Off-line and on-line scheduling on heterogeneous master-slave platforms

Size: px
Start display at page:

Download "Off-line and on-line scheduling on heterogeneous master-slave platforms"

Transcription

1 Laboratore de Informatque du Paraésme Écoe Normae Supéreure de Lyon Unté Mxte de Recherche CNRS-INRIA-ENS LYON-UCBL n o 5668 Off-ne and on-ne schedung on heterogeneous master-save patforms Jean-Franços Pneau, Yves Robert, Frédérc Vven Juy 2005 Research Report N o Écoe Normae Supéreure de Lyon 46 Aée d Itae, Lyon Cedex 07, France Tééphone : +33(0) Téécopeur : +33(0) Adresse éectronque : p@ens-yon.fr

2 Off-ne and on-ne schedung on heterogeneous master-save patforms Jean-Franços Pneau, Yves Robert, Frédérc Vven Juy 2005 Abstract In ths wor, we dea wth the probem of schedung ndependent tass on heterogeneous master-save patforms. We target both off-ne and on-ne probems, wth severa obectve functons (maespan, maxmum response tme, tota competon tme). On the theoretca sde, our resuts are two-fod: () For off-ne schedung, we prove severa optmaty resuts for probems wth reease dates; () For on-ne schedung, we estabsh ower bounds on the compettve rato of any determnstc agorthm. On the practca sde, we have mpemented severa heurstcs, some cassca and some new ones derved n ths paper. We studed expermentay these heurstcs on a sma but fuy heterogeneous MPI patform. Our resuts show the superorty of those heurstcs whch fuy tae nto account the reatve capacty of the communcaton ns. Keywords: Schedung, Master-save patforms, Heterogeneous computng, On-ne, Reease dates. Résumé Nous nous ntéressons c au probème de ordonnancement d un ensembe de tâches ndépendantes sur une pate-forme maître escave hétérogène. Nous consdérons es probèmes en-gne (ou à a voée) et hors-gne, pour des fonctons obectves dfférentes (durée totae d exécuton, temps de réponse maxmum, temps de réponse moyen). D un pont de vue théorque, nous obtenons deux types de résutats : () pour e probème hors-gne, nous avons étab puseurs résutats d optmaté pour des probèmes avec dates d arrvée ; () pour e probème en-gne, nous avons étab des bornes nféreures sur e facteur de compéttvté des agorthmes détermnstes. D un pont de vue pratque, nous avons mpémenté puseurs heurstques, certanes cassques, d autres ssues du présent trava. Nous avons étudé expérmentaement ces heurstques sur une pette pate-forme MPI totaement hétérogène. Les résutats expérmentaux montrent a supérorté des heurstques qu prennent compètement en compte es capactés reatves des dfférents ens de communcaton. Mots-cés: Ordonnancement en gne, Ordonnancement hors-gne, Cacu hétérogène, Pate-forme maître-escave

3 Off-ne and on-ne schedung on heterogeneous master-save patforms 1 1 Introducton In ths paper, we dea wth the probem of schedung ndependent tass on a heterogeneous mastersave patform. We assume that ths patform s operated under the one-port mode, where the master can communcate wth a snge save at any tme-step. Ths mode s much more reastc than the standard mode from the terature, where the number of smutaneous messages nvovng a processor s not bounded. However, very few compexty resuts are nown for ths mode (see Secton 7 for a short survey). The maor obectve of ths paper s to assess the dffcuty of off-ne and on-ne schedung probems under the one-port mode. We dea wth probems where a tass have the same sze. Otherwse, even the smpe probem of schedung wth two dentca saves, wthout payng any cost for the communcatons from the master, s NP-hard [12]. Assume that the patform s composed of a master and m saves P 1, P 2,..., P m. Let c be the tme needed by the master to send a tas to P, and et p be the tme needed by P to execute a tas. Our man resuts are the foowng: When the patform s fuy homogeneous (c = c and p = p for a ), we desgn an agorthm whch s optma for the on-ne probem and for three dfferent obectve functons (maespan, maxmum response tme, tota competon tme). When the communcatons are homogeneous (c = c for a, but dfferent vaues of p ), we desgn an optma maespan mnmzaton agorthm for the off-ne probem wth reease dates. Ths agorthm generazes, and provdes a new proof of, a resut of Smons [27]. When the computatons are homogeneous (p = p for a, but dfferent vaue of c ), we faed to derve an optma maespan mnmzaton agorthm for the off-ne probem wth reease dates, but we provde an effcent heurstc for ths probem. For these ast two scenaros (homogeneous communcatons and homogeneous computatons), we show that there does not exst any optma on-ne agorthm. Ths hods true for the prevous three obectve functons (maespan, maxmum response tme, tota competon tme). We even estabsh ower bounds on the compettve rato of any determnstc agorthm. The man contrbutons of ths paper are mosty theoretca. However, on the practca sde, we have mpemented severa heurstcs, some cassca and some new ones derved n ths paper, on a sma but fuy heterogeneous MPI patform. Our (premnary) resuts show the superorty of those heurstcs whch fuy tae nto account the reatve capacty of the communcaton ns. The rest of the paper s organzed as foows. In Secton 2, we state some notatons for the schedung probems under consderaton. Secton 3 deas wth fuy homogeneous patforms. We study communcaton-homogeneous patforms n Secton 4, and computaton-homogeneous patforms n Secton 5. We provde an expermenta comparson of severa schedung heurstcs n Secton 6. Secton 7 s devoted to an overvew of reated wor. Fnay, we state some concudng remars n Secton 8. 2 Framewor To be consstent wth the terature [16, 9], we use the notaton α β γ where: α: the patform As n the standard, we use P for patforms wth dentca processors, and Q for patforms wth dfferent-speed processors 1. We add MS to ths fed to ndcate that we wor wth master-save patforms. β: the constrants We wrte on-ne for on-ne probems, and r when there are reease dates. We wrte c = c for communcaton-homogeneous patforms, and p = p for computatonhomogeneous patforms. 1 As we ony target sets of same-sze tass, we aways fa under the unform processors framewor. In other words, the executon tme of a tas on a processor w ony depend on the processor runnng t and not on the tas.

4 2 J.-F. Pneau, Y. Robert, F. Vven γ: the obectve We et C denote the end of the executon of tas. We dea wth three obectve functons: the maespan (tota executon tme) max C ; the maxmum response tme (or maxmum fow) max C r : ndeed, C r s the tme spent by tas n the system; the tota competon tme C, whch s equvaent to the sum of the response tmes (C r ). 3 Fuy homogeneous patforms For fuy homogeneous patforms, we are abe to prove the optmaty of the Round-Robn agorthm whch processes the tass n the order of ther arrva, and whch assgns them n a cycc fashon to processors: Theorem 1. The Round-Robn agorthm s optma for the probem P, MS onne, r, p = p, c = c (C r ), as we as for the mnmzaton of the maespan and of the maxmum response tme. We pont out that the compexty of the Round-Robn agorthm s near n the number of tass and does not depend upon the patform sze. Proof. To prove that the greedy agorthm Round-Robn s optma for our probem, we show that there s an optma schedue under whch the executon of each tas starts at the exact same date than under Round-Robn. To prove ths, we frst show two resuts statng that we can focus on certan partcuar optma schedues. 1. There s an optma schedue such that the master sends the tass to the saves n the order of ther arrva. We prove ths resut wth permutaton arguments. Let S be an optma schedue not verfyng the desred property. Remember that the master use ts communcaton ns n a sequenta fashon. Then we denote by r the date at whch the tas arrves on a save. By hypothess on S, there are two tass, and, such that arrves on the master before, but s sent to a processor save after. So: r < r and r < r. We then defne from S a new schedue S as foows: If the tas was nevertheess treated earer than the tas (.e., f C C ), then we smpy reverse the dspatch dates of tass and, but do not change the processors where they are computed. Ths s ustrated on Fgure 1. In ths case, the remander of the schedue s et unaffected, and the tota fow remans the same (ust as the maespan, and the maxmum fow). If the tas was processed ater than the tas,.e., f C > C, then we send the tas to the processor that was recevng under S, at the tme tas was sent to that processor, and conversey. Ths s ustrated on Fgure 2. Snce the tass and have the same sze, the use of the processors w be the same, and the remander of the schedue w reman unchanged. One obtans a new schedue S, havng as tota fow: n, (C r ) + (C r ) + (C r ) = n (C r ) (1)

5 Off-ne and on-ne schedung on heterogeneous master-save patforms 3 Therefore, ths s aso an optma schedue. In the same way, the maespan as we as the maxmum fow are unchanged. P P P P (a) Before permutaton (b) After permutaton Fgure 1: Permutaton on the optma schedue S (case C C ). P P P P (a) Before permutaton (b) After permutaton Fgure 2: Permutaton on the optma schedue S (case C > C ). By teratng ths process, we obtan an optma schedue where the master sends the tass accordng to ther arrva dates,.e., by ncreasng r s. Indeed, f one consders the set of the coupes {(, ) r < r & r < r }, we notce that each teraton of the process strcty ncreases the sze of ths set. 2. There s an optma schedue such that the master sends the tass to the saves n the order of ther arrva, and such that the tass are executed n the order of ther arrva. We w permute tass to bud an optma schedue satsfyng ths property from a schedue satsfyng the property stated n pont 1. Let S be an optma schedue n whch tass are sent by the master n the order of ther arrva. From the above study, we now that such a schedue exsts. Let us suppose that S does not satsfy the desred property. Then, there are two tass and, such that r r, r < r, and C > C. Then we defne a new schedue S by ust exchangng the processors to whch the tass and were sent. Then, the tas s computed under S at the tme when was computed under S, and conversey. Ths way, we obtan the same tota fow ((C r ) + (C r ) = (C r )+(C r )), the same maespan (snce the worng tmes of the processors remans unchanged), whereas the maxmum fow can decrease. Among the optma schedues whch respect the property stated n pont 2, we now oo at the subset of the soutons computng the frst tas as soon as possbe. Then, among ths subset, we

6 4 J.-F. Pneau, Y. Robert, F. Vven oo at the soutons computng the second tas as soon as possbe. And so on. Ths way, we defne from the set of a optma schedues an optma souton, denoted ASAP, whch processes the tass n the order of ther arrva, and whch processes each of them as soon as possbe. We w now compare ASAP wth the schedue Round-Robn, formay defned as foows: under Round-Robn the tas s sent to the processor mod m as soon as possbe, whe respectng the order of arrva of the tass. 3. The computaton of any tas starts at the same tme under the schedues ASAP and Round-Robn. The demonstraton s done by nducton on the number of tass. Round-Robn sends the frst tas as soon as possbe, ust as ASAP does. Let us suppose now that the frst tass satsfy the property. Let us oo at the behavor of Round-Robn on the arrva of the ( + 1)-th tas. The computaton of the ( + 1)-th tas starts at tme: RR( + 1) = max { r +1, RR( + 1 m) + p}. Indeed, ether the processor s avaabe at the tme the tas arrves on the save, and the tas executon starts as soon as the tas arrves,.e., at tme r +1, or the processor s busy when the tas arrves. In the atter case, the processor w be avaabe when the ast tas t prevousy receved (.e., the ( + 1 m)-th tas accordng to the Round-Robn strategy) w be competed, at tme RR( + 1 m) + p. Therefore, f RR( + 1) = r +1, Round-Robn remans optma, snce the tas s processed as soon as t s avaabe on a save, and snce t was sent as soon as possbe. Otherwse, RR( + 1) = RR( + 1 m) + p. But, by nducton hypothess, we now that λ, 1 λ m, RR( + 1 λ) = ASAP ( + 1 λ). Furthermore, thans to the Round-Robn schedung pocy, we now that, RR() RR( + 1). Therefore: λ, 1 λ m, RR( + 1 m) RR( + 1 λ) < RR( + 1 m) + p = RR( + 1) Ths mpes that, between RR( + 1 m) and RR(), m tass of sze p were started, under Round-Robn, and aso under ASAP because of the nducton hypothess. Therefore, durng that tme nterva, m saves were seected. Then, unt the date RR( + 1 m) + p, a the saves are used and, thus, the tas + 1 s aunched as soon as possbe by Round-Robn, nowng that ASAP coud not have aunched t earer. Therefore, ASAP (+1) = RR(+1). We can concude. We have aready stated that the demonstratons of ponts 1 and 2 are vad for schedues mnmzng ether maespan, tota fow, or maxmum fow. The reasonng foowed n the demonstraton of pont 3 s ndependent from the obectve functon. Therefore, we demonstrated the optmaty of Round-Robn for these three obectve functons. 4 uncaton-homogeneous patforms In ths secton, we have c = c but dfferent-speed processors. We order them so that P 1 s the fastest processor (p 1 s the smaest computng tme p ), whe P m s the sowest processor. 4.1 On-ne schedung Theorem 2. There s no schedung agorthm for the probem P, MS onne, r, p, c = c max C wth a compettve rato ess than Proof. Suppose the exstence of an on-ne agorthm A wth a compettve rato ρ = ɛ, wth ɛ > 0. We w bud a patform and study the behavor of A opposed to our adversary. The patform conssts of two processors, where p 1 = 2, p 2 = , and c = 1.

7 Off-ne and on-ne schedung on heterogeneous master-save patforms 5 Intay, the adversary sends a snge tas at tme 0. A sends the tas ether on P 1, achevng a maespan at east equa to 3, or on P 2, wth a maespan at east equa to At tme-step 1, we chec f A made a decson concernng the schedung of, and the adversary reacts consequenty: 1. If A dd not begn the sendng of the tas, the adversary does not send other tass. The best maespan s then 4. As the optma maespan s 3, we have a compettve rato of 4 3 > Ths refutes the assumpton on ρ. Thus the agorthm A must have schedued the tas at tme If A schedued the tas on P 2 the adversary does not send other tass. The best possbe maespan s then equa to , whch s even worse than the prevous case. Consequenty, agorthm A does not have another choce than to schedue the tas on P 1. At tme-step 1, the adversary sends another tas,. In ths case, we oo, at tme-step 2, at the assgnment A made for : 1. If s sent on P 2, the adversary does not send any more tas. The best achevabe maespan s then , whereas the optma s 5. The compettve rato s then > ρ. 2. If s sent on P 1 the adversary sends a ast tas at tme-step 2. The best possbe maespan s then , whereas the optma s The compettve rato s st , hgher than ρ. Remar 1. Smary, we can show that there s no on-ne schedung for the probem P, MS onne, r, p, c = c C whose compettve rato ρ s strcty ower than , and that there s no on-ne schedung for the probem P, MS onne, r, p, c = c max (C r ) whose compettve rato ρ s strcty ower than Off-ne schedung In ths secton, we am at desgnng an optma agorthm for the off-ne verson of the probem, wth reease dates. We target the obectve max C. Intutvey, to mnmze the competon date of the tas arrvng ast, t s necessary to aocate ths tas to the fastest processor (whch w fnsh t the most rapdy). However, the other tass shoud aso be assgned so that ths fastest processor w be avaabe as soon as possbe for the tas arrvng ast. We defne the greedy agorthm SLJF (Schedung Last Jobs Frst) as foows: Intazaton Tae the ast tas whch arrves n the system and aocate t to the fastest processor (Fgure 3(a)). Schedung bacwards Among the not-yet-aocated tass, seect the one whch arrved atest n the system. Assgn t, wthout tang ts arrva date nto account, to the processor whch w begn ts executon at the atest, but wthout exceedng the competon date of the prevousy schedued tas (Fgure 3(b)). Memorzaton Once a tass are aocated, record the assgnment of the tass to the processors (Fgure 3(c)). Assgnment The master sends the tass accordng to ther arrva dates, as soon as possbe, to the processors whch they have been assgned to n the prevous step (Fgure 3(d)). Theorem 3. SLJF s an optma agorthm for the probem Q, MS r, p, c = c max C.

8 6 J.-F. Pneau, Y. Robert, F. Vven P 2 : p = 4 P 2 : p = 3 P 1 : p = 2 P 2 : p = 4 P 2 : p = 3 P 1 : p = 2 (a) Intazaton (b) Schedung bacwards P 3 : p = 4 P 2 : p = 3 P 1 : p = 2 P 2 : p = 4 P 2 : p = 3 P 1 : p = 2 P 1 P 3 P 2 P 1 Arrva:, (c) Memorzaton (d) Assgnment Fgure 3: Dfferent steps of the SLJF agorthm, wth four tass,,, and. Proof. The frst three phases of the SLJF agorthm are ndependent of the reease dates, and ony depend on the number of tass whch w arrve n the system. The proof proceeds n three steps. Frst we study the probem wthout communcaton costs, nor reease dates. Next, we tae reease dates nto account. Fnay, we extend the resut to the case wth communcatons. The second step s the most dffcut. For the frst step, we have to mnmze the maespan durng the schedung of dentca tass wth heterogeneous processors, wthout reease dates. Wthout communcaton costs, ths s a wenown oad baancng, probem, whch can be soved by a greedy agorthm [6]. The schedung bacwards phase of SLJF soves ths oad baancng probem optmay. Snce the probem s wthout reease dates, the memorzaton phase does not ncrease the maespan, whch thus remans optma. Next we add the constrants of reease dates. To show that SLJF s optma, we proceed by nducton on the number of tass. For a snge tas, t s obvous that the addton of a reease date does not change anythng about the optmaty of the souton. Let us suppose the agorthm optma for n tass, or ess. Then oo at the behavor of the agorthm to process n + 1 tass. If the addton of the reease dates does not ncrease the maespan compared to that obtaned durng the memorzaton step, then an optma schedung s obtaned. If not, et us oo once agan at the probem startng from the end. Compare the competon tmes of the tass n the schedung of the memorzaton phase (denoted as (C n C ) memo ), and n the assgnment phase (denoted as (C n C ) fna ). If both maespans are equa, we are fnshed. Otherwse, there are tass such that (C n C ) memo < (C n C ) fna. Let be the ast tas satsfyng ths property. In ths case, the schedung of the (n 1) ast tass corresponds to SLJF n the case of (n 1) tass, when the frst tas arrves at tme r +1 (see Fgure 4). And snce s the ast tas satsfyng the above property, we are sure that the processors are free at the expected tmes. Usng the nducton hypothess, schedung s thus optma from r +1, and tas + 1 cannot begn ts computaton

9 Off-ne and on-ne schedung on heterogeneous master-save patforms 7 Arrva:, P 3 : p = 4 P 2 : p = 3 P 1 : p = 2 P 1 C C = 2 P 3 C C = 0 Maespan = 4 P 2 C C = 0 (a) Schedung bacwards P 1 C C = 0 P 2 : p = 4 P 2 : p = 3 P 1 : p = 2 SLJF (3) C C > 2 C C > 0 C C = 0 C C = 0 (b) Assgnment Fgure 4: Detang the ast two phases of the SLJF agorthm. earer. The whoe schedung s thus optma. Fnay, SLJF s optma to mnmze the maespan n the presence of reease dates. Tang communcatons nto account s now easy. Under the one-port mode, wth a unform communcaton tme for a tass and processors, the optma pocy of the master conssts n sendng the tass as soon as they arrve. Now, we can consder the dates at whch the tass are avaabe on the saves, and consder them as reease dates for a probem wthout communcatons. Remar 2. It shoud be stressed that, by posng c = 0, our approach aows to provde a new proof to the resut of Barbara Smons [27]. 5 Computaton-homogeneous patforms In ths secton, we have p = p but processor ns wth dfferent capactes. We order them, so that P 1 s the fastest communcatng processor (c 1 s the smaest computng tme c ). 5.1 On-ne schedung Just as n Secton 4, we can bound the compettve rato of any determnstc agorthm: Theorem 4. There s no schedung agorthm for the probem P, MS onne, r, p = p, c max C whose compettve rato ρ s strcty ower than 6 5. Proof. Assume that there exsts a determnstc on-ne agorthm A whose compettve rato s ρ = 6 5 ɛ, wth ɛ > 0. We w bud a patform and an adversary to derve a contradcton. The patform s made up wth two processors P 1 and P 2 such that p 1 = p 2 = p = max{5, 12 25ɛ }, c 1 = 1 and c 2 = p 2. Intay, the adversary sends a snge tas at tme 0. A executes the tas, ether on P 1 wth a maespan at east 2 equa to 1 + p, or on P 2 wth a maespan at east equa to 3p 2. At tme-step p 2, we chec whether A made a decson concernng the schedung of, and whch one: 1. If A schedued the tas on P 2 the adversary does not send other tass. The best possbe maespan s then 3p 2. The optma schedung beng of maespan 1+p, we have a compettve 2 Nothng forces A to send the tas as soon as possbe.

10 8 J.-F. Pneau, Y. Robert, F. Vven rato of ρ 3p p = (p + 1) > 6 5 because p 5 by assumpton. Ths contradcts the hypothess on ρ. Thus the agorthm A cannot schedue tas on P If A dd not begn to send the tas, the adversary does not send other tass. The best maespan that can be acheved s then equa to p 2 + (1 + p) = 1 + 3p 2, whch s even worse than the prevous case. Consequenty, the agorthm A does not have any other choce than to schedue tas on P 1. At tme-step p 2, the adversary sends three tass,, and. No schedue whch executes three of the four tass on the same processor can have a maespan ower than 1+3p (mnmum duraton of a communcaton and executon wthout deay of the three tass). We now consder the schedues whch compute two tass on each processor. Snce s computed on P 1, we have three cases to study, dependng upon whch other tas (,, or ) s computed on P 1 : 1. If s computed on P 1 : (a) Tas s sent to P 1 durng the nterva [0, 1] and s computed durng the nterva [1, 1+p]. (b) Tas s sent to P 1 durng the nterva [ p 2, 1 + p 2 ] and s computed durng the nterva [1 + p, 1 + 2p]. (c) Tas s sent to P 2 durng the nterva [1+ p 2, 1+p] and s computed durng the nterva [1 + p, 1 + 2p]. (d) Tas s sent to P 2 durng the nterva [1+p, 1+ 3p 2 ] and s computed durng the nterva [1 + 2p, 1 + 3p]. The maespan of ths schedue s then 1 + 3p. 2. If s computed on P 1 : (a) Tas s sent to P 1 durng the nterva [0, 1] and s computed durng the nterva [1, 1+p]. (b) Tas s sent to P 2 durng the nterva [ p 2, p] and s computed durng the nterva [p, 2p]. (c) Tas s sent to P 1 durng the nterva [p, 1 + p] and s computed durng the nterva [1 + p, 1 + 2p]. (d) Tas s sent to P 2 durng the nterva [1+p, 1+ 3p 2 ] and s computed durng the nterva [2p, 3p]. The maespan of ths schedung s then 3p. 3. If s computed on P 1 : (a) Tas s sent to P 1 durng the nterva [0, 1] and s computed durng the nterva [1, 1+p]. (b) tas s sent to P 2 durng the nterva [ p 2, p] and s computed durng the nterva [p, 2p]. (c) Tas s sent to P 2 durng the nterva [p, 3p 2 ] and s computed durng the nterva [2p, 3p]. (d) Tas s sent to P 1 durng the nterva [ 3p 2, 1 + 3p 2 ] and s computed durng the nterva [1 + 3p 2, 1 + 5p 2 ]. The maespan of ths schedue s then 3p.

11 Off-ne and on-ne schedung on heterogeneous master-save patforms 9 Consequenty, the ast two schedues are equvaent and are better than the frst. Atogether, the best achevabe maespan s 3p. But a better schedue s obtaned when computng on P 2, then on P 1, then on P 2, and fnay on P 1. The maespan of the atter schedue s equa to 1 + 5p 2. The compettve rato of agorthm A s necessary arger than the rato of the best reachabe maespan (namey 3p) and the optma maespan, whch s not arger than 1 + 5p 2. Consequenty: ρ 3p 1 + 5p 2 = (5p + 2) > p 6 5 ɛ 2 whch contradcts the assumpton ρ = 6 5 ɛ wth ɛ > Off-ne schedung In the easy case where p c p, and wthout reease dates, Round-Robn s optma for maespan mnmzaton. But n the genera case, not a saves w be enroed n the computaton. Intutvey, the dea s to use the fastest m ns, where m s computed so that the tme p to execute a tas es between the tme necessary to send a tas on each of the fastest m 1 ns and the tme necessary to send a tas on each of the fastest m ns. Formay, m 1 c < p and c p. m Wth ony m ns seected n the patform, we am at dervng an agorthm smar to Round- Robn. But we dd not succeed n provng the optmaty of our approach. Hence the agorthm beow shoud rather be seen as a heurstc. The dffcuty es n decdng when to use the m -th processor. In addton to be the one havng the sowest communcaton n, ts use can cause a moment of nactvty on another processor, snce m 1 c + c m p. Our greedy agorthm w smpy compare the performances of two strateges, the one sendng tass ony on the m 1 frst processors, and the other usng the m -th processor at the best possbe moment. Let RRA be the agorthm sendng the tass to the m 1 fastest processors n a cycc way, startng wth the fastest processor, and schedung the tass n the reverse order, from the ast one to the frst one. Let RRB be the agorthm sendng the ast tas to processor m, then foowng the RRA pocy. We see that RRA sees to contnuousy use the processors, even though de tme may occur on the communcaton n, and on the processor P m. On the contrary, RRB tres to contnuousy use the communcaton n, despte eavng some processors de. The goba behavor of the greedy agorthm, SLJFWC (Schedung the Last Job Frst Wth uncaton) s as foows: Intazaton: Aocate the m 1 ast tass to the fastest m 1 processors, from the fastest to the sowest. Comparson: Compare the schedues RRA and RRB. If there are not enough tass to enforce the foowng stop and save condton, then eep the fastest pocy (see Fgure 5). Stop and save: After (m 1) + 1 aocated tass ( 2), f (see Fgure 6) { m 1 c + c m > p ( + 1) m 1 c + c m ( + 1)p then eep the tas assgnment of RRB for the ast (m 1) + 1 tass, and start agan the comparson phase for the remanng tass. If not, proceed wth the comparson step. End: When the ast tas s treated, eep the fastest pocy.

12 10 J.-F. Pneau, Y. Robert, F. Vven P 4 : c = 5 Intazaton P 3 : c = 4 q q n n P 2 : c = 2 p p m m P 1 : c = 1 o o p = 8 (a) Agorthm RRA Intazaton P 4 : c = 5 P 3 : c = 4 o o P 2 : c = 2 q n q n P 1 : c = 1 p m p m p = 8 (b) Agorthm RRB Fgure 5: Agorthms RRA and RRB wth 9 tass. The ntuton under ths agorthm s smpe. We now that f we ony have the m 1 fastest processors, then RRA s optma to mnmze the maespan. However, the tme necessary for sendng a tas on each of the m 1 processors s ower than p. Ths means that the sendng of the tass taes advances compared to ther executon. Ths advance, whch accumuates for a the m 1 tass, can become suffcenty arge to aow the sendng of a tas on another m-th processor, for free,.e. wthout deayng the treatment of the next tass to come on the other processors. 6 MPI experments 6.1 The expermenta patform We bud a sma heterogeneous master-save patform wth fve dfferent computers, connected to each other by a fast Ethernet swtch (100 Mbt/s). The fve machnes are a dfferent, both n terms of CPU speed and n the amount of avaabe memory. The heterogenety of the communcaton ns s many due to the dfferences between the networ cards. Each tas w be a matrx, and each save w have to cacuate the determnant of the matrces that t w receve. Whenever needed, we pay wth matrx szes so as to acheve more heterogenety n the CPU speeds or communcaton bandwdths. Beow we report experments for the foowng confguraton (n an arbtrary unt): c 1 = et p 1 = c 2 = et p 2 =

13 Off-ne and on-ne schedung on heterogeneous master-save patforms 11 m X 1 ( + 1) c + c m m X 2 c m X 1 c + cn P 4 : c = P 3 : c = P 2 : c = P 1 : c = p = 8 p m X 2 c ( + 1)p Fgure 6: The stop and save condton. c 3 = et p 3 = c 4 = et p 4 = Resuts Fgure 7 shows the maespan obtaned wth cassca schedung agorthms, such as SRPT (Shortest Remanng Processng Tme), Lst Schedung, and severa varants of Round-Robn, as we as wth SLJF and SLJF W C. In ths expernebt, a the tass to be schedued arrved at tme 0 (off-ne framewor wthout reease dates). Each pont on the fgure, representng the maespan of a schedue, corresponds n reaty to an average obtaned whe aunchng severa tmes the experment. We see that SLJF W C obtans good resuts. SLJF remans compettve, even f t was not desgned for a patform wth dfferent communcatons ns. Fgure 8 aso represents the average maespan of varous agorthms, but on a dfferent patform. Ths tme, the parameters were modfed by software n order to render the processors homogeneous. In ths case, SLJF W C s st better, and SLJF obtans poor performances. Fnay, Fgure 9 represents the average maespan n the presence of reease-dates. Agan, SLJF W C performs we, even though t was not desgned for probems wth reease-dates. 7 Reated wor We cassfy severa reated papers aong the foowng four man nes: Modes for heterogeneous patforms In the terature, one-port modes come n two varants. In the undrectona varant, a processor cannot be nvoved n more than one communcaton at a gven tme-step, ether a send or a receve. In the bdrectona mode, a processor can send and receve n parae, but at most to a gven neghbor n each drecton. In both varants, f P u sends a message to P v, both P u and P v are boced throughout the communcaton. The bdrectona one-port mode s used by Bhat et a [7, 8] for fxed-sze messages. They advocate ts use because current hardware and software do not easy enabe mutpe messages to be transmtted smutaneousy. Even f non-bocng mut-threaded communcaton brares aow for ntatng mutpe send and receve operatons, they cam that a these operatons are eventuay serazed by the snge hardware port to the networ. Expermenta evdence of ths fact has recenty been reported by Saf and Parashar [24], who report

14 12 J.-F. Pneau, Y. Robert, F. Vven 2.4e e+07 2e e+07 SLJFWC SLJF SRPT Lst Schedung Round Robn Round Robn Round Robn Proc Maespan 1.6e e e+07 1e+07 8e+06 6e Number of tass Fgure 7: Comparng the maespan of severa agorthms. that asynchronous MPI sends get serazed as soon as message szes exceed a few megabytes. Ther resuts hod for two popuar MPI mpementatons, MPICH on Lnux custers and IBM MPI on the SP2. The one-port mode fuy accounts for the heterogenety of the patform, as each n has a dfferent bandwdth. It generazes a smper mode studed by Banazem et a. [1], Lu [19], and Khuer and Km [15]. In ths smper mode, the communcaton tme ony depends on the sender, not on the recever: n other words, the communcaton speed from a processor to a ts neghbors s the same. Fnay, we note that some papers [2, 3] depart from the one-port mode as they aow a sendng processor to ntate another communcaton whe a prevous one s st on-gong on the networ. However, such modes nsst that there s an overhead tme to pay before beng engaged n another operaton, so there are not aowng for fuy smutaneous communcatons. Tas graph schedung Tas graph schedung s usuay studed usng the so-caed macrodatafow mode [20, 26, 10, 11], whose maor faw s that communcaton resources are not mted. In ths mode, a processor can send (or receve) any number of messages n parae, hence an unmted number of communcaton ports s assumed (ths expans the name macro-datafow for the mode). Aso, the number of messages that can smutaneousy crcuate between processors s not bounded, hence an unmted number of communcatons can smutaneousy occur on a gven n. In other words, the communcaton networ s assumed to be contenton-free, whch of course s not reastc as soon as the processor number exceeds a few unts. More recent papers [29, 21, 23, 4, 5, 28] tae communcaton resources nto account. Hoermann et a. [13] and Hsu et a. [14] ntroduce the foowng mode for tas graph schedung: each processor can ether send or receve a message at a gven tme-step (bd-

15 Off-ne and on-ne schedung on heterogeneous master-save patforms 13 7e+06 6e+06 5e+06 SLJFWC SLJF Round Robn Proc SRPT Lst Schedung Round Robn Round Robn Maespan 4e+06 3e+06 2e+06 1e Number of tass Fgure 8: Maespan on a patform wth homogeneous saves. rectona communcaton s not possbe); aso, there s a fxed atency between the ntaton of the communcaton by the sender and the begnnng of the recepton by the recever. St, the mode s rather cose to the one-port mode dscussed n ths paper. On-ne schedung A good survey of on-ne schedung can be found n [25, 22]. Two papers focus on the probem of on-ne schedung for master-saves patforms. In [17], Leung and Zhao proposed severa compettve agorthms mnmzng the tota competon tme on a master-save patform, wth or wthout pre- and post-processng. In [18], the same authors studed the compexty of mnmzng the maespan or the tota response tme, and proposed some heurstcs. However, none of these wors tae nto consderaton communcaton costs. 8 Concuson In ths paper, we have deat wth the probem of schedung ndependent, same-sze tass on mastersave patforms. We enforce the one-port mode, and we study the mpact of the communcatons on the desgn and anayss of the proposed agorthms. On the theoretca sde, we have derved severa new resuts, ether for on-ne schedung, or for off-ne schedung wth reease dates. There are two mportant drectons for future wor. Frst, the bounds on the compettve rato that we have estabshed for on-ne schedung on communcatonhomogeneous, and computaton-homogeneous patforms, are ower bounds: t woud be very nterestng to see whether these bounds can be met, and to desgn the correspondng approxmaton agorthms. Second, there remans to derve an optma agorthm for off-ne schedung wth reease dates on computaton-homogeneous patforms. On the practca sde, we have to wden the scope of the MPI experments. A detaed comparson of a the heurstcs that we have mpemented needs to be conducted on sgnfcanty arger patforms (wth severa tens of saves). Such a comparson woud, we beeve, further demonstrate the superorty of those heurstcs whch fuy tae nto account the reatve capacty of the

16 14 J.-F. Pneau, Y. Robert, F. Vven 3.5e+07 3e+07 SLJFWC SLJF SRPT Lst Schedung Round Robn Round Robn Proc Round Robn 2.5e+07 Maespan 2e e+07 1e Number of tass Fgure 9: Maespan wth reease dates. communcaton ns. References [1] M. Banazem, V. Moorthy, and D. K. Panda. Effcent coectve communcaton on heterogeneous networs of worstatons. In Proceedngs of the 27th Internatona Conference on Parae Processng (ICPP 98). IEEE Computer Socety Press, [2] M. Banazem, J. Sampathumar, S. Prabhu, D.K. Panda, and P. Sadayappan. uncaton modeng of heterogeneous networs of worstatons for performance characterzaton of coectve operatons. In HCW 99, the 8th Heterogeneous Computng Worshop, pages IEEE Computer Socety Press, [3] Amotz Bar-Noy, Sudpto Guha, Joseph (Seff) Naor, and Baruch Scheber. Message mutcastng n heterogeneous networs. SIAM Journa on Computng, 30(2): , [4] Over Beaumont, Vncent Boudet, and Yves Robert. A reastc mode and an effcent heurstc for schedung wth heterogeneous processors. In HCW 2002, the 11th Heterogeneous Computng Worshop. IEEE Computer Socety Press, [5] Over Beaumont, Arnaud Legrand, and Yves Robert. A poynoma-tme agorthm for aocatng ndependent tass on heterogeneous for-graphs. In ISCIS XVII, Seventeenth Internatona Symposum On Computer and Informaton Scences, pages CRC Press, [6] Over Beaumont, Arnaud Legrand, and Yves Robert. The master-save paradgm wth heterogeneous processors. IEEE Trans. Parae Dstrbuted Systems, 14(9): , 2003.

17 Off-ne and on-ne schedung on heterogeneous master-save patforms 15 [7] P.B. Bhat, C.S. Raghavendra, and V.K. Prasanna. Effcent coectve communcaton n dstrbuted heterogeneous systems. In ICDCS 99 19th Internatona Conference on Dstrbuted Computng Systems, pages IEEE Computer Socety Press, [8] P.B. Bhat, C.S. Raghavendra, and V.K. Prasanna. Effcent coectve communcaton n dstrbuted heterogeneous systems. Journa of Parae and Dstrbuted Computng, 63: , [9] J. Bazewcz, J.K. Lenstra, and A.H. Kan. Schedung subect to resource constrants. Dscrete Apped Mathematcs, 5:11 23, [10] P. Chrétenne, E. G. Coffman Jr., J. K. Lenstra, and Z. Lu, edtors. Schedung Theory and ts Appcatons. John Wey and Sons, [11] H. E-Rewn, H. H. A, and T. G. Lews. Tas schedung n mutprocessng systems. Computer, 28(12):27 37, [12] M. R. Garey and D. S. Johnson. Computers and Intractabty, a Gude to the Theory of NP-Competeness. W. H. Freeman and Company, [13] L. Hoermann, T. S. Hsu, D. R. Lopez, and K. Vertanen. Schedung probems n a practca aocaton mode. J. Combnatora Optmzaton, 1(2): , [14] T. S. Hsu, J. C. Lee, D. R. Lopez, and W. A. Royce. Tas aocaton on a networ of processors. IEEE Trans. Computers, 49(12): , [15] S. Khuer and Y.A. Km. On broadcastng n heterogenous networs. In Proceedngs of the ffteenth annua ACM-SIAM symposum on Dscrete agorthms, pages Socety for Industra and Apped Mathematcs, [16] J.K. Lenstra, R. Graham, E. Lawer, and A.H. Kan. Optmzaton and approxmaton n determnstc sequencng and schedung: a survey. Annas of Dscrete Mathematcs, 5: , [17] Joseph Y-T. Leung and Harong Zhao. Mnmzng tota competon tme n master-save systems, Avaabe at [18] Joseph Y-T. Leung and Harong Zhao. Mnmzng mean fowtme and maespan on mastersave systems. J. Parae and Dstrbuted Computng, 65(7): , [19] P. Lu. Broadcast schedung optmzaton for heterogeneous custer systems. Journa of Agorthms, 42(1): , [20] M. G. Norman and P. Thansch. Modes of machnes and computaton for mappng n mutcomputers. ACM Computng Surveys, 25(3): , [21] J. M. Orduna, F. Sa, and J. Duato. A new tas mappng technque for communcaton-aware schedung strateges. In T. M. Pnston, edtor, Worshop for Schedung and Resource Management for Custer Computng (ICPP 01), pages IEEE Computer Socety Press, [22] Ir Pruhs, Jr Sga, and Erc Torng. On-ne schedung. In J. Leung, edtor, Handboo of Schedung: Agorthms, Modes, and Performance Anayss, pages CRC Press, [23] C. Rog, A. Rpo, M. A. Senar, F. Gurado, and E. Luque. Improvng statc schedung usng nter-tas concurrency measures. In T. M. Pnston, edtor, Worshop for Schedung and Resource Management for Custer Computng (ICPP 01), pages IEEE Computer Socety Press, 2001.

18 16 J.-F. Pneau, Y. Robert, F. Vven [24] T. Saf and M. Parashar. Understandng the behavor and performance of non-bocng communcatons n MPI. In Proceedngs of Euro-Par 2004: Parae Processng, LNCS 3149, pages Sprnger, [25] J. Sga. On ne schedung-a survey. In On-Lne Agorthms, Lecture Notes n Computer Scence 1442, pages Sprnger-Verag, Bern, [26] B. A. Shraz, A. R. Hurson, and K. M. Kav. Schedung and oad baancng n parae and dstrbuted systems. IEEE Computer Scence Press, [27] Barbara Smons. Mutprocessor schedung of unt-tme obs wth arbtrary reease tmes and deadnes. SIAM Journa on Computng, 12(2): , [28] Over Snnen and Leone Sousa. uncaton contenton n tas schedung. IEEE Trans. Parae Dstrbuted Systems, 16(6): , [29] M. Tan, H. J. Sege, J. K. Antono, and Y. A. L. Mnmzng the apcaton executon tme through schedung of subtass and communcaton traffc n a heterogeneous computng system. IEEE Transactons on Parae and Dstrbuted Systems, 8(8): , 1997.

Approximation Algorithms for Data Distribution with Load Balancing of Web Servers

Approximation Algorithms for Data Distribution with Load Balancing of Web Servers Approxmaton Agorthms for Data Dstrbuton wth Load Baancng of Web Servers L-Chuan Chen Networkng and Communcatons Department The MITRE Corporaton McLean, VA 22102 chen@mtreorg Hyeong-Ah Cho Department of

More information

A Simple Congestion-Aware Algorithm for Load Balancing in Datacenter Networks

A Simple Congestion-Aware Algorithm for Load Balancing in Datacenter Networks A Smpe Congeston-Aware Agorthm for Load Baancng n Datacenter Networs Mehrnoosh Shafee, and Javad Ghader, Coumba Unversty Abstract We study the probem of oad baancng n datacenter networs, namey, assgnng

More information

An Efficient Job Scheduling for MapReduce Clusters

An Efficient Job Scheduling for MapReduce Clusters Internatona Journa of Future Generaton ommuncaton and Networkng, pp. 391-398 http://dx.do.org/10.14257/jfgcn.2015.8.2.32 An Effcent Job Schedung for MapReduce usters Jun Lu 1, Tanshu Wu 1, and Mng We Ln

More information

Expressive Negotiation over Donations to Charities

Expressive Negotiation over Donations to Charities Expressve Negotaton over Donatons to Chartes Vncent Contzer Carnege Meon Unversty 5000 Forbes Avenue Pttsburgh, PA 523, USA contzer@cs.cmu.edu Tuomas Sandhom Carnege Meon Unversty 5000 Forbes Avenue Pttsburgh,

More information

Dynamic Virtual Network Allocation for OpenFlow Based Cloud Resident Data Center

Dynamic Virtual Network Allocation for OpenFlow Based Cloud Resident Data Center 56 IEICE TRANS. COMMUN., VOL.E96 B, NO. JANUARY 203 PAPER Speca Secton on Networ Vrtuazaton, and Fuson Patform of Computng and Networng Dynamc Vrtua Networ Aocaton for OpenFow Based Coud Resdent Data Center

More information

A Resources Allocation Model for Multi-Project Management

A Resources Allocation Model for Multi-Project Management A Resources Aocaton Mode for Mut-Proect Management Hamdatou Kane, Aban Tsser To cte ths verson: Hamdatou Kane, Aban Tsser. A Resources Aocaton Mode for Mut-Proect Management. 9th Internatona Conference

More information

Asymptotically Optimal Inventory Control for Assemble-to-Order Systems with Identical Lead Times

Asymptotically Optimal Inventory Control for Assemble-to-Order Systems with Identical Lead Times Asymptotcay Optma Inventory Contro for Assembe-to-Order Systems wth Identca ead Tmes Martn I. Reman Acate-ucent Be abs, Murray H, NJ 07974, marty@research.be-abs.com Qong Wang Industra and Enterprse Systems

More information

TCP/IP Interaction Based on Congestion Price: Stability and Optimality

TCP/IP Interaction Based on Congestion Price: Stability and Optimality TCP/IP Interacton Based on Congeston Prce: Stabty and Optmaty Jayue He Eectrca Engneerng Prnceton Unversty Ema: jhe@prncetonedu Mung Chang Eectrca Engneerng Prnceton Unversty Ema: changm@prncetonedu Jennfer

More information

Recurrence. 1 Definitions and main statements

Recurrence. 1 Definitions and main statements Recurrence 1 Defntons and man statements Let X n, n = 0, 1, 2,... be a MC wth the state space S = (1, 2,...), transton probabltes p j = P {X n+1 = j X n = }, and the transton matrx P = (p j ),j S def.

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

Predictive Control of a Smart Grid: A Distributed Optimization Algorithm with Centralized Performance Properties*

Predictive Control of a Smart Grid: A Distributed Optimization Algorithm with Centralized Performance Properties* Predctve Contro of a Smart Grd: A Dstrbuted Optmzaton Agorthm wth Centrazed Performance Propertes* Phpp Braun, Lars Grüne, Chrstopher M. Keett 2, Steven R. Weer 2, and Kar Worthmann 3 Abstract The authors

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

1 Example 1: Axis-aligned rectangles

1 Example 1: Axis-aligned rectangles COS 511: Theoretcal Machne Learnng Lecturer: Rob Schapre Lecture # 6 Scrbe: Aaron Schld February 21, 2013 Last class, we dscussed an analogue for Occam s Razor for nfnte hypothess spaces that, n conjuncton

More information

DEFINING %COMPLETE IN MICROSOFT PROJECT

DEFINING %COMPLETE IN MICROSOFT PROJECT CelersSystems DEFINING %COMPLETE IN MICROSOFT PROJECT PREPARED BY James E Aksel, PMP, PMI-SP, MVP For Addtonal Informaton about Earned Value Management Systems and reportng, please contact: CelersSystems,

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

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by 6 CHAPTER 8 COMPLEX VECTOR SPACES 5. Fnd the kernel of the lnear transformaton gven n Exercse 5. In Exercses 55 and 56, fnd the mage of v, for the ndcated composton, where and are gven by the followng

More information

Predicting Advertiser Bidding Behaviors in Sponsored Search by Rationality Modeling

Predicting Advertiser Bidding Behaviors in Sponsored Search by Rationality Modeling Predctng Advertser Bddng Behavors n Sponsored Search by Ratonaty Modeng Hafeng Xu Centre for Computatona Mathematcs n Industry and Commerce Unversty of Wateroo Wateroo, ON, Canada hafeng.ustc@gma.com Dy

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

The Dynamics of Wealth and Income Distribution in a Neoclassical Growth Model * Stephen J. Turnovsky. University of Washington, Seattle

The Dynamics of Wealth and Income Distribution in a Neoclassical Growth Model * Stephen J. Turnovsky. University of Washington, Seattle The Dynamcs of Weath and Income Dstrbuton n a Neocassca Growth Mode * Stephen J. Turnovsy Unversty of Washngton, Seatte Ceca García-Peñaosa CNRS and GREQAM March 26 Abstract: We examne the evouton of the

More information

Increasing Supported VoIP Flows in WMNs through Link-Based Aggregation

Increasing Supported VoIP Flows in WMNs through Link-Based Aggregation Increasng Supported VoIP Fows n WMNs through n-based Aggregaton J. Oech, Y. Hamam, A. Kuren F SATIE TUT Pretora, South Afrca oechr@gma.com T. Owa Meraa Insttute Counc of Scentfc and Industra Research (CSIR)

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

Multi-agent System for Custom Relationship Management with SVMs Tool

Multi-agent System for Custom Relationship Management with SVMs Tool Mut-agent System for Custom Reatonshp Management wth SVMs oo Yanshan Xao, Bo Lu, 3, Dan Luo, and Longbng Cao Guangzhou Asan Games Organzng Commttee, Guangzhou 5063, P.R. Chna Facuty of Informaton echnoogy,

More information

An Ensemble Classification Framework to Evolving Data Streams

An Ensemble Classification Framework to Evolving Data Streams Internatona Journa of Scence and Research (IJSR) ISSN (Onne): 39-7064 An Ensembe Cassfcaton Framework to Evovng Data Streams Naga Chthra Dev. R MCA, (M.Ph), Sr Jayendra Saraswathy Maha Vdyaaya, Coege of

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

SIMPLIFYING NDA PROGRAMMING WITH PROt SQL

SIMPLIFYING NDA PROGRAMMING WITH PROt SQL SIMPLIFYING NDA PROGRAMMING WITH PROt SQL Aeen L. Yam, Besseaar Assocates, Prnceton, NJ ABSRACf The programmng of New Drug Appcaton (NDA) Integrated Summary of Safety (ISS) usuay nvoves obtanng patent

More information

GRADIENT METHODS FOR BINARY INTEGER PROGRAMMING

GRADIENT METHODS FOR BINARY INTEGER PROGRAMMING Proceedns of the 4st Internatona Conference on Computers & Industra Enneern GRADIENT METHODS FOR BINARY INTEGER PROGRAMMING Chen-Yuan Huan¹, Ta-Chun Wan² Insttute of Cv Avaton, Natona Chen Kun Unversty,

More information

What is Candidate Sampling

What is Candidate Sampling What s Canddate Samplng Say we have a multclass or mult label problem where each tranng example ( x, T ) conssts of a context x a small (mult)set of target classes T out of a large unverse L of possble

More information

Branch-and-Price and Heuristic Column Generation for the Generalized Truck-and-Trailer Routing Problem

Branch-and-Price and Heuristic Column Generation for the Generalized Truck-and-Trailer Routing Problem REVISTA DE MÉTODOS CUANTITATIVOS PARA LA ECONOMÍA Y LA EMPRESA (12) Págnas 5 38 Dcembre de 2011 ISSN: 1886-516X DL: SE-2927-06 URL: http://wwwupoes/revmetcuant/artphp?d=51 Branch-and-Prce and Heurstc Coumn

More information

"Research Note" APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES *

Research Note APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES * Iranan Journal of Scence & Technology, Transacton B, Engneerng, ol. 30, No. B6, 789-794 rnted n The Islamc Republc of Iran, 006 Shraz Unversty "Research Note" ALICATION OF CHARGE SIMULATION METHOD TO ELECTRIC

More information

denote the location of a node, and suppose node X . This transmission causes a successful reception by node X for any other node

denote the location of a node, and suppose node X . This transmission causes a successful reception by node X for any other node Fnal Report of EE359 Class Proect Throughput and Delay n Wreless Ad Hoc Networs Changhua He changhua@stanford.edu Abstract: Networ throughput and pacet delay are the two most mportant parameters to evaluate

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

Efficient Bandwidth Management in Broadband Wireless Access Systems Using CAC-based Dynamic Pricing

Efficient Bandwidth Management in Broadband Wireless Access Systems Using CAC-based Dynamic Pricing Effcent Bandwdth Management n Broadband Wreless Access Systems Usng CAC-based Dynamc Prcng Bader Al-Manthar, Ndal Nasser 2, Najah Abu Al 3, Hossam Hassanen Telecommuncatons Research Laboratory School of

More information

INSTITUT FÜR INFORMATIK

INSTITUT FÜR INFORMATIK INSTITUT FÜR INFORMATIK Schedulng jobs on unform processors revsted Klaus Jansen Chrstna Robene Bercht Nr. 1109 November 2011 ISSN 2192-6247 CHRISTIAN-ALBRECHTS-UNIVERSITÄT ZU KIEL Insttut für Informat

More information

Loop Parallelization

Loop Parallelization - - Loop Parallelzaton C-52 Complaton steps: nested loops operatng on arrays, sequentell executon of teraton space DECLARE B[..,..+] FOR I :=.. FOR J :=.. I B[I,J] := B[I-,J]+B[I-,J-] ED FOR ED FOR analyze

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

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

Clustering based Two-Stage Text Classification Requiring Minimal Training Data

Clustering based Two-Stage Text Classification Requiring Minimal Training Data OI: 10.2298/CSIS120130044Z Custerng based Two-Stage Text Cassfcaton Requrng Mnma Tranng ata Xue Zhang 1,2 and Wangxn Xao 3,4 1 Key Laboratory of Hgh Confdence Software Technooges, Mnstry of Educaton, Pekng

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

On the Interaction between Load Balancing and Speed Scaling

On the Interaction between Load Balancing and Speed Scaling On the Interacton between Load Balancng and Speed Scalng Ljun Chen, Na L and Steven H. Low Engneerng & Appled Scence Dvson, Calforna Insttute of Technology, USA Abstract Speed scalng has been wdely adopted

More information

Optimization of network mesh topologies and link capacities for congestion relief

Optimization of network mesh topologies and link capacities for congestion relief Optmzaton of networ mesh topologes and ln capactes for congeston relef D. de Vllers * J.M. Hattngh School of Computer-, Statstcal- and Mathematcal Scences Potchefstroom Unversty for CHE * E-mal: rwddv@pu.ac.za

More information

On the Interaction between Load Balancing and Speed Scaling

On the Interaction between Load Balancing and Speed Scaling On the Interacton between Load Balancng and Speed Scalng Ljun Chen and Na L Abstract Speed scalng has been wdely adopted n computer and communcaton systems, n partcular, to reduce energy consumpton. An

More information

MAC Layer Service Time Distribution of a Fixed Priority Real Time Scheduler over 802.11

MAC Layer Service Time Distribution of a Fixed Priority Real Time Scheduler over 802.11 Internatonal Journal of Software Engneerng and Its Applcatons Vol., No., Aprl, 008 MAC Layer Servce Tme Dstrbuton of a Fxed Prorty Real Tme Scheduler over 80. Inès El Korb Ecole Natonale des Scences de

More information

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ). REVIEW OF RISK MANAGEMENT CONCEPTS LOSS DISTRIBUTIONS AND INSURANCE Loss and nsurance: When someone s subject to the rsk of ncurrng a fnancal loss, the loss s generally modeled usng a random varable or

More information

Conferencing protocols and Petri net analysis

Conferencing protocols and Petri net analysis Conferencng protocols and Petr net analyss E. ANTONIDAKIS Department of Electroncs, Technologcal Educatonal Insttute of Crete, GREECE ena@chana.tecrete.gr Abstract: Durng a computer conference, users desre

More information

Sngle Snk Buy at Bulk Problem and the Access Network

Sngle Snk Buy at Bulk Problem and the Access Network A Constant Factor Approxmaton for the Sngle Snk Edge Installaton Problem Sudpto Guha Adam Meyerson Kamesh Munagala Abstract We present the frst constant approxmaton to the sngle snk buy-at-bulk network

More information

Open Access A Load Balancing Strategy with Bandwidth Constraint in Cloud Computing. Jing Deng 1,*, Ping Guo 2, Qi Li 3, Haizhu Chen 1

Open Access A Load Balancing Strategy with Bandwidth Constraint in Cloud Computing. Jing Deng 1,*, Ping Guo 2, Qi Li 3, Haizhu Chen 1 Send Orders for Reprnts to reprnts@benthamscence.ae The Open Cybernetcs & Systemcs Journal, 2014, 8, 115-121 115 Open Access A Load Balancng Strategy wth Bandwdth Constrant n Cloud Computng Jng Deng 1,*,

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

Distributed Optimal Contention Window Control for Elastic Traffic in Wireless LANs

Distributed Optimal Contention Window Control for Elastic Traffic in Wireless LANs Dstrbuted Optmal Contenton Wndow Control for Elastc Traffc n Wreless LANs Yalng Yang, Jun Wang and Robn Kravets Unversty of Illnos at Urbana-Champagn { yyang8, junwang3, rhk@cs.uuc.edu} Abstract Ths paper

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

Dynamic Fleet Management for Cybercars

Dynamic Fleet Management for Cybercars Proceedngs of the IEEE ITSC 2006 2006 IEEE Intellgent Transportaton Systems Conference Toronto, Canada, September 17-20, 2006 TC7.5 Dynamc Fleet Management for Cybercars Fenghu. Wang, Mng. Yang, Ruqng.

More information

ANALYTICAL CHARACTERIZATION OF WLANS FOR QUALITY-OF-SERVICE WITH ACTIVE QUEUE MANAGEMENT

ANALYTICAL CHARACTERIZATION OF WLANS FOR QUALITY-OF-SERVICE WITH ACTIVE QUEUE MANAGEMENT M USHA AND K MADHANA: ANALYTICAL CHARACTERIZATION OF WLANS FOR QUALITY-OF-SERVICE WITH ACTIVE QUEUE MANAGEMENT ANALYTICAL CHARACTERIZATION OF WLANS FOR QUALITY-OF-SERVICE WITH ACTIVE QUEUE MANAGEMENT M.

More information

A New Paradigm for Load Balancing in Wireless Mesh Networks

A New Paradigm for Load Balancing in Wireless Mesh Networks A New Paradgm for Load Balancng n Wreless Mesh Networks Abstract: Obtanng maxmum throughput across a network or a mesh through optmal load balancng s known to be an NP-hard problem. Desgnng effcent load

More information

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK Sample Stablty Protocol Background The Cholesterol Reference Method Laboratory Network (CRMLN) developed certfcaton protocols for total cholesterol, HDL

More information

XAC08-6 Professional Project Management

XAC08-6 Professional Project Management 1 XAC08-6 Professona Project anagement Ths Lecture: Tte s so manager shoud ncude a s management pan a document that gudes any experts agree Some faure project to Ba, ba, ba, ba Communcaton anagement Wee

More information

A Fast Incremental Spectral Clustering for Large Data Sets

A Fast Incremental Spectral Clustering for Large Data Sets 2011 12th Internatonal Conference on Parallel and Dstrbuted Computng, Applcatons and Technologes A Fast Incremental Spectral Clusterng for Large Data Sets Tengteng Kong 1,YeTan 1, Hong Shen 1,2 1 School

More information

VRT012 User s guide V0.1. Address: Žirmūnų g. 27, Vilnius LT-09105, Phone: (370-5) 2127472, Fax: (370-5) 276 1380, Email: info@teltonika.

VRT012 User s guide V0.1. Address: Žirmūnų g. 27, Vilnius LT-09105, Phone: (370-5) 2127472, Fax: (370-5) 276 1380, Email: info@teltonika. VRT012 User s gude V0.1 Thank you for purchasng our product. We hope ths user-frendly devce wll be helpful n realsng your deas and brngng comfort to your lfe. Please take few mnutes to read ths manual

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

Ants Can Schedule Software Projects

Ants Can Schedule Software Projects Ants Can Schedule Software Proects Broderck Crawford 1,2, Rcardo Soto 1,3, Frankln Johnson 4, and Erc Monfroy 5 1 Pontfca Unversdad Católca de Valparaíso, Chle FrstName.Name@ucv.cl 2 Unversdad Fns Terrae,

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

Logical Development Of Vogel s Approximation Method (LD-VAM): An Approach To Find Basic Feasible Solution Of Transportation Problem

Logical Development Of Vogel s Approximation Method (LD-VAM): An Approach To Find Basic Feasible Solution Of Transportation Problem INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME, ISSUE, FEBRUARY ISSN 77-866 Logcal Development Of Vogel s Approxmaton Method (LD- An Approach To Fnd Basc Feasble Soluton Of Transportaton

More information

Multi-Source Video Multicast in Peer-to-Peer Networks

Multi-Source Video Multicast in Peer-to-Peer Networks ult-source Vdeo ultcast n Peer-to-Peer Networks Francsco de Asís López-Fuentes*, Eckehard Stenbach Technsche Unverstät ünchen Insttute of Communcaton Networks, eda Technology Group 80333 ünchen, Germany

More information

2008/8. An integrated model for warehouse and inventory planning. Géraldine Strack and Yves Pochet

2008/8. An integrated model for warehouse and inventory planning. Géraldine Strack and Yves Pochet 2008/8 An ntegrated model for warehouse and nventory plannng Géraldne Strack and Yves Pochet CORE Voe du Roman Pays 34 B-1348 Louvan-la-Neuve, Belgum. Tel (32 10) 47 43 04 Fax (32 10) 47 43 01 E-mal: corestat-lbrary@uclouvan.be

More information

Analysis of Energy-Conserving Access Protocols for Wireless Identification Networks

Analysis of Energy-Conserving Access Protocols for Wireless Identification Networks From the Proceedngs of Internatonal Conference on Telecommuncaton Systems (ITC-97), March 2-23, 1997. 1 Analyss of Energy-Conservng Access Protocols for Wreless Identfcaton etworks Imrch Chlamtac a, Chara

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

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

APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT

APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT Toshhko Oda (1), Kochro Iwaoka (2) (1), (2) Infrastructure Systems Busness Unt, Panasonc System Networks Co., Ltd. Saedo-cho

More information

8 Algorithm for Binary Searching in Trees

8 Algorithm for Binary Searching in Trees 8 Algorthm for Bnary Searchng n Trees In ths secton we present our algorthm for bnary searchng n trees. A crucal observaton employed by the algorthm s that ths problem can be effcently solved when the

More information

Real-Time Process Scheduling

Real-Time Process Scheduling Real-Tme Process Schedulng ktw@cse.ntu.edu.tw (Real-Tme and Embedded Systems Laboratory) Independent Process Schedulng Processes share nothng but CPU Papers for dscussons: C.L. Lu and James. W. Layland,

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

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

How To Make A Co-Ocaton Work For Free

How To Make A Co-Ocaton Work For Free Enabng Far Prcng on HPC Systems wth Node Sharng Aex D. Bresow Ananta Twar Martn Schuz Laura Carrngton Lngja Tang Jason Mars Unversty of Caforna, San Dego, CA, USA, abresow@cs.ucsd.edu San Dego Supercomputer

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

Performance Analysis and Comparison of QoS Provisioning Mechanisms for CBR Traffic in Noisy IEEE 802.11e WLANs Environments

Performance Analysis and Comparison of QoS Provisioning Mechanisms for CBR Traffic in Noisy IEEE 802.11e WLANs Environments Tamkang Journal of Scence and Engneerng, Vol. 12, No. 2, pp. 143149 (2008) 143 Performance Analyss and Comparson of QoS Provsonng Mechansms for CBR Traffc n Nosy IEEE 802.11e WLANs Envronments Der-Junn

More information

An Integrated Approach for Maintenance and Delivery Scheduling in Military Supply Chains

An Integrated Approach for Maintenance and Delivery Scheduling in Military Supply Chains An Integrated Approach for Mantenance and Delvery Schedulng n Mltary Supply Chans Dmtry Tsadkovch 1*, Eugene Levner 2, Hanan Tell 2 and Frank Werner 3 2 1 Bar Ilan Unversty, Department of Management, Ramat

More information

Research on Single and Mixed Fleet Strategy for Open Vehicle Routing Problem

Research on Single and Mixed Fleet Strategy for Open Vehicle Routing Problem 276 JOURNAL OF SOFTWARE, VOL 6, NO, OCTOBER 2 Research on Snge and Mxed Feet Strategy for Open Vehce Routng Probe Chunyu Ren Heongjang Unversty /Schoo of Inforaton scence and technoogy, Harbn, Chna Ea:

More information

A New Quality of Service Metric for Hard/Soft Real-Time Applications

A New Quality of Service Metric for Hard/Soft Real-Time Applications A New Qualty of Servce Metrc for Hard/Soft Real-Tme Applcatons Shaoxong Hua and Gang Qu Electrcal and Computer Engneerng Department and Insttute of Advanced Computer Study Unversty of Maryland, College

More information

To Fill or not to Fill: The Gas Station Problem

To Fill or not to Fill: The Gas Station Problem To Fll or not to Fll: The Gas Staton Problem Samr Khuller Azarakhsh Malekan Julán Mestre Abstract In ths paper we study several routng problems that generalze shortest paths and the Travelng Salesman Problem.

More information

Packet Dispersion and the Quality of Voice over IP Applications in IP networks

Packet Dispersion and the Quality of Voice over IP Applications in IP networks acet Dsperson and the Qualty of Voce over I Applcatons n I networs Ham Zlatorlov, Hanoch Levy School of Computer Scence Tel Avv Unversty Tel Avv, Israel Abstract- Next Generaton Networs (NGN and the mgraton

More information

On-Line Trajectory Generation: Nonconstant Motion Constraints

On-Line Trajectory Generation: Nonconstant Motion Constraints 2012 IEEE Internatona Conference on Robotcs and Automaton RverCentre, Sant Pau, Mnnesota, USA May 14-18, 2012 On-Lne Trajectory Generaton: Nonconstant Moton Constrants Torsten Kröger Abstract A concept

More information

v a 1 b 1 i, a 2 b 2 i,..., a n b n i.

v a 1 b 1 i, a 2 b 2 i,..., a n b n i. SECTION 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS 455 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS All the vector spaces we have studed thus far n the text are real vector spaces snce the scalars are

More information

A Novel Auction Mechanism for Selling Time-Sensitive E-Services

A Novel Auction Mechanism for Selling Time-Sensitive E-Services A ovel Aucton Mechansm for Sellng Tme-Senstve E-Servces Juong-Sk Lee and Boleslaw K. Szymansk Optmaret Inc. and Department of Computer Scence Rensselaer Polytechnc Insttute 110 8 th Street, Troy, Y 12180,

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

Sangam - Efficient Cellular-WiFi CDN-P2P Group Framework for File Sharing Service

Sangam - Efficient Cellular-WiFi CDN-P2P Group Framework for File Sharing Service Sangam - Effcent Cellular-WF CDN-P2P Group Framework for Fle Sharng Servce Anjal Srdhar Unversty of Illnos, Urbana-Champagn Urbana, USA srdhar3@llnos.edu Klara Nahrstedt Unversty of Illnos, Urbana-Champagn

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

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

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

Profit-Aware DVFS Enabled Resource Management of IaaS Cloud

Profit-Aware DVFS Enabled Resource Management of IaaS Cloud IJCSI Internatonal Journal of Computer Scence Issues, Vol. 0, Issue, No, March 03 ISSN (Prnt): 694-084 ISSN (Onlne): 694-0784 www.ijcsi.org 37 Proft-Aware DVFS Enabled Resource Management of IaaS Cloud

More information

An Interest-Oriented Network Evolution Mechanism for Online Communities

An Interest-Oriented Network Evolution Mechanism for Online Communities An Interest-Orented Network Evoluton Mechansm for Onlne Communtes Cahong Sun and Xaopng Yang School of Informaton, Renmn Unversty of Chna, Bejng 100872, P.R. Chna {chsun,yang}@ruc.edu.cn Abstract. Onlne

More information

A Performance Analysis of View Maintenance Techniques for Data Warehouses

A Performance Analysis of View Maintenance Techniques for Data Warehouses A Performance Analyss of Vew Mantenance Technques for Data Warehouses Xng Wang Dell Computer Corporaton Round Roc, Texas Le Gruenwald The nversty of Olahoma School of Computer Scence orman, OK 739 Guangtao

More information

Swing-Free Transporting of Two-Dimensional Overhead Crane Using Sliding Mode Fuzzy Control

Swing-Free Transporting of Two-Dimensional Overhead Crane Using Sliding Mode Fuzzy Control Swng-Free Transportng of Two-Dmensona Overhead Crane Usng Sdng Mode Fuzzy Contro Dantong Lu, Janqang, Dongn Zhao, and We Wang Astract An adaptve sdng mode fuzzy contro approach s proposed for a two-dmensona

More information

Complete Fairness in Secure Two-Party Computation

Complete Fairness in Secure Two-Party Computation Complete Farness n Secure Two-Party Computaton S. Dov Gordon Carmt Hazay Jonathan Katz Yehuda Lndell Abstract In the settng of secure two-party computaton, two mutually dstrustng partes wsh to compute

More information

Relay Secrecy in Wireless Networks with Eavesdropper

Relay Secrecy in Wireless Networks with Eavesdropper Relay Secrecy n Wreless Networks wth Eavesdropper Parvathnathan Venktasubramanam, Tng He and Lang Tong School of Electrcal and Computer Engneerng Cornell Unversty, Ithaca, NY 14853 Emal : {pv45, th255,

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 DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION. Michael E. Kuhl Radhamés A. Tolentino-Peña

A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION. Michael E. Kuhl Radhamés A. Tolentino-Peña Proceedngs of the 2008 Wnter Smulaton Conference S. J. Mason, R. R. Hll, L. Mönch, O. Rose, T. Jefferson, J. W. Fowler eds. A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION

More information

A generalized hierarchical fair service curve algorithm for high network utilization and link-sharing

A generalized hierarchical fair service curve algorithm for high network utilization and link-sharing Computer Networks 43 (2003) 669 694 www.elsever.com/locate/comnet A generalzed herarchcal far servce curve algorthm for hgh network utlzaton and lnk-sharng Khyun Pyun *, Junehwa Song, Heung-Kyu Lee Department

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

CLoud computing technologies have enabled rapid

CLoud computing technologies have enabled rapid 1 Cost-Mnmzng Dynamc Mgraton of Content Dstrbuton Servces nto Hybrd Clouds Xuana Qu, Hongxng L, Chuan Wu, Zongpeng L and Francs C.M. Lau Department of Computer Scence, The Unversty of Hong Kong, Hong Kong,

More information

IWFMS: An Internal Workflow Management System/Optimizer for Hadoop

IWFMS: An Internal Workflow Management System/Optimizer for Hadoop IWFMS: An Internal Workflow Management System/Optmzer for Hadoop Lan Lu, Yao Shen Department of Computer Scence and Engneerng Shangha JaoTong Unversty Shangha, Chna lustrve@gmal.com, yshen@cs.sjtu.edu.cn

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