research article a dynamic pricing reverse auction-based...

Research ArticleA Dynamic Pricing Reverse Auction-Based Resource AllocationMechanism in Cloud Workflow Systems

Xuejun Li12 Ruimiao Ding1 Xiao Liu3 Xiangjun Liu1 Erzhou Zhu1 and Yunxiang Zhong1

1School of Computer Science and Technology Anhui University Hefei China2School of Software and Electrical Engineering Swinburne University of Technology Melbourne Australia3School of Information Technology Deakin University Melbourne Australia

Correspondence should be addressed to Xiao Liu xiaoliudeakineduau

Received 22 July 2016 Accepted 3 October 2016

Academic Editor Wenbing Zhao

Copyright copy 2016 Xuejun Li et alThis is an open access article distributed under theCreativeCommonsAttribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Market-oriented reverse auction is an efficient and cost-effective method for resource allocation in cloud workflow systems sinceit can dynamically allocate resources depending on the supply-demand relationship of the cloud market However during theauction the price of cloud resource is usually fixed and the current resource allocation mechanisms cannot adapt to the changeablemarket properly which results in the low efficiency of resource utilization To address such a problem a dynamic pricing reverseauction-based resource allocation mechanism is proposed During the auction resource providers can change prices according tothe trading situation so that our novel mechanism can increase the chances of making a deal and improve efficiency of resourceutilization In addition resource providers can improve their competitiveness in the market by lowering prices and thus userscan obtain cheaper resources in shorter time which would decrease monetary cost and completion time for workflow executionExperiments with different situations and problem sizes are conducted for dynamic pricing-based allocation mechanism (DPAM)on resource utilization and themeasurement of TimelowastCost (TC)The results show that ourDPAMcanoutperform its representativein resource utilization monetary cost and completion time and also obtain the optimal price reduction rates

1 Introduction

Workflow model is often used to manage complex businessapplications A workflow is defined as a collection of taskswhich are handled in a specific order [1 2] A workflowman-agement system needs to allocate and execute tasks efficientlyto meet usersrsquo needs Cloud computing uses a pay-as-you-go model which provides virtually unlimited computationalresources at lower costs with better reliability and delivers theresources by means of virtualization technologies [3] Cloudworkflow systems are workflow systems deployed on cloudcomputing environment to gain unlimited resources includ-ing computation storage and network [4]

Resource allocation for cloud workflow systems hasreceivedmuch attention Allocating cloud resources to work-flow is an NP-hard problem and needs to consider the overallperformance of system especially monetary cost and com-pletion time [5] Resource allocation mechanism includesconventional methods and market-oriented methods [6 7]

Conventional methods require global knowledge and com-plete information Users pay for the resources based onreserved price In contrast market-oriented methods canoffer incentives to participants and the methods decide theprice based on the values that users can get from the resources[8]

Different from conventional counterparts market-ori-ented methods assume that providers and users are rationaland intelligent And resource allocation depends on manyfactors including supply-demand relationship and resourceprice Auction is a powerful tool to allocate resources in themarket Generally speaking auction is a protocol that allowsparticipants to indicate their interests in different resourcesand use these indications of interest to determine bothresource allocation and price [9] Reverse auctionmethod is atypical auction In conventional auction there are one sellerandmultiple buyers As for reverse auction there aremultiplesellers and only one buyer [10] The user sends the specifica-tion of resource requirement to the cloud broker and requests

Hindawi Publishing CorporationScientific ProgrammingVolume 2016 Article ID 7609460 13 pageshttpdxdoiorg10115520167609460

2 Scientific Programming

resources The cloud broker transfers the specification toall cloud providers The cloud providers sell resources withproper price and capabilities And then the users select theoptimal resources according to some criteria for exampleQuality of Service [11 12] In general reverse auction isused to prevent the occurrence of trading fraud and achievedynamic pricing and automatic procurement [11] In themost relevant literature [12] the authors propose BiobjectiveScheduling Strategy (BOSS) based on reverse auction toallocate resources for tasks of workflows and each task startsan auction and gets a resource to minimize the monetarycost and completion time However the resource price isfixed during the auction so that some providers with weakercompetitiveness may lose the auction all the time This leadsto low efficiency of resource utilization which decreases theallocation equilibrium of tasks on resources in the cloudmarket

In this paper we firstly present a dynamic pricing strategyto change resource prices according to the trading situationand then present a DPAM mechanism to improve theefficiency of resource utilization Changing resource price isa common and efficient way to increase competitiveness [13]In the dynamic pricing strategy the resource price changesaccording to the trading situationThose providers who havestrong competitiveness (ie higher chance of winning inhistoric auctions) in the market often keep resource pricesunchanged during an auction But for those who have weakcompetitiveness decreasing the resource price with a certainrate is an effective way to increase the winning chancesTherefore changing prices can increase the chance of win-ning an auction and gaining more revenue for providerswith weaker competitiveness In the meantime the users cangain cheaper resources In DPAMmany providers with weakcompetitiveness use dynamic pricing strategy to increasewinning chances and revenue so that resource utilizationof the cloud market increases Meanwhile cloud workflowswill be executed timely with less monetary cost In ourexperiment the measurement of TC is employed to evaluatethe performance of our proposed strategy

In our previous preliminary work a dynamic pricingstrategy in reserve auction was presented to change resourceprices according to the trading situation and then a novelDPAM was proposed to improve the efficiency of resourceutilization [14] Based on this work we have further inves-tigated the resource allocation based on reverse auction andmade the following substantial extensions in this paper

(i) In problem analysis a real world stock trading work-flow is given and resource allocation processes ofBOSS and DPAM are described based on this exam-ple The process illustrates the difference of pricingmechanism between BOSS and DPAM

(ii) In evaluation firstly the performance of BOSS andDPAM on resource utilization and the measurementof TC with different problem sizes are tested Thensimulation on DPAM with different price reductionrates evaluates its performance on resource utilizationand TC

Dynamic pricing based allocationmechanism is designedto improve resource utilization and decrease monetary costand completion time In summary this mechanism has threemajor advantages

(i) In reverse auction dynamic pricing strategy is moreeffective to increase the revenue of providers withweak competitiveness and decrease usersrsquo cost thanfixed pricing Providers with weak competitivenessdecrease the price to increase the chance of winningauction and gaining more revenue Simultaneouslyusers who choose the resource with lower price willhave less monetary cost

(ii) Dynamic pricing based allocation mechanism canimprove resource utilization for providers Providerschange resource prices to sell more resources accord-ing to dynamic pricing strategy especially for thosewithweak competitivenessThis brings higher resour-ces utilization because more resources are chosen

(iii) Dynamic pricing based allocation mechanism candecrease monetary cost and completion time forusers More price-competitive resources will appearin the market because of increasing competitionamong providers So users can easily obtain cheaperand more resources and hence decrease monetarycost and completion time

The rest of the paper is organized as follows Section 2describes some related research In Section 3 we showan example to analyze the problem Section 4 proposes adynamic pricing strategy and Section 5 presents dynamicpricing-based allocationmechanism In Section 6 traditionalmechanism and ours are evaluated Finally Section 7 con-cludes the paper and discusses our future work

2 Related Work

Resource allocation has become an important task to pro-vide efficient and economical resources in cloud computingenvironment [15] Wood et al [16] propose an approach fordynamic allocation of resources by defining a unique metricbased on the consumption of the three resources CPU net-work andmemory Gorlach and Leymann propose amethodfor dynamic provisioning of services in clouds in order tooptimize the distribution of services [17] However theirproposed approaches are not efficient and economic becausethe allocation of resource does not consider market situation

Market-oriented resource allocation has received muchattention as it is a significant problem of large-scale dis-tributed systems In [18] authors present amodel for resourceallocation in grid using market-oriented concepts includingcommodity market posted price modeling and contract netmodels bargaining modeling Mao and Humphrey explorethe cloud autoscaling framework for resource allocationThegoal is to ensure that all jobs finish before their respectivedeadlines while running on these resources which consumethe least amount of money [19] In [20] in order to effectivelymanage resource allocation and workflow execution Wanget al design a mechanism which responds to the userrsquos

Scientific Programming 3

continuous workflow requests and schedules their executionIn [21 22] a lot of heuristic methods are presented to solvethe problem in services systems These heuristic methodsconsider the optimization algorithm from the aspects of costdeadline and reliability which can improve the performanceof the algorithm Ludwig presents a heuristic program forresource allocation on utility computing infrastructure Thisheuristic program optimizes the number of resources allo-cated to tasks of workflow and speeds up the executionwithina limitation of budget [23] In [24] the authors propose aresource allocation approach to better match the resourceallocated to the job with the cloudrsquos residual resource

Auction [25 26] is a popular method to solve resourceallocation problem In [27 28] the authors present auction-based mechanisms to determine optimal resource pricetaking into account the userrsquos budget and time constraintsPrasad G et al present a combinatorial auction mechanismto allocate multiple resources in one auction [6] Howeverthey consider the pricing model of only one seller Reverseauction is a popular auction in which the roles of buyerand seller are reversed [29] In an ordinary auction buyerscompete to obtain goods or service by offering increasinglyhigher prices [30] In the reverse auction [31] the sellerstypically decrease prices to compete against each other andobtain business from the buyer The authors employ reverseauction to select the optimal resource based on availableinformation to maximize their own profits [32] A cloudresource allocation approach based reverse auction [33] ispresented to select suitable cloud resource providers for users

However these methods do not focus on the pricingmechanism of the resource allocation Resource pricing isan important aspect in resource allocation In [34] authorsmention Commodity Market which means that the sellersset the price for merchandise and the buyers pay money toget it The price is predetermined by the seller and does notchange over time based on supply-demand relationship Butfixed pricing is not suitable for the changeablemarket of cloudresources

Dynamic pricing has gained wide attention from bothindustry and academia in the cloud computing Amazon EC2[35] has introduced a ldquospot pricingrdquo scheme where the spotprice is set according to resource supply anddemand Becausefixed pricing does not reflect the dynamic changes of supplyand demand a dynamic scheme for allocation of multiple-type resources [36] is proposed to increase the percentagesof successful buying and selling In [12] authors introducea pricing model and a truthful mechanism for schedulingsingle tasks considering two objectives completion time andmonetary cost based on reverse auction However they donot consider the competitiveness among providers whichleads to the fact that the losers may always lose auctionbecause they do not try to improve their competitivenessTo solve this problem we propose dynamic pricing basedscheduling mechanism for allocating resources efficiently inwhich providers who lose an auction will decrease resourceprice in order to win so as to gain more revenue

A stock issuanceB fixed auction

C continuous auctionD formation of price chart

DA

B

C

Figure 1 Stock trading workflow

Table 1 Characteristic of tasks

Task WorkloadA 6B 4C 5D 7

Table 2 Characteristics of resources

RN CA RP SP1 1 009 0142 14 014 0203 12 012 017RN resource number CA computation ability RP reserve price and SPstarting price

3 Problem Analysis

In this section a stock trading workflow is given to explainthe difference of pricing mechanism between the BOSSmechanism and ours Stock trading is a typical process in themarket At first stock exchange starts with stock issuance andthen price formation process follows During this processfixed auction and continuous auction happen simultaneouslyAt last price chart is generated

As shown in Figure 1 stock trading workflow containsfour tasks The execution sequence of tasks partially dependson relation The succeeding tasks start only when their pre-decessor tasks finish Table 1 indicates the workload of eachtask Table 2 shows some characteristics of three resourcesincluding resource number computation ability reserveprice and starting price Here computation ability is thecomputation speed of CPUs the reserve price is the lowestprice of resource during the auction and the starting price isthe first resource price when auction starts

31 Resource Allocation Process of DPAM For the BOSSmechanism [12] tasks start an auction according to the spe-cific order to select the resource with the minimum productof completion time and total monetary cost And providersgive their bids 119887119894119889119895 = (CA119895 SP119895) which indicate computa-tion ability and starting price of resource 119895 The workflow


Table 3 Resource allocation process of BOSS

RN ST ET CT MC TC WR(a) First auction

1 0 600 600 084 50422 0 429 429 086 367

3 0 500 500 085 425(b) Second auction

1 429 400 829 056 46422 429 286 715 057 408

3 429 333 762 057 432(c) Third auction

1 429 500 929 070 65032 715 357 1072 071 766

3 429 417 846 071 599(d) Last auction

1 846 700 1546 098 151522 846 500 1346 100 1346

3 846 583 1429 099 1417RN resource number ST start time ET execution time CT completiontime MC monetary cost TC Time lowast cost and WR winner

completion time is the time required for executing the wholeworkflow under the partially ordered relation of tasks Thetotal monetary cost of workflow is the sum of all tasksrsquo mon-etary costThe taskrsquos monetary cost only covers the executioncost on the allocated resource It is assumed that there isno additional cost while moving execution from one cloudprovider to another Once a task is assigned to one resourceit will be executed on this resource until its completion andcannot be reallocated to a cheaper or better resource duringits execution The resource allocation process of BOSS isshown in Table 3

In Table 3 the start time is the larger one between theresourcersquos free time and predecessor tasksrsquo finish time Theexecution time is calculated as the workload divided by com-putation ability The completion time is the sum of the starttime and the execution time The measurement of TC is theproduct of completion time and monetary cost Tasks selectthe resource with the minimum TC In Table 3(a) task Aselects resource 2 as the winner because its TC is the mini-mum Similarly task B selects resource 2 as the winner in thesecond auction and task C selects resource 3 as the winner inthe third auction In the last auction task D selects resource 2as the winner From the tables the completion time is 1346which is the completion time of last task D on resource 2Thetotal monetary cost of all tasks is 086 + 057 + 071 + 100 =314The product of completion time and total monetary costis 422644

32 Resource Allocation Process of DPAM In ourmechanismthe resource price can be dynamically changed to improveprovidersrsquo competitiveness If one provider wins an auctionresource price will be kept unchanged Otherwise resourceprice will be decreased at a certain rate in the next auctionHowever during any auction the resource price cannot be

Table 4 Resource allocation process of DPAM

RN CP ST ET CT MC TC WR(a) First auction

1 014 000 600 600 084 50422 020 000 429 429 086 367

3 017 000 500 500 085 425(b) Second auction

1 011 429 400 829 045 37132 020 429 286 715 057 408

3 014 429 333 762 045 346(c) Third auction

1 009 429 500 929 045 41612 016 429 357 786 057 449

3 014 762 417 1179 057 668(d) Last auction

1 009 929 700 1629 063 102222 014 929 500 1429 070 1000

3 012 929 583 1512 070 1059RN resource number CP current price ST start time ET execution timeCT completion time MC monetary cost TC Timelowast cost andWR winner

lower than its reserve price Resource allocation process ofDPAM is shown in Table 4 Here the price reduction rate isset as 20

At first task A starts an auction and three resources givetheir bids (CACP) As shown in Table 4(a) the currentprice is the resource price of current auction Task A selectsresource 2 as the winner because its TC is the minimum Soresource 1 and resource 3 lose the auction and they decreasethe current prices by 20 inTable 4(b) In the second auctiontask B selects resource 3 as the winner Then resource 1 andresource 2 decrease their current price by 20 as shown inTable 4(c) Similarly task C selects resource 1 as the winnerand task D selects resource 2 From Table 4 the completiontime of the workflow is 1429 which is the completion time oflast task The total monetary cost of all tasks is 086 + 045 +045 + 070 = 246 The product of completion time and totalmonetary cost is 351534

33 Comparison of BOSS and DPAM In this subsectionresource prices and the winner of each auction are comparedbetween BOSS and DPAM as shown in Table 5

In Table 5 the winners of BOSS are 2 2 3 and 2 But thewinners of DPAM are 2 3 1 and 2 In BOSS resource price isalways fixed and resource 1 with low competitiveness is neverused However in DPAM resource 1 becomes new winner bydynamic pricing strategy which brings higher resource uti-lization Resource utilization shows the allocation of tasks onresources (see Formula (2)) Therefore resource utilizationof DPAM is 1[(2 minus 43)2 + (1 minus 43)2 + (1 minus 43)2]3 =92 which is bigger than resource utilization of BOSS (1[(3 minus 43)2 + (1 minus 43)2 + (0 minus 43)2]3 = 914) The reasonis that these providers which never become winner in BOSSmay win the auction in DPAM This means more providerswin the auction and sell their resources


429

715

134

6

Reso

urce

1

2

3

846

T1 T2 T4

T3

Completion time

(a) Gantt chart of BOSS

Completion time

Reso

urce

1

2

3

429

762

929

T3

T1 T4

T2

142

9

(b) Gantt chart of DPAM

Figure 2 Gantt charts of resource allocation and task execution

Table 5 Comparison of BOSS and DPAM

RN PB PD WB WD(a) First auction

1 014 0142 22 020 020

3 017 017(b) Second auction

1 014 0112 32 020 020

3 017 014(c) Third auction

1 014 0093 12 020 016

3 017 014(d) Last auction

1 014 0092 22 020 014

3 017 012RN resource number PB price of BOSS PD price of DPAM WB winnerof BOSS and WD winner of DPAM

The Gantt charts of BOSS and DPAM are depicted in Fig-ure 2The charts show tasks execution order and the resourceexecutes on which task In Figure 2(a) only resources 2 and3 are used While in Figure 2(b) all resources are used Itis easy to draw that DPAM has higher resource utilizationthan BOSS In DPAM resource price is dynamic so resourcewith weak competitiveness decreases price to improve com-petitiveness until it wins one auction However in BOSS theresource with low competiveness may never win any auction

4 Dynamic Pricing Strategy

As the number of cloud resource providers increases inreverse auction they compete against each other tomaximizetheir revenue So an effective pricing strategy is necessary

for providers to increase their competitiveness Firstly twopropositions are described to prove that the dynamic pricingstrategy can improve the revenue of providers and alsodecrease the monetary cost of users Then dynamic pricingstrategy is proposed

Proposition 1 Dynamic pricing strategy can increase therevenue of provider with weak competitiveness

Proof Assume that provider A and provider B have theresources with same computation ability Resource price of Ais119901A and resource price of B is119901B where119901A lt 119901B So providerB will lose auction because his competitiveness is weakerthan A If the resource price is fixed competitiveness of B isalways weaker than A and then provider B would never winan auction Otherwise if resource price is dynamic providerB can decrease the price from119901B to1199011015840B which is lower than119901AThen B can win auction and hence increase revenue becauseits competitiveness is higher than that of A Hence thedynamic pricing strategy can increase the revenue of providerwith weak competitiveness

Proposition 2 Dynamic pricing strategy can decrease userrsquosmonetary cost

Proof Assume that provider A and provider B have theresources with same computation ability Resource price ofA is 119901A and resource price of B is 119901B where 119901A lt 119901B Userwill select Arsquos resource because its resource price is lower Ifresource price is fixed user will always select Arsquos resource andmonetary cost is 119901A Otherwise if resource price is dynamicprovider B must decrease the price from 119901B to 1199011015840B to winthe auction Here 1199011015840B is lower than 119901A So user will select Brsquosresource and monetary cost is 1199011015840B Hence dynamic pricingstrategy can decrease userrsquos monetary cost

Each provider sets reserve price starting price and pricereduction rate for a resource When one task starts auctionproviders join the auction and give their bids with computa-tion ability and price After one auction finishes providerschange or do not change the resource price according to


transaction situation If providers want to increase com-petitiveness and win auctions they will change their priceaccording to the dynamic pricing strategy in Formula (1)

119901cur1015840A

=

119901curA if A is winner

119901curA sdot (1 minus 120574) if A is loser and 119901cur

A sdot (1 minus 120574) gt 119901resA

119901resA if A is loser and 119901cur

A sdot (1 minus 120574) lt 119901resA

(1)

where 119901curA refers to the current resource price of provider A119901res

A refers to the reserve price of resource 120574 denotes the pricereduction rate In the strategy if provider A is the winner itsresource price will still be 119901cur

A in the next auction Otherwiseif provider A is the loser and 119901cur

A sdot (1 minus 120574) gt 119901resA its resource

price will be 119901curA sdot (1 minus 120574) in the next auction The resource

price will be 119901resA if 119901cur


In conclusion dynamic pricing strategy is efficient duringthe auction Providers decrease the prices to increase thechance of winning auctions in order to gain more revenueSimultaneously users choose the resources with lower pricesand spend less monetary cost

5 Dynamic Pricing Based AllocationMechanism (DPAM)

In this section firstly resource utilization (Formula (2))and evaluation value TC (Formula (3)) are defined andthen novel dynamic pricing based allocation mechanism isproposed In the auction-based cloud market the purposeof providers is to sell resources at the most proper price soas to gain the highest revenue And the purpose of usersis to execute workflows with shortest completion time andlowestmonetary cost In thismechanism users select the bestresource according to the product of completion time andmonetary cost And the provider with the minimum productwill be the winner After each auction providers change theirresource price according to current trading situation If theircompetitiveness is weak and loses the auction they usuallydecrease the price in certain rate to increase competitivenessOtherwise if they win it is effectively to keep the priceunchanged or increased

Resource utilization shows the allocation equilibrium oftasks on resources It is described by variance of winning auc-tion times for each provider Resource utilization is inverselyproportional to variance Especiallywhen the variance is zeroresource utilization is optimal

119877119890119904119900119906119903119888119890119880119905119894119897119894119911119886119905119894119900119899 = 1sum1198991 (119899119906119898119895 minus 119899119906119898)2119899 (2)

where 119899 is the amount of resources 119899119906119898119895 refers to winningtimes of resource 119895 and 119899119906119898 is the average value of winningauction times of all providers

During auction a task selects the resource with theminimumTCas thewinner TC119894119895 is the product of completiontime andmonetary cost of task 119894 on resource 119895 In [12] authorsuse the measurement TC to measure the BOSS with other

mechanisms There are two reasons for using TC as a mea-surement (1) it presents the whole evaluation of completiontime and monetary cost for workflow execution (2) thetruthfulness of the BOSS mechanism depends on TC So weuse the measurement TC in order to make a more accuratecomparison with BOSS

TC119894119895 = (119905119894119895 + 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 )

lowast (119901119903119894119888119890119895 lowast 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 ) (3)

Each task starts execution only when its predecessor taskshave finished according to the partially ordered relation oftasks 119905119894119895 refers to the start time of task 119894 executing on resource119895 It equals the latest time when its predecessor tasks havefinished and simultaneously resource 119895 is idle 119908119900119903119896119897119900119886119889119894 isthe workload of task 119894 119886119887119894119897119894119905119910119895 and 119901119903119894119888119890119895 are computationability and price per time unit of resource 119895 respectively So119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 is the time required for task 119894 on resource119895 And (119905119894119895 + 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895) refers to the finishing timeof task 119894 on resource 119895 (119901119903119894119888119890119895 lowast 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895) is themonetary cost required for task 119894

In Algorithm 1 there are 119899 tasks and 119898 resources (lines(1)-(2)) Each user starts an auction in order and calculatesthe product of completion time and monetary cost forevery resource (lines (3)ndash(10)) and then selects the resourcewith the minimum product (line (11)) Then user pays tothe winner (line (12)) At last all providers change priceaccording to dynamic pricing strategy and join the nextauction (line (13))

When workflows are submitted and tasks start auctionsproviders give their bids to compete for the opportunity ofproviding resources In the auction providers change priceand increase chances of selling resource so that they can gainmore revenue and higher resource utilization In additionusers always select the optimal resource so the product of thecompletion time and monetary cost of executing tasks is theminimum

6 Evaluation

In this section experiments are conducted for evaluation ofthe performance of BOSS and DPAM on resource utilizationand the measurement of TC with different situations andproblem sizes Moreover the performance of DPAM onresource utilization (see Formula (2)) and TC with differentprice reduction rates is verified Firstly experiment setup isgiven (see Section 61) Secondly simulation of specific work-flow is described for evaluating BOSS and DPAM (see Sec-tion 62) Thirdly we conduct experiments with the mediumproblem size and evaluate BOSS and DPAM with differentsituations and the performance of DPAMwith different pricereduction rates (see Section 63) At last both from differentsituations and different problem sizes simulation resultsshow the performance of BOSS and DPAM and the per-formance of DPAM with different price reduction rates (seeSection 64)


Input workflows and resourcesOutput allocation of tasks on resources(1) 119905119886119904119896119904 larr [119899] lowast Assign the tasks to 119905119886119904119896119904 list with partially relation lowast(2) 119903119890119904119900119906119903119888119890119904 larr [119898] lowast Assign the resources to 119903119890119904119900119906119903119888119890119904 list lowast(3) 119894 = 1(4) While 119894 le 119899 do(5) 119905119890119898119901119879119886119904119896 larr 119894119905ℎ 119905119886119904119896(6) 119895 = 1(7) While 119895 le 119898 do(8) 119905119890119898119901119877119890119904119900119906119903119888119890 larr 119895119905ℎ 119903119890119904119900119906119903119888119890119904(9) 119879119862119904 larr 119862119886119897119888119906119897119886119905119890119879119862119904(119905119890119898119901119879119886119904119896 119905119890119898119901119877119890119904119900119906119903119888119890)

lowast calculate TC of 119905119890119898119901119879119886119904119896 on 119905119890119898119901119877119890119904119900119906119903119888119890 (Formula (3)) lowast(10) End(11) 119908119894119899119899119890119903 larr 119903119890119904119900119906119903119888119890119882119894119905ℎ119872119894119899119879119862(119879119862119904)

lowast select the optimal resource with the minimum TC lowast(12) 119905119890119898119901119879119886119904119896 pays to 119908119894119899119899119890119903(13) all providers 119888ℎ119886119899119892119890119875119903119894119888119890

lowast change resource price (Formula (1)) lowast(14) End

Algorithm 1 Dynamic pricing based allocation mechanism

Table 6 Problem size classification

Small Medium Large1 le 119899 le 40 50 le 119899 le 100 200 le 119899 le 3001 le 119898 le 10 10 le 119898 le 50 80 le 119898 le 120

61 Experiment Setup The simulation environment runs ona PC with the following configurations 2 CPU cores 4GBRAM and Microsoft Windows 7 OS The workflows areclassified into three situations balanced semibalanced andunbalanced [12] Task workload follows normal distribution119873(1000000 1000)The resource ability is set from 200 to 1200with an arithmetic sequence and the common difference isquotient of 1000 divided by task amountThe resource price isset from the real Amazon Web Services price (httpsawsamazoncom) In BOSS resource price is set from 014 to084 per time unit In DPAM to implement dynamic pricingstrategy all resources have starting prices and reserve pricesThe starting price is set from 014 to 084 and reserve price isset from 01 to 06 respectively

In simulation of specific cloud workflows the workflowhas 10 tasks and the amount of resources is 7 The pricereduction rate for DPAM is 10 In simulation of generalcloud workflows they are classified into small medium andlarge by problem size besides different situations Problemsize classification is shown in Table 6 where 119899 is amount oftasks and 119898 is amount of resources In addition the pricereduction rate is set from 0 to 1 in step of 01

62 Simulation of Specific Workflows In specific experimentspecific workflows are used to verify whether DPAM per-forms better than BOSS on resource utilization and TC

As shown in Figure 3(a) resource utilization of DPAMis always higher than that of BOSS DPAM can improveresource utilization compared with BOSS This is because

providers with low competitiveness change their resourceprices and then these resources have more chances to be sold

In Figure 3(b) three different situations of TCs of DPAMare all lower than those of BOSS This means that it takesshorter time and lower monetary cost for workflow execu-tion In DPAM providers decrease their resource prices toimprove the competitiveness So users can get the resourcewith shorter completion time or lower monetary cost

63 Simulation of General Workflows with Different BalancedSituations In this section two experiments are conductedon general workflows with different balanced situations Theproblem size is medium The first experiment simulatesBOSS and DPAM to evaluate their performance on recourseutilization and TC (see Figure 4) The second experimentsimulates DPAM with different price reduction rates toevaluate its performance on resource utilization and TC (seeFigure 5)

631 Resource Utilization and TC of BOSS versus DPAMAs shown in Figure 4(a) resource utilization of DPAM isalways higher thanBOSSThis indicates thatDPAMperformsbetter in resource utilization In DPAM more resources aresold by changing prices especially for resources with lowercompetitiveness These resources are never sold in BOSSFigure 4(b) shows that TC of DPAM is lower than thatof BOSS DPAM brings shorter completion time and lowermonetary cost The reason is that resource price is dynamicand then there are more resources with higher computationability and lower price

632 Resource Utilization and TC of DPAM with DifferentPrice Reduction Rates Figure 5 shows resource utilizationand TC of DPAM with different price reduction ratesIn Figure 5(a) resource utilization is constant when price


0

10

20

30

Balanced Semibalanced Unbalanced

Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM

(a) Resource utilization

0

2

4

6

Balanced Semibalanced UnbalancedBalanced situation

TC(times1012)

BOSSDPAM

(b) TC

Figure 3 Resource utilization and TC of BOSS versus DPAM for specific workflow

0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC

Figure 4 Resource utilization and TC of BOSS versus DPAM for general workflows

00

20

40

0 02 04 06 08 1

Reso

urce

util

izat

ion

()

Price reduction rate

SemibalancedUnbalancedBalanced (times01)



0123456

0 02 04 06 08 1

TC(times1012)

Balanced SemibalancedUnbalanced

(b) TC

Figure 5 Resource utilization and TC of DPAM with different price reduction rates


Small Medium LargeProblem size

0

10

20

30Re

sour

ce u

tiliz

atio

n (

)

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM

Large (times01)MediumSmall (times1000)

(b) TC

Figure 6 Resource utilization and TC of BOSS versus DPAM in balanced situation


0

10

5

20

15

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6TC

(times1012)

BOSSDPAM


(b) TC

Figure 7 Resource utilization and TC of BOSS versus DPAM in semibalanced situation

reduction rate is bigger than 02 This is because the resourceprice is equal to the reserve price when price reduction rate ishigh enough As shown in Figure 5(b) TC of workflows withall situations decreases when price reduction rate is not zeroIt is easy to draw that DPAM is better than BOSS

64 Simulation of General Workflows with Different ProblemSizes In this subsection another two sets of experimentsconducted on general workflows with different problemsizes and balanced situations are described The first set ofexperiments simulates BOSS and DPAM to evaluate theirperformance on recourse utilization and TC (see Figures6ndash8) The second set of experiments simulates DPAM withdifferent price reduction rates to evaluate the performance onresource utilization and TC (see Figures 9ndash11)

641 Resource Utilization and TC of BOSS versus DPAMFigures 6ndash8 present the performance of BOSS and DPAMon resource utilization and TC from different balancedsituations and different problem size In Figures 6(a) 7(a)and 8(a) resource utilization of balanced workflow is higherthat of unbalanced workflow This is because more tasks

in balanced workflow are executed in parallel and manyresources are used In three situations resource utilizationsof DPAM are all higher than that of BOSS Figures 6(b) 7(b)and 8(b) show that TC of DPAM is always lower than thatof BOSS The reason is that the resource with lower price orhigher computation ability is selected as winner

642 Resource Utilization and TC of DPAM with DifferentPrice Reduction Rates Figures 9(a) 10(a) and 11(a) show thatresource utilization changes only when price reduction rate islower than 03 This indicates that it is not necessary to makeprice reduction rate too highThe reason is that resource pricecannot be smaller than reserve price Figures 9(b) 10(b) and11(b) show that TC decreases when resource price reduces insome rates AndTCof large problem sizeworkflows decreasesapparently than other sizes This is because dynamic pricingbrings more competitive resources with lower price andhigher ability

In overall terms the performance of DPAM on resourceutilization and TC with different situations is better thanBOSS shown in Figure 4 The performance of DPAM onresource utilization and TC with different problem sizes is



00

05

15

10

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC

Figure 8 Resource utilization and TC of BOSS versus DPAM in unbalanced situation

00

100

300

200

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC

Figure 9 Resource utilization and TC with different rates in balanced situation

00

05

20

15

10

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


LargeSmallMedium (times10)



0

5

10

15

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC

Figure 10 Resource utilization and TC with different rates in semibalanced situation


00

05

10

15

20

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

10

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC

Figure 11 Resource utilization and TC with different rates in unbalanced situation

shown in Figures 6ndash8 In DPAM many providers with weakcompetitiveness use dynamic pricing strategy to increasechances of making a deal and gain more revenue so resourceutilization of market increases Meanwhile workflows canexecute timely with less cost So the performance of DPAMon resource utilization and TC is better than that of BOSSMoreover the performance of DPAM on resource utilizationand TC with different price reduction rates is shown inFigures 5 and 9ndash11 Resource utilization and TC are invariantwhen price reduction rate is higher than 02 This is becauseresource price cannot be lower than the reserve price Inaddition performance of TC and resource utilization isalways better when price reduction rate is bigger than zero

7 Conclusion and Future Work

In this paper we proposed a dynamic pricing strategy toimprove resource providersrsquo competitiveness in the cloudmarket A novel dynamic pricing based allocation mecha-nismwas presented to allocate resources for cloudworkflowsWith our mechanism resource providers can change theprice to increase the possibility of selling resources and gainmore revenue which improves resources utilization Theusers select the best resource with the minimum TC (Time lowastCost) which ensures shorter completion time and lowermonetary cost Finally we evaluated our mechanism andcompared with the representative BOSS strategy The resultsshowed that our mechanism can achieve high resources uti-lization shorter completion time and lower monetary costWith the dynamic pricing strategy providers can decreasetheir resource price to improve competitiveness

In future increasing price will be involved in dynamicpricing strategy It is a good way for those resource providerswho have sharply higher competitiveness to increase price togain more revenue At the same time we will use the stan-dard scientific datasets to run experiments besides randomdata This will increase the credibility of the results of theexperiment and be more scientific to reflect the performanceof the DPAM mechanism In addition besides completiontime and monetary cost we will consider adding other QoS

criteria such as reliability response time and service provid-ersrsquo reputation

Competing Interests

There is no conflict of interests related to this paper

Acknowledgments

This work is partially supported by Natural Science Founda-tion of China under nos 61672034 61300042 and 61300169MOE Project of Humanities and Social Sciences under no16YJCZH048 and the Key Natural Science Foundation ofEducation Bureau of Anhui Province Project KJ2016A024The authors are grateful for Professor Yun Yang from Swin-burneUniversity of Technology Australia for providing con-structive feedback to improve this paperThe price reductionrate is set by empirical knowledgeTherefore the rational ratedeserved to be researched

References

[1] J Wang M AbdelBaky J Diaz-Montes S Purawat MParashar and I Altintas ldquoKepler + cometcloud dynamic scien-tific workflow execution on federated cloud resourcesrdquo ProcediaComputer Science vol 80 pp 700ndash711 2016

[2] G Juve and E Deelman ldquoScientific workflows and cloudsrdquoCrossroads vol 16 no 3 pp 14ndash18 2010

[3] A Prasad PGreen and JHeales ldquoOn governance structures forthe cloud computing services and assessing their effectivenessrdquoInternational Journal of Accounting Information Systems vol 15no 4 pp 335ndash356 2014

[4] C Lin and S Lu ldquoScheduling scientific workflows elastically forcloud computingrdquo in Proceedings of the 2011 IEEE 4th Interna-tional Conference on Cloud Computing (CLOUD rsquo11) pp 746ndash747 Washington DC USA July 2011

[5] T T Huu and C K Tham ldquoAn auction-based resource alloca-tion model for green cloud computingrdquo in Proceedings of theIEEE International Conference on Cloud Engineering (IC2E rsquo13)pp 269ndash278 San Francisco Calif USA March 2013


[6] V Prasad G S Rao and A S Prasad ldquoA combinatorial auc-tion mechanism for multiple resource procurement in cloudcomputingrdquo in Proceedings of the 12th International Conferenceon Intelligent Systems Design and Applications (ISDA rsquo12) pp337ndash344 Kochi India November 2012

[7] M A Rahman and R M Rahman ldquoCAPMAuction reputationindexed auction model for resource allocation in Grid com-putingrdquo in Proceedings of the 7th International Conference onElectrical and Computer Engineering (ICECE rsquo12) pp 651ndash654IEEE Dhaka Bangladesh December 2012

[8] XWeng XWang C-LWang K Li andMHuang ldquoResourceallocation in cloud environment a model based on doublemulti-attribute auction mechanismrdquo in Proceedings of the 6thIEEE International Conference on Cloud Computing Technologyand Science (CloudCom rsquo14) pp 599ndash604 December 2014

[9] C N Boyer and B W Brorsen ldquoImplications of a reserve pricein an agent-based common-value auctionrdquo Computational Eco-nomics vol 43 no 1 pp 33ndash51 2014

[10] H Qu I O Ryzhov and M C Fu ldquoLearning logistic demandcurves in business-to-business pricingrdquo in Proceedings of the43rd Winter Simulation Conference Simulation Making Deci-sions in a Complex World (WSC rsquo13) pp 29ndash40 WashingtonDC USA December 2013

[11] A S Prasad and S Rao ldquoA mechanism design approach toresource procurement in cloud computingrdquo IEEE Transactionson Computers vol 63 no 1 pp 17ndash30 2014

[12] H M Fard R Prodan and T Fahringer ldquoA truthful dynamicworkflow scheduling mechanism for commercial multicloudenvironmentsrdquo IEEE Transactions on Parallel and DistributedSystems vol 24 no 6 pp 1203ndash1212 2013

[13] B Sharma R K Thulasiram P Thulasiraman S K Garg andR Buyya ldquoPricing cloud compute commodities a novel finan-cial economic modelrdquo in Proceedings of the 12th IEEEACMInternational Symposium on Cluster Cloud and Grid Computing(CCGrid rsquo12) pp 451ndash457 IEEE Ottawa Canada May 2012

[14] X Li X Liu and E Zhu ldquoAn efficient resource allocationmechanism based on dynamic pricing reverse auction for cloudworkflow systemsrdquo in Proceedings of the Asia-Pacific Conferenceon Business Process Management pp 59ndash69 2015

[15] H Xu and B Li ldquoResource allocation with flexible channelcooperation in cognitive radio networksrdquo IEEE Transactions onMobile Computing vol 12 no 5 pp 957ndash970 2013

[16] T Wood P J Shenoy A Venkataramani and M S YousifldquoBlack-box and gray-box strategies for virtual machine migra-tionrdquo in Proceedings of the 4th USENIX Conference on Net-worked Systems Design amp Implementation pp 229ndash242 2007

[17] K Gorlach and F Leymann ldquoDynamic service provisioning forthe cloudrdquo in Proceedings of the IEEE 9th International Confer-ence on Services Computing (SCC rsquo12) pp 555ndash561 June 2012

[18] X Shi and Y Zhao ldquoDynamic resource scheduling and work-flow management in cloud computingrdquo in Proceedings of theInternational Conference on Web Information Systems Engineer-ing pp 440ndash448 2010

[19] M Mao andM Humphrey ldquoAuto-scaling to minimize cost andmeet application deadlines in cloud workflowsrdquo in Proceedingsof the International Conference for High Performance Comput-ing Networking Storage and Analysis (SC rsquo11) pp 1ndash12 ACMSeattle Wash USA November 2011

[20] J Wang P Korambath I Altintas J Davis and D CrawlldquoWorkflow as a service in the cloud architecture and scheduling

algorithmsrdquo Procedia Computer Science vol 29 pp 546ndash5562014

[21] L Wang J Shen and J Yong ldquoA survey on bio-inspired algo-rithms for web service compositionrdquo in Proceedings of the 2012IEEE 16th International Conference on Computer SupportedCooperativeWork in Design (CSCWD rsquo12) pp 569ndash574WuhanChina May 2012

[22] L Wang and J Shen ldquoMulti-phase ant colony system for multi-party data-intensive service provisionrdquo IEEE Transactions onServices Computing vol 9 no 2 pp 264ndash276 2016

[23] S A Ludwig ldquoParticle swarmoptimization approachwith para-meter-wise hill-climbing heuristic for task allocation of work-flow applications on the cloudrdquo in Proceedings of the 25th IEEEInternational Conference on Tools with Artificial Intelligence(ICTAI rsquo13) pp 201ndash206 IEEE Herndon Va USA November2013

[24] D Li C Chen J Guan Y Zhang J Zhu and R Yu ldquoDClouddeadline-aware resource allocation for cloud computing jobsrdquoIEEE Transactions on Parallel and Distributed Systems vol 27no 8 pp 2248ndash2260 2016

[25] H Wang Z Kang and L Wang ldquoPerformance-aware cloudresource allocation via fitness-enabled auctionrdquo IEEE Transac-tions on Parallel and Distributed Systems vol 27 no 4 pp 1160ndash1173 2016

[26] M M Nejad L Mashayekhy and D Grosu ldquoTruthful greedymechanisms for dynamic virtual machine provisioning andallocation in cloudsrdquo IEEE Transactions on Parallel and Dis-tributed Systems vol 26 no 2 pp 594ndash603 2015

[27] F Teng and F Magoules ldquoResource pricing and equilibriumallocation policy in cloud computingrdquo in Proceedings of the 10thIEEE International Conference on Computer and InformationTechnology pp 195ndash202 2010

[28] M Mihailescu and Y M Teo ldquoOn economic and computa-tional-efficient resource pricing in large distributed systemsrdquo inProceedings of the 10th IEEEACM International Symposium onCluster Cloud and Grid Computing pp 838ndash843 MelbourneAustralia May 2010

[29] L Pham J Teich H Wallenius and J Wallenius ldquoMulti-attri-bute online reverse auctions recent research trendsrdquo EuropeanJournal of Operational Research vol 242 no 1 pp 1ndash9 2015

[30] M Takeda D Takahashi andM Shobayashi ldquoCollective actionvs conservation auction lessons from a social experiment ofa collective auction of water conservation contracts in JapanrdquoLand Use Policy vol 46 pp 189ndash200 2015

[31] C Xu L Song Z Han et al ldquoEfficiency resource allocation fordevice-to-device underlay communication systems a reverseiterative combinatorial auction based approachrdquo IEEE Journalon Selected Areas in Communications vol 31 no 9 pp 348ndash3582013

[32] P Setia and C Speier-Pero ldquoReverse auctions to innovate pro-curement processes effects of bid information presentationdesign on a supplierrsquos bidding outcomerdquo Decision Sciences vol46 no 2 pp 333ndash366 2015

[33] J R Fooks K D Messer and J M Duke ldquoDynamic entryreverse auctions and the purchase of environmental servicesrdquoLand Economics vol 91 no 1 pp 57ndash75 2015

[34] W Depoorter K Vanmechelen and J Broeckhove ldquoAdvancereservation co-allocation and pricing of network and computa-tional resources in gridsrdquo Future Generation Computer Systemsvol 41 pp 1ndash15 2014


[35] Y Zhao Y Li I Raicu S Lu W Tian and H Liu ldquoEnablingscalable scientific workflow management in the Cloudrdquo FutureGeneration Computer Systems vol 46 pp 3ndash16 2015

[36] MMihailescu and YM Teo ldquoStrategy-proof dynamic resourcepricing of multiple resource types on federated cloudsrdquo inAlgorithms and Architectures for Parallel Processing C-H HsuL T Yang J H Park and S-S Yeo Eds vol 6081 of LectureNotes in Computer Science pp 337ndash350 Springer Berlin Ger-many 2010

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Distributed Sensor Networks


Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014


ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014


Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications


Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia


Biomedical Imaging


ArtificialNeural Systems

Advances in


RoboticsJournal of



Computational Intelligence and Neuroscience

Industrial EngineeringJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in



resources The cloud broker transfers the specification toall cloud providers The cloud providers sell resources withproper price and capabilities And then the users select theoptimal resources according to some criteria for exampleQuality of Service [11 12] In general reverse auction isused to prevent the occurrence of trading fraud and achievedynamic pricing and automatic procurement [11] In themost relevant literature [12] the authors propose BiobjectiveScheduling Strategy (BOSS) based on reverse auction toallocate resources for tasks of workflows and each task startsan auction and gets a resource to minimize the monetarycost and completion time However the resource price isfixed during the auction so that some providers with weakercompetitiveness may lose the auction all the time This leadsto low efficiency of resource utilization which decreases theallocation equilibrium of tasks on resources in the cloudmarket

In this paper we firstly present a dynamic pricing strategyto change resource prices according to the trading situationand then present a DPAM mechanism to improve theefficiency of resource utilization Changing resource price isa common and efficient way to increase competitiveness [13]In the dynamic pricing strategy the resource price changesaccording to the trading situationThose providers who havestrong competitiveness (ie higher chance of winning inhistoric auctions) in the market often keep resource pricesunchanged during an auction But for those who have weakcompetitiveness decreasing the resource price with a certainrate is an effective way to increase the winning chancesTherefore changing prices can increase the chance of win-ning an auction and gaining more revenue for providerswith weaker competitiveness In the meantime the users cangain cheaper resources In DPAMmany providers with weakcompetitiveness use dynamic pricing strategy to increasewinning chances and revenue so that resource utilizationof the cloud market increases Meanwhile cloud workflowswill be executed timely with less monetary cost In ourexperiment the measurement of TC is employed to evaluatethe performance of our proposed strategy

In our previous preliminary work a dynamic pricingstrategy in reserve auction was presented to change resourceprices according to the trading situation and then a novelDPAM was proposed to improve the efficiency of resourceutilization [14] Based on this work we have further inves-tigated the resource allocation based on reverse auction andmade the following substantial extensions in this paper

(i) In problem analysis a real world stock trading work-flow is given and resource allocation processes ofBOSS and DPAM are described based on this exam-ple The process illustrates the difference of pricingmechanism between BOSS and DPAM

(ii) In evaluation firstly the performance of BOSS andDPAM on resource utilization and the measurementof TC with different problem sizes are tested Thensimulation on DPAM with different price reductionrates evaluates its performance on resource utilizationand TC

Dynamic pricing based allocationmechanism is designedto improve resource utilization and decrease monetary costand completion time In summary this mechanism has threemajor advantages

(i) In reverse auction dynamic pricing strategy is moreeffective to increase the revenue of providers withweak competitiveness and decrease usersrsquo cost thanfixed pricing Providers with weak competitivenessdecrease the price to increase the chance of winningauction and gaining more revenue Simultaneouslyusers who choose the resource with lower price willhave less monetary cost

(ii) Dynamic pricing based allocation mechanism canimprove resource utilization for providers Providerschange resource prices to sell more resources accord-ing to dynamic pricing strategy especially for thosewithweak competitivenessThis brings higher resour-ces utilization because more resources are chosen

(iii) Dynamic pricing based allocation mechanism candecrease monetary cost and completion time forusers More price-competitive resources will appearin the market because of increasing competitionamong providers So users can easily obtain cheaperand more resources and hence decrease monetarycost and completion time

The rest of the paper is organized as follows Section 2describes some related research In Section 3 we showan example to analyze the problem Section 4 proposes adynamic pricing strategy and Section 5 presents dynamicpricing-based allocationmechanism In Section 6 traditionalmechanism and ours are evaluated Finally Section 7 con-cludes the paper and discusses our future work

2 Related Work

Resource allocation has become an important task to pro-vide efficient and economical resources in cloud computingenvironment [15] Wood et al [16] propose an approach fordynamic allocation of resources by defining a unique metricbased on the consumption of the three resources CPU net-work andmemory Gorlach and Leymann propose amethodfor dynamic provisioning of services in clouds in order tooptimize the distribution of services [17] However theirproposed approaches are not efficient and economic becausethe allocation of resource does not consider market situation

Market-oriented resource allocation has received muchattention as it is a significant problem of large-scale dis-tributed systems In [18] authors present amodel for resourceallocation in grid using market-oriented concepts includingcommodity market posted price modeling and contract netmodels bargaining modeling Mao and Humphrey explorethe cloud autoscaling framework for resource allocationThegoal is to ensure that all jobs finish before their respectivedeadlines while running on these resources which consumethe least amount of money [19] In [20] in order to effectivelymanage resource allocation and workflow execution Wanget al design a mechanism which responds to the userrsquos








DA

B

C






3 Problem Analysis







1 0 600 600 084 50422 0 429 429 086 367

3 0 500 500 085 425(b) Second auction

1 429 400 829 056 46422 429 286 715 057 408

3 429 333 762 057 432(c) Third auction

1 429 500 929 070 65032 715 357 1072 071 766

3 429 417 846 071 599(d) Last auction

1 846 700 1546 098 151522 846 500 1346 100 1346







1 014 000 600 600 084 50422 020 000 429 429 086 367

3 017 000 500 500 085 425(b) Second auction

1 011 429 400 829 045 37132 020 429 286 715 057 408

3 014 429 333 762 045 346(c) Third auction

1 009 429 500 929 045 41612 016 429 357 786 057 449

3 014 762 417 1179 057 668(d) Last auction

1 009 929 700 1629 063 102222 014 929 500 1429 070 1000







429

715

134

6

Reso

urce

1

2

3

846

T1 T2 T4

T3

Completion time


Completion time

Reso

urce

1

2

3

429

762

929

T3

T1 T4

T2

142

9





1 014 0142 22 020 020


1 014 0112 32 020 020


1 014 0093 12 020 016


1 014 0092 22 020 014













119901cur1015840A

=






(1)












119877119890119904119900119906119903119888119890119880119905119894119897119894119911119886119905119894119900119899 = 1sum1198991 (119899119906119898119895 minus 119899119906119898)2119899 (2)




TC119894119895 = (119905119894119895 + 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 )

lowast (119901119903119894119888119890119895 lowast 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 ) (3)




6 Evaluation




















0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


00

20

40

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0123456

0 02 04 06 08 1

TC(times1012)


(b) TC




0

10

20

30Re

sour

ce u

tiliz

atio

n (

)

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC



0

10

5

20

15

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6TC

(times1012)

BOSSDPAM


(b) TC










00

05

15

10

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC


00

100

300

200

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC


00

05

20

15

10

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0

5

10

15

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC



00

05

10

15

20

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

10

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC







Competing Interests


Acknowledgments


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in










DA

B

C






3 Problem Analysis







1 0 600 600 084 50422 0 429 429 086 367

3 0 500 500 085 425(b) Second auction

1 429 400 829 056 46422 429 286 715 057 408

3 429 333 762 057 432(c) Third auction

1 429 500 929 070 65032 715 357 1072 071 766

3 429 417 846 071 599(d) Last auction

1 846 700 1546 098 151522 846 500 1346 100 1346







1 014 000 600 600 084 50422 020 000 429 429 086 367

3 017 000 500 500 085 425(b) Second auction

1 011 429 400 829 045 37132 020 429 286 715 057 408

3 014 429 333 762 045 346(c) Third auction

1 009 429 500 929 045 41612 016 429 357 786 057 449

3 014 762 417 1179 057 668(d) Last auction

1 009 929 700 1629 063 102222 014 929 500 1429 070 1000







429

715

134

6

Reso

urce

1

2

3

846

T1 T2 T4

T3

Completion time


Completion time

Reso

urce

1

2

3

429

762

929

T3

T1 T4

T2

142

9





1 014 0142 22 020 020


1 014 0112 32 020 020


1 014 0093 12 020 016


1 014 0092 22 020 014













119901cur1015840A

=






(1)












119877119890119904119900119906119903119888119890119880119905119894119897119894119911119886119905119894119900119899 = 1sum1198991 (119899119906119898119895 minus 119899119906119898)2119899 (2)




TC119894119895 = (119905119894119895 + 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 )

lowast (119901119903119894119888119890119895 lowast 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 ) (3)




6 Evaluation




















0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


00

20

40

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0123456

0 02 04 06 08 1

TC(times1012)


(b) TC




0

10

20

30Re

sour

ce u

tiliz

atio

n (

)

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC



0

10

5

20

15

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6TC

(times1012)

BOSSDPAM


(b) TC










00

05

15

10

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC


00

100

300

200

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC


00

05

20

15

10

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0

5

10

15

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC



00

05

10

15

20

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

10

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC







Competing Interests


Acknowledgments


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in






1 0 600 600 084 50422 0 429 429 086 367

3 0 500 500 085 425(b) Second auction

1 429 400 829 056 46422 429 286 715 057 408

3 429 333 762 057 432(c) Third auction

1 429 500 929 070 65032 715 357 1072 071 766

3 429 417 846 071 599(d) Last auction

1 846 700 1546 098 151522 846 500 1346 100 1346







1 014 000 600 600 084 50422 020 000 429 429 086 367

3 017 000 500 500 085 425(b) Second auction

1 011 429 400 829 045 37132 020 429 286 715 057 408

3 014 429 333 762 045 346(c) Third auction

1 009 429 500 929 045 41612 016 429 357 786 057 449

3 014 762 417 1179 057 668(d) Last auction

1 009 929 700 1629 063 102222 014 929 500 1429 070 1000







429

715

134

6

Reso

urce

1

2

3

846

T1 T2 T4

T3

Completion time


Completion time

Reso

urce

1

2

3

429

762

929

T3

T1 T4

T2

142

9





1 014 0142 22 020 020


1 014 0112 32 020 020


1 014 0093 12 020 016


1 014 0092 22 020 014













119901cur1015840A

=






(1)












119877119890119904119900119906119903119888119890119880119905119894119897119894119911119886119905119894119900119899 = 1sum1198991 (119899119906119898119895 minus 119899119906119898)2119899 (2)




TC119894119895 = (119905119894119895 + 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 )

lowast (119901119903119894119888119890119895 lowast 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 ) (3)




6 Evaluation




















0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


00

20

40

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0123456

0 02 04 06 08 1

TC(times1012)


(b) TC




0

10

20

30Re

sour

ce u

tiliz

atio

n (

)

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC



0

10

5

20

15

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6TC

(times1012)

BOSSDPAM


(b) TC










00

05

15

10

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC


00

100

300

200

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC


00

05

20

15

10

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0

5

10

15

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC



00

05

10

15

20

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

10

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC







Competing Interests


Acknowledgments


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




429

715

134

6

Reso

urce

1

2

3

846

T1 T2 T4

T3

Completion time


Completion time

Reso

urce

1

2

3

429

762

929

T3

T1 T4

T2

142

9





1 014 0142 22 020 020


1 014 0112 32 020 020


1 014 0093 12 020 016


1 014 0092 22 020 014













119901cur1015840A

=






(1)












119877119890119904119900119906119903119888119890119880119905119894119897119894119911119886119905119894119900119899 = 1sum1198991 (119899119906119898119895 minus 119899119906119898)2119899 (2)




TC119894119895 = (119905119894119895 + 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 )

lowast (119901119903119894119888119890119895 lowast 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 ) (3)




6 Evaluation




















0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


00

20

40

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0123456

0 02 04 06 08 1

TC(times1012)


(b) TC




0

10

20

30Re

sour

ce u

tiliz

atio

n (

)

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC



0

10

5

20

15

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6TC

(times1012)

BOSSDPAM


(b) TC










00

05

15

10

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC


00

100

300

200

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC


00

05

20

15

10

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0

5

10

15

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC



00

05

10

15

20

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

10

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC







Competing Interests


Acknowledgments


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





119901cur1015840A

=






(1)












119877119890119904119900119906119903119888119890119880119905119894119897119894119911119886119905119894119900119899 = 1sum1198991 (119899119906119898119895 minus 119899119906119898)2119899 (2)




TC119894119895 = (119905119894119895 + 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 )

lowast (119901119903119894119888119890119895 lowast 119908119900119903119896119897119900119886119889119894119886119887119894119897119894119905119910119895 ) (3)




6 Evaluation




















0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


00

20

40

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0123456

0 02 04 06 08 1

TC(times1012)


(b) TC




0

10

20

30Re

sour

ce u

tiliz

atio

n (

)

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC



0

10

5

20

15

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6TC

(times1012)

BOSSDPAM


(b) TC










00

05

15

10

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC


00

100

300

200

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC


00

05

20

15

10

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0

5

10

15

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC



00

05

10

15

20

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

10

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC







Competing Interests


Acknowledgments


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





















0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


0

10

20

30


Reso

urce

util

izat

ion

()

Balanced situation

BOSSDPAM


0

2

4

6


TC(times1012)

BOSSDPAM

(b) TC


00

20

40

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0123456

0 02 04 06 08 1

TC(times1012)


(b) TC




0

10

20

30Re

sour

ce u

tiliz

atio

n (

)

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC



0

10

5

20

15

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6TC

(times1012)

BOSSDPAM


(b) TC










00

05

15

10

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC


00

100

300

200

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC


00

05

20

15

10

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0

5

10

15

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC



00

05

10

15

20

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

10

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC







Competing Interests


Acknowledgments


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





0

10

20

30Re

sour

ce u

tiliz

atio

n (

)

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC



0

10

5

20

15

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6TC

(times1012)

BOSSDPAM


(b) TC










00

05

15

10

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC


00

100

300

200

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC


00

05

20

15

10

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0

5

10

15

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC



00

05

10

15

20

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

10

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC







Competing Interests


Acknowledgments


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





00

05

15

10

Reso

urce

util

izat

ion

()

BOSSDPAM


Problem size

0

2

4

8

6

TC(times1012)

BOSSDPAM


(b) TC


00

100

300

200

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC


00

05

20

15

10

0 02 04 06 08 1

Reso

urce

util

izat

ion

()





0

5

10

15

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC



00

05

10

15

20

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

10

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC







Competing Interests


Acknowledgments


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




00

05

10

15

20

0 02 04 06 08 1

Reso

urce

util

izat

ion

()


MediumLargeSmall



0

2

4

6

10

8

0 02 04 06 08 1

TC(times1012)

MediumLargeSmall

(b) TC







Competing Interests


Acknowledgments


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in












































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in



research article a dynamic pricing reverse auction-based...

Documents