporary transient storage is flowed.

But this situation appears in above-mentioned classification sometimes, and a resource had both comprised processing power and also comprised storage capacity.If this moment, the storage capacity of this resource was served for processing power, it is considered herein that this resource is to handle resource; And if the storage capacity of this resource can be independent of processing power, or the storage capacity that provides has exceeded subsidiary book and has been in the required quantity of reason ability, it is considered herein that so this resource is to handle the synthesis of resource and storage resources.

The present invention's supposition is in a certain class resource transaction, if some resource is also often concluded the business because of demand is big, association area can be made the standard criterion of such resource requirement so.Whether meet these standard criterions according to gridding resource, the present invention is divided into pervasive resource and special resource with gridding resource.This point-score also show simultaneously the gridding resource that meets some pervasive resource criterion can be according to owner's wish resource is included into a certain pervasive resources-type in, thereby the same standard of value with such pervasive resource of similar pervasive resource with other is concluded the business; Or this kind resource concluded the business with the buyer's oneself the standard of value as a kind of specific special resource.When the seller is used as ability as the special ability processing, the special value place that can point out its resource institute providing capability by extra descriptor; And when ability is used as the transaction of pervasive ability, be confined in the scope of standard code for the descriptor of this ability.

Use above two kinds of rules to divide simultaneously, the present invention is divided into pervasive processing resource, special processing resource, pervasive storage resources and special storage resources with resource.The goods to auction of concluding the business among the present invention is exactly in fact by the right to use of a certain class resource in a certain period in these resources, and the actuals that final victor obtains by transaction is the corresponding ability that interior during this period of time this resource provides.Because the time corresponding of having rented of resources use right limits, so its ability that provides has time attribute naturally.The present invention is divided into short-term ability and long-term ability according to the length of the shortest lasting rental period of ability with ability.

When the pervasive ability of transaction short-term, the ability that the auction system of the present invention's design takes block ability description to come each seller is provided is carried out the standardization statistics, and each ability piece has following three attributes: start time, duration, capacity.The present invention will be called the term of validity of ability piece from the start time of ability piece continuous time through this section after the duration.When the commodity of transaction were processing power, what capacity was represented was the task amount that can finish in the term of validity.When the commodity of transaction were storage capacity, what capacity was represented was the maximum storage that can admit in the random time in the term of validity.The essence that block ability is divided be to use performance curve portrayal resource capability over time after, use the quantity delivered of ability in method statistic units chunk time of discretize.And the duration size of ability piece should reflect that not only the chronomere of associated user's demand measures, and also should make the seller add up resource capability as a kind of time range of reasonable tolerance.The bulk of ability is divided referring to Fig. 1.

When the long-term ability of transaction, the auction platform of the present invention's design takes banded ability description to portray the buyer's demand and the seller's supply.Each ability has following three attributes: start time, duration and instantaneous ability value.Equally, the present invention will be called the term of validity of ability band from the start time of ability band continuous time through this section after the duration.When the commodity of transaction were processing power, what instantaneous ability value was represented was the average task amount that can finish in the time per unit in the term of validity.When the commodity of transaction were storage capacity, what instantaneous ability value was represented was the maximum storage that any time is distributed to this buyer in the term of validity.

2 user's requests and the market segmentation

To the classification of goods to auction, the present invention is divided into the auction marketplace and is used to the conclude the business market of the pervasive ability of short-term, the market of the long-term pervasive resource capability of transaction and the market of transaction special ability according to above.By market being divided into three classes, the present invention has designed relevant auction mechanism respectively.

As shown in Figure 2, the buyer that may take part in auction can be divided into large-scale user, application service provider and special user according to the purpose difference that takes part in auction.Wherein, the purpose that the application service provider takes part in auction is in order to buy a large amount of widely used by the user, processing power and storage capacities with mass market, will these abilities integrates, through selling medium and small sized enterprises or domestic consumer in the mode of price after the specific configuration.It is need rent a large amount of, long-term resource because of the demand of itself that large-scale user takes part in auction.And the purpose that the special user takes part in auction is to rent some special resource ability for the individual demand needs that satisfy oneself in a period of time.The seller who participates in these auctions then is a zonal gridding resource provider.

As shown in Figure 3, large-scale user and application service provider's ability need may be a time dependent curve.Generally, these grid users can according to the demand cycle of each period rent needs in following a period of time ability and actual conditions of demand make comparisons, over-evaluating shown in the figure, average and underestimate three kinds of situations may appear.The long-term ability of buying as the buyer is in average or when underestimating, the demand that a certain period can take place has exceeded the situation that resource can be supplied in user's hand.The sudden ability shortage that produces needs the buyer to obtain a certain amount of short-term ability.

For two kinds of different demands of long and short phase, the present invention uses continuous two way auction to cooperate block ability description to satisfy the short-term ability need of the buyer for pervasive ability at the buyer.

3 continuous two way auction mechanism

The present invention makes amendment to original continuous two way auction mechanism, and cooperates block ability division to make it to be applicable to the pervasive ability transaction situation of short-term.

3.1 design main points

In two way auction, buyer and seller equates on the status, and just the purpose that takes part in auction of a side is in order to buy resource, and the purpose that the opposing party takes part in auction is in order to sell resource.At this moment, auction platform is as the third party, and its prime responsibility is to receive the target information of both parties' submission and according to certain rule both parties are concluded the transaction.Therefore, how auction platform is handled the target that both parties submit to and how relevant auxiliary mechanism is set becomes the subject matter that designs two way auction, and its problem that may should be noted that is as follows:

Handle in real time and batch processing.After auction platform receives buyer's target, can check the formation of seller's target immediately and select suitable seller set and make knockdown price and both sides are concluded the transaction with certain rule; Carry out the batch coupling after also this buyer's target can being delayed a period of time and waiting for the arrival of other buyer's target together.Be not difficult to find out that if be criterion to satisfy buyer's target number, coupling may have better effect than real-time coupling so in batches.

Supply-demand information and trading rules.If auction can enough predict supply and demand situation in this section period and regulation when mating target with real-time processing, auction can make real-time coupling reach effect as well as possible with corresponding regulation rule.Thus, if auction can be selected corresponding trading rules in view of the above to judge the general trend of supplydemand relationship according to relevant information.For example: can stipulate when supply exceed demand, seller's target be arranged that this moment, those sellers that establish reserve price lower more likely sold the resource that it provides with the order that reserve price is ascending.And when supply-less-than-demand, with the order that reserve price is descending seller's target is arranged, can satisfy the buyer's demand this moment as much as possible.

Priority and supplementary.When auction platform need be handled several target simultaneously, auction platform needed a kind of auxiliary mechanism to solve the situation that conflict occurs.In addition, also can adopt some supplementarys to obtain a ratio, use this ratio to distribute this demand according to formulating good rule.

3.2 both sides' target portrayal

3.2.1 short-term ability need portrayal

The present invention uses block ability description that the buyer's demand, ability that the seller provides and the actual schedule of operation are portrayed when handling the short-term ability.

The present invention uses operation to describe the short-term processing capability requirements that grid user may occur, the association attributes of operation has: workload, homework type, operation submission time, expected performance time and off period, the term of validity that during this period of time be called this operation of the present invention till will be from the submission time of operation to the off period.Different and several situations that may occur in the position of the expected performance time scope that homework type has been represented operation in the operation term of validity comprise that The faster the better, preferably and good more more slowly on time.On behalf of the expected performance time of operation, The faster the better be positioned at the operation term of validity, apart from the nearest ability block end time of operation submission time; More slowly easily represent more the expected performance time of operation be positioned at operation by the phase; And preferably represented other situation except that above-mentioned two kinds of situations on time.In addition, use following form to describe the short-term storage ability need that each grid user may occur, the association attributes of demand has: demand, rent zero-time and off period.Demand represents the buyer from renting zero-time to the constant demand of this section of off period in the time.

As can be seen, processing capability requirements is because definite corresponding relation of life period and demand not is actually a kind of elastic demand from foregoing description.Auction platform can be according to the buyer's associated description, according to the actual provision amount in the operation term of validity relevant ability is dispatched and satisfies this demand.And the storage capacity demand is actually a kind of rigid demand, if there is supply deficiency in a certain unit interval in the official hour scope, then this demand can't be satisfied.

This shows that the disposition that auction platform is compared pervasive storage capacity for the processing of the pervasive processing power of short-term is more flexible.Therefore the main relevant design of describing about the pervasive processing power of concluding the business of following explanation is correspondingly mentioned between the two design difference according to actual conditions.

3.2.2 buyer's target portrayal

With the transaction processing ability is example, should comprise in the target that ability need person (buyer) submits to:

User's identify label (User Identification): uid.The identify label center that this identify label is subordinate to can be judged the whether legal of this user ID.After the legitimacy authentication, in auction platform, use unique corresponding with it resource requirement person to identify bid _jRepresent this user.

The workload (Length) that needs processing: L _jRequired ability number of blocks after this variable has represented operation that buyer j submits to according to the capacity conversion of ability piece, this variable is 1 with integer representation and minimum value.The capacity of ability piece is stipulated by auction platform.

The type of preferences of operation deadline (Preference Type): pt _jPt _j∈ S_pt, wherein S_pt is the type of preferences set of deadline, S_pt={pt ₁, pt ₂, pt ₃, pt ₁, pt ₂, pt ₃Respectively representative The faster the better, preferably and good more more slowly type of preferences on time.

Operation submission time (Job Submitting Time): T _Js ^jT _Js ^jRepresenting buyer j can guarantee to sell in this time point can obtain required information and begin to handle this operation.

Expected performance time (Expected Finished Time):

The representative when the seller when this fulfils assignment constantly, buyer j is the most satisfied for the performance of operation.

Off period (Deadline): Dl _jDl _jRepresent the receptible operation Late Finish of buyer j.

Expectation expenditure (Expected Cost):

Representative is when finishing the amount of money that this operation spends and be worth smaller or equal to this, and buyer j pleases oneself.

Budget (Budget): B _jB _jRepresent buyer j to finish the Maximum Amount that this operation can be paid, can draw buyer j thus is to finish the units chunk best bid that this operation can be paid

When handling the storage capacity transaction, should comprise in the target that ability need person submits to: user's identify label, rent zero-time, off period, unit interval demand, expectation expenditure and budget.

3.2.3 seller's target portrayal

With the processing power is example, and the ability description that resource provider provides should comprise:

User's identify label (User Identification): uid.The identify label center that this identify label is subordinate to can be judged the whether legal of this user ID.After the legitimacy authentication, in auction platform, use unique corresponding with it resource provider sign pid _iRepresent this resource provider.

The expectation commercial value of each ability piece (Expected Share Price): Esp _iEsp _iRepresent the ability piece commercial value of resource provider i expectation.

The reservation price of every ability (Reserved Share Price): Rsp _iRsp _iRepresent the receptible lowest capability piece commercial value of resource provider i.

The supply formation of ability piece, each element wherein comprise following two attributes:

The start time (Start Time) of this group ability piece:

This group ability piece time corresponding scope of representing resource provider i to provide is

Arrive

Should keep with the zero-time of auction platform regulation promptly should satisfying synchronously

(0≤k≤N _Ms), T wherein _SpRepresent the time span of every ability of auction platform regulation.

The supply of ability (Block Number):

Represent resource provider i to exist

Arrive Time range in the ability number of blocks that provides.

When handling the storage capacity transaction, every and every implication is identical therewith in the target that the seller submits to.

Two, technical scheme of the present invention

Grid resource scheduling method based on continuous two way auction mechanism of the present invention, by inch of candle mechanism is carried out grid resource scheduling, changes continuous two way auction into cycle system, and the adjacent auction cycle joins end to end; But the total duration in an auction cycle equates with the time range of its trading capacity and is corresponding one by one; Ability in the corresponding with it time range of in this auction cycle, can only concluding the business, and the ability in scope sometime can only be concluded the business in the auction cycle in correspondence; Each auction cycle comprises the steps:

Step 1, preparatory stage: finish the trust stage in the traditional auction, be responsible for receiving seller's target and add up and predict supply-demand information as required;

Step 1.1: receive seller's target;

Step 1.2: upgrade each user's integration information, integration is represented the ability piece number that this user has struck a bargain;

Step 1.3: prediction dealing side supplydemand relationship:, predict reasonable knockdown price Rsap and reasonable demand amount Ardm in this auction cycle according to average knockdown price Sap of piece in the periodicity transaction record in the past and reasonable demand amount Rdm; First cycle do not predict, supposes the auction platform state that is in that supply exceed demand; The change of supposing these data is clocklike, uses m historical data to go to estimate m+1 data, promptly known A ₁, A ₂, A ₃Λ, A _m, ask A _M+1Select for use linear prediction and model prediction as finding the solution A _M+1Two kinds of methods;

(a) linear prediction

Suppose A _kBe from A ₁To A _K-1The weights of this k number and, A _K+1Be from A ₂To A _kThis k number and ..., A _mBe from A _M-k+1To A _M-1Weights and, and the weights correspondent equal of all groups;

$\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">w</figure-callout>}}_1{\&times;\mathrm{<figure-callout\; id="1"\; label="A"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">A</figure-callout>}}_1+w_2{\&times;A}_2+\mathrm{\&Lambda;}+w_{k-1}\&times;A_{k-1}=A_k\\ w_1\&times;A_2+w_2\&times;A_3+\mathrm{\&Lambda;}+w_{k-1}\&times;A_k=A_{k+1}\\ M\\ w_1{\&times;A}_{m-k}+w_2{\&times;A}_{m-k+1}+\mathrm{\&Lambda;}+w_{k-1}\&times;A_{m-2}=A_{m-1}\\ {\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">w</figure-callout>}}_1{\&times;A}_{m-k+1}+w_2{\&times;A}_{m-k+2}+\mathrm{\&Lambda;}+w_{k-1}{\&times;A}_{m-1}=A_m\end{array}\right.---\left(3.4\right)$

Get

Behind the solving equation, use w ₁* A _M-k+2+ w ₂* A _M-k+3+ Λ+w _K-1* A _mValue as A _M+1Predicted value;

(b) model prediction

Suppose (A _L+1, A _L+2, Λ, A _L+n-1) this continuous n-1 number pattern and the A that form _L+nForm a kind of corresponding relation; In like manner, with A _M+1That relevant is pattern M ₁(A _M-n+2, A _M-n+3, Λ, A _m); In order to estimate A _M+1Value, need from historical data, find out and pattern M ₁The most close pattern M ₂(A _J+1, A _J+2, Λ, A _J+n-1); According to pattern M ₂Corresponding value A _J+nCan estimate A _M+1Fuzzy value; From the time consideration, from pattern M ₁Nearest pattern is pattern M ₃(A _M-n+1, A _M-n+2, Λ, A _M-1), by comparing the difference between these patterns, can be at A _J+nObtain a more suitable value as A around the value _M+1Estimated value;

Select the module of Euclidean distance (Euclidean distance) for use as the similarity degree between the pattern; Suppose in all patterns, with pattern M ₁Nearest pattern be M ₂, distance between the two is dist (M ₁, M ₂), pattern M ₁With pattern M ₃Between distance be dist (M ₁, M ₃), use following formula (3.5) to estimate A _M+1Value;

$A_{m+1}=A_{j+n}+\frac{\mathrm{dist}({\mathrm{<figure-callout\; id="1"\; label="M"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">M</figure-callout>}}_1,M_2)}{\mathrm{dist}({\mathrm{<figure-callout\; id="1"\; label="M"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">M</figure-callout>}}_1,M_3)}\&times;(A_{j+n}-A_m)---\left(3.5\right)$

All predict reasonable knockdown price Rsap and reasonable demand amount Ardm by above dual mode at first, compare two kinds of methods error size in nearly N the auction cycle, the less Forecasting Methodology of use error is auctioned the variable Forecasting Methodology in cycle as this;

According to supply and the reserve price in the seller's target that receives, calculate reasonable supply (Reasonable Supple) Rsup in this auction cycle, on behalf of all sellers' that receive in the preparatory stage reservation price, Rsup be not higher than the supply sum of the reasonable knockdown price Rsap that estimates;

Step 1.4: calculate supply and demand scale parameter ω, computing formula is: ω=Rsup/Ardm (3.3)

When ω＞1, think that overall supplies that the seller provides can satisfy most of buyer's demand, and can be in the operation term of validity overall supplies be to select optimal that part of ability piece to improve the buyer's consumption satisfaction the ability set of blocks of ω times of workload; At this moment, the piece knockdown price Sp of each ability piece of buyer j and seller i transaction _Ij, use following formula (3.1) to calculate:

${\mathrm{Sp}}_{\mathrm{ij}}=\left\{\begin{array}{cc}{\mathrm{Rsp}}_i+({\mathrm{Hsp}}_j-{\mathrm{Rsp}}_i)\&times;1/(1+\mathrm{\&omega;})& \mathrm{\&omega;}<{\mathrm{\&omega;}}_{\max }\\ {\mathrm{Rsp}}_i& \mathrm{\&omega;}\&GreaterEqual;{\mathrm{\&omega;}}_{\max }\end{array}\right.---\left(3.1\right)$

When ω≤1, think that the overall supplies that the seller provides may be less than demand, this moment, auction platform just merely checked whether can fulfil assignment and do not consider both sides' satisfaction in the operation term of validity; At this moment, the piece knockdown price Sp of each ability piece of buyer j and seller i transaction _Ij, use following formula (3.2) to calculate:

${\mathrm{Sp}}_{\mathrm{ij}}=\left\{\begin{array}{cc}{\mathrm{Rsp}}_i+({\mathrm{Hsp}}_j-{\mathrm{Rsp}}_i)\&times;1/(1+\mathrm{\&omega;})& \mathrm{\&omega;}<{\mathrm{\&omega;}}_{\min }\\ {\mathrm{Rsp}}_j& \mathrm{\&omega;}\&GreaterEqual;{\mathrm{\&omega;}}_{\min }\end{array}\right.---(3.2)$

Step 2, auction phase: auction platform receives buyer's target, finishes coupling and accounts settling phase according to demand; The seller can revise the reservation price of the current ability of sale at any time or supply more ability in this stage; Each transaction clearing back auction platform writes down this transaction and issues the buyer and show scheduling time;

Step 2.1: when ω＞1, think that overall supplies that the seller provides can satisfy most of buyer's demand; When ω≤1, think that the overall supplies that the seller provides may be less than demand;

During this auction cycle appearance situation that supply exceed demand, auction platform uses the matching algorithm of the information of considering both sides' satisfaction to choose the resource capability that those meet buyer's demand most; This auction cycle, it should be purpose in the ability of day part supply fully to sell out the seller that shortage of resources can cause auction platform when the situation of supply-less-than-demand occurring;

During this auction cycle appearance situation that supply exceed demand, the non real-time trigger condition is set to after a target arrives auction platform, target quantity and preset value in the formation of current buyer's target are compared, if the target number in the formation reaches preset value, then trigger coupling immediately, otherwise the setting timer is when timer mates the residue target in the target formation to after date; When supply-less-than-demand, what auction platform used is the matching algorithm of not considering both sides' satisfaction, and the trigger condition of algorithm is set to real-time triggering;

Step 2.2: auction platform is set following variable:

The capacity of ability piece (Block Capability): Bcp;

The time span of each ability piece (Block Time Span): T _Sp

But the maximum time span of trading capacity time range: N _Ms* T _SpBut the time range of then auctioning trading ability piece in the phase is [T _Ast, T _Ast+ N _Ms* T _Sp], T _AstRepresent the start time of the piece of ability the earliest that can conclude the business in this auction phase;

Supply and demand ratio lower limit: ω _MinWhen the supply and demand ratio in an auction cycle is not higher than this preset value, the knockdown price of ability piece is set at the buyer's piece best bid;

The supply and demand ratio upper limit: ω _MaxWhen the supply and demand ratio in an auction cycle is not less than this preset value, the knockdown price of ability piece is set at the seller's piece reservation price;

Step 2.3:, mate if use to mate in batches then change step 2.5 if mate if auction platform uses real-time coupling then changes step 2.4 ω＞1; Otherwise supply-less-than-demand uses step 2.6 to mate;

Step 2.4: the Real Time Matching Algorithm when supply exceed demand;

Auction platform triggers matching algorithm after receiving the target of buyer's transmission immediately;

When the transaction processing ability, for the target request that the buyer proposes, taking-up residue overall supplies in buyer's operation term of validity is not less than the seller of ω times of workload as seller's set of satisfying this demand from seller's formation;

When the transaction storage capacity, taking-up residue overall supplies in the required time section is not less than the seller of ω times of demand as seller's set of satisfying this demand from seller's formation;

According to the satisfaction function that statistics draws, the workload in buyer's demand is suitably distributed to each seller according to the principle of total satisfaction maximum;

Step 2.4.1: what the seller gathered determines:

When transaction is when handling resource, what suppose that current platform receives is the target that buyer j sends, according to the workload L in the target _jDetermine for the required supply of the current target of optimization process

According to operation submission time T _Js ^jWith off period Dl _jDetermine the term of validity scope of operation;

Step 2.4.1.1: variable is set The overall supplies of representing current seller's set to provide in the operation term of validity scope of target j correspondence, the seller gathers PS _jSeller's set that representative is chosen for this operation of optimization process, seller's target number that on behalf of inspected, n cross; Will Initial value be made as 0, the initial value of n is made as 1, the set PS _jBe set to sky;

Step 2.4.1.2: from seller's formation, duplicate the information of n seller's target, relatively highest-capacity piece bid Hsp _jWith the piece reserve price Rsp in current seller's target _iIf, Rsp _i≤ Hsp _jAnd the available ability piece of this seller i number in the operation term of validity

Then

Seller i is joined set PS _jIn;

Step 2.4.1.3: if

Or n forwards rapid 2.4.1.4 to during greater than the target number in seller's formation; Otherwise n=n+1 forwards step 2.4.1.2 to;

Step 2.4.1.4: if

Explanation can provide abundant ability to satisfy this demand from supplier's set; Otherwise because the surplus capacity deficiency, this demand of auction platform notice buyer j can't be satisfied;

When the resource of transaction is storage capacity, among the step 2.4.1.2

What represent is the sustainable ability piece number that provide of seller i in the required time section, and other step is same as described above;

Step 2.4.2: the mathematic(al) representation of problem correspondence:

When transaction be processing power the time, suppose to have in the supplier set n _OriThe individual seller, they are respectively at the ability set of blocks that buyer j provides

The ability set of blocks

Deadline be The actual finish time time of this operation _jIt is the concluding time of the ability piece that begins the latest in these set, that is:

${\mathrm{time}}_j=\underset{0<i<n_{\mathrm{ori}}}{\max }\left(\mathrm{endTime}\left({\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\right)\right)---\left(3.17\right)$

The cost amount of money amount that buyer j fulfils assignment _jAvailable following formula calculates:

${\mathrm{amount}}_j=\underset{i=1}{\overset{n_{\mathrm{ori}}}{\mathrm{\&Sigma;}}}({\mathrm{Sp}}_{\mathrm{ij}}\&times;|{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\left|\right)---\left(3.18\right)$

Use formula (3.19), can calculate the satisfaction of buyer j

$S_{\mathrm{bid}}^j={\mathrm{\&omega;}}_{\mathrm{bp}}\&times;S_{\mathrm{bp}}^j\left({\mathrm{amount}}_j\right)+{\mathrm{\&omega;}}_t\&times;S_t^j\left({\mathrm{time}}_j\right)---\left(3.19\right)$

Use formula (3.20), can calculate the satisfaction of seller i

$S_{\mathrm{sp}}^i=S_{\mathrm{sp}}^i\left({\mathrm{Sp}}_{\mathrm{ij}}\right)---\left(3.20\right)$

At this moment, problem is converted into:

Maximize：S _total

$S_{\mathrm{total}}=\frac{1}{(1+\mathrm{\&omega;})}{S}_{\mathrm{bid}}^j+\frac{\mathrm{\&omega;}}{(1+\mathrm{\&omega;})}\frac{\underset{1\&le;i\&le;n_{\mathrm{ori}}}{\mathrm{\&Sigma;}}\left(\right|{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}|\&times;S_{\mathrm{sp}}^i)}{L_j}---\left(3.21\right)$

$s.t.{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\&SubsetEqual;P_{\mathrm{left}}^i$

$\left|\underset{1\&le;j\&le;n_{\mathrm{bid}}}{\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">Y</figure-callout>}}{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\right|=L_j$

Wherein

Represent the ability set of blocks of seller i current residual;

When transaction be storage capacity the time, use formula (3.22) can calculate the satisfaction of buyer j

Remaining is the same;

$S_{\mathrm{bid}}^j=S_{\mathrm{bp}}^j\left({\mathrm{amount}}_j\right)---\left(3.22\right)$

Step 2.4.3: use this problem of genetic algorithm for solving:

(a) chromosome coding

Use the one-dimension array structure, each chromosome representative is corresponding to a kind of scheduling scheme of buyer j resource requirement; Gene figure place in the chromosome is corresponding to above n _Ori, each gene in the array is corresponding to each seller in the supplier's set that obtains when determining the hunting zone, i position genic value

The ability piece number that i the seller of representative in supplier set provides for buyer j: when conclude the business be processing power the time, the ability set of blocks

Relevant with homework type; When the transaction be storage capacity the time, the ability set of blocks

Be in the time period of buyer's demand scope

The supply of representative; According to transaction content (computational resource, storage resources), homework type (The faster the better, best, good more more slowly on time) and the ability piece reserved, the decision seller distributes to the ability set of blocks of this operation;

(b) initialization population

After setting population scale, produce each chromosome successively; When determining that chromosomal genic value is big or small, need to guarantee genic value $\left|{\mathrm{PS}}_{\mathrm{use}}^i\right|\&le;P_{\mathrm{ori}}^{\mathrm{ij}},$ And $\underset{1\&le;i\&le;n_{\mathrm{ori}}}{\mathrm{\&Sigma;}}\left|{\mathrm{PS}}_{\mathrm{use}}^i\right|=L_j;$

(c) suitable value function

The total satisfaction function of both sides of selecting formula (3.21) representative for use is as chromosomal suitable value function;

(d) select, intersect and variation

Selection operation uses the roulette back-and-forth method in the traditional genetic algorithm;

Interlace operation is used evenly and is intersected, and produces n _OriIndividual 0-1 integer wherein 0 is represented corresponding gene position exchange, and on behalf of corresponding gene position, 1 do not exchange; Mutation operation is the gene position respective value to be adjusted into 0 at random arrive

Between a positive integer value;

Because intersection and mutation operation may cause

This moment is with workload L _jProportional distribution according to existing gene place value is given each seller; If adjusting the back occurs

Then the value of this gene position correspondence is set to

And

Workload provide by other seller in seller's formation;

Use different crossover probabilities and variation probability as the evolutionary operator probability of algorithm at different times; In early stage, algorithm is set to bigger crossover probability and less variation probability; In the later stage, algorithm is set to less crossover probability and bigger variation probability;

Step 2.4.4: reserved resource

After finishing, algorithm chooses optimum chromosome separating as this problem, respective markers is reserved and made to each genic value according to this chromosome correspondence to the resource capability in seller's target, according to time sequencing joins in job scheduling table at the ability piece that day part provides each seller in reservation procedure; Fulfil assignment behind the dispatch list, write down this transaction and transaction record and job scheduling table are issued the buyer;

Step 2.5: the batch matching algorithm when supply exceed demand

After auction platform triggered the batch matching algorithm, auction platform can be in each the inspection till processing target as much as possible all targets in handling the formation of buyer's target; Suppose that the buyer's target quantity in the formation of current buyer's target is n _Bid, algorithm flow is as follows:

Step 2.5.1: determine that the seller in the matching algorithm gathers PS

With this n _BidIndividual buyer's target is taken out buyer's target successively according to priority queueing; What suppose current taking-up is the target of buyer j, at first calculates to guarantee that matching algorithm necessarily has the minimum supply of separating

With the required supply of the current target of optimization process

When transaction is that step is as follows when handling resource:

Step 2.5.1.1: be provided with

With

Initial value, make

Step 2.5.1.2: check the buyer's target in the formation to be matched successively, check whether the operation term of validity of current target k correspondence in the formation and the operation term of validity of target j have common factor; When two operation terms of validity have common factor, $D_{\max }^j=D_{\max }^j+L_k,$ $D_{\min }^j=D_{\min }^j+L_k;$

When the transaction be storage capacity the time, among the step 2.5.1.2

With

Calculating need check its place time period, when the time section is positioned at the common factor of two operation terms of validity,

When the time section is positioned at other period,

With Value remain unchanged;

Obtain

With

Value after, when transaction be processing power the time, can the step below using be judged and target j is joined in the formation to be matched:

Step 2.5.1.2.1: variable is set

The overall supplies of representing current seller's set to provide in the operation term of validity scope of target j correspondence, the seller gathers PS _jSeller's set that representative is chosen for this operation of optimization process, seller's target number that on behalf of inspected, n cross; Will

Initial value be made as 0, the initial value of n is made as 1, set PS _jBe set to sky;

Step 2.5.1.2.2: from seller's formation, duplicate the information of n seller's target, relatively the ability piece best bid Hsp of target j _jWith the piece reserve price Rsp among current seller's target i _iIf Rsp _i≤ Hsp _jAnd the available ability piece of this seller i number in the operation term of validity Then

Seller i is joined set PS _jIn;

Step 2.5.1.2.3: if

Or n forwards step 2.5.1.2.4 to during greater than the target number in seller's formation; Otherwise n=n+1 forwards step 2.5.1.2.2 to;

Step 2.5.1.2.4: if

Target j is joined in the formation to be matched, the seller is gathered PS _jMerging to original seller gathers among the PS; Otherwise, if Target j joined wait to look in the formation; If

The supply deficiency of the then current seller in the operation term of validity of target j correspondence can't satisfy this demand;

When the transaction be storage capacity the time, among the step 2.5.1.2.2

Calculating need check

The time period distribution situation; In like manner, in the comparison operation of step 2.5.1.2.3, have only

In all time periods all more than or equal to

The time, just think $D_{\mathrm{now}}^j\&GreaterEqual;D_{\max }^j;$

The mathematic(al) representation of step 2.5.2 problem correspondence

When the transaction be processing power the time, suppose that the buyer's target number in the formation to be matched is n _Bid, the current seller gathers PS n _AskThe individual seller, n _AskThe individual seller is respectively for the actual ability set of blocks that provides of demand that satisfies buyer j

The ability set of blocks Deadline be

The deadline time of this operation _jIt is the concluding time of the ability piece that begins the latest in these set, that is:

${\mathrm{time}}_j=\underset{\&ForAll;0<i<n_{\mathrm{aski}}}{\max }\left(\mathrm{endTime}\left({\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\right)\right)---\left(3.23\right)$

The cost amount of money amount that buyer j fulfils assignment _jAvailable following formula calculates:

${\mathrm{amount}}_j=\underset{i=1}{\overset{n_{\mathrm{ask}}}{\mathrm{\&Sigma;}}}({\mathrm{Sp}}_{\mathrm{ij}}\&times;|{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\left|\right)---\left(3.24\right)$

Use formula (3.19), can calculate the satisfaction of buyer j

Use formula (3.20), can calculate the satisfaction of seller i

At this moment, problem is converted into:

Maximize：S _total

$S_{\mathrm{total}}=\frac{\underset{1\&le;j\&le;n_{\mathrm{bid}}}{\mathrm{\&Sigma;}}{S}_j^{\&prime;}}{n_{\mathrm{bid}}}---\left(3.25\right)$

$S_j^{\&prime;}=\frac{1}{(1+\mathrm{\&omega;})}{S}_{\mathrm{bid}}^j+\frac{\mathrm{\&omega;}}{(1+\mathrm{\&omega;})}\frac{\underset{1\&le;i\&le;n_{\mathrm{ask}}}{\mathrm{\&Sigma;}}\left(\right|{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}|\&times;S_{\mathrm{sp}}^i)}{\underset{1\&le;j\&le;n_{\mathrm{bid}}}{\mathrm{\&Sigma;}}{L}_j}---\left(3.26\right)$

$s.t.{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\&SubsetEqual;P_{\mathrm{left}}^i$

$\left|\underset{1\&le;j\&le;n_{\mathrm{bid}}}{\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">Y</figure-callout>}}{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\right|=L_j$

Wherein

Represent the ability set of blocks of seller i current residual;

When transaction be storage capacity the time, use formula (3.22) to calculate the satisfaction of buyer j

Remainder formula is the same;

Step 2.5.3 uses this problem of genetic algorithm for solving

(a) chromosome coding

Use the two-dimensional matrix structure, line number is buyer's target number of this time coupling, and columns is to gather the element number of determining among the PS the seller; Each row of matrix is represented the scheduling scheme of buyer's demand, wherein the element of the capable j row of i

The ability set of blocks of representing seller j to provide for buyer i, the particular location of ability set of blocks in original seller's ability description are to use the strategy identical with step 2.4.3 to determine after the ability that i-1 is capable before reserving;

(b) initialization population

After setting population scale, produce each chromosome successively; When producing chromosome, determine the element value of each row in the chromosome successively according to the order of buyer's target; When determining each row element value, need make a mark avoiding the ability piece to be taken once more to relevant ability piece, and guarantee by the operation of back

(c) suitable value function

The total satisfaction function of both sides of selecting formula (3.25) representative for use is as chromosomal suitable value function;

(d) select, intersect and variation

Selection operation uses the roulette back-and-forth method in the traditional genetic algorithm; Interlace operation is used evenly and is intersected, and produces n _BidIndividual 0-1 integer, the wherein row element collective exchange of the corresponding operation of 1 representative, 0 representative does not exchange; Mutation operation is that the delegation's element value with the operation correspondence carries out random initializtion again;

Because interlace operation may cause

The maximum capacity piece number of representing seller i under present case, to provide for buyer j; Corresponding method of adjustment is that the value of this gene position correspondence is set to

And

Workload provide according to priority by other seller in seller's formation; Use different crossover probabilities and variation probability as the evolutionary operator probability of algorithm at different times; In early stage, algorithm is set to bigger crossover probability and less variation probability; In the later stage, algorithm is set to less crossover probability and bigger variation probability;

Step 2.5.4: reserved resource

After finishing, algorithm chooses optimum chromosome separating as this problem, respective markers is reserved and made to each row element value according to this chromosome correspondence to the resource capability in seller's target successively, in reservation procedure each seller joined according to time sequencing in the corresponding job scheduling table at the ability piece that day part provides; Fulfil assignment behind the dispatch list, write down each transaction and transaction record and job scheduling table are issued the corresponding buyer;

Step 2.6: the matching algorithm during supply-less-than-demand

Under the situation of supply-less-than-demand, because the ability short supply of whole period can't be satisfied each request that all buyeies submit to; So in order to satisfy buyer's demand as much as possible, can algorithm not be considered the satisfaction information of the buyer in asking, only check in correlation time satisfy the demands in the section; After auction platform is received buyer's target, trigger matching algorithm immediately;

Suppose that current that receive is buyer's target j, variable is set

The overall supplies of representing current seller's set to provide in the required time scope of target j, the seller gathers PS _jSeller's set that representative is chosen in order to satisfy this demand, seller's target number that on behalf of inspected, n cross, algorithm flow is as follows;

Step 2.6.1: will

Initial value be made as 0, the initial value of n is made as 1, set PS _jBe set to sky;

Step 2.6.2: from seller's formation, duplicate the information of n seller's target, relatively highest-capacity piece bid Hsp _jPiece reserve price Rsp with current seller i _iIf Rsp _i≤ Hsp _jAnd the available supply in the operation term of validity

Then

Seller i is joined set PS _jIn;

Step 2.6.3: relatively

And L _jIf,

The demand actual supply that provide of seller i then for satisfying target j

If The actual supply that provides of seller i then

Step 2.6.4: if

Or n forwards step 2.6.5 to when equaling target number in seller's formation; Otherwise n=n+1 forwards step 2.6.2 to;

Step 2.6.5: if

According to

Value respective markers is reserved and made to the resource capability in seller's target in reservation procedure, according to time sequencing join in job scheduling table at the ability piece that day part provides each seller; Fulfil assignment behind the dispatch list, write down this transaction and transaction record and job scheduling table are issued the buyer; If

Then the supply deficiency of all sellers in the required time section of target j correspondence can't satisfy this demand;

Step 2.7: each transaction clearing back auction platform writes down this transaction and issues the buyer and show scheduling time;

Step 3, the waiting period: finish the periodicity supply-demand information record in auction cycle, the concluding time by making ability after waiting for a period of time and the zero-time in next auction cycle are synchronous;

Step 3.1: count average knockdown price Sap of piece and reasonable demand amount Rdm according to the interim transaction record of auction;

Calculate the mean value of the knockdown price of All Activity in this auction cycle, promptly the merchant of total business volume and total amount of transactions is the average knockdown price Sap of this period blocks;

Calculate the demand and do not strike a bargain but the piece highest quotation is higher than the demand sum that equals average knockdown price Sap in this cycle is Rdm of striking a bargain of all buyeies in this auction cycle.

The beneficial effect of the grid resource scheduling method based on continuous two way auction mechanism of the present invention: with the pricing method of auction technique as the gridding resource ability, and carry out the configuration of gridding resource with this, can reflect market information well and give full play to users slack resources is joined grid, improve the scheduling of resource efficient under the grid environment; Improve original continuous two way auction mechanism, made its scheduling that is fit to gridding resource that block ability is divided, and designed gridding resource supply exceed demand state matching process down, better be fit to the short-term grid resource scheduling situation of reality.

Description of drawings

Fig. 1 is the synoptic diagram that the bulk of ability is divided;

Fig. 2 is the synoptic diagram of the application background of auction;

Fig. 3 is the synoptic diagram of large-scale user's ability need fluctuation;

Fig. 4 is the synoptic diagram of the corresponding relation of auction duration in the continuous two way auction and ability time range;

Fig. 5 is the synoptic diagram of continuous two way auction;

Fig. 6 is the synoptic diagram of the worth curve in the prospect theory;

Fig. 7 is the synoptic diagram of the buyer's transaction utility curve;

Fig. 8 is the synoptic diagram of the consumption utility curve of the buyer when The faster the better for homework type;

The synoptic diagram of three types the buyer's consumption utility curve when Fig. 9 is good more more slowly for homework type;

Figure 10 is the synoptic diagram of three types the buyer's consumption utility curve preferably time the on time for homework type;

Figure 11 is three types the seller's the synoptic diagram of satisfaction curve;

Figure 12 is the synoptic diagram of utility curve algorithm flow;

Figure 13 consumes two kinds of situation synoptic diagram that may occur in the satisfaction process for the match difficulty satisfies the type buyer;

Figure 14 is the synoptic diagram of the ability piece reservation mode of three kinds of homework type correspondences;

Figure 15 is the synoptic diagram of batch matching algorithm flow process.

Embodiment

Below in conjunction with accompanying drawing a kind of grid resource scheduling method based on continuous two way auction mechanism of the present invention is done and to be described in further detail.

One, based on the relevant rudimentary theory of the grid resource scheduling of auction mechanism

1 auction technique

Auction need be by carrying out with the corresponding auction technique of goods to auction.Auction technique is meant the concrete operation regulation when auction facility is carried out auction activity, and this operation regulation should fully take into account the kind and the characteristics of goods to auction, by selecting or revise existing auction technique targetedly after auction items is determined.

Along with the theoretical development of auction, the participation quantity of both parties in auction both may be the relation of one-to-many, also may be the relation of multi-to-multi, auction can be divided into unidirectional auction and two way auction again according to the difference of this participation quantity of both sides.

1.1 unidirectional auction technique

According to different rules, auction technique can have different types of dividing mode, as Wiggans unidirectional auction technique is divided into three kinds: sealing auction, advance price auction and Dutch Auction; Whether Rothkopf etc. then openly can be divided into public auction and sealing auction to auction to everyone according to buyer's target information in the auction process.At present, the basic auction technique of Gong Rening has following four kinds: advance price auction, Dutch auction, sealing first price auction and seal second price auction.Wherein, divide according to the standard of Rothkopf etc., preceding two kinds of basic auction techniques belong to public auction, and back two kinds of basic auction techniques belong to the sealing auction.

(1) advance price auction

The advance price auction also claims English auction, is a kind of auction technique the most used in the actual life.In English auction, a plurality of bidder can be arranged, each bidder can offer to goods to auction, and price is soaring gradually in auction process, and auction lasts till do not have new bid price to be higher than till the current bid price always in a period of time.

(2) Dutch auction

Dutch auction is driving in the wrong direction of advance price auction, i.e. Dutch auction use is auctioned the knockdown price that opposite mode is determined goods to auction with advance price, because of its auction that is applied to Dutch fresh flower the earliest well-known, so also claim Dutch Auction.

(3) sealing auction

Sealing auction is meant that the seller after goods to auction is set reserve price seals it, then by the appraisal of auction side declaration auction items, indicates the time, place, goods to auction quantity, conditions of auction of auction bid etc., gives the suitor by auction side.The suitor giving auction side in the quotation of oneself, does not know how many quotations separately is with the mode that seals letter between the suitor.Examination one by one and is mutually relatively received after all suitors' the quotation by auction side, and the suitor who meets seller's condition most is the winning bidder, obtains the article of auctioning simultaneously.When if the price of winning bidder's payment is his bid price, then this auction is called as sealing first price auction; When if the winning bidder only need pay highest bid except that him, be called sealing second price auction.

More than the essential structure of four kinds of auction techniques be identical, both parties all are the market structures of " one-to-many ".In this structure, if the buyer has only one, then the seller is with regard to more than one; If the seller has only one, then the buyer is with regard to more than one.Both parties always have a side grasping scarce resource in the market, have the advantage of resource, and this advantage makes this side have the option dealing mode and formulates the right of trading rules.

1.2 two way auction mode

Different with unidirectional auction, participate in all more than family of both parties of two way auction, the sellers and the buyeies submit bid to, and (the target price that the seller submits to is called " charge ", the target price that the buyer carries is called " asked price "), the buyer is according to the tender price formation that causes the demand of sorting from high to low, and the seller sorts from low to high according to tender price and produces the supply formation; By the charge (lowest price from seller's formation backward move) of the coupling seller from two formations and the buyer's asked price (ceiling price from seller's formation is mobile backward) but determine the maximum number of transaction; Determine equilibrium price and knockdown price separately according to certain market clearing rule then.

The trading time period of two way auction, any buyer can declare publicly the article of being willing to be intended to buy on which kind of price how many units of quantity; Any seller also can declare publicly the article of being willing to be intended to sell on which kind of price how many units.In case the buyer's asked price is accepted (or the seller's charge is accepted by the buyer) by the seller, just have article and strike a bargain; If the buyer's asked price is not accepted (or the seller's charge is not accepted by the buyer) by the seller, the buyer can improve its asked price gradually, and the seller also can reduce its charge gradually, till a side is accepted by the opposing party; This process lasted till till the closing the transaction time that no longer includes transaction generation or the prior regulation of arrival in a period of time always.

2 genetic algorithms

2.1 key concept and thought

Genetic algorithm copies the production process of occurring in nature life and biological evolutionary process to produce.Each is separated corresponding to an individuality with certain features in the biotic population in the solution space of problem correspondence, and each individuality has corresponding with it chromosome, and Different Individual presents different characteristics and embodied by different chromosome.In by " physical environment " that problem determined, different characteristics are corresponding to the adaptedness different to environment, individual formed population procreation from generation to generation in this " physical environment ", survive, and run to up to algorithm till the termination condition of regulation.

In algorithm, chromosome is by genomic constitution, and the coloured differently body structure is corresponding to different coding forms.The coding here refers to and uses specific data structure and corresponding numerical value to come separating of problem of representation.Individually represent by the pairing chromosomal fitness size of individuality and weigh for the adaptedness of environment, and fitness to be the fitness function that is determined by problem and coded system determined.

In this microprocess of generation of life, a new life's individuality may be that to inherit fully also may be to produce after having changed the part or all of characteristics of first generation individuality.Therefore a newborn individual adaptedness for " physical environment " depends on the individual quality of generation earlier that produces it to a certain extent, if but this does not represent elder generation's generation individuality bad for the adaptedness of environment, the adaptedness of the son individuality of their generations is also necessarily bad so.This is because formerly in individual all characteristics that had, have a part of characteristics and help the ideal adaptation environment, but owing to individual adaptedness for environment is that all characteristics that possessed by individuality are determined, so formerly may be covered by other unfavorable characteristics, make individual undesirable for the whole adaptedness of " physical environment " for the favourable characteristics in the individuality.Then can and integrate these favourable characteristics as the sub-individuality that these elder generation's generation individualities are produced, thereby have than the whole characteristics that have more advantage earlier for individuality with certain probability succession.

In the algorithm with it respective operations be THE REPLICATION OF CHROMOSOME, intersection and variation.Duplicate and refer to the individual elder generation of inheriting fully of son for all individual characteristics, it is identical for individual chromosome with an elder generation that produces it to show as sub individual homologue.Thereby intersect refer to according to generation earlier individual the part characteristics reconfigure and produce a new individuality, this be presented as chromosomal counter structure according to certain rule earlier generation chromosomal portion gene reconfigure and form a new chromosome.Variation refers to through duplicating or intersecting and after producing new sub-individuality, thereby the characteristics that change in the sub-individuality with certain small probability form new characteristics, and this is presented as according to certain rule and the probability existing genic value that son is individual and changes over new genic value.

Individual to produce son individual by first generation, earlier generation individual place initial population the newborn population at the individual place of quilt is substituted.In this course, owing to have better adaptedness than other individuality with population for environment for the part in the population is individual earlier, these individualities just may not eliminated rapidly by " physical environment " institute, thereby have had right and the chance that produces offspring.This is representing for the adaptedness of environment strong more, then should produce offspring with regard to being chosen by other individuality more by individuality.This design of algorithm has reflected the thought of " survival of the fittest in natural selection " in the biological evolution process.By the introducing of this thought, thereby the deviser of algorithm wishes that the sub-population that produces can more approach the optimum individual that " physical environment " can produce than having the characteristics that better adaptation characteristics embody the part individuality of population for population for environment earlier.

Corresponding with it in algorithm is selection operation, selection refers to selecting the operation of the individuality of a certain quantity in certain probability population of before generation, and each individual selected probability size is determined by the whole adaptedness of all individualities in the adaptedness of the corresponding environment of this individuality and the population to environment.

2.2 algorithm flow

The individual pairing chromosome of in the population each is represented by a related data structure, solution space dimension corresponding when the dimension of data structure is found the solution problem with algorithm designer is identical, and each chromosome all can be calculated own corresponding adaptive value size according to the pairing fitness function of problem.Generally, algorithm needs to preserve the pairing chromosome congression of previous generation population in the process of implementation, and produces new population with this, carries out this process repeatedly and write down more excellent the separating that algorithm obtains as algorithm in the current optimized individual that obtains when finishing.

The process of genetic algorithm is as follows:

2.2.1: determined to produce initial population at random behind the population scale, each individuality is expressed as the pairing gene code value of chromosome;

2.2.2: calculate each individual fitness in the population, judge whether to satisfy the termination condition of algorithm,, then separating as the net result of algorithm of current optimized individual representative exported, and finish algorithm flow if satisfy; Otherwise continue to carry out 2.2.3;

2.2.3: the individuality of in current population, selecting requirement according to fitness;

2.2.4: crossover probability according to the rules and cross method produce new individual;

2.2.5: according to the rules variation probability and variation method, individuality is made amendment;

2.2.6: after producing the population of individuality as a new generation of specified quantity, return top 2.2.2.

Termination condition in the genetic algorithm, can be according to the different choice of problem different modes.For example, can select the termination condition of one of following condition for use as algorithm:

(1) fitness of current optimized individual has surpassed preset value;

(2) individual average fitness has surpassed preset value in the current population;

(3) proceed to current genetic algebra and surpassed preset value.

Two, the grid resource scheduling method based on continuous two way auction mechanism of the present invention

As shown in Figure 4, the grid resource scheduling method based on continuous two way auction mechanism of the present invention, by inch of candle mechanism is carried out grid resource scheduling, changes continuous two way auction into cycle system, and the adjacent auction cycle joins end to end; But the total duration in an auction cycle equates with the time range of its trading capacity and is corresponding one by one; Ability in the corresponding with it time range of in this auction cycle, can only concluding the business, and the ability in scope sometime can only be concluded the business in the auction cycle in correspondence;

As shown in Figure 5, an auction cycle is divided into three phases, be respectively preparatory stage, auction phase and the waiting period.

Should be based on the grid resource scheduling method of continuous two way auction mechanism, it is characterized in that: by inch of candle mechanism is carried out grid resource scheduling, changes continuous two way auction into cycle system, and the adjacent auction cycle joins end to end; But the total duration in an auction cycle equates with the time range of its trading capacity and is corresponding one by one; Ability in the corresponding with it time range of in this auction cycle, can only concluding the business, and the ability in scope sometime can only be concluded the business in the auction cycle in correspondence; Each auction cycle comprises the steps:

Step 1, preparatory stage: finish the trust stage in the traditional auction, be responsible for receiving seller's target and add up and predict supply-demand information as required;

Step 1.1: receive seller's target;

Step 1.2: upgrade each user's integration information, integration is represented the ability piece number that this user has struck a bargain;

Step 1.3: prediction dealing side supplydemand relationship:, predict reasonable knockdown price Rsap and reasonable demand amount Ardm in this auction cycle according to average knockdown price Sap of piece in the periodicity transaction record in the past and reasonable demand amount Rdm;

(a) linear prediction

(b) model prediction

Step 1.4: calculate supply and demand scale parameter ω, computing formula is: ω=Rsup/Ardm (3.3)

When ω＞1, think that overall supplies that the seller provides can satisfy most of buyer's demand, and can be in the operation term of validity overall supplies be to select optimal that part of ability piece to improve the buyer's consumption satisfaction the ability set of blocks of ω times of workload;

When ω≤1, think that the overall supplies that the seller provides may be less than demand, this moment, auction platform just merely checked whether can fulfil assignment and do not consider both sides' satisfaction in the operation term of validity;

Step 2, auction phase: auction platform receives buyer's target, finishes coupling and accounts settling phase according to demand; The seller can revise the reservation price of the current ability of sale at any time or supply more ability in this stage; Each transaction clearing back auction platform writes down this transaction and issues the buyer and show scheduling time;

Step 2.1: when ω＞1, think that overall supplies that the seller provides can satisfy most of buyer's demand; When ω≤1, think that the overall supplies that the seller provides may be less than demand;

Step 2.2: auction platform is set variable:

The capacity of ability piece (Block Capability): Bcp;

The time span of each ability piece (Block Time Span): T _Sp

But the maximum time span of trading capacity time range: N _Ms* T _SpBut the time range of then auctioning trading ability piece in the phase is [T _Ast, T _Ast+ N _Ms* T _Sp], T _AstRepresent the start time of the piece of ability the earliest that can conclude the business in this auction phase;

Supply and demand ratio lower limit: ω _MinWhen the supply and demand ratio in an auction cycle is not higher than this preset value, the knockdown price of ability piece is set at the buyer's piece best bid;

The supply and demand ratio upper limit: ω _MaxWhen the supply and demand ratio in an auction cycle is not less than this preset value, the knockdown price of ability piece is set at the seller's piece reservation price;

Step 2.3:, mate if use to mate in batches then change step 2.5 if ω＞1 is changeed step (2.4) if auction platform uses real-time coupling and mated; Otherwise supply-less-than-demand uses step 2.6 to mate;

Step 2.4: the Real Time Matching Algorithm when supply exceed demand;

Auction platform triggers matching algorithm after receiving the target of buyer's transmission immediately;

When the transaction processing ability, for the target request that the buyer proposes, taking-up residue overall supplies in buyer's operation term of validity is not less than the seller of ω times of workload as seller's set of satisfying this demand from seller's formation;

When the transaction storage capacity, taking-up residue overall supplies in the required time section is not less than the seller of ω times of demand as seller's set of satisfying this demand from seller's formation;

According to the satisfaction function that statistics draws, the workload in buyer's demand is suitably distributed to each seller according to the principle of total satisfaction maximum;

Step 2.4.1: what the seller gathered determines:

Step 2.4.2: the mathematic(al) representation of problem correspondence:

Step 2.4.3: use this problem of genetic algorithm for solving:

(a) chromosome coding

Use the one-dimension array structure, each chromosome representative is corresponding to a kind of scheduling scheme of buyer j resource requirement;

(b) initialization population

(c) suitable value function

(d) select, intersect and variation

Step 2.4.3: reserved resource

After finishing, algorithm chooses optimum chromosome separating as this problem, respective markers is reserved and made to each genic value according to this chromosome correspondence to the resource capability in seller's target, according to time sequencing joins in job scheduling table at the ability piece that day part provides each seller in reservation procedure; Fulfil assignment behind the dispatch list, write down this transaction and transaction record and job scheduling table are issued the buyer;

Step 2.5: the batch matching algorithm when supply exceed demand

Step 2.5.1: determine that the seller in the matching algorithm gathers PS

Step 2.5.2: the mathematic(al) representation of problem correspondence

Step 2.5.3: use this problem of genetic algorithm for solving

(a) chromosome coding

Use the two-dimensional matrix structure, line number is buyer's target number of this time coupling, and columns is to gather the element number of determining among the PS the seller;

(b) initialization population

(c) suitable value function

(d) select, intersect and variation

Step 2.5.4: reserved resource

After finishing, algorithm chooses optimum chromosome separating as this problem, respective markers is reserved and made to each row element value according to this chromosome correspondence to the resource capability in seller's target successively, in reservation procedure each seller joined according to time sequencing in the corresponding job scheduling table at the ability piece that day part provides; Fulfil assignment behind the dispatch list, write down each transaction and transaction record and job scheduling table are issued the corresponding buyer;

Step 2.6: the matching algorithm during supply-less-than-demand

Under the situation of supply-less-than-demand, because the ability short supply of whole period can't be satisfied each request that all buyeies submit to; So in order to satisfy buyer's demand as much as possible, can algorithm not be considered the satisfaction information of the buyer in asking, only check in correlation time satisfy the demands in the section; After auction platform is received buyer's target, trigger matching algorithm immediately;

Step 2.7: each transaction clearing back auction platform writes down this transaction and issues the buyer and show scheduling time;

Step 3, the waiting period: finish the periodicity supply-demand information record in auction cycle, the concluding time by making ability after waiting for a period of time and the zero-time in next auction cycle are synchronous;

Step 3.1: count average knockdown price Sap of piece and reasonable demand amount Rdm according to the interim transaction record of auction;

Calculate the mean value of the knockdown price of All Activity in this auction cycle, promptly the merchant of total business volume and total amount of transactions is the average knockdown price Sap of this period blocks;

Calculate the demand and do not strike a bargain but the piece highest quotation is higher than the demand sum that equals average knockdown price Sap in this cycle is Rdm of striking a bargain of all buyeies in this auction cycle.

Grid resource scheduling method of the present invention is mainly used in the grid resource scheduling platform, by the demand of gridding resource both parties being obtained and being coordinated, distribute by carry out resource supply and demand based on the mode of continuous two way auction mechanism, finally form the scheduling of resource timetable for grid resource scheduling platform actual schedule.

Grid resource scheduling method of the present invention has also been considered the content of following several respects in specific implementation process:

The selection of 1 matching algorithm

Auction platform needs to predict according to historical transactional information in the past the supply and demand ratio in this auction cycle after the preparatory stage finishes.According to this ratio, the auction platform decision is at the interim target matching algorithm of selecting for use of auction.

When the supply and demand ratio shows this auction cycle appearance situation that supply exceed demand, the satisfaction information that auction platform can propose in target with respect to the buyer as much as possible, thus use the matching algorithm of the information of considering both sides' satisfaction to choose the resource capability that those meet buyer's demand most.When the supply and demand ratio showed that the situation of supply-less-than-demand appears in this auction cycle, it was purpose fully to sell out the seller in the ability of day part supply only that shortage of resources can cause auction platform.Continue to use the matching algorithm of considering satisfaction information as auction platform this moment, may cause the minimizing of total volume, so auction platform adopts the algorithm of not considering satisfaction information.

The trigger condition of 2 matching algorithms

The buyer can send target to auction platform in any time in the auction phase, and the trigger condition of auction platform matching algorithm may be real-time triggering (triggering immediately) or non real-time triggering.

For relatively two kinds of algorithm triggers modes are for the difference of scheduling of resource under supply exceed demand situation, the present invention uses this dual mode as the trigger condition of matching algorithm and its experimental result relatively under same case respectively in the simulated auction operational process.Wherein the non real-time trigger condition is set to after a target arrives auction platform, and target quantity and preset value in the formation of current buyer's target are compared, if the target number in the formation reaches preset value, then triggers coupling immediately; Otherwise the setting timer is when timer mates the residue target in the target formation to after date.

When supply-less-than-demand, what auction platform used is the matching algorithm of not considering both sides' satisfaction, and the trigger condition of the present invention's algorithm is set to real-time triggering.

3 auction platforms need the variable setting, add up and predict

The variable that auction platform need be set has:

The capacity of ability piece (Block Capability): Bcp.

The time span of each ability piece (Block Time Span): T _Sp

But the maximum time span of trading capacity time range: N _Ms* T _SpBut the time range of then auctioning trading ability piece in the phase is [T _Ast, T _Ast+ N _Ms* T _Sp], T _AstRepresent the start time of the piece of ability the earliest that can conclude the business in this auction phase.

Supply and demand ratio lower limit: ω _MinWhen the supply and demand ratio in an auction cycle is not higher than this preset value, the knockdown price of ability piece is set at the buyer's piece best bid.

The supply and demand ratio upper limit: ω _MaxWhen the supply and demand ratio in an auction cycle is not less than this preset value, the knockdown price of ability piece is set at the seller's piece reservation price.

Auction platform needs to add up the relevant information in the last auction cycle in the preparatory stage, need the variable of record to have:

The average knockdown price of piece (Share Average Price): Sap.Sap represents the average knockdown price of piece in the last auction cycle, uses the total business volume of this time period and the merchant of total amount of transactions to represent.

Reasonable demand amount (Reasonable Demand): Rdm.Rdm represents the demand and do not strike a bargain but the piece highest quotation is higher than the demand sum that equals average knockdown price Sap in this cycle of striking a bargain of all buyeies in the last auction cycle.

Two variablees below auction platform can be estimated according to the supply in the seller's target that receives and reserve price and periodicity transaction record in the past after the preparatory stage finishes:

Reasonable supply (Reasonable Supple) in this auction cycle: Rsup.On behalf of all sellers' that receive in the preparatory stage reservation price, Rsup be not higher than the supply sum of the reasonable knockdown price Rsap that estimates.

The reasonable demand amount of estimating in this auction cycle (Anticipated Reasonable Demand): Ardm.Issuable reasonable demand amount in that Ardm representative is estimated according to historical information, this auction phase, wherein the reasonable demand amount refers to the rational demand of quotation in all demands.

The definition of 4 both sides' satisfactions and tolerance

In the step 2.1, when this auction cycle appearance situation that supply exceed demand, auction platform uses the matching algorithm of the information of considering both sides' satisfaction to choose the resource capability that those meet buyer's demand most.Therefore, the present invention needs to define the satisfaction of both parties and how to measure in specific implementation.Next, the satisfaction notion that uses among the present invention is described and measures:

In the two way auction of gridding resource ability, the buyer understood the service level that forethought can provide to ability before submitting target to, and the value of required ability is estimated that this assessed value can be defined within limits usually.When supply was sufficient, platform can select the seller of specific quantity to satisfy this demand according to the supply and demand situation and the trading rules of reality after receiving buyer's target.

When transaction during storage capacity, may spend the money of varying number with the different seller transaction buyeies.When the transaction processing ability, with different seller transaction, the buyer may spend the money of varying number, and also difference to some extent of the deadline of operation.In addition, the different sellers are for different ability piece knockdown prices, and its satisfaction also can be different.In order to portray the satisfaction difference of different user, the use of the present invention satisfaction curve corresponding with the user portrayed satisfaction.

4.1 satisfaction portrayal

According to the description in the Customer Praxiology, the consumption effectiveness that the total utility that the user obtains in whole process of consumption produces during by the transaction effectiveness of the generation in the purchasing process and consumer lines or service is formed jointly.For grid user, the deviation between the psychology expection before amount of money of paying during purchasing power and the transaction has formed transaction effectiveness, and the service quality level that obtains after the use ability is then corresponding to consumption effectiveness.

For the better satisfaction of portrayal both sides in process of exchange, when considering the satisfaction curve of effectiveness correspondence, the present invention has quoted the following feature of cost function in the prospect theory that Kahneman and Tversky propose and has carried out the satisfaction portrayal:

Must be a relative notion with mistake, be a certain subjective reference point set at people oneself.People pay close attention to be actual conditions with respect to this reference point apart from size rather than the pairing absolute figure of actual conditions.

Must present the rule that sensitivity is successively decreased with mistake.The curve of cost function is the curve of approximate " s ".In the two-dimensional coordinate system of Fig. 6 correspondence, what be positioned at first quartile is the profit curve, and what be positioned at third quadrant is the loss curve, and true origin is corresponding to subjective reference point.Near more from reference point, difference makes people all the more responsive; Away from the difference of reference point, people are insensitive more for.With money is example, if the money that on behalf of people, initial point obtain is 0, the gap between obtaining 10 yuan and 20 yuan on the subjective sensation is greater than the gap of 1000 yuan and 1010 yuan so.

Loss is evaded.It is bigger for people's impact than the interests that obtain isodose that this refers to a certain amount of interests of loss.With money is example, and the misery of losing 100 yuan is much stronger than the happy psychological feelings that causes that obtains 100 yuan.

The present invention has used foregoing as the basis when the portrayal user satisfaction, and the feedback information according to the user uses correlation technique to try to achieve the satisfaction function simultaneously.Adopt the rational people's supposition in the microeconomics simultaneously, think that the buyer can not exceed the psychology expection of oneself in the process of quotation, if in the time of can obtaining maximum benefit during the valuation range of buyer's truly expressed oneself, the buyer can express the valuation range of oneself truly so.

When carrying out the processing power transaction, the present invention is divided into two parts with total satisfaction of the buyer: use the satisfaction representative transaction effectiveness of the buyer aspect the cost amount of money that fulfils assignment, the buyer is in the representative consumption of the satisfaction aspect deadline effectiveness, utilizes both weighted sum to represent total satisfaction of the buyer.When handling the storage capacity transaction, total the buyer's of the present invention satisfaction is exactly the satisfaction that spends in amount of money aspect.

4.2 correlation type is divided

According to the difformity that both sides' satisfaction curve may occur, the present invention is divided three classes the user: easily satisfy type, difficulty satisfies type and rational type.

Fig. 7 represents is three types the satisfaction curve of the buyer aspect the cost amount of money.The satisfaction of curve representative descends till the low limit value of such curve to cost amount of money direction gradually along initial point, and this low limit value is represented is when the conclusion of the business price equals budget that the buyer quotes, the satisfaction value of the buyer's minimum transaction effectiveness correspondence.Even the buyer who it is considered herein that three types quotes identical budget, the satisfaction when equaling budget for knockdown price also may be different.The expectation expenditure of satisfaction when in like manner, being positioned at to(for) knockdown price also may be different.

To pay this a part of called after expectation conclusion of the business zone of budget in the curve from expectation, the transaction effectiveness satisfaction that difficulty satisfies the type buyer is a concave function in expectation conclusion of the business zone.This is because such buyer is a reference point with expectation expenditure usually, when the cost amount of money gradually away from the expectation expenditure to budget near the time, the buyer can think and oneself lost a part of interests that oneself should obtain, so satisfaction descends rapidly.When curve gradually to budget near the time, according in this section to the associated description of worth curve, the fall of curve can be slowed down gradually.

The rational type buyer's transaction effectiveness satisfaction is a linear function in expectation conclusion of the business zone, and such buyer thinks that it is natural working as the change between expectation expenditure and budget of the cost amount of money, so the satisfaction and the cost amount of money present linear relationship.

The transaction effectiveness satisfaction that easily satisfies the type buyer is a convex function in expectation conclusion of the business zone, such buyer is reference point usually with the budget, when the cost amount of money gradually away from budget to expectation expenditure near the time, the buyer can think and oneself obtained a part of extra interests, this moment, satisfaction can rise rapidly, but along with the cost amount of money approaches to the expectation knockdown price gradually, the rising degree of satisfaction can be slowed down gradually.

The present invention is divided three classes the homework type of operation according to different deadline preferences: The faster the better, good more and best on time more slowly.Fig. 8, Fig. 9 represent respectively is that three types the buyer is that The faster the better and good more more slowly satisfaction curve for homework type.Figure 10 has just showed best on time a kind of situation, it is considered herein that when actual finish time departed from the expected performance time of buyer's appointment, the buyer might pay extra cost when this situation occurring, and therefore consuming satisfaction can descend.Figure 10 representative be that the buyer's satisfaction type is in the identical situation in expected performance time both sides.

When the seller's satisfaction is set at seller's output capacity piece on sale for the satisfaction value of piece knockdown price.The same with the buyer, the seller is divided three classes according to the difference of curve shape, corresponding satisfaction curve is as shown in figure 11.

When handling the storage resources transaction, the ability that it is considered herein that the process standard that all sellers provide is an indifference for the buyer, therefore the present invention thinks that when considering the buyer's satisfaction consumption effectiveness equates with transaction effectiveness, at this moment the buyer's the satisfaction effectiveness satisfaction that equals to conclude the business.When seller's satisfaction is set at seller's output capacity piece on sale for the satisfaction value of piece knockdown price.

4.3 the computing method of satisfaction

Because only when supply exceed demand, the present invention just considered both sides' satisfaction size, so the narration of this section all is with supply exceed demand as default situations.As shown in figure 12, the flow process of calculating satisfaction function is as follows.

After a transaction is reached, if the expenditure that fulfils assignment during less than budget, it is considered herein that buyer's feedback of this transaction is effective greater than the buyer's expectation expenditure for the consumption satisfaction of calculating this buyer.

When effective transaction count of the buyer less than preset value N _MinThe time, think that this buyer's transaction count is not enough to count required relevant information, suppose that this buyer is rational type this moment.

When effective transaction count of user has surpassed N _MinAfter, use following three coordinate points to come the match quafric curve: (C _Exp, Sat _Type(C _Exp)), (B, Sat _TypeAnd (C (B)) _Fuz, Sat _Type(C _Fuz)).Horizontal stroke, the ordinate of these three points represented (expectation expenditure respectively, the satisfaction of corresponding buyer's type correspondence), (budget, the satisfaction of corresponding buyer's type correspondence), (fuzzy expenditure, the satisfaction of corresponding buyer's type correspondence), the satisfaction preset value of the same reference point of wherein different buyer's type correspondences may be different.

In order to obtain the buyer's transaction effectiveness satisfaction curve, need calculate C _FuzBecause C _FuzValue often change with the difference of buyer's target, so the present invention needs a kind of can the expenditure according to the expectation in buyer's target With budget B _jCalculate C _FuzMethod.In order to attempt to determine C _FuzWith the expectation expenditure

With budget B _jBetween relation, the present invention makes and tries to achieve C with the following method _FuzThe fuzzyyest corresponding intellectual type satisfaction

Value:

Use following form to represent for the buyer about the evaluation information of the gross expenditure (amount) that fulfils assignment:

Finish the gross expenditure amount of every transaction according to buyer j _j, the expenditure of the expectation in buyer's target

Budget B _j, use formula (3.7) to calculate this expenditure for the satisfaction S when buyer j is the intellectual type buyer _r(amount _j) (annotate: this formula is the simplification version of intellectual type satisfaction curve mentioned above, has changed broken line in that part formula below outside the expectation turnover).Wherein, Sat _r(C _Exp) represent the transaction effectiveness satisfaction value of intellectual type buyer when gross expenditure equals to expect to pay, Sat _r(B) represent the transaction effectiveness satisfaction value of intellectual type buyer when gross expenditure equals budget.Calculate S _r(amount _j) after, with S _r(amount _j) and the buyer's evaluation E (amount _j) formation two dimensional surface coordinate (S _r(amount _j), E (amount _j)).

Use nearest N _MinThe coordinate that inferior transaction obtains

The horizontal ordinate of these coordinates satisfies

Obtain the fuzzyyest intellectual type satisfaction according to these coordinates

Make separator bar

Coordinate plane is divided into two parts, and a part is assembled the buyer as much as possible and is evaluated as satisfied coordinate points, and another part is assembled the buyer as much as possible and is evaluated as unsatisfied coordinate points.

Because separator bar x=S _r(amount ' _j) purpose be to tell as much as possible to be evaluated as satisfied coordinate and to be evaluated as unsatisfied coordinate, therefore introduce penalty F _PunAdd up separator bar and distinguish the fine or not degree of these two kinds of coordinates, wherein F _PunValue more little, then the differentiation degree of this separator bar is good more.Penalty F _PunComputing method be:

Initialization F _Pun=0.Travel through this N _MinIndividual coordinate supposes that current coordinate is

If

And

Then think to satisfy condition 1, if

And

Then think and satisfy condition 2.Use following formula to calculate F _Pun:

Therefore, the present invention makes and calculates the fuzzyyest intellectual type satisfaction with the following method

Value:

Respectively with N _MinHorizontal ordinate in the individual coordinate is obtained this N as separator bar _MinMake the separator bar of penalty value minimum in the bar separator bar

With make the penalty value obtain time little separator bar

Use

As separator bar obtain corresponding penalty value and with

Corresponding penalty value is made comparisons, and the point of selecting to make the penalty minimum in these two points is as final separator bar, and with the fuzzy satisfactory degree of its abscissa value as intellectual type

Obtain

Value after, need with Fuzzy satisfactory degree Sat with the default intellectual type user of auction platform _r(P _f) compare and judge buyer's type.If

Then this buyer is considered to difficulty and satisfies the type buyer; If

Then this buyer is considered to easily satisfy the type buyer; If

Then this buyer is considered to the rational type buyer.

When platform receives the buyer's target, pay according to the expectation in the target pricing information Budget B _j, the correspondence expenditure when using formula (3.9) can calculate consumption satisfaction corresponding to this buyer's target to reach intermediate value

$P_f^j=C_{\exp }^j+S_r\left({\mathrm{amount}}_j^k\right)\&times;(B_j-C_{\exp }^j)---\left(3.9\right)$

The assumptive close price is the expectation expenditure

Budget B _jThe fuzzyyest knockdown price

Corresponding satisfaction is respectively the higher satisfaction of corresponding buyer's type

Low satisfaction

Fuzzy satisfactory degree

Obtained thus about three coordinate points on the satisfaction curve of knockdown price

With The present invention uses quafric curve to come the satisfaction function of piecewise fitting buyer j

Concrete grammar is as follows:

Calculate in the consumption satisfaction curve from the expectation expenditure

To budget B _jPart the time, be calculated as follows ternary linear function group, try to achieve relevant quafric curve parameter a ₁, b ₁, c ₁

$\left\{\begin{array}{c}{a}_1{\&times;C}_{\exp }^j\&times;C_{\exp }^j+{\mathrm{<figure-callout\; id="1"\; label="b"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">b</figure-callout>}}_1\&times;C_{\exp }^j+c_1=S_b^j\\ a_1{\&times;B}_j{\&times;B}_j+{\mathrm{<figure-callout\; id="1"\; label="b"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">b</figure-callout>}}_1{\&times;B}_j+c_1=S_w^j\\ a_1{\&times;P}_f^j{\&times;P}_f^j+{\mathrm{<figure-callout\; id="1"\; label="b"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">b</figure-callout>}}_1{\&times;P}_f^j+c_1=S_f^j\end{array}\right.---\left(3.10\right)$

As shown in figure 13, because the inherent characteristics of quafric curve, whether the curve shape that the present invention obtains judgement with the following method is identical with expection.Calculate the line of symmetry of quafric curve

Judge

With

Whether these three points are in the same side of line of symmetry.When if three points are positioned at a side of curve simultaneously, then this curve shape is identical with expection, otherwise by adjusting the quafric curve parameter curve shape is adjusted.

In order to guarantee to consume the monotone decline of satisfaction function, the present invention use straight line replace in the curve from

Arrive

This part, be calculated as follows the linear equation in two unknowns group, try to achieve the correlation parameter a of straight line ₂, b ₂

$\left\{\begin{array}{c}{a}_2{\&times;B}_j+b_2=S_w^j\\ a_1\&times;P_f^j+b_2=S_f^j\end{array}\right.---\left(3.11\right)$

Calculate in the consumption satisfaction curve and pay from 0 to the expectation expenditure

Part the time, it is considered herein that adjacent two sections curves exist

The first order derivative of point is identical, and second derivative changes.The second derivative variation factor of supposing this buyer's corresponding types is a ", be calculated as follows ternary linear function group, try to achieve relevant quafric curve parameter a ₃, b ₃, c ₃

$\left\{\begin{array}{c}{a}_3=a_1\&times;a^{\&prime;\&prime;}\\ 2a_3\&times;C_{\exp }^j+b_3={2a}_1\&times;C_{\exp }^j+b_1\\ a_3\&times;C_{\exp }^j\&times;C_{\exp }^j+b_3\&times;C_{\exp }^j+c_3=S_b^j\end{array}\right.---\left(3.12\right)$

Obtain consuming and pay in the satisfaction curve from 0 to the expectation expenditure

Part after, need to calculate this section curve and whether comprise the most satisfied expenditure that makes functional value equal 1 If comprise, then stipulate curve from 0 to the most satisfied expenditure

That part of functional value be 1.

Therefore, when irregularly shaped and the most satisfied expenditure shown in Figure 13 occurring

The time,

Respective function concern that segmentation is expressed as follows.

$S_{\mathrm{bp}}^j\left(\mathrm{amount}\right)=\left\{\begin{array}{cc}1& 0\&le;\mathrm{amount}\&le;C_{\mathrm{best}}^j\\ a_3\&times;{\mathrm{amount}}^2+b_3\&times;\mathrm{amount}+c_3& C_{\mathrm{best}}^j\&le;\mathrm{amount}\&le;C_{\exp }^j\\ a_1\&times;{\mathrm{amount}}^2+{\mathrm{<figure-callout\; id="1"\; label="b"\; filenames="HDA0000079728740000033.png"\; state="\{\{state\}\}">b</figure-callout>}}_1\&times;\mathrm{amount}+c_1& C_{\exp }^j\&le;\mathrm{amount}\&le;P_f^j\\ a_2\&times;\mathrm{amount}+b_2& P_f^j\&le;\mathrm{amount}\&le;B_j\end{array}\right.---\left(3.13\right)$

Use above method and related data information, can obtain the transaction satisfaction function S of the buyer respectively about the reality expenditure _Bp(amount) with about the consumption satisfaction function S of operation deadline _t(time), the seller is about the satisfaction function S of knockdown price _Sp(Sp).The weight of supposing transaction satisfaction in buyer's satisfaction is ω _Bp, the weight of consumption satisfaction is ω _t, and ω _Bp+ ω _t=1.The seller's satisfaction S then _Prv, the buyer satisfaction S _BidBe respectively:

S _prv＝S _sp (3.14)

S _bid＝ω _bp×S _bp+ω _t×S _t (3.15)

When calculating the total satisfaction of both sides, do not protect the interests of Reference Group simultaneously in order to make the supply and demand situation, in formula (3.16), used the supply and demand proportion omegab to regulate the weight of both sides' satisfaction separately in total satisfaction, total satisfaction S _TotalComputing formula as follows:

$S_{\mathrm{total}}=\frac{1}{(1+\mathrm{\&omega;})}{S}_{\mathrm{prv}}+\frac{\mathrm{\&omega;}}{(1+\mathrm{\&omega;})}{S}_{\mathrm{bid}}---\left(3.16\right)$

4.4 User Priority

For the problem that manages conflict in the demand matching process of transaction, the present invention uses the ordering reference frame of the priority of Three Estate as target.The size that at first is to use integration is as first priority, and the present invention uses integration to represent the ability piece number that this user has struck a bargain, and corresponding relation is that ability piece of transaction increases and pluses fifteen.Using integration is in order to distinguish the percentage contribution of each user for auction platform.

When user's integration was identical, in order to distinguish dissimilar users, the rational type fuzzy satisfactory degree of trying to achieve when using the transaction effectiveness satisfaction of calculating the user in the present invention was as second priority.This value more little representative this buyer is easy more to be satisfied, the under equal conditions request that auction platform should these buyeies of priority processing, thus should the more little priority of value high more.

When top two values are all identical, according to distinguishing the priority size time of arrival of user's target.

After preparatory stage in this auction cycle finishes, according to descending order seller's target is sorted and use the seller's formation after the ordering that buyer's target is mated according to the seller's priority.

The ability piece reservation mode of 5 three kinds of homework type correspondences

When using this problem of genetic algorithm for solving among the step 2.4.3, in to chromosome coding, need be according to transaction content, homework type (The faster the better, best, good more more slowly on time) and the ability piece of having reserved, the decision seller distributes to the ability set of blocks of this operation.

Use the one-dimension array structure, each chromosome representative is corresponding to a kind of scheduling scheme of buyer j resource requirement; Gene figure place in the chromosome is corresponding to above n _Ori, each gene in the array is corresponding to each seller in the supplier's set that obtains when determining the hunting zone, i position genic value

The ability piece number that i the seller of representative in supplier set provides for buyer j: when conclude the business be processing power the time, the ability set of blocks

Relevant with homework type, as shown in figure 14 diagonal line hatches for reserving, tiltedly the seller who represents for gene of grid shade distributes to the ability set of blocks of this operation, supposes that this genic value is 5, then the ability set of blocks of three types of representatives

As shown in figure 14; When the transaction be storage capacity the time, the ability set of blocks

Be in the time period of buyer's demand scope

The supply of representative, the ability piece reservation mode of three kinds of homework type correspondences as shown in figure 14.

6 batch matching algorithm flow processs

The batch matching treatment flow process of step 2.5.1 as shown in figure 15.

Grid resource scheduling of the present invention is estimated the price of current common process ability according to the related data of large-scale door fictitious host computer in specific implementation process.By grid resource scheduling method of the present invention and traditional comparison, find that grid resource scheduling method of the present invention all shows good at aspects such as improving grid resource allocation efficient, the maximization of resource provider interests based on one-sided grid resource scheduling method.

Claims (1)

1. grid resource scheduling method based on continuous two way auction mechanism, it is characterized in that: by inch of candle mechanism is carried out grid resource scheduling, changes continuous two way auction into cycle system, and the adjacent auction cycle joins end to end; But the total duration in an auction cycle equates with the time range of its trading capacity and is corresponding one by one; Ability in the corresponding with it time range of in this auction cycle, can only concluding the business, and the ability in scope sometime can only be concluded the business in the auction cycle in correspondence; Each auction cycle comprises the steps:

Step 1, preparatory stage: finish the trust stage in the traditional auction, be responsible for receiving seller's target and add up and predict supply-demand information as required;

Step 1.1: receive seller's target;

Step 1.2: upgrade each user's integration information, integration is represented the ability piece number that this user has struck a bargain;

Step 1.3: prediction dealing side supplydemand relationship:, predict reasonable knockdown price Rsap and reasonable demand amount Ardm in this auction cycle according to average knockdown price Sap of piece in the periodicity transaction record in the past and reasonable demand amount Rdm; First cycle do not predict, supposes the auction platform state that is in that supply exceed demand; The change of supposing these data is clocklike, uses m historical data to go to estimate m+1 data, promptly known A ₁, A ₂, A ₃Λ, A _m, ask A _M+1Select for use linear prediction and model prediction as finding the solution A _M+1Two kinds of methods;

(a) linear prediction

Suppose A _kBe from A ₁To A _K-1The weights of this k number and, A _K+1Be from A ₂To A _kThis k number and ..., A _mBe from A _M-k+1To A _M-1Weights and, and the weights correspondent equal of all groups;

$\left\{\begin{array}{c}{w}_1{\&times;A}_1+w_2{\&times;A}_2+\mathrm{\&Lambda;}+w_{k-1}\&times;A_{k-1}=A_k\\ w_1\&times;A_2+w_2\&times;A_3+\mathrm{\&Lambda;}+w_{k-1}\&times;A_k=A_{k+1}\\ M\\ w_1{\&times;A}_{m-k}+w_2{\&times;A}_{m-k+1}+\mathrm{\&Lambda;}+w_{k-1}\&times;A_{m-2}=A_{m-1}\\ w_1{\&times;A}_{m-k+1}+w_2{\&times;A}_{m-k+2}+\mathrm{\&Lambda;}+w_{k-1}{\&times;A}_{m-1}=A_m\end{array}\right.---\left(3.4\right)$

Get

Behind the solving equation, use w ₁* A _M-k+2+ w ₂* A _M-k+3+ Λ+w _K-1* A _mValue as A _M+1Predicted value;

(b) model prediction

Suppose (A _L+1, A _L+2, Λ, A _L+n-1) this continuous n-1 number pattern and the A that form _L+nForm a kind of corresponding relation; In like manner, with A _M+1That relevant is pattern M ₁(A _M-n+2, A _M-n+3, Λ, A _m); In order to estimate A _M+1Value, need from historical data, find out and pattern M ₁The most close pattern M ₂(A _J+1, A _J+1, Λ, A _J+n-1); According to pattern M ₂Corresponding value A _J+nCan estimate A _M+1Fuzzy value; From the time consideration, from pattern M ₁Nearest pattern is pattern M ₃(A _M-n+1, A _M-n+2, Λ, A _M-1), by comparing the difference between these patterns, can be at A _J+nObtain a more suitable value as A around the value _M+1Estimated value;

Select the module of Euclidean distance (Euclidean distance) for use as the similarity degree between the pattern; Suppose in all patterns, with pattern M ₁Nearest pattern be M ₂, distance between the two is dist (M ₁, M ₂), pattern M ₁With pattern M ₃Between distance be dist (M ₁, M ₃), use following formula (3.5) to estimate A _M+1Value;

$A_{m+1}=A_{j+n}+\frac{\mathrm{dist}(M_1,M_2)}{\mathrm{dist}(M_1,M_3)}\&times;(A_{j+n}-A_m)---\left(3.5\right)$

All predict reasonable knockdown price Rsap and reasonable demand amount Ardm by above dual mode at first, compare two kinds of methods error size in nearly N the auction cycle, the less Forecasting Methodology of use error is auctioned the variable Forecasting Methodology in cycle as this;

According to supply and the reserve price in the seller's target that receives, calculate reasonable supply (Reasonable Supple) Rsup in this auction cycle, on behalf of all sellers' that receive in the preparatory stage reservation price, Rsup be not higher than the supply sum of the reasonable knockdown price Rsap that estimates;

Step 1.4: calculate supply and demand scale parameter ω, computing formula is: ω=Rsup/Ardm (3.3)

When ω＞1, think that overall supplies that the seller provides can satisfy most of buyer's demand, and can be in the operation term of validity overall supplies be to select optimal that part of ability piece to improve the buyer's consumption satisfaction the ability set of blocks of ω times of workload; At this moment, the piece knockdown price Sp of each ability piece of buyer j and seller i transaction _Ij, use following formula (3.1) to calculate:

${\mathrm{Sp}}_{\mathrm{ij}}=\left\{\begin{array}{cc}{\mathrm{Rsp}}_i+({\mathrm{Hsp}}_j-{\mathrm{Rsp}}_i)\&times;1/(1+\mathrm{\&omega;})& \mathrm{\&omega;}<{\mathrm{\&omega;}}_{\max }\\ {\mathrm{Rsp}}_i& \mathrm{\&omega;}\&GreaterEqual;{\mathrm{\&omega;}}_{\max }\end{array}\right.---\left(3.1\right)$

When ω≤1, think that the overall supplies that the seller provides may be less than demand, this moment, auction platform just merely checked whether can fulfil assignment and do not consider both sides' satisfaction in the operation term of validity; At this moment, the piece knockdown price Sp of each ability piece of buyer j and seller i transaction _Ij, use following formula (3.2) to calculate:

${\mathrm{Sp}}_{\mathrm{ij}}=\left\{\begin{array}{cc}{\mathrm{Rsp}}_i+({\mathrm{Hsp}}_j-{\mathrm{Rsp}}_i)\&times;1/(1+\mathrm{\&omega;})& \mathrm{\&omega;}<{\mathrm{\&omega;}}_{\min }\\ {\mathrm{Rsp}}_j& \mathrm{\&omega;}\&GreaterEqual;{\mathrm{\&omega;}}_{\min }\end{array}\right.---(3.2)$

Step 2, auction phase: auction platform receives buyer's target, finishes coupling and accounts settling phase according to demand; The seller can revise the reservation price of the current ability of sale at any time or supply more ability in this stage; Each transaction clearing back auction platform writes down this transaction and issues the buyer and show scheduling time;

Step 2.1: when ω＞1, think that overall supplies that the seller provides can satisfy most of buyer's demand; When ω≤1, think that the overall supplies that the seller provides may be less than demand;

During this auction cycle appearance situation that supply exceed demand, auction platform uses the matching algorithm of the information of considering both sides' satisfaction to choose the resource capability that those meet buyer's demand most; This auction cycle, it should be purpose in the ability of day part supply fully to sell out the seller that shortage of resources can cause auction platform when the situation of supply-less-than-demand occurring;

During this auction cycle appearance situation that supply exceed demand, the non real-time trigger condition is set to after a target arrives auction platform, target quantity and preset value in the formation of current buyer's target are compared, if the target number in the formation reaches preset value, then trigger coupling immediately, otherwise the setting timer is when timer mates the residue target in the target formation to after date; When supply-less-than-demand, what auction platform used is the matching algorithm of not considering both sides' satisfaction, and the trigger condition of algorithm is set to real-time triggering;

Step 2.2: auction platform is set following variable:

The capacity of ability piece (Block Capability): Bcp;

The time span of each ability piece (Block Time Span): T _Sp

But the maximum time span of trading capacity time range: N _Ms* T _SpBut the time range of then auctioning trading ability piece in the phase is [T _Ast, T _Ast+ N _Ms* T _Sp], T _AstRepresent the start time of the piece of ability the earliest that can conclude the business in this auction phase;

Supply and demand ratio lower limit: ω _MinWhen the supply and demand ratio in an auction cycle is not higher than this preset value, the knockdown price of ability piece is set at the buyer's piece best bid;

The supply and demand ratio upper limit: ω _MaxWhen the supply and demand ratio in an auction cycle is not less than this preset value, the knockdown price of ability piece is set at the seller's piece reservation price;

Step 2.3:, mate if use to mate in batches then change step 2.5 if mate if auction platform uses real-time coupling then changes step 2.4 ω＞1; Otherwise supply-less-than-demand uses step 2.6 to mate;

Step 2.4: the Real Time Matching Algorithm when supply exceed demand;

Auction platform triggers matching algorithm after receiving the target of buyer's transmission immediately;

When the transaction processing ability, for the target request that the buyer proposes, taking-up residue overall supplies in buyer's operation term of validity is not less than the seller of ω times of workload as seller's set of satisfying this demand from seller's formation;

When the transaction storage capacity, taking-up residue overall supplies in the required time section is not less than the seller of ω times of demand as seller's set of satisfying this demand from seller's formation;

According to the satisfaction function that statistics draws, the workload in buyer's demand is suitably distributed to each seller according to the principle of total satisfaction maximum;

Step 2.4.1: what the seller gathered determines:

When transaction is when handling resource, what suppose that current platform receives is the target that buyer j sends, according to the workload L in the target _jDetermine for the required supply of the current target of optimization process

According to operation submission time T _Js ^jWith off period Dl _jDetermine the term of validity scope of operation;

Step 2.4.1.1: variable is set

The overall supplies of representing current seller's set to provide in the operation term of validity scope of target j correspondence, the seller gathers PS _jSeller's set that representative is chosen for this operation of optimization process, seller's target number that on behalf of inspected, n cross; Will

Initial value be made as 0, the initial value of n is made as 1, the set PS _jBe set to sky;

Step 2.4.1.2: from seller's formation, duplicate the information of n seller's target, relatively highest-capacity piece bid Hsp _jWith the piece reserve price Rsp in current seller's target _iIf, Rsp _i≤ Hsp _jAnd the available ability piece of this seller i number in the operation term of validity

Then Seller i is joined set PS _jIn;

Step 2.4.1.3: if

Or n forwards rapid 2.4.1.4 to during greater than the target number in seller's formation; Otherwise n=n+1 forwards step 2.4.1.2 to;

Step 2.4.1.4: if Explanation can provide abundant ability to satisfy this demand from supplier's set; Otherwise because the surplus capacity deficiency, this demand of auction platform notice buyer j can't be satisfied;

When the resource of transaction is storage capacity, among the step 2.4.1.2

What represent is the sustainable ability piece number that provide of seller i in the required time section, and other step is same as described above;

Step 2.4.2: the mathematic(al) representation of problem correspondence:

When transaction be processing power the time, suppose to have in the supplier set n _OriThe individual seller, they are respectively at the ability set of blocks that buyer j provides

The ability set of blocks

Deadline be

The actual finish time time of this operation _jIt is the concluding time of the ability piece that begins the latest in these set, that is:

${\mathrm{time}}_j=\underset{0<i<n_{\mathrm{ori}}}{\max }\left(\mathrm{endTime}\left({\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\right)\right)---\left(3.17\right)$

The cost amount of money amount that buyer j fulfils assignment _jAvailable following formula calculates:

${\mathrm{amount}}_j=\underset{i=1}{\overset{n_{\mathrm{ori}}}{\mathrm{\&Sigma;}}}({\mathrm{Sp}}_{\mathrm{ij}}\&times;|{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\left|\right)---\left(3.18\right)$

Use formula (3.19), can calculate the satisfaction of buyer j

$S_{\mathrm{bid}}^j={\mathrm{\&omega;}}_{\mathrm{bp}}\&times;S_{\mathrm{bp}}^j\left({\mathrm{amount}}_j\right)+{\mathrm{\&omega;}}_t\&times;S_t^j\left({\mathrm{time}}_j\right)---\left(3.19\right)$

Use formula (3.20), can calculate the satisfaction of seller i

$S_{\mathrm{sp}}^i=S_{\mathrm{sp}}^i\left({\mathrm{Sp}}_{\mathrm{ij}}\right)---\left(3.20\right)$

At this moment, problem is converted into:

Maximize：S _total

$S_{\mathrm{total}}=\frac{1}{(1+\mathrm{\&omega;})}{S}_{\mathrm{bid}}^j+\frac{\mathrm{\&omega;}}{(1+\mathrm{\&omega;})}\frac{\underset{1\&le;i\&le;n_{\mathrm{ori}}}{\mathrm{\&Sigma;}}\left(\right|{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}|\&times;S_{\mathrm{sp}}^i)}{L_j}---\left(3.21\right)$

$s.t.{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\&SubsetEqual;P_{\mathrm{left}}^i$

$\left|\underset{1\&le;j\&le;n_{\mathrm{bid}}}{Y}{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\right|=L_j$

Wherein

Represent the ability set of blocks of seller i current residual;

When transaction be storage capacity the time, use formula (3.22) can calculate the satisfaction of buyer j

Remaining is the same;

$S_{\mathrm{bid}}^j=S_{\mathrm{bp}}^j\left({\mathrm{amount}}_j\right)---\left(3.22\right)$

Step 2.4.3: use this problem of genetic algorithm for solving:

(a) chromosome coding

Use the one-dimension array structure, each chromosome representative is corresponding to a kind of scheduling scheme of buyer j resource requirement; Gene figure place in the chromosome is corresponding to above n _Ori, each gene in the array is corresponding to each seller in the supplier's set that obtains when determining the hunting zone, i position genic value

The ability piece number that i the seller of representative in supplier set provides for buyer j: when conclude the business be processing power the time, the ability set of blocks

Relevant with homework type; When the transaction be storage capacity the time, the ability set of blocks

Be in the time period of buyer's demand scope

The supply of representative; According to transaction content (computational resource, storage resources), homework type (The faster the better, best, good more more slowly on time) and the ability piece reserved, the decision seller distributes to the ability set of blocks of this operation;

(b) initialization population

After setting population scale, produce each chromosome successively; When determining that chromosomal genic value is big or small, need to guarantee genic value $\left|{\mathrm{PS}}_{\mathrm{use}}^i\right|\&le;P_{\mathrm{ori}}^{\mathrm{ij}},$ And $\underset{1\&le;i\&le;n_{\mathrm{ori}}}{\mathrm{\&Sigma;}}\left|{\mathrm{PS}}_{\mathrm{use}}^i\right|=L_j;$

(c) suitable value function

The total satisfaction function of both sides of selecting formula (3.21) representative for use is as chromosomal suitable value function;

(d) select, intersect and variation

Selection operation uses the roulette back-and-forth method in the traditional genetic algorithm;

Interlace operation is used evenly and is intersected, and produces n _OriIndividual 0-1 integer wherein 0 is represented corresponding gene position exchange, and on behalf of corresponding gene position, 1 do not exchange; Mutation operation is the gene position respective value to be adjusted into 0 at random arrive

Between a positive integer value;

Because intersection and mutation operation may cause

This moment is with workload L _jProportional distribution according to existing gene place value is given each seller; If adjusting the back occurs

Then the value of this gene position correspondence is set to

And Workload provide by other seller in seller's formation;

Use different crossover probabilities and variation probability as the evolutionary operator probability of algorithm at different times; In early stage, algorithm is set to bigger crossover probability and less variation probability; In the later stage, algorithm is set to less crossover probability and bigger variation probability;

Step 2.4.4: reserved resource

After finishing, algorithm chooses optimum chromosome separating as this problem, respective markers is reserved and made to each genic value according to this chromosome correspondence to the resource capability in seller's target, according to time sequencing joins in job scheduling table at the ability piece that day part provides each seller in reservation procedure; Fulfil assignment behind the dispatch list, write down this transaction and transaction record and job scheduling table are issued the buyer;

Step 2.5: the batch matching algorithm when supply exceed demand

After auction platform triggered the batch matching algorithm, auction platform can be in each the inspection till processing target as much as possible all targets in handling the formation of buyer's target; Suppose that the buyer's target quantity in the formation of current buyer's target is n _Bid, algorithm flow is as follows:

Step 2.5.1: determine that the seller in the matching algorithm gathers PS

With this n _BidIndividual buyer's target is taken out buyer's target successively according to priority queueing; What suppose current taking-up is the target of buyer j, at first calculates to guarantee that matching algorithm necessarily has the minimum supply of separating

With the required supply of the current target of optimization process

When transaction is that step is as follows when handling resource:

Step 2.5.1.1: be provided with With

Initial value, make

Step 2.5.1.2: check the buyer's target in the formation to be matched successively, check whether the operation term of validity of current target k correspondence in the formation and the operation term of validity of target j have common factor; When two operation terms of validity have common factor, $D_{\max }^j=D_{\max }^j+L_k,$ $D_{\min }^j=D_{\min }^j+L_k;$

When the transaction be storage capacity the time, among the step 2.5.1.2

With

Calculating need check its place time period, when the time section is positioned at the common factor of two operation terms of validity,

When the time section is positioned at other period,

With

Value remain unchanged;

Obtain

With

Value after, when transaction be processing power the time, can the step below using be judged and target j is joined in the formation to be matched:

Step 2.5.1.2.1: variable is set

The overall supplies of representing current seller's set to provide in the operation term of validity scope of target j correspondence, the seller gathers PS _jSeller's set that representative is chosen for this operation of optimization process, seller's target number that on behalf of inspected, n cross; Will

Initial value be made as 0, the initial value of n is made as 1, set PS _jBe set to sky;

Step 2.5.1.2.2: from seller's formation, duplicate the information of n seller's target, relatively the ability piece best bid Hsp of target j _jWith the piece reserve price cave Rsp among current seller's target i _iIf Rsp _i≤ Hsp _jAnd the available ability piece of this seller i number in the operation term of validity

Then

Seller i is joined set PS _jIn;

Step 2.5.1.2.3: if Or n forwards step 2.5.1.2.4 to during greater than the target number in seller's formation; Otherwise n=n+1 forwards step 2.5.1.2.2 to;

Step 2.5.1.2.4: if Target j is joined in the formation to be matched, the seller is gathered PS _jMerging to original seller gathers among the PS; Otherwise, if

Target j joined wait to look in the formation; If

The supply deficiency of the then current seller in the operation term of validity of target j correspondence can't satisfy this demand;

When the transaction be storage capacity the time, among the step 2.5.1.2.2

Calculating need check

The time period distribution situation; In like manner, in the comparison operation of step 2.5.1.2.3, have only

In all time periods all more than or equal to

The time, just think $D_{\mathrm{now}}^j\&GreaterEqual;D_{\max }^j;$

The mathematic(al) representation of step 2.5.2 problem correspondence

When the transaction be processing power the time, suppose that the buyer's target number in the formation to be matched is n _Bid, the current seller gathers PS n _AskThe individual seller, n _AskThe individual seller is respectively for the actual ability set of blocks that provides of demand that satisfies buyer j

The ability set of blocks

Deadline be

The deadline time of this operation _jIt is the concluding time of the ability piece that begins the latest in these set, that is:

${\mathrm{time}}_j=\underset{\&ForAll;0<i<n_{\mathrm{aski}}}{\max }\left(\mathrm{endTime}\left({\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\right)\right)---\left(3.23\right)$

The cost amount of money amount that buyer j fulfils assignment _jAvailable following formula calculates:

${\mathrm{amount}}_j=\underset{i=1}{\overset{n_{\mathrm{ask}}}{\mathrm{\&Sigma;}}}({\mathrm{Sp}}_{\mathrm{ij}}\&times;|{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\left|\right)---\left(3.24\right)$

Use formula (3.19), can calculate the satisfaction of buyer j Use formula (3.20), can calculate the satisfaction of seller i

At this moment, problem is converted into:

Maximize：S _total

$S_{\mathrm{total}}=\frac{\underset{1\&le;j\&le;n_{\mathrm{bid}}}{\mathrm{\&Sigma;}}{S}_j^{\&prime;}}{n_{\mathrm{bid}}}---\left(3.25\right)$

$S_j^{\&prime;}=\frac{1}{(1+\mathrm{\&omega;})}{S}_{\mathrm{bid}}^j+\frac{\mathrm{\&omega;}}{(1+\mathrm{\&omega;})}\frac{\underset{1\&le;i\&le;n_{\mathrm{ask}}}{\mathrm{\&Sigma;}}\left(\right|{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}|\&times;S_{\mathrm{sp}}^i)}{\underset{1\&le;j\&le;n_{\mathrm{bid}}}{\mathrm{\&Sigma;}}{L}_j}---\left(3.26\right)$

$s.t.{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\&SubsetEqual;P_{\mathrm{left}}^i$

$\left|\underset{1\&le;j\&le;n_{\mathrm{bid}}}{Y}{\mathrm{PS}}_{\mathrm{use}}^{\mathrm{ij}}\right|=L_j$

Wherein

Represent the ability set of blocks of seller i current residual;

When transaction be storage capacity the time, use formula (3.22) to calculate the satisfaction of buyer j

Remainder formula is the same;

Step 2.5.3 uses this problem of genetic algorithm for solving

(a) chromosome coding

Use the two-dimensional matrix structure, line number is buyer's target number of this time coupling, and columns is to gather the element number of determining among the PS the seller; Each row of matrix is represented the scheduling scheme of buyer's demand, wherein the element of the capable j row of i The ability set of blocks of representing seller j to provide for buyer i, the particular location of ability set of blocks in original seller's ability description are to use the strategy identical with step 2.4.3 to determine after the ability that i-1 is capable before reserving;

(b) initialization population

After setting population scale, produce each chromosome successively; When producing chromosome, determine the element value of each row in the chromosome successively according to the order of buyer's target; When determining each row element value, need make a mark avoiding the ability piece to be taken once more to relevant ability piece, and guarantee by the operation of back

(c) suitable value function

The total satisfaction function of both sides of selecting formula (3.25) representative for use is as chromosomal suitable value function;

(d) select, intersect and variation

Selection operation uses the roulette back-and-forth method in the traditional genetic algorithm; Interlace operation is used evenly and is intersected, and produces n _BidIndividual 0-1 integer, the wherein row element collective exchange of the corresponding operation of 1 representative, 0 representative does not exchange; Mutation operation is that the delegation's element value with the operation correspondence carries out random initializtion again;

Because interlace operation may cause

The maximum capacity piece number of representing seller i under present case, to provide for buyer j; Corresponding method of adjustment is that the value of this gene position correspondence is set to

And Workload provide according to priority by other seller in seller's formation; Use different crossover probabilities and variation probability as the evolutionary operator probability of algorithm at different times; In early stage, algorithm is set to bigger crossover probability and less variation probability; In the later stage, algorithm is set to less crossover probability and bigger variation probability;

Step 2.5.4: reserved resource

After finishing, algorithm chooses optimum chromosome separating as this problem, respective markers is reserved and made to each row element value according to this chromosome correspondence to the resource capability in seller's target successively, in reservation procedure each seller joined according to time sequencing in the corresponding job scheduling table at the ability piece that day part provides; Fulfil assignment behind the dispatch list, write down each transaction and transaction record and job scheduling table are issued the corresponding buyer;

Step 2.6: the matching algorithm during supply-less-than-demand

Under the situation of supply-less-than-demand, because the ability short supply of whole period can't be satisfied each request that all buyeies submit to; So in order to satisfy buyer's demand as much as possible, can algorithm not be considered the satisfaction information of the buyer in asking, only check in correlation time satisfy the demands in the section; After auction platform is received buyer's target, trigger matching algorithm immediately;

Suppose that current that receive is buyer's target j, variable is set

The overall supplies of representing current seller's set to provide in the required time scope of target j, the seller gathers PS _jSeller's set that representative is chosen in order to satisfy this demand, seller's target number that on behalf of inspected, n cross, algorithm flow is as follows;

Step 2.6.1: will

Initial value be made as 0, the initial value of n is made as 1, set PS _jBe set to sky;

Step 2.6.2: from seller's formation, duplicate the information of n seller's target, relatively highest-capacity piece bid Hsp _jPiece reserve price Rsp with current seller i _iIf Rsp _i≤ Hsp _jAnd the available supply in the operation term of validity

Then Seller i is joined set PS _jIn;

Step 2.6.3: relatively

And L _jIf,

The demand actual supply that provide of seller i then for satisfying target j If

The actual supply that provides of seller i then

Step 2.6.4: if

Or n forwards step 2.6.5 to when equaling target number in seller's formation; Otherwise n=n+1 forwards step 2.6.2 to;

Step 2.6.5: if

According to

Value respective markers is reserved and made to the resource capability in seller's target in reservation procedure, according to time sequencing join in job scheduling table at the ability piece that day part provides each seller; Fulfil assignment behind the dispatch list, write down this transaction and transaction record and job scheduling table are issued the buyer; If Then the supply deficiency of all sellers in the required time section of target j correspondence can't satisfy this demand;

Step 2.7: each transaction clearing back auction platform writes down this transaction and issues the buyer and show scheduling time;

Step 3, the waiting period: finish the periodicity supply-demand information record in auction cycle, the concluding time by making ability after waiting for a period of time and the zero-time in next auction cycle are synchronous;

Step 3.1: count average knockdown price Sap of piece and reasonable demand amount Rdm according to the interim transaction record of auction;

Calculate the mean value of the knockdown price of All Activity in this auction cycle, promptly the merchant of total business volume and total amount of transactions is the average knockdown price Sap of this period blocks;

Calculate the demand and do not strike a bargain but the piece highest quotation is higher than the demand sum that equals average knockdown price Sap in this cycle is Rdm of striking a bargain of all buyeies in this auction cycle.

Images