A General and Practical Datacenter Selection Framework for Cloud Services

Size: px
Start display at page:

Download "A General and Practical Datacenter Selection Framework for Cloud Services"

Transcription

1 212 IEEE Ffth Internatonal Conference on Clou Computng A General an Practcal Datacenter Selecton Framework for Clou Servces Hong Xu, Baochun L henryxu, bl@eecg.toronto.eu Department of Electrcal an Computer Engneerng Unversty of Toronto Abstract Many clou servces nowaays are runnng on top of geographcally strbute nfrastructures for better relablty an performance. They nee an effectve way to rect the user requests to a sutable atacenter, epenng on factors nclung performance, cost, etc. Prevous work focuse on effcency an nvarably consere the smple objectve of maxmzng aggregate utlty. These approaches favor users closer to the nfrastructure. In ths paper, we argue that farness shoul be consere to ensure users at savantageous locatons also enjoy reasonable performance, an performance s balance across the entre system. We aopt a general farness crteron base on Nash barganng solutons, an present a general optmzaton framework that moels the realstc envronment an practcal constrants that a clou faces. We evelop an effcent strbute algorthm base on ual ecomposton an the subgraent metho, an evaluate ts effectveness an practcalty usng realworl traffc traces an electrcty prces. I. INTRODUCTION Internet-scale onlne servces are becomng ncreasngly essental to our everyay lfe, wth mportant applcatons nclung web search, veo streamng, an onlne gamng. The burgeonng of clou computng platforms, such as Amazon AWS, further enables small enterprses to raply eploy clou servces at scale. Almost all of these clou servces are bult atop geographcally strbute nfrastructures,.e. atacenters locate n fferent regons as shown n Fg. 1, to prove relablty an performance. They nee an effectve way to rect clents across the we area to an approprate atacenter. Usually, clou servces hanle atacenter selecton by eployng mappng noes, whch are typcally DNS servers, to customze the IP aress(es) returne to fferent clents. Alternatvely, they can also outsource atacenter selecton to thr-partes [1], [2] or the clou prover [3]. Clents Mappng noes (DNS) Mappng Requests Datacenters Fg. 1. An example of a clou servce runnng atop a geographcally strbute clou nfrastructure. An effcent atacenter selecton algorthm s mperatve to the operaton of clou servces. Many prevous works exst n ths area. The problem can be cast as an optmzaton that maxmzes the total utlty or mnmze the total cost, both of whch can be efne n fferent ways epenng on how the clou servces efne performance. The formulate problems an ther solutons are focuse on the effcency ssue. These approaches thus ten to favor clents closer to the nfrastructure, because prortzng them n assgnng atacenter capacty mproves system utlty. Ths results n poor performance for savantage users far away from the nfrastructure, an can potentally lea to substantal revenue losses. For nstance, Amazon reports that every 1 ms elay n page loa tme ecreases sales by 1 percent [4]. Many farness crtera have been consere n the tratonal context of traffc engneerng n wre networks, an resource allocaton n cellular networks. Max-mn an proportonal farness moels are arguably the most wely use n the lterature [5] [7]. However, n the context of clou computng, clou servces usually have stnct Servce Level Agreements (SLAs) that nee to be satsfe. Ths requrement cannot be accommoate by nether the max-mn nor the proportonal farness moel. Moreover, the max-mn approach eals wth the worse-case scenaro an penalzes clents wth better contons, thus unnecessarly reucng the system effcency. In ths paper, our man contrbuton s a general optmzaton framework for atacenter selecton base on the farness crteron stemmng from the concept of Nash barganng games n game theory [8]. Essentally, we can vew the problem as a cooperatve barganng game amongst the mappng noes. Each noe, servng requests aggregate from a specfc area, has a mnmum utlty requrement for clents t serves base on the SLA, an they compete ( bargan ) cooperatvely for the clou resources. The soluton of ths game, calle the Nash barganng solutons (NBS), s a unque Nash equlbrum pont wth NBS farness among mappng noes, whch s a generalze proportonal farness noton. It strves to satsfy the mnmum utlty requrements frst, an allocates the remanng atacenter capacty proportonally among mappng noes accorng to ther contons. Therefore t s able to acheve hgh system effcency wth goo farness. Our framework s general n the sense that t moels practcal envronments that clou operators face. The utlty /12 $ IEEE DOI 1.119/CLOUD

2 abstracton encompasses many possble performance conseratons, nclung throughput an latency, as well as polcy preferences, such as atacenter loa, user localty, etc. It s a functon of user-atacenter tuples n orer to moel the locaton versty of performance. We also conser the cost of servng requests an moel the prce versty of atacenters, snce recent works have recognze the beneft of locatonepenent electrcty prce n terms of mnmzng the energy bll of atacenters [9]. By usng ual ecomposton, our optmzaton formulaton can be ecentralze to the mappng noe level. Specfcally, the problem can be ecompose nto many subproblems, each solvable by an nvual mappng noe tself. Ths enables us to evelop effcent strbute mplementatons of our atacenter selecton algorthm to fn the optmal noeatacenter assgnment, base on the subgraent metho. We also pont out that our algorthms reman relevant n an are applcable to other request recton scenaros, such as a commercal CDN. We evaluate the effectveness an practcalty of our ecentralze mplementaton usng the real-worl traffc traces collecte from UUSee [1], a commercal Veo-on-Deman prover n Chna, as well as real-worl electrcty prces collecte from the U.S. Feeral Energy Regulatory Commsson [11]. Results show that our algorthm acheves better farness wth satsfactory performance n terms of total utlty an cost, an s amenable to practcal mplementatons, snce t converges wthn 2 teratons. The rest of the paper s structure as follows. In Sec. II we ntrouce the concept of NBS an present our formulaton of the atacenter selecton problem. In Sec. III we evelop strbute algorthms to solve the optmzaton problem base on ual ecomposton. Numercal results are prove n Sec. IV. In Sec. V we summarze relate work. Fnally, we conclue the paper n Sec. VI. II. AN OPTIMIZATION FRAMEWORK BASED ON NBS In ths secton, we present our optmzaton framework base on NBS. A. System Moel We start by ntroucng the system moel. We conser a clou nfrastructure wth M geographcally strbute atacenters. The clou eploys N mappng noes (e.g. DNS servers) at fferent locatons to rect clent requests to the approprate atacenters as etermne by the atacenter selecton algorthm. Snce the request traffc fluctuates ynamcally, the atacenter selecton algorthm has to be run perocally to optmze performance. We assume that the clou operator employs learnng technques [12], [13] to prect the traffc eman of each noe D n each epoch wth satsfactory accuracy. We also assume that the electrcty prce at each atacenter W s avalable at the begnnng of an epoch, an remans statc throughout the entre epoch. Ths s a practcal assumpton n toay s electrcty market. If the local electrcty market of atacenter s a regulate utlty regon, the electrcty prce s fxe. If on the other han the atacenter s n a eregulate market regon, such as Calforna an Texas, there s a forwar market wth settlements of varous kns, such as ay-ahea an hourahea, for customers to lock n the prce [11]. W =[W ] s calle the cost matrx. We use an abstract utlty noton U to capture the performance of the clou servce, when a request from noe s recte to atacenter. Ths noton allows us a conserable amount of expressveness. For example, f the clou servce s an nteractve applcaton an seeks mnmal latency, U can be a ecreasng functon of the roun trp tme (RTT), rectly measure or estmate by varous means. If the clou servce s a bulk transfer applcaton an seeks goo throughput, U can be a ecreasng functon of the network congeston level or the lnk utlzaton. Ths utlty functon can be any shape e.g. a convex functon of the latency or throughput. For more scusson of the generalty of the utlty noton one can refer to [2]. Fnally, U =[U ] s calle the performance matrx n the followng. B. Bascs of Nash barganng solutons We present the salent concepts an results from Nash barganng solutons here, whch are use n the sequel. For etals we refer reaers to [8]. The basc settng s as follows: Let N be the set of mappng noes. Let S be a close an convex subset of R N to represent the set of feasble utlty allocatons. Let U mn be the mnmal utlty that the -th noe requre. In our atacenter selecton problem, ths s obtane from clents SLAs. Suppose {U S U U mn, N}s a nonempty boune set. Defne U mn =(U1 mn,...,un mn), then the par (S, U mn ) s calle a N-noe barganng problem. Wthn the feasble set S, we frst efne the noton of Pareto optmalty as a selecton crteron n a typcal game. Defnton 1: The pont (U 1,...,U N ) s sa to be Pareto optmal f an only f there s no other allocaton U such that U U, N, an U >U, N. That s, there exsts no other allocaton that leas to superor performance for some noe wthout nferor performance for some other noe. Our next selecton crteron s the farness of resource sharng. In ths paper, we use the NBS farness axoms from game theory. Defnton 2: r s a NBS,.e. r = φ(s, R mn ), f the followng axoms are satsfe: Feasblty, Pareto Optmalty, Inepenence of Irrelevant Alternatves, Inepenence of Lnear Transformatons, an Symmetry [8]. Theorem 1: There s a unque soluton functon φ(s, U mn ) that satsfes all axoms n Defnton 2 such that [8] φ(s, U mn ( ) argmax U U mn ). (1) U S,U U mn N NBS farness naturally fts to the atacenter selecton problem. U mn correspon to the specfc SLA of the mappng 1

3 noe to guarantee performance of clents. After the mnmal requrements are met for all noes, the rest of the resources are allocate proportonally to noes accorng to ther contons, so that every noe obtans a far share. When U mn =for all, NBS farness reuces to proportonal farness [7]. C. An Optmzaton Framework base on NBS Now we formally ntrouce our optmzaton framework. For our atacenter selecton problem, we wsh to conser long-term NBS farness whch epens on the average utlty. Long-term farness not only fathfully reflects clents performance, but also gves more flexblty to explot locaton versty of the clou nfrastructure. As scusse, the atacenter selecton problem can be vewe as a barganng game. Each mappng noe has the average utlty Ū as ts objectve. The goal s to maxmze all Ū cooperatvely. Each noe also has U mn that represents the mnmal utlty requrement base on the SLAs wth the clou. In the general case, fferent noes can have fferent SLAs epenng on ther locatons an thus fferent U mn. Note that the SLAs have to be satsfe n each epoch, nstea of statstcally over tme. The problem at epoch t, then, s to fn the NBS,.e. to solve for the optmzaton problem max U(t) U mn N (Ū (t) U mn ). (2) Mathematcally t s equvalent to the followng objectve: N ln ( Ū (t) U mn ). (3) Note that Ū(t) s a functon of the nstantaneous utlty U (t), typcally obtane by usng an exponentally weghte lowpass tme wnow: Ū (t) =(1 ρ w )Ū(t 1) + ρ w U (t). ρ w =(T s /T w ) where T s s the slot length an T w s the length of the tme wnow to calculate the average. Ths s ffcult to solve because the logarthm functon s not lnear. It has been shown n the semnal work [14] that maxmzng the total margnal gan of V ( X (t 1))X (t) at each t acheves long-term maxmzaton of V ( X (t)) asymptotcally. Thus, we can greatly reuce the computatonal complexty by transformng (3) to the followng lnear objectve: N U (t) U mn Ū (t 1) U mn. (4) Note that wthout conserng long-term performance, the optmzaton must guarantee farness n each epoch. However, when a tme wnow s use, the farness requrement s relaxe to the tme wnow length. Ths proves more flexblty to mprove the system effcency, by makng the current atacenter selecton relate to prevous ones. The term Ū (t 1) U mn n the enomnator of (4) serves as a weght factor to ajust the prorty of noe. If the noe has an unfarly large utlty gan from prevous epochs, ts prorty s lower n obtanng a goo atacenter n the current epoch. Therefore the long-term farness moel encourages noes to share the clou resources cooperatvely, an n turn gan more when t nees more help. In general t helps to acheve better system performance whle enforcng the farness noton over long run. Wthout loss of generalty, we use Ū to enote Ū(t 1), an rop the tme nex t n all notatons an the constant term U mn n the numerator of the objectve functon (4). The optmzaton problem at each epoch then can be formulate as DC-FAIR: N U max U Ū (5) U mn M s.t. U = p U,, (6) M p U U mn,, (7) M p =1,, p,,, (8) N p P,, (9) N p D C,, (1) M N W p D B. (11) The ecson varables are p,.e. the proporton of requests recte to atacenter from noe. The equalty (6) calculates the total utlty U gven by the atacenter selecton p an the performance matrx U. Constrant (7) s the basc SLA requrement for each noe. Constrant (8) correspons to the smple facts that all the requests at noe shoul be serve an that the ecson varable p s non-negatve. Constrant (9) moels the possble loa balancng requrement of the clou servce such that atacenter shoul at least hanle P [, 1] out of the total requests. (1) s the capacty constrant at, an (11) captures the cost constrant that the total cost of servng all the requests shoul not excee the buget B. Some may argue that some QoS parameters efne n a SLA, such as fee, responsblty, an securty level, cannot be capture by utlty. We comment that though ths s the case, ts effect on our framework s mnmal. Typcal clou servces o not have QoS guarantees on securty an prvacy, whch cannot be feasbly realze currently an stll remans an actve research topc. Even when these QoS parameters o nee to be consere, they can be easly ncorporate as atonal constrants on the request recton matrx, because the requests of a mappng noe can only be recte to atacenters that satsfy the responsblty an securty requrements. 11

4 III. A DECENTRALIZED IMPLEMENTATION The optmzaton problem DC-FAIR s essentally a LP, an can be solve n polynomal tme. However, ths requres a central coornator whch ntrouces a sngle pont of falure an s vulnerable to attacks. Further, the complexty of solvng the LP also ncreases sgnfcantly when the problem sze scales up, snce the number of varables s NM an the number of constrants s 2N +3M. A centralze soluton also makes t less aaptve to suen changes n traffc eman n a flash crow scenaro. Thus, for reasons of relablty, securty, scalablty, an performance, we are motvate to evelop strbute solutons n whch the mappng noes teratvely solve the DC-FAIR problem. A. Dual Decomposton Substtutng (6) nto the objectve functon (5), an relax the constrants (9) (11), we can obtan the Lagrangan of DC- FAIR: L(p, λ, μ,ν)= p U Ū + ( ) λ U mn p P + ( μ C ) ( p D + ν B ) W p D, where λ, μ, ν are the Lagrange multplers assocate wth the loa balancng, capacty, an cost constrants, respectvely. The ual functon s then { max g(λ, μ,ν)= p s.t. L(p, λ, μ,ν) p, U U mn (12) To solve g(λ, μ,ν), t s equvalent to solvng the problem wth the followng objectve ( ) U p Ū + λ U mn μ D νw D where the constant terms n L(p, λ, μ,ν) can be safely remove. The key observaton here s that t can be ecompose nto N per-noe maxmzaton sub-problems ( ) U p + λ μ D νw D max p Ū U mn s.t. p =1, p U U mn, (13) The per-noe sub-problem naturally emboes an economc nterpretaton. Each mappng noe strves to maxmze the total utlty of servng the requests, scounte by the costs of volatng the loa balancng, capacty, an buget constrants, as prce by the Lagrange multplers λ, μ,ν. The SLA constrant prevents the noe from rectng all ts requests to the most economcal atacenter, an forces t to strbute the requests among atacenters wth goo performance. Ths balances the clou operator s nterest n mnmzng ts costs an the clent s nterest n maxmzng ts performance. Note that the ecomposton greatly reuces the complexty of the optmzaton. The per-user sub-problem has only M varables an 2 constrants. In a typcal proucton clou, the number of atacenters M s much smaller than the number of mappng noes N, whch can be hunres. It s essentally a small-scale lnear program that can be solve effcently by stanar optmzaton solvers. B. A Dstrbute Algorthm We have shown that the ual functon of DC-FAIR can be ecompose nto N per-noe maxmzaton problem, whch s a smple lnear program. Now we nee to solve the ual problem mn λ,μ,ν g(λ, μ,ν) s.t. λ, μ,ν. (14) Subgraent metho [15] can be use to solve the ual problem. The upatng rules for the ual varables are as follows: [ ( λ (l+1) = μ (l+1) = [ λ (l) μ (l) + δ(l) + ɛ(l) P )] + p,, (15) ( )] + p D C,, (16) [ ( + ν (l+1) = ν (l) + σ (l) W p D B)], (17) where [x] + represents max{,x}, an δ, ɛ,σ are the step szes. Accorng to [15], the above proceure s guarantee to converge as long as the followng conton s satsfe. Proposton 1: The subgraent upates as n (15) (17) converge to the optmal ual varables f a mnshng step sze rule s followe for choosng δ, ɛ, σ [15]. Observe that, because of the ual ecomposton, ual optmzaton by subgraent metho can be one n a strbute fashon. Frst, n each teraton, the per-noe problems (13) can be solve smultaneously by the mappng noes, wth λ, μ, ν gven by atacenters. Secon, subgraent upates can also be strbutvely performe by each mappng noe. The algorthm can be perceve as an teratve barganng process. The ual varables λ, μ are transmtte from atacenters to all noes. They serve as prce sgnals to coornate the resource consumpton an loa balancng. For example, when the total traffc route to atacenter excees ts capacty,.e. p D >C, ncreases ts prce μ for the next roun of barganng to suppress the excessve eman. Smlarly, f has not reache ts mnmum loa P,.e. p <P, t ncreases the rewar prce λ to attract more traffc an therefore balance the loa 1. The process contnues untl t converges to the optmal resource allocaton. The upate metho of the other ual varable ν,.e. the buget prce, s also worth mentonng. Whle λ an μ can be nepenently upate by each atacenter wth only local nformaton, ν nees to be upate wth global nformaton 1 Note that λ s a postve term n (13). 12

5 from all atacenters. Ths can be one n a strbute way as follows. Intally, the prevous ν (l) s mae common knowlege among the atacenters. Frst, a atacenter s ranomly chosen an gven a token wth the total buget B. It calculates ts own monetary cost of servng the requests W p D, an euct ths amount from B. It puts a mark n the token, an pass t on to the next atacenter, who also upates the remanng buget, marks the token, an passes t further own. A atacenter etermnes t s the last one n the loop by examnng that except tself, everyone else has marke the token. It thus upates the remanng buget, calculates the upate buget prce ν (l+1), an broacasts to every mappng noe an atacenter. Algorthm 1 Optmal Dstrbute Datacenter Selecton 1. Each atacenter ntalzes λ (),μ(). ν() s ntalze to. 2. Each mappng noe collects λ (l),μ (l),ν (l), an nepenently solves the per-noe problem (13) usng stanar optmzaton solvers an obtan p. 3. Each atacenter collects ts loa p from all noes, an perform a subgraent upate for the loa an capacty prce λ (l),μ(l) as n (15) an (16). The upate λ (l+1),μ (l+1) are broacast to every mappng noe. 4. A atacenter s ranomly chosen an gven a token wth the buget B. 5. The atacenter eucts ts cost W p D from the remanng buget n the token, marks t, an passes t own. 6. Repeat step 5 untl the last atacenter calculates the fnal remanng buget, upates ν (l) as n (17), an broacasts to every atacenter an mappng noe. 7. Return to step 2 untl convergence. The complete barganng algorthm s shown n Algorthm 1. Snce t optmally solves the ual problem (14), t optmally solves the prmal problem DC-FAIR because the ualty gap for lnear programs s zero. Theorem 2: The strbute barganng algorthm as shown n Algorthm 1 always converges, an when t converges ts soluton optmally solves the atacenter selecton problem DC- FAIR. C. Dscussons We scuss some mplementaton ssues relate to our strbute atacenter selecton algorthm. Frst, step (2) an (3) of Algorthm 1 nee to be performe n a synchronze fashon across the mappng noes, whch may be of concern to some reaers for practcal mplementaton. We comment that ths can actually be one effcently. Note that the synchronzaton requrement s loose n the sense that we only requre each noe to have the complete ual varables before solvng (13) n step (2), an that step (3) to be performe after step (2) s complete for each noe. Ths can be realy acheve wthout any nee of explct or mplct tme synchronzaton, by havng each noe to wat for all the upate ual varables λ (l),μ (l),ν (l) at roun l to be receve completely before attemptng to solve the per-noe optmzaton, an by havng each atacenter to wat for all the upate solutons p (l) to be receve completely before upatng ts own ual varables. Therefore the entre proceure s naturally a self-synchronzng soluton. The concept of roun oes not nee any synchronzaton or central coornaton ether, snce each noe an atacenter knows for how many tmes t has solve the per-noe problem or upate the ual varables. Secon, we examne the message exchangng overhea of Algorthm 1. Each atacenter nees to share ts loa an capacty prce to all noes, whch mples O(MN) messages n total, each of sze 2. Each noe nees to share ts selecton ecson of sze M wth all atacenters, whch mples O(MN) messages, each of sze O(M). Thus n each roun, O(MN) messages are exchange each of sze O(M). The overhea scales lnearly wth the number of noes an atacenters, whch s practcal. IV. EVALUATION We present our smulaton stues n ths secton. A. Setup 1) Deman matrx: To represent the request traffc for a clou servce, we use the real-worl traces collecte from UUSee Inc. [1], a major commercal Veo-on-Deman prover wth servers strbute geographcally n Chna. The ataset contans, among other thngs, the banwth eman for UUSee veo programs sample every 1 mnutes, n a 12-ay pero urng the 28 Bejng Olympcs, from 14:51:58, Fray, August 8, 28 (GMT+8) to 23:43:56, Tuesay, August 19, 28 (GMT+8). Although the scale of the UUSee nfrastructure s not as large as that of a clou prover, we beleve the traces fathfully reflect the traffc eman strbuton for a clou servce, an t s approprate to use them for the purpose of benchmarkng the performance of our atacenter selecton algorthm. The precton of traffc eman can be one accurately as emonstrate by prevous work [12], [13], an n the smulaton we smply aopt the measure traffc eman as the precte eman matrx D. We use the traffc eman of stnct veo channels to represent eman of stnct mappng noes. We smulate a clou wth 1 mappng noes. Fg. 2 shows a sample of traffc eman for 3 mappng noes. Snce the ata s collecte every 1 mnutes, the optmzaton epoch s also set to 1 mnutes. 2) Datacenter placement an cost matrx: To capture the locaton versty of the clou nfrastructure an electrcty market, we assume the atacenters are eploye across the contnental U.S. Accorng to the Feeral Energy Regulatory Commsson (FERC), the U.S. electrcty market s consste of multple regonal markets as shown n Fg. 3 [11]. Each regonal market has several hubs wth ther own prcng. Therefore for the ease of exploraton, we assume that there s one atacenter eploye n a ranomly chosen hub n each 13

6 Deman (Mbps) Fg. 2. Noe 1 Noe 2 Noe 3 Tme (epochs) A 1-epoch sample of traffc eman for 3 mappng noes. regonal market. We use the 211 annual average ay-ahea on peak prce ($/MWh) publshe onlne by FERC as the electrcty prce for each atacenter,.e. W, as summarze n Table I [11]. The capacty of each atacenter s ranomly set such that ther total capacty s 15 Mbps. Fg. 3. The U.S. electrcty market an our clou atacenter map. Source: FERC [11]. TABLE I 211 ANNUAL AVERAGE DAY AHEAD ON PEAK PRICE ($/MWH) IN DIFFERENT REGIONAL MARKETS. SOURCE: FERC [11]. Regon Hub Prce Calforna NP15 $35.83 Mwest Mchgan Hub $42.73 New Englan Mass Hub $52.64 New York NY Zone J $62.71 Northwest Calforna-Oregon Borer (COB) $32.57 PJM PJM West $51.99 Southeast VACAR $44.44 Southwest Four Corners $36.36 SPP SPP North $36.41 Texas ERCOT North $ ) Performance matrx: Fnally, we conser a latencycrtcal clou servce, whose utlty can be efne by the negatve Euclean stance between the mappng noes an the atacenters. To calculate the performance matrx, we frst obtan the longtue an lattue of ten ranomly chosen countes n the area covere by each of the ten hubs as the exact locatons of our atacenters n the U.S. We then ranomly choose another 1 countes as the locatons of the 1 mappng noes. All the locaton nformaton s obtane from [16]. The Euclean stance between any gven par of mappng noe an atacenter then can be realy calculate, whch consttutes the performance matrx U. Wthout loss of generalty, we assume that servng 1 Gbps per epoch,.e. 1 mnutes, consumes 1 kwh electrcty. The SLA constrant for each noe U mn s set to be 3% lower than the best utlty may obtan among all atacenters. B. Effectveness We frst evaluate the effectveness of our strbute atacenter selecton algorthm. The buget B s set to $5 per epoch. Fg. 4 shows the per-noe average utlty wth total eman for a 1-epoch pero of tme. We observe that when the average utlty stans at 52 most of the tme,.e. on average the request s recte to a atacenter 52 km away, unless the total eman shoots beyon aroun 12 Mbps. Ths emonstrates that when eman s low, our algorthm matches mappng noes to ther closet atacenters, thus rectng requests to the best avalable atacenters. The consstent performance can be better explane by Fg. 5 that shows the total cost of servng requests versus total eman uner the same settng. Clearly the total cost fgure closely follows the total eman, an s below the $5 buget lmt all the tme. Thus, the buget constrant s nactve, whch correspons to a scenaro where the clou operator has ample fnancal resources to optmze performance wthout havng to conser the extra cost nvolve n ong so. Ths results n a constant average utlty curve am fluctuatng eman. When the eman becomes sgnfcant, t becomes necessary to rect some requests to stant atacenters to conform the buget constrant, whch explans the performance egraaton n epochs To see the effectveness of our algorthm on guaranteeng SLAs, we plot U U mn for each mappng noe at three sample epochs, 1, 75, an 77 n Fg. 6. Epoch 1 correspons to a low eman epoch, an epoch 75 an 77 correspon to extremely hgh eman peros. We observe that no matter the eman, the SLA s always satsfe for each noe because the curves stay above. It can also be seen that each noe enjoys a better utlty when eman s low n epoch 1, whch s ntutve to unerstan. Now to see the effectveness of our algorthm on guaranteeng the buget, we evaluate the effect of buget on the performance of atacenter selecton. We run our algorthm wth a reuce buget of $4, whle keepng the other settngs unchange. Compare to Fg. 4, Fg. 7 shows that now the performance swngs wely along wth the eman curve, an egraes to less than 6 when eman pkes over 1 Mbps. Fg. 8 also shows that the total cost s reuce compare to Fg. 5. It s thus event that a tghter buget negatvely affects the performance, but helps reucng the operatng costs of the atacenters. Reaers may notce that urng the epochs from 74 to 84 wth a hgh volume of eman, total cost s actually beyon the $4 buget. The strngent buget constrant causes the problem 14

7 1 Mbps Total eman 66 Average utlty Tme (epoch) km 1 Mbps Total eman 2.8 Total cost Tme (epoch) 6 $ per 1 mnutes (km) U U mn Epoch 1 Epoch 75 Epoch 78 Noe Fg. 4. Average utlty an total eman, B =5. Fg. 5. Total cost an total eman, B =5. Fg. 6. U U mn at sample epochs, B =5. 1 Mbps Total eman 66 Average utlty Tme (epoch) km 1 Mbps Total eman 2.8 Total cost Tme (epoch) 6 $ per 1 mnutes (km) U U mn Epoch 1 Epoch 75 Epoch 78 Noe Fg. 7. Average utlty an total eman, B =4. Fg. 8. Total cost an total eman, B =4. Fg. 9. U U mn at sample epochs, B =4. to be nfeasble for those epochs wth extremely hgh eman. In our mplementaton, we set a stoppng crteron to be relate to the absolute change of the objectve values, so that our algorthm satsfes the buget constrant whenever t s feasble to o so, an stll prouces sensble results wth an nfeasble buget constrant. The SLA constrant s always satsfe as shown n Fg. 9, however, because t s not relaxe n the per-noe problem (13). C. Effcency an farness To emonstrate the effcency an farness acheve wth our algorthm, we use a smple LP, calle DC-OPT, that shares all the same constrants wth our DC-FAIR problem, but uses the total utlty U as the objectve functon nstea of (5) as the benchmark. Therefore t oes not conser farness. DC- OPT can also be solve by the same metho of strbute subgraent upates. Fg. 1 frst shows the farness comparson of DC-FAIR an the smple DC-OPT. We use the stanar evaton of Ū (t) U mn,.e. the tme average surplus utlty, across noes at each epoch t as the farness measure. A far selecton algorthm yels a proportonal surplus utlty allocaton among noes wth a smaller stanar evaton, whereas a poor algorthm wthout conserng farness prouces an allocaton wth a larger stanar evaton. We can observe from Fg. 1 that n most of the tme, DC-FAIR acheves a much smaller stanar evaton for the surplus utlty, whch translates to much better farness among noes. Farness s generally mprove over tme except for the hgh eman epochs, whch emonstrates the effectveness of our long-term NBS base farness moel. The performance an cost comparsons between our DC- FAIR an the DC-OPT algorthms are shown n Fg. 11 an Fg. 12. DC-OPT performs better than our DC-FAIR n terms of per-noe average utlty along the tme lne, whch s expecte snce the sole objectve of DC-OPT s to maxmze the aggregate utlty wthout conserng farness. In terms of cost, our DC-FAIR algorthm acheves slghtly better results. Overall, DC-FAIR attans a fferent performance-farness trae-off wth much better farness among noes, at the cost of system-we performance. As we scusse, farness s a crtcal requrement of the atacenter selecton problem, snce clents serve by a partcular mappng noe wsh to obtan a far share of the avalable clou resources after the mnmum SLAs are satsfe. Our algorthm thus acheves the farness requrement wth satsfactory performance. D. Spee of convergence Number of teratons Number of teratons Fg. 13. Hstogram of the convergence spee, B =5. gence spee, B Fg. 14. Hstogram of the conver- = Fnally, we evaluate the convergence spee of the DC-FAIR algorthm. Fg. 13 an Fg. 14 show the hstograms of the number of teratons the algorthm takes to converge to the optmal soluton for fferent bugets. Clearly, we observe that t usually takes 19 2 teratons, an never takes more than 2 teratons to fnsh. For a problem wth 1 atacenters an 1 noes, the spee of convergence shoul be consere 15

8 Stanar evaton FAIR OPT 75 Tme (epoch) km FAIR OPT 65 Tme (epoch) $ FAIR OPT Tme (epoch) Fg. 1. Farness comparson, B =4. Fg. 11. Average utlty comparson, B =4. Fg. 12. Total cost comparson, B =4. very satsfactory. Thus we beleve that our algorthm s also practcal for real-worl problems wth larger scales. V. RELATED WORK The topc of atacenter selecton an loa recton for a geo-strbute clou has starte to gan attenton n the research communty. Quresh et al. [9] ntrouce the ea of utlzng the locaton versty of electrcty spot prce to ntellgently rect requests to atacenters wth lower prces. Wenell et al. [2] evelope a ecentralze atacenter selecton algorthm for clou servces, an evaluate ts performance usng a prototype an realstc traffc traces. Rao et al. [17] consere a jont loa balancng an power control problem for Internet atacenters to explot the tme an locaton versty of electrcty prce. [18] specfcally consere the effect of geographcal loa balancng on provng envronmental gans by encouragng the use of green energy. [19] stue a complementary problem of ata placement n a geo-strbute clou, conserng the ata localty, WAN banwth costs, an atacenter capacty. These works, however, o not conser farness n ther problem formulaton. The concept of Nash barganng games has been apple to networkng problems n other omans. [2] apples t to ensure farness n a network flow control problem. Kelly n [6] has shown that Nash barganng ensures proportonal farness n a TCP settng. [7] stue a far multuser channel allocaton scheme for OFDMA networks base on Nash barganng solutons an coaltons. [21] apple NBS to a resource allocaton problem n cooperatve cogntve rao networks. We conser a new context of clou computng, the specfcs of whch are not capture n these works. VI. CONCLUDING REMARKS In ths paper, we presente a general optmzaton framework to solve the atacenter selecton problem for clou servces. Our framework s base on the concept of NBS farness to proportonally allocate the avalable clou resources across the mappng noes, after the mnmum SLA requrements are met. We aopte a ual ecomposton approach to evelop a strbute algorthm base on the subgraent metho. Our work can be extene n many rectons. One possble recton s to conser the onlne atacenter selecton wth farness conseraton, whch s more ffcult than the offlne problem we solve here. The other recton s to take nto account the stochastc nature of the request traffc, an prove stochastc performance guarantee am such ranom traffc. REFERENCES [1] DynDNS, [2] P. Wenell, J. W. Jang, M. J. Freeman, an J. Rexfor, DONAR: Decentralze server selecton for clou servces, n Proc. ACM SIG- COMM, 21. [3] Amazon AWS elastc loa balancng, elastcloabalancng/. [4] R. Kohav an R. Longbotham, Onlne experments: Lessons learne, IEEE Computer, vol. 4, no. 9, pp , September 27. [5] D. K. Golenberg, L. Qu, H. Xe, Y. R. Yang, an Y. Zhang, Optmzng cost an performance for multhomng, n Proc. ACM SIGCOMM, 24. [6] F. P. Kelly, A. K. Maulloo, an D. K. H. Tan, Rate control for communcaton networks: Shaow prces, proportonal farness an stablty, J. Operat. Res. Soc., vol. 49, no. 3, pp , March [7] Z. Han, Z. J. J, an K. J. R. Lu, Far multuser channel allocaton for OFDMA networks usng Nash barganng solutons an coaltons, IEEE Trans. Commun., vol. 53, no. 8, pp , August 25. [8] G. Owen, Game Theory. New York: Acaemc, 21. [9] A. Quresh, R. Weber, H. Balakrshnan, J. Guttag, an B. Maggs, Cuttng the electrcty bll for Internet-scale systems, n Proc. ACM SIGCOMM, 29. [1] UUSee Inc. [11] Feeral Energy Regulatory Commsson, U.S. electrc power markets, [12] D. Nu, H. Xu, B. L, an S. Zhao, Rsk management for veo-oneman servers leveragng eman forecast, n Proc. ACM Multmea, 211. [13], Qualty-assure clou banwth auto-scalng for veo-oneman applcatons, n Proc. IEEE INFOCOM, 212. [14] G. Song an Y. G. L, Cross-layer optmzaton for OFDM wreless network Part II: Algorthm evelopment, IEEE Trans. Wreless Commun., vol. 4, no. 2, pp , March 25. [15] S. Boy an A. Mutapcc, Subgraent methos, Lecture notes of EE364b, Stanfor Unversty, Wnter Quarter stanfor.eu/class/ee364b/notes/subgra metho notes.pf. [16] WebGIS, [17] L. Rao, X. Lu, L. Xe, an W. Lu, Mnmzng electrcty cost: Optmzaton of strbute Internet ata centers n a mult-electrctymarket envronment, n Proc. IEEE INFOCOM, 21. [18] Z. Lu, M. Ln, A. Werman, S. H. Low, an L. L. Anrew, Greenng geographcal loa balancng, n Proc. ACM Sgmetrcs, 211. [19] S. Agarwal, J. Dunagan, N. Jan, S. Sarou, A. Wolman, an H. Bhogan, Volley: Automate ata placement for geo-strbute clou servces, n Proc. USENIX NSDI, 21. [2] R. Mazumar, L. G. Mason, an C. Doulglgers, Farness n network optmal flow control: Optmalty of prouct forms, IEEE Trans. Commun., vol. 39, no. 5, pp , May [21] H. Xu an B. L, Effcent resource allocaton wth flexble channel cooperaton n OFDMA cogntve rao networks, n Proc. IEEE INFOCOM,

Cost Efficient Datacenter Selection for Cloud Services

Cost Efficient Datacenter Selection for Cloud Services Cost Efficient Datacenter Selection for Clou Services Hong u, Baochun Li henryxu, bli@eecg.toronto.eu Department of Electrical an Computer Engineering University of Toronto Abstract Many clou services

More information

EXAMPLE PROBLEMS SOLVED USING THE SHARP EL-733A CALCULATOR

EXAMPLE PROBLEMS SOLVED USING THE SHARP EL-733A CALCULATOR EXAMPLE PROBLEMS SOLVED USING THE SHARP EL-733A CALCULATOR 8S CHAPTER 8 EXAMPLES EXAMPLE 8.4A THE INVESTMENT NEEDED TO REACH A PARTICULAR FUTURE VALUE What amount must you nvest now at 4% compoune monthly

More information

An Efficient Recovery Algorithm for Coverage Hole in WSNs

An Efficient Recovery Algorithm for Coverage Hole in WSNs An Effcent Recover Algorthm for Coverage Hole n WSNs Song Ja 1,*, Wang Balng 1, Peng Xuan 1 School of Informaton an Electrcal Engneerng Harbn Insttute of Technolog at Weha, Shanong, Chna Automatc Test

More information

Blind Estimation of Transmit Power in Wireless Networks

Blind Estimation of Transmit Power in Wireless Networks Bln Estmaton of Transmt Power n Wreless Networks Murtaza Zafer (IBM Research), Bongjun Ko (IBM Research), Chatschk Bskan (IBM Research) an Ivan Ho (Imperal College, UK) Transmt-power Estmaton: Problem

More information

On the Optimal Control of a Cascade of Hydro-Electric Power Stations

On the Optimal Control of a Cascade of Hydro-Electric Power Stations On the Optmal Control of a Cascade of Hydro-Electrc Power Statons M.C.M. Guedes a, A.F. Rbero a, G.V. Smrnov b and S. Vlela c a Department of Mathematcs, School of Scences, Unversty of Porto, Portugal;

More information

DEGREES OF EQUIVALENCE IN A KEY COMPARISON 1 Thang H. L., Nguyen D. D. Vietnam Metrology Institute, Address: 8 Hoang Quoc Viet, Hanoi, Vietnam

DEGREES OF EQUIVALENCE IN A KEY COMPARISON 1 Thang H. L., Nguyen D. D. Vietnam Metrology Institute, Address: 8 Hoang Quoc Viet, Hanoi, Vietnam DEGREES OF EQUIVALECE I A EY COMPARISO Thang H. L., guyen D. D. Vetnam Metrology Insttute, Aress: 8 Hoang Quoc Vet, Hano, Vetnam Abstract: In an nterlaboratory key comparson, a ata analyss proceure for

More information

A Binary Quantum-behaved Particle Swarm Optimization Algorithm with Cooperative Approach

A Binary Quantum-behaved Particle Swarm Optimization Algorithm with Cooperative Approach IJCSI Internatonal Journal of Computer Scence Issues, Vol., Issue, No, January 3 ISSN (Prnt): 694-784 ISSN (Onlne): 694-84 www.ijcsi.org A Bnary Quantum-behave Partcle Swarm Optmzaton Algorthm wth Cooperatve

More information

On the Optimal Marginal Rate of Income Tax

On the Optimal Marginal Rate of Income Tax On the Optmal Margnal Rate of Income Tax Gareth D Myles Insttute for Fscal Stues an Unversty of Exeter June 999 Abstract: The paper shows that n the quas-lnear moel of ncome taxaton, the optmal margnal

More information

Fault tolerance in cloud technologies presented as a service

Fault tolerance in cloud technologies presented as a service Internatonal Scentfc Conference Computer Scence 2015 Pavel Dzhunev, PhD student Fault tolerance n cloud technologes presented as a servce INTRODUCTION Improvements n technques for vrtualzaton and performance

More information

THE LOAD PLANNING PROBLEM FOR LESS-THAN-TRUCKLOAD MOTOR CARRIERS AND A SOLUTION APPROACH. Professor Naoto Katayama* and Professor Shigeru Yurimoto*

THE LOAD PLANNING PROBLEM FOR LESS-THAN-TRUCKLOAD MOTOR CARRIERS AND A SOLUTION APPROACH. Professor Naoto Katayama* and Professor Shigeru Yurimoto* 7th Internatonal Symposum on Logstcs THE LOAD PLAIG PROBLEM FOR LESS-THA-TRUCKLOAD MOTOR CARRIERS AD A SOLUTIO APPROACH Professor aoto Katayama* an Professor Shgeru Yurmoto* * Faculty of Dstrbuton an Logstcs

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Max Wellng Department of Computer Scence Unversty of Toronto 10 Kng s College Road Toronto, M5S 3G5 Canada wellng@cs.toronto.edu Abstract Ths s a note to explan support vector machnes.

More information

Joint Request Mapping and Response Routing for Geo-distributed Cloud Services

Joint Request Mapping and Response Routing for Geo-distributed Cloud Services Jont Request Mappng and Response Routng for Geo-dstrbuted Cloud Servces Hong Xu, Baochun L henryxu, bl@eecg.toronto.edu Department of Electrcal and Computer Engneerng Unversty of Toronto Abstract Many

More information

An algorithm of choosing LSPs in the MPLS network with unreliable links

An algorithm of choosing LSPs in the MPLS network with unreliable links Ireneusz OLSZESKI Unversty of Technology an Lfe Scences n Bygoszcz, Faculty of Telecommuncatons an Electrcal Engneerng An algorthm of choosng LSPs n the MPLS network wth unrelable lnks Streszczene. pracy

More information

A Lyapunov Optimization Approach to Repeated Stochastic Games

A Lyapunov Optimization Approach to Repeated Stochastic Games PROC. ALLERTON CONFERENCE ON COMMUNICATION, CONTROL, AND COMPUTING, OCT. 2013 1 A Lyapunov Optmzaton Approach to Repeated Stochastc Games Mchael J. Neely Unversty of Southern Calforna http://www-bcf.usc.edu/

More information

Answer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy

Answer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy 4.02 Quz Solutons Fall 2004 Multple-Choce Questons (30/00 ponts) Please, crcle the correct answer for each of the followng 0 multple-choce questons. For each queston, only one of the answers s correct.

More information

The Design of Efficiently-Encodable Rate-Compatible LDPC Codes

The Design of Efficiently-Encodable Rate-Compatible LDPC Codes The Desgn of Effcently-Encoable Rate-Compatble LDPC Coes Jaehong Km, Atya Ramamoorthy, Member, IEEE, an Steven W. McLaughln, Fellow, IEEE Abstract We present a new class of rregular low-ensty party-check

More information

Decomposition Methods for Large Scale LP Decoding

Decomposition Methods for Large Scale LP Decoding Decomposton Methos for Large Scale LP Decong Sharth Barman Xshuo Lu Stark Draper Benjamn Recht Abstract Felman et al. IEEE Trans. Inform. Theory, Mar. 2005) showe that lnear programmng LP) can be use to

More information

PAS: A Packet Accounting System to Limit the Effects of DoS & DDoS. Debish Fesehaye & Klara Naherstedt University of Illinois-Urbana Champaign

PAS: A Packet Accounting System to Limit the Effects of DoS & DDoS. Debish Fesehaye & Klara Naherstedt University of Illinois-Urbana Champaign PAS: A Packet Accountng System to Lmt the Effects of DoS & DDoS Debsh Fesehaye & Klara Naherstedt Unversty of Illnos-Urbana Champagn DoS and DDoS DDoS attacks are ncreasng threats to our dgtal world. Exstng

More information

Enabling P2P One-view Multi-party Video Conferencing

Enabling P2P One-view Multi-party Video Conferencing Enablng P2P One-vew Mult-party Vdeo Conferencng Yongxang Zhao, Yong Lu, Changja Chen, and JanYn Zhang Abstract Mult-Party Vdeo Conferencng (MPVC) facltates realtme group nteracton between users. Whle P2P

More information

Incentive Compatible Mechanisms for Group Ticket Allocation in Software Maintenance Services

Incentive Compatible Mechanisms for Group Ticket Allocation in Software Maintenance Services 14th Asa-Pacfc Software Engneerng Conference Incentve Compatble Mechansms for Group Tcket Allocaton n Software Mantenance Servces Karthk Subban, Ramakrshnan Kannan IBM R Ina Software Lab, EGL D Block,

More information

How To Improve Power Demand Response Of A Data Center Wth A Real Time Power Demand Control Program

How To Improve Power Demand Response Of A Data Center Wth A Real Time Power Demand Control Program Demand Response of Data Centers: A Real-tme Prcng Game between Utltes n Smart Grd Nguyen H. Tran, Shaole Ren, Zhu Han, Sung Man Jang, Seung Il Moon and Choong Seon Hong Department of Computer Engneerng,

More information

Distributed Strategic Learning with Application to Network Security

Distributed Strategic Learning with Application to Network Security Amercan Control Conference on O'Farrell Street San Francsco CA USA June 9 - July Dstrbute Strategc Learnng wth Applcaton to Network Securty Quanyan Zhu Hamou Tembne an Tamer Başar Abstract We conser n

More information

Cost Minimization using Renewable Cooling and Thermal Energy Storage in CDNs

Cost Minimization using Renewable Cooling and Thermal Energy Storage in CDNs Cost Mnmzaton usng Renewable Coolng and Thermal Energy Storage n CDNs Stephen Lee College of Informaton and Computer Scences UMass, Amherst stephenlee@cs.umass.edu Rahul Urgaonkar IBM Research rurgaon@us.bm.com

More information

A Practical Study of Regenerating Codes for Peer-to-Peer Backup Systems

A Practical Study of Regenerating Codes for Peer-to-Peer Backup Systems A Practcal Stuy of Regeneratng Coes for Peer-to-Peer Backup Systems Alessanro Dumnuco an Ernst Bersack EURECOM Sopha Antpols, France {umnuco,bersack}@eurecom.fr Abstract In strbute storage systems, erasure

More information

The Greedy Method. Introduction. 0/1 Knapsack Problem

The Greedy Method. Introduction. 0/1 Knapsack Problem The Greedy Method Introducton We have completed data structures. We now are gong to look at algorthm desgn methods. Often we are lookng at optmzaton problems whose performance s exponental. For an optmzaton

More information

QoS-Aware Spectrum Sharing in Cognitive Wireless Networks

QoS-Aware Spectrum Sharing in Cognitive Wireless Networks QoS-Aware Spectrum Sharng n Cogntve reless Networks Long Le and Ekram Hossan Abstract e consder QoS-aware spectrum sharng n cogntve wreless networks where secondary users are allowed to access the spectrum

More information

Hosting Virtual Machines on Distributed Datacenters

Hosting Virtual Machines on Distributed Datacenters Hostng Vrtual Machnes on Dstrbuted Datacenters Chuan Pham Scence and Engneerng, KyungHee Unversty, Korea pchuan@khu.ac.kr Jae Hyeok Son Scence and Engneerng, KyungHee Unversty, Korea sonaehyeok@khu.ac.kr

More information

How To Solve An Onlne Control Polcy On A Vrtualzed Data Center

How To Solve An Onlne Control Polcy On A Vrtualzed Data Center Dynamc Resource Allocaton and Power Management n Vrtualzed Data Centers Rahul Urgaonkar, Ulas C. Kozat, Ken Igarash, Mchael J. Neely urgaonka@usc.edu, {kozat, garash}@docomolabs-usa.com, mjneely@usc.edu

More information

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts Power-of-wo Polces for Sngle- Warehouse Mult-Retaler Inventory Systems wth Order Frequency Dscounts José A. Ventura Pennsylvana State Unversty (USA) Yale. Herer echnon Israel Insttute of echnology (Israel)

More information

Feasibility of Using Discriminate Pricing Schemes for Energy Trading in Smart Grid

Feasibility of Using Discriminate Pricing Schemes for Energy Trading in Smart Grid Feasblty of Usng Dscrmnate Prcng Schemes for Energy Tradng n Smart Grd Wayes Tushar, Chau Yuen, Bo Cha, Davd B. Smth, and H. Vncent Poor Sngapore Unversty of Technology and Desgn, Sngapore 138682. Emal:

More information

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic Lagrange Multplers as Quanttatve Indcators n Economcs Ivan Mezník Insttute of Informatcs, Faculty of Busness and Management, Brno Unversty of TechnologCzech Republc Abstract The quanttatve role of Lagrange

More information

A Secure Password-Authenticated Key Agreement Using Smart Cards

A Secure Password-Authenticated Key Agreement Using Smart Cards A Secure Password-Authentcated Key Agreement Usng Smart Cards Ka Chan 1, Wen-Chung Kuo 2 and Jn-Chou Cheng 3 1 Department of Computer and Informaton Scence, R.O.C. Mltary Academy, Kaohsung 83059, Tawan,

More information

Cluster Analysis. Cluster Analysis

Cluster Analysis. Cluster Analysis Cluster Analyss Cluster Analyss What s Cluster Analyss? Types of Data n Cluster Analyss A Categorzaton of Maor Clusterng Methos Parttonng Methos Herarchcal Methos Densty-Base Methos Gr-Base Methos Moel-Base

More information

Multi-Resource Fair Allocation in Heterogeneous Cloud Computing Systems

Multi-Resource Fair Allocation in Heterogeneous Cloud Computing Systems 1 Mult-Resource Far Allocaton n Heterogeneous Cloud Computng Systems We Wang, Student Member, IEEE, Ben Lang, Senor Member, IEEE, Baochun L, Senor Member, IEEE Abstract We study the mult-resource allocaton

More information

A RELIABLE SEMI-DISTRIBUTED LOAD BALANCING ARCHITECTURE OF HETEROGENEOUS WIRELESS NETWORKS

A RELIABLE SEMI-DISTRIBUTED LOAD BALANCING ARCHITECTURE OF HETEROGENEOUS WIRELESS NETWORKS Internatonal Journal of Computer Networks & Communcatons (IJCNC) Vol.4, No., January 0 A RELIABLE SEMI-DISTRIBUTED LOAD BALANCING ARCHITECTURE OF HETEROGENEOUS WIRELESS NETWORKS M. Golam Rabul Alam, Chayan

More information

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services When Network Effect Meets Congeston Effect: Leveragng Socal Servces for Wreless Servces aowen Gong School of Electrcal, Computer and Energy Engeerng Arzona State Unversty Tempe, AZ 8587, USA xgong9@asuedu

More information

Formulating & Solving Integer Problems Chapter 11 289

Formulating & Solving Integer Problems Chapter 11 289 Formulatng & Solvng Integer Problems Chapter 11 289 The Optonal Stop TSP If we drop the requrement that every stop must be vsted, we then get the optonal stop TSP. Ths mght correspond to a ob sequencng

More information

An MILP model for planning of batch plants operating in a campaign-mode

An MILP model for planning of batch plants operating in a campaign-mode An MILP model for plannng of batch plants operatng n a campagn-mode Yanna Fumero Insttuto de Desarrollo y Dseño CONICET UTN yfumero@santafe-concet.gov.ar Gabrela Corsano Insttuto de Desarrollo y Dseño

More information

High Performance Latent Dirichlet Allocation for Text Mining

High Performance Latent Dirichlet Allocation for Text Mining Hgh Performance Latent Drchlet Allocaton for Text Mnng A thess submtte for Degree of Doctor of Phlosophy By Department of Electronc an Computer Engneerng School of Engneerng an Desgn Brunel Unversty September

More information

Dynamic Pricing for Smart Grid with Reinforcement Learning

Dynamic Pricing for Smart Grid with Reinforcement Learning Dynamc Prcng for Smart Grd wth Renforcement Learnng Byung-Gook Km, Yu Zhang, Mhaela van der Schaar, and Jang-Won Lee Samsung Electroncs, Suwon, Korea Department of Electrcal Engneerng, UCLA, Los Angeles,

More information

How To Solve A Problem In A Powerline (Powerline) With A Powerbook (Powerbook)

How To Solve A Problem In A Powerline (Powerline) With A Powerbook (Powerbook) MIT 8.996: Topc n TCS: Internet Research Problems Sprng 2002 Lecture 7 March 20, 2002 Lecturer: Bran Dean Global Load Balancng Scrbe: John Kogel, Ben Leong In today s lecture, we dscuss global load balancng

More information

On the computation of the capital multiplier in the Fortis Credit Economic Capital model

On the computation of the capital multiplier in the Fortis Credit Economic Capital model On the computaton of the captal multpler n the Forts Cret Economc Captal moel Jan Dhaene 1, Steven Vuffel 2, Marc Goovaerts 1, Ruben Oleslagers 3 Robert Koch 3 Abstract One of the key parameters n the

More information

Temperature Aware Workload Management in Geo-distributed Data Centers

Temperature Aware Workload Management in Geo-distributed Data Centers IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, 1X 1 Temperature Aware Workload Management n Geo-dstrbuted Data Centers Hong Xu, Member, IEEE, Chen Feng, Member, IEEE, and Baochun L, Senor Member,

More information

Downlink Power Allocation for Multi-class. Wireless Systems

Downlink Power Allocation for Multi-class. Wireless Systems Downlnk Power Allocaton for Mult-class 1 Wreless Systems Jang-Won Lee, Rav R. Mazumdar, and Ness B. Shroff School of Electrcal and Computer Engneerng Purdue Unversty West Lafayette, IN 47907, USA {lee46,

More information

Stock Profit Patterns

Stock Profit Patterns Stock Proft Patterns Suppose a share of Farsta Shppng stock n January 004 s prce n the market to 56. Assume that a September call opton at exercse prce 50 costs 8. A September put opton at exercse prce

More information

How To Improve Delay Throughput In Wireless Networks With Multipath Routing And Channel Codeing

How To Improve Delay Throughput In Wireless Networks With Multipath Routing And Channel Codeing Delay-Throughput Enhancement n Wreless Networs wth Mult-path Routng and Channel Codng Kevan Ronas, Student Member, IEEE, Amr-Hamed Mohsenan-Rad, Member, IEEE, Vncent W.S. Wong, Senor Member, IEEE, Sathsh

More information

Fair Virtual Bandwidth Allocation Model in Virtual Data Centers

Fair Virtual Bandwidth Allocation Model in Virtual Data Centers Far Vrtual Bandwdth Allocaton Model n Vrtual Data Centers Yng Yuan, Cu-rong Wang, Cong Wang School of Informaton Scence and Engneerng ortheastern Unversty Shenyang, Chna School of Computer and Communcaton

More information

Learning the Best K-th Channel for QoS Provisioning in Cognitive Networks

Learning the Best K-th Channel for QoS Provisioning in Cognitive Networks 000 001 002 003 004 005 006 007 008 009 010 011 012 013 014 015 016 017 018 019 020 021 022 023 024 025 026 027 028 029 030 031 032 033 034 035 036 037 038 039 040 041 042 043 044 045 046 047 048 049 050

More information

Availability-Based Path Selection and Network Vulnerability Assessment

Availability-Based Path Selection and Network Vulnerability Assessment Avalablty-Based Path Selecton and Network Vulnerablty Assessment Song Yang, Stojan Trajanovsk and Fernando A. Kupers Delft Unversty of Technology, The Netherlands {S.Yang, S.Trajanovsk, F.A.Kupers}@tudelft.nl

More information

An Alternative Way to Measure Private Equity Performance

An Alternative Way to Measure Private Equity Performance An Alternatve Way to Measure Prvate Equty Performance Peter Todd Parlux Investment Technology LLC Summary Internal Rate of Return (IRR) s probably the most common way to measure the performance of prvate

More information

J. Parallel Distrib. Comput.

J. Parallel Distrib. Comput. J. Parallel Dstrb. Comput. 71 (2011) 62 76 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. journal homepage: www.elsever.com/locate/jpdc Optmzng server placement n dstrbuted systems n

More information

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network 700 Proceedngs of the 8th Internatonal Conference on Innovaton & Management Forecastng the Demand of Emergency Supples: Based on the CBR Theory and BP Neural Network Fu Deqang, Lu Yun, L Changbng School

More information

Performance Analysis of Energy Consumption of Smartphone Running Mobile Hotspot Application

Performance Analysis of Energy Consumption of Smartphone Running Mobile Hotspot Application Internatonal Journal of mart Grd and lean Energy Performance Analyss of Energy onsumpton of martphone Runnng Moble Hotspot Applcaton Yun on hung a chool of Electronc Engneerng, oongsl Unversty, 511 angdo-dong,

More information

Joint Scheduling of Processing and Shuffle Phases in MapReduce Systems

Joint Scheduling of Processing and Shuffle Phases in MapReduce Systems Jont Schedulng of Processng and Shuffle Phases n MapReduce Systems Fangfe Chen, Mural Kodalam, T. V. Lakshman Department of Computer Scence and Engneerng, The Penn State Unversty Bell Laboratores, Alcatel-Lucent

More information

On File Delay Minimization for Content Uploading to Media Cloud via Collaborative Wireless Network

On File Delay Minimization for Content Uploading to Media Cloud via Collaborative Wireless Network On Fle Delay Mnmzaton for Content Uploadng to Meda Cloud va Collaboratve Wreless Network Ge Zhang and Yonggang Wen School of Computer Engneerng Nanyang Technologcal Unversty Sngapore Emal: {zh0001ge, ygwen}@ntu.edu.sg

More information

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module LOSSLESS IMAGE COMPRESSION SYSTEMS Lesson 3 Lossless Compresson: Huffman Codng Instructonal Objectves At the end of ths lesson, the students should be able to:. Defne and measure source entropy..

More information

Luby s Alg. for Maximal Independent Sets using Pairwise Independence

Luby s Alg. for Maximal Independent Sets using Pairwise Independence Lecture Notes for Randomzed Algorthms Luby s Alg. for Maxmal Independent Sets usng Parwse Independence Last Updated by Erc Vgoda on February, 006 8. Maxmal Independent Sets For a graph G = (V, E), an ndependent

More information

Fair and Efficient User-Network Association Algorithm for Multi-Technology Wireless Networks

Fair and Efficient User-Network Association Algorithm for Multi-Technology Wireless Networks Far and Effcent User-Network Assocaton Algorthm for Mult-Technology Wreless Networks Perre Coucheney, Cornne Touat and Bruno Gaujal INRIA Rhône-Alpes and LIG, MESCAL project, Grenoble France, {perre.coucheney,

More information

Heuristic Static Load-Balancing Algorithm Applied to CESM

Heuristic Static Load-Balancing Algorithm Applied to CESM Heurstc Statc Load-Balancng Algorthm Appled to CESM 1 Yur Alexeev, 1 Sher Mckelson, 1 Sven Leyffer, 1 Robert Jacob, 2 Anthony Crag 1 Argonne Natonal Laboratory, 9700 S. Cass Avenue, Argonne, IL 60439,

More information

On Multi-controller Placement Optimization in Software Defined Networking - based WANs

On Multi-controller Placement Optimization in Software Defined Networking - based WANs On Mult-controller Placement Optmzaton n Software Defned Networkng - based WANs Eugen Borcoc, Radu Badea, Serban Georgca Obreja, Marus Vochn Unversty POLITEHNICA of Bucharest - UPB Bucharest, Romana eugen.borcoc@elcom.pub.ro,

More information

RSA Cryptography using Designed Processor and MicroBlaze Soft Processor in FPGAs

RSA Cryptography using Designed Processor and MicroBlaze Soft Processor in FPGAs RSA Cryptography usng Desgne Processor an McroBlaze Soft Processor n FPGAs M. Nazrul Islam Monal Dept. of CSE, Rajshah Unversty of Engneerng an Technology, Rajshah-6204, Banglaesh M. Al Mamun Dept. of

More information

BERNSTEIN POLYNOMIALS

BERNSTEIN POLYNOMIALS On-Lne Geometrc Modelng Notes BERNSTEIN POLYNOMIALS Kenneth I. Joy Vsualzaton and Graphcs Research Group Department of Computer Scence Unversty of Calforna, Davs Overvew Polynomals are ncredbly useful

More information

PROBABILISTIC DECISION ANALYSIS FOR SEISMIC REHABILITATION OF A REGIONAL BUILDING SYSTEM

PROBABILISTIC DECISION ANALYSIS FOR SEISMIC REHABILITATION OF A REGIONAL BUILDING SYSTEM 3 th Worl Conference on Earthquake Engneerng Vancouver, B.C., Canaa August -6, 4 Paper No. 54 PROBABILISTIC DECISION ANALYSIS FOR SEISMIC REHABILITATION OF A REGIONAL BILDING SYSTEM Joonam PARK, Barry

More information

ON CYCLOTOMIC POLYNOMIALS WITH ±1 COEFFICIENTS

ON CYCLOTOMIC POLYNOMIALS WITH ±1 COEFFICIENTS ON CYCLOTOMIC POLYNOMIALS WITH ±1 COEFFICIENTS PETER BORWEIN AND KWOK-KWONG STEPHEN CHOI Abstract. We characterze all cyclotomc polynomals of even egree wth coeffcents restrcte to the set {+1, 1}. In ths

More information

Agile Traffic Merging for Data Center Networks. Qing Yi and Suresh Singh Portland State University, Oregon June 10 th, 2014

Agile Traffic Merging for Data Center Networks. Qing Yi and Suresh Singh Portland State University, Oregon June 10 th, 2014 Agle Traffc Mergng for Data Center Networks Qng Y and Suresh Sngh Portland State Unversty, Oregon June 10 th, 2014 Agenda Background and motvaton Power optmzaton model Smulated greedy algorthm Traffc mergng

More information

Fisher Markets and Convex Programs

Fisher Markets and Convex Programs Fsher Markets and Convex Programs Nkhl R. Devanur 1 Introducton Convex programmng dualty s usually stated n ts most general form, wth convex objectve functons and convex constrants. (The book by Boyd and

More information

Exact GP Schema Theory for Headless Chicken Crossover and Subtree Mutation

Exact GP Schema Theory for Headless Chicken Crossover and Subtree Mutation Exact GP Schema Theory for Healess Chcken Crossover an Subtree Mutaton Rccaro Pol School of Computer Scence The Unversty of Brmngham Brmngham, B5 TT, UK R.Pol@cs.bham.ac.uk Ncholas F. McPhee Dvson of Scence

More information

Energy Efficient Routing in Ad Hoc Disaster Recovery Networks

Energy Efficient Routing in Ad Hoc Disaster Recovery Networks Energy Effcent Routng n Ad Hoc Dsaster Recovery Networks Gl Zussman and Adran Segall Department of Electrcal Engneerng Technon Israel Insttute of Technology Hafa 32000, Israel {glz@tx, segall@ee}.technon.ac.l

More information

行 政 院 國 家 科 學 委 員 會 補 助 專 題 研 究 計 畫 成 果 報 告 期 中 進 度 報 告

行 政 院 國 家 科 學 委 員 會 補 助 專 題 研 究 計 畫 成 果 報 告 期 中 進 度 報 告 行 政 院 國 家 科 學 委 員 會 補 助 專 題 研 究 計 畫 成 果 報 告 期 中 進 度 報 告 畫 類 別 : 個 別 型 計 畫 半 導 體 產 業 大 型 廠 房 之 設 施 規 劃 計 畫 編 號 :NSC 96-2628-E-009-026-MY3 執 行 期 間 : 2007 年 8 月 1 日 至 2010 年 7 月 31 日 計 畫 主 持 人 : 巫 木 誠 共 同

More information

Cloud-based Social Application Deployment using Local Processing and Global Distribution

Cloud-based Social Application Deployment using Local Processing and Global Distribution Cloud-based Socal Applcaton Deployment usng Local Processng and Global Dstrbuton Zh Wang *, Baochun L, Lfeng Sun *, and Shqang Yang * * Bejng Key Laboratory of Networked Multmeda Department of Computer

More information

How To Plan A Network Wide Load Balancing Route For A Network Wde Network (Network)

How To Plan A Network Wide Load Balancing Route For A Network Wde Network (Network) Network-Wde Load Balancng Routng Wth Performance Guarantees Kartk Gopalan Tz-cker Chueh Yow-Jan Ln Florda State Unversty Stony Brook Unversty Telcorda Research kartk@cs.fsu.edu chueh@cs.sunysb.edu yjln@research.telcorda.com

More information

General Auction Mechanism for Search Advertising

General Auction Mechanism for Search Advertising General Aucton Mechansm for Search Advertsng Gagan Aggarwal S. Muthukrshnan Dávd Pál Martn Pál Keywords game theory, onlne auctons, stable matchngs ABSTRACT Internet search advertsng s often sold by an

More information

Schedulability Bound of Weighted Round Robin Schedulers for Hard Real-Time Systems

Schedulability Bound of Weighted Round Robin Schedulers for Hard Real-Time Systems Schedulablty Bound of Weghted Round Robn Schedulers for Hard Real-Tme Systems Janja Wu, Jyh-Charn Lu, and We Zhao Department of Computer Scence, Texas A&M Unversty {janjaw, lu, zhao}@cs.tamu.edu Abstract

More information

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12 14 The Ch-squared dstrbuton PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 1 If a normal varable X, havng mean µ and varance σ, s standardsed, the new varable Z has a mean 0 and varance 1. When ths standardsed

More information

Revenue Management for a Multiclass Single-Server Queue via a Fluid Model Analysis

Revenue Management for a Multiclass Single-Server Queue via a Fluid Model Analysis OPERATIONS RESEARCH Vol. 54, No. 5, September October 6, pp. 94 93 ssn 3-364X essn 56-5463 6 545 94 nforms do.87/opre.6.35 6 INFORMS Revenue Management for a Multclass Sngle-Server Queue va a Flud Model

More information

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(7):1884-1889 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A hybrd global optmzaton algorthm based on parallel

More information

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network *

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network * JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 24, 819-840 (2008) Data Broadcast on a Mult-System Heterogeneous Overlayed Wreless Network * Department of Computer Scence Natonal Chao Tung Unversty Hsnchu,

More information

M3S MULTIMEDIA MOBILITY MANAGEMENT AND LOAD BALANCING IN WIRELESS BROADCAST NETWORKS

M3S MULTIMEDIA MOBILITY MANAGEMENT AND LOAD BALANCING IN WIRELESS BROADCAST NETWORKS M3S MULTIMEDIA MOBILITY MANAGEMENT AND LOAD BALANCING IN WIRELESS BROADCAST NETWORKS Bogdan Cubotaru, Gabrel-Mro Muntean Performance Engneerng Laboratory, RINCE School of Electronc Engneerng Dubln Cty

More information

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis The Development of Web Log Mnng Based on Improve-K-Means Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna wangtngzhong2@sna.cn Abstract.

More information

Optimal Bidding Strategies for Generation Companies in a Day-Ahead Electricity Market with Risk Management Taken into Account

Optimal Bidding Strategies for Generation Companies in a Day-Ahead Electricity Market with Risk Management Taken into Account Amercan J. of Engneerng and Appled Scences (): 8-6, 009 ISSN 94-700 009 Scence Publcatons Optmal Bddng Strateges for Generaton Companes n a Day-Ahead Electrcty Market wth Rsk Management Taken nto Account

More information

Traffic State Estimation in the Traffic Management Center of Berlin

Traffic State Estimation in the Traffic Management Center of Berlin Traffc State Estmaton n the Traffc Management Center of Berln Authors: Peter Vortsch, PTV AG, Stumpfstrasse, D-763 Karlsruhe, Germany phone ++49/72/965/35, emal peter.vortsch@ptv.de Peter Möhl, PTV AG,

More information

Multiple-Period Attribution: Residuals and Compounding

Multiple-Period Attribution: Residuals and Compounding Multple-Perod Attrbuton: Resduals and Compoundng Our revewer gave these authors full marks for dealng wth an ssue that performance measurers and vendors often regard as propretary nformaton. In 1994, Dens

More information

J. Parallel Distrib. Comput. Environment-conscious scheduling of HPC applications on distributed Cloud-oriented data centers

J. Parallel Distrib. Comput. Environment-conscious scheduling of HPC applications on distributed Cloud-oriented data centers J. Parallel Dstrb. Comput. 71 (2011) 732 749 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. ournal homepage: www.elsever.com/locate/pdc Envronment-conscous schedulng of HPC applcatons

More information

Project Networks With Mixed-Time Constraints

Project Networks With Mixed-Time Constraints Project Networs Wth Mxed-Tme Constrants L Caccetta and B Wattananon Western Australan Centre of Excellence n Industral Optmsaton (WACEIO) Curtn Unversty of Technology GPO Box U1987 Perth Western Australa

More information

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College Feature selecton for ntruson detecton Slobodan Petrovć NISlab, Gjøvk Unversty College Contents The feature selecton problem Intruson detecton Traffc features relevant for IDS The CFS measure The mrmr measure

More information

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS 21 22 September 2007, BULGARIA 119 Proceedngs of the Internatonal Conference on Informaton Technologes (InfoTech-2007) 21 st 22 nd September 2007, Bulgara vol. 2 INVESTIGATION OF VEHICULAR USERS FAIRNESS

More information

A Replication-Based and Fault Tolerant Allocation Algorithm for Cloud Computing

A Replication-Based and Fault Tolerant Allocation Algorithm for Cloud Computing A Replcaton-Based and Fault Tolerant Allocaton Algorthm for Cloud Computng Tork Altameem Dept of Computer Scence, RCC, Kng Saud Unversty, PO Box: 28095 11437 Ryadh-Saud Araba Abstract The very large nfrastructure

More information

Present Values and Accumulations

Present Values and Accumulations Present Values an Accumulatons ANGUS S. MACDONALD Volume 3, pp. 1331 1336 In Encyclopea Of Actuaral Scence (ISBN -47-84676-3) Ete by Jozef L. Teugels an Bjørn Sunt John Wley & Sons, Lt, Chchester, 24 Present

More information

Period and Deadline Selection for Schedulability in Real-Time Systems

Period and Deadline Selection for Schedulability in Real-Time Systems Perod and Deadlne Selecton for Schedulablty n Real-Tme Systems Thdapat Chantem, Xaofeng Wang, M.D. Lemmon, and X. Sharon Hu Department of Computer Scence and Engneerng, Department of Electrcal Engneerng

More information

Forecasting the Direction and Strength of Stock Market Movement

Forecasting the Direction and Strength of Stock Market Movement Forecastng the Drecton and Strength of Stock Market Movement Jngwe Chen Mng Chen Nan Ye cjngwe@stanford.edu mchen5@stanford.edu nanye@stanford.edu Abstract - Stock market s one of the most complcated systems

More information

A Resource-trading Mechanism for Efficient Distribution of Large-volume Contents on Peer-to-Peer Networks

A Resource-trading Mechanism for Efficient Distribution of Large-volume Contents on Peer-to-Peer Networks A Resource-tradng Mechansm for Effcent Dstrbuton of Large-volume Contents on Peer-to-Peer Networks SmonG.M.Koo,C.S.GeorgeLee, Karthk Kannan School of Electrcal and Computer Engneerng Krannet School of

More information

Coordinated Denial-of-Service Attacks in IEEE 802.22 Networks

Coordinated Denial-of-Service Attacks in IEEE 802.22 Networks Coordnated Denal-of-Servce Attacks n IEEE 82.22 Networks Y Tan Department of ECE Stevens Insttute of Technology Hoboken, NJ Emal: ytan@stevens.edu Shamk Sengupta Department of Math. & Comp. Sc. John Jay

More information

Dominant Resource Fairness in Cloud Computing Systems with Heterogeneous Servers

Dominant Resource Fairness in Cloud Computing Systems with Heterogeneous Servers 1 Domnant Resource Farness n Cloud Computng Systems wth Heterogeneous Servers We Wang, Baochun L, Ben Lang Department of Electrcal and Computer Engneerng Unversty of Toronto arxv:138.83v1 [cs.dc] 1 Aug

More information

Dynamic Online-Advertising Auctions as Stochastic Scheduling

Dynamic Online-Advertising Auctions as Stochastic Scheduling Dynamc Onlne-Advertsng Auctons as Stochastc Schedulng Isha Menache and Asuman Ozdaglar Massachusetts Insttute of Technology {sha,asuman}@mt.edu R. Srkant Unversty of Illnos at Urbana-Champagn rsrkant@llnos.edu

More information

An Energy-Efficient Data Placement Algorithm and Node Scheduling Strategies in Cloud Computing Systems

An Energy-Efficient Data Placement Algorithm and Node Scheduling Strategies in Cloud Computing Systems 2nd Internatonal Conference on Advances n Computer Scence and Engneerng (CSE 2013) An Energy-Effcent Data Placement Algorthm and Node Schedulng Strateges n Cloud Computng Systems Yanwen Xao Massve Data

More information

Price Competition in an Oligopoly Market with Multiple IaaS Cloud Providers

Price Competition in an Oligopoly Market with Multiple IaaS Cloud Providers Prce Competton n an Olgopoly Market wth Multple IaaS Cloud Provders Yuan Feng, Baochun L, Bo L Department of Computng, Hong Kong Polytechnc Unversty Department of Electrcal and Computer Engneerng, Unversty

More information

Online Auctions in IaaS Clouds: Welfare and Profit Maximization with Server Costs

Online Auctions in IaaS Clouds: Welfare and Profit Maximization with Server Costs Onlne Auctons n IaaS Clouds: Welfare and roft Maxmzaton wth Server Costs aox Zhang Dept. of Computer Scence The Unvety of Hong Kong xxzhang@cs.hku.hk Zongpeng L Dept. of Computer Scence Unvety of Calgary

More information

Chapter 4 ECONOMIC DISPATCH AND UNIT COMMITMENT

Chapter 4 ECONOMIC DISPATCH AND UNIT COMMITMENT Chapter 4 ECOOMIC DISATCH AD UIT COMMITMET ITRODUCTIO A power system has several power plants. Each power plant has several generatng unts. At any pont of tme, the total load n the system s met by the

More information

SIMULATION OPTIMIZATION: APPLICATIONS IN RISK MANAGEMENT

SIMULATION OPTIMIZATION: APPLICATIONS IN RISK MANAGEMENT Internatonal Journal of Informaton Technology & Decson Makng Vol. 7, No. 4 (2008) 571 587 c World Scentfc Publshng Company SIMULATION OPTIMIZATION: APPLICATIONS IN RISK MANAGEMENT MARCO BETTER and FRED

More information

Survey on Virtual Machine Placement Techniques in Cloud Computing Environment

Survey on Virtual Machine Placement Techniques in Cloud Computing Environment Survey on Vrtual Machne Placement Technques n Cloud Computng Envronment Rajeev Kumar Gupta and R. K. Paterya Department of Computer Scence & Engneerng, MANIT, Bhopal, Inda ABSTRACT In tradtonal data center

More information