Scheduling within Temporal Partitions: Response-time Analysis and Server Design

Size: px
Start display at page:

Download "Scheduling within Temporal Partitions: Response-time Analysis and Server Design"

Transcription

1 chedulng wthn Temporal Parttons: Response-tme Analyss and erver Desgn Lus Almeda LE-IEETA/DET Unversdade de Avero Avero, Portugal Paulo Pedreras LE-IEETA/DET Unversdade de Avero Avero, Portugal ABTRACT As the bandwdth of CPUs and networks contnues to grow, t becomes more attractve, for effcency reasons, to share such resources among several applcatons wth the mnmum level of nterference. Ths can be acheved usng temporal parttons, wth each applcaton assgned to ts own partton and executng as f t was executng alone on a resource wth lower bandwdth. The parttons are assocated to servers that execute the applcaton tasks accordng to a gven applcaton-level scheduler. On the other hand, the set of servers s scheduled by a system-level scheduler. Ths paper addresses the partcular case of fxed prortes-based applcaton-level schedulers together wth a perodc server model at the system level. It starts wth an adequate response tme analyss based on the noton of server avalablty for a known server. Then t addresses the nverse problem of desgnng a server wth mnmum system-level resource requrements to fulfll the applcaton tme constrants. In ths context, the paper shows that response tme based schedulablty tests wth lnear tme bounds do not need to consder all tasks but ust a small subset, whch may lead to substantal speed-ups. The proposed method goes a step further wth respect to other recent works n the lterature by consderng a more complete task model, effectvely computng the server parameters and establshng a better trade-off concernng complexty and tghtness. Categores and ubect Descrptors D..7 [Operatng ystems]: Organzaton and Desgn Real-tme systems and embedded systems. General Terms Algorthms, Desgn, Theory. Keywords Real-tme systems, real-tme schedulng, herarchcal schedulng, response-tme analyss. Permsson to make dgtal or hard copes of all or part of ths work for personal or classroom use s granted wthout fee provded that copes are not made or dstrbuted for proft or commercal advantage and that copes bear ths notce and the full ctaton on the frst page. To copy otherwse, or republsh, to post on servers or to redstrbute to lsts, requres pror specfc permsson and/or a fee. EMOFT, eptember 7--9,, Psa, Italy. Copyrght ACM //9...$5... INTRODUCTION Herarchcal schedulng has been generatng a consderable nterest, recently, due to ts ablty to separate the concerns of schedulng at the system and applcaton levels. It s a fundamental brck n the current trend towards hgher ntegraton and flexblty n embedded systems [8], whch opens the way to hgher effcency and lower costs by means of resource sharng, as well as to hgher reslence to hardware falures by means of dynamc reallocaton of computng or communcaton enttes [7]. Herarchcal schedulng s ntmately connected wth resource temporal parttonng accordng to whch a shared resource, e.g. CPU or network, s used by several complex applcatons each of whch s composed of a set of enttes, e.g. tasks or streams. These enttes must be scheduled nternally to the applcaton nsde one specfc resource partton to whch they were allocated. At a hgher level, all resource parttons are scheduled usng a gven system-level polcy. The concept of server s well adapted to ths level, supportng temporal solaton among parttons (Fgure ). ystem scheduler Applcaton n ystem level Applcaton Applcaton level Task Task Task m Applcaton ervers (capacty Cs, perod Ts ) Applcaton scheduler Fxed prortes Fgure. Herarchcal schedulng framework. However, two problems arse: how to derve real-tme guarantees for the applcatons runnng wthn each server and, conversely, how to desgn the server for a gven applcaton so that t fulfls the applcaton requrements wth the least resource utlzaton. Ths paper addresses both problems. It relates drectly to recent work avalable n the lterature and consttutes a further contrbuton to the analyss and desgn of temporally parttoned 95

2 systems. A prelmnary work-n-progress verson was presented n []. The related work s dscussed n secton, whch also hghlghts the contrbutons of ths paper, secton presents the task model, secton descrbes the response-tme analyss and secton 5 shows the server desgn approach, wth an example n secton 6. Fnally, secton 7 extends the prevous results ncludng ntra-server blockng and secton 8 concludes the paper.. RELATED WORK The problem of herarchcal schedulng bears many resemblances wth other schedulng problems that have been tackled n the past, such as those regardng exclusons [5] and nserted dle-tme [9][]. In fact, lookng to the system from the perspectve of one applcaton executng wthn a server,.e. a temporal partton, the perods of tme n whch the respectve server cannot execute, e.g. because another server has been scheduled at the system level, can be seen as excluson perods or perods of nserted dle-tme. Moreover, wthn specfc scopes, such as real-tme communcaton over shared meda, some forms of herarchcal schedulng/ temporal parttonng have long been used. Ths s the case wth TDMA (as n TTP/C), n whch each node has one or more dedcated slots n the TDMA round to transmt ts traffc, as well as wth the mult-phased cyclc framework (as n WorldFIP, FlexRay or FTT-CAN []), wth several phases used n sequence wthn a mcro-cycle, each for a gven type of traffc (Fgure ). In both cases, ether slots or phases can be taken as perodc servers wthn whch several message streams must be scheduled. TDMA round wth nodes Bus tme Mcro-cycle wth phases sync async sync async sync Bus tme Fgure. Common parttonng schemes for shared buses used n real-tme communcaton. However, recent work has brought to lght new results that are more general and abstract away the specfc applcaton scope. In [7] the authors use the concepts of vrtual resource, vrtual tme, bounded-delay resource parttonng and server supply functon to deduce real-tme guarantees for herarchcal partton schedulng. In [6] the focus s on herarchcal resource parttonng usng fxed prortes local schedulers and both deferrable and sporadc servers. It presents a response-tme analyss as well as utlzaton bounds. In [], the authors deduce a tghter schedulablty test for parttons usng fxed prortes local schedulers and address the ssue of server desgn n order to meet the applcaton tme constrants. In [], the authors present a response tme analyss for a composed model n whch EDF local schedulng s executed wthn a system-level fxed prortes framework. The task model consders several practcal ssues such as release tter and ntertask synchronzaton blockng. Fnally, n [] the authors present an analyss for herarchcal partton schedulng, consderng both fxed prortes and EDF local schedulng, together wth a generc perodc resource model. A general schedulng nterface s also presented that facltates the composton of parttons and dervaton of real-tme guarantees. The paper also addresses the nverse problem of defnng a partton n order to meet a gven set of real-tme requrements. Ths paper relates closely to the works referred above. It s nterestng to note that these works are very recent and some of them were submtted n parallel, leadng to overlappng of short parts. Our motvaton was to extend the work started n [] and later mproved n [], n whch we developed a response-tme analyss for the asynchronous traffc wthn a double phase cyclc framework as depcted n Fgure. In ths paper we use the same reasonng but wth a more general model that fts well n task schedulng. We wll consder a fxed prortes local scheduler together wth a perodc server model to manage a resource partton. Then, we derve upper bounds to the worst-case response tme of tasks executng wthn a gven server. We wll also use the concept of server supply functon as n [7], or resource supply n [], but we wll refer to t as server avalablty functon for coherence wth our prevous work. The response-tme analyss we propose s smlar to the one for fxed prortes presented n [] but we extend t to cover a more realstc task model that ncludes deadlnes shorter than or equal to perods, release tter and synchronzaton blockng. Our response-tme analyss s also equvalent, despte dfferent, to the one n [6] but we beleve ours s more flexble snce t can be easly adapted to rregular parttons as n [] and [], by descrbng them analytcally n the respectve server avalablty functon (see further on). The server desgn problem s addressed n both [] and []. When compared to the former one, our method s smpler but less tght and, as shown later, may represent a more favorable compromse between tghtness and complexty n certan crcumstances. On the other hand, [] follows the same reasonng as we dd n [] of matchng the applcaton demand wth the server supply to derve the worst-case response tme. The correspondng result presented n [] s equvalent to the one we presented n []. In ths paper we extend that work by generatng the server perod that mnmzes the server bandwdth together wth the overhead caused by context swtchng at the system level, usng the cost functon proposed n []. Fnally, we consder release tter as well as ntra-server blockng n our task model, whch none of the other approaches does. In summary, ths paper bulds upon the work n [],[], [] and [] and proposes the followng: extenson of the task model to a more realstc one; alternatve way of deducng the soluton space for the server parameters; new method to search the server soluton space based on the new concept of applcaton external ponts; full computaton of the server parameters that mnmze a gven cost functon.. TAK MODEL In ths work, we consder that a gven actve resource, say a CPU, s used to execute a set of ndependent applcatons. Each applcaton Ω s composed by a set Γ Ω of N Ω tasks, Γ Ω {τ (C,T,D,J,P ), =..N Ω }, n whch each task, at ths pont, wll be consdered ndependent and fully preemptve. Each task τ 96

3 s characterzed by a perod T, a worst-case executon tme C, a relatve deadlne D that s shorter than or equal to the respectve perod, a maxmum release tter of J and a fxed prorty P that may possbly be derved from the perod, the deadlne, or any suboptmal crteron. Also, the set of tasks Γ Ω executes by prorty order wthn a perodc server Ω, whch s characterzed, at a system level, by a perod T and a capacty C. The server can be of any type as long as, under contnuous demand, t behaves lke a perodc task executng for C tme unts wthn every T tme unts nterval. Ths s also the server model consdered n [] and [] ([6] consders deferrable and sporadc models, only). The system-level scheduler that schedules parttons, and the current system load, determne when the server s avalable to execute applcaton tasks. We thus defne the server avalablty functon, A (t), whch returns for each nstant t the cumulatve CPU tme avalable for the applcaton to execute snce an arbtrary tme orgn. For statc system level schedulng such as TDMA (Fgure ), A (t) may be exactly characterzed a pror. However, f an on-lne system-level schedulng polcy s used, the specfc pattern of A (t) may be dffcult to predct. Therefore, for the sake of ndependence wth respect to the system-level schedulng polcy and load, t s helpful to use a lower bound A_ (t) that assures that the cumulatve CPU tme avalable for an applcaton s never smaller than a gven value. In order to determne a lower bound to the avalablty functon, we can use the same reasonng as n [] or []. Bascally, t consders the worst-case server avalablty pattern wth respect to the arbtrary tme orgn n whch the server suffers maxmum latency ( ) n the begnnng and then follows a perodc pattern wth ts capacty avalable at the end of each perodc nstance (Fgure ). The value of depends drectly on the maxmum fnshng tter of the server executon, as determned by the system scheduler. In certan cases, e.g. n regular TDMA schedules, such tter s elmnated because the server executes n a strct perodc fashon, leadng to = T -C. Thus, wll vary between ths best-case value and the worst-case depcted n Fgure n whch = *(T -C ). For the sake of generalty we wll consder such ntal latency as gven by Equaton. Cs T s possble avalablty pattern avalablty lower bound A s(t) A_ s(t) tme T s -C s C s Fgure. erver avalablty functons A (t) and A_ (t). Fnshng tter refers to the absolute tter that affects the nstant n whch the server exhausts ts capacty wthn each perod. = (+β )*(T -C ) () wth β = (R -C ) / (T -C ) Notce that β ( β ) s a normalzed measure of the maxmum server fnshng tter, whle R (C R T ) s the maxmum relatve fnshng nstant of all server nstances. To mantan ndependency from the system schedulng polcy β = should be consdered (worst-case). For smplcty, n the remander of the paper we wll use the same expresson avalablty functon for the lower bound functon, and refer to t as A_ (t) (Equaton ). In [] ths functon s called server characterstc functon, n [7] server supply functon and n [] resource supply bound functon. A _, ( t) = t < t ( + k * ( T C ), + k * T t < + k ( k + ) * C, + k * T + C t < + ( k + ) k = ( t ) / T * T + C * T. REPONE-TIME ANALYI A relatvely smple but effectve way of upper boundng the response-tme for each task τ wthn a gven server s to use the same reasonng explaned n [] and []. The fact that we are now usng a fully preemptve task model together wth a perodc server further smplfes the analyss theren presented, whch was based on non-preempton wth nserted dle-tme together wth a background server. Therefore, we compute for each task τ and for each nstant t the maxmum load submtted to the server by the task tself after ts release together wth all hgher prorty tasks. We call ths the level submtted load functon, H (t) (Equaton ). It can be determned by the usual methods n fxed-prortes response tme analyss [] snce the crtcal nstant for each task s not changed by the presence of the server [6]. Equaton consders Γ Ω sorted by decreasng prortes,, < P >P. ( t + J ) / T = () H ( t) = * C () The worst-case response tme for task τ, referred to as R, can thus be obtaned as expressed n Lemma. Lemma. Gven the task set Γ Ω executed wthn a server wth avalablty functon A_ (t), the worst-case response tme R for task τ Γ Ω s obtaned by determnng the earlest nstant n whch the maxmum level submtted load H (t) matches the least server avalablty A_ (t). Proof: Lemma can be proved by consderng the defntons of both A_ (t) and H (t). In fact, A_ (t) stands for the mnmum executon tme that the server can delver to the applcaton counted from t=. On the other hand, H (t) stands for the maxmum executon tme requred to execute τ to completon, when released at t= and consderng the maxmum nterference t may suffer by hgher prorty tasks wthn the applcaton. Thus, the worst-case response tme s gven by the nstant when the least avalablty s ust enough to cover the longest requested 97

4 executon tme (Fgure ). If the server, n one or more nstances, executes before than consdered n A_ (t) the response tme can only be shorter. level subm tted load 8 6 A_ (t) s H (t) 5 5 tm e Fgure. Worst-case response tme of task τ. Wth Lemma, we can perform a trval schedulablty test as stated n theorem. Theorem. The task set Γ Ω executed wthn a server wth an avalablty functon A_ (t) s schedulable f (and only f) τ Γ Ω R D Proof: The f s trvally proved but the only f requres R to be an accurate value. Ths depends on the accuracy of the avalablty lower bound A_ (t). In practce, ths lower bound wll be pessmstc because A (t) wll not suffer maxmum delay n all nstances after startup, thus there wll always be a gven load for whch some exact worst-case response tmes are lower than the computed R. In ths case, the test n Theorem wll be suffcent, only. However, n partcular stuatons, such as regular TDMA schedules, A_ (t) = A (t) and thus the test wll be necessary (ths s why we kept the only f wthn parenthess). The value of R can be determned usng Equaton or, more effcently, the equvalent Equaton 5, whch makes use of the nverse of the avalablty functon A_ (t), referred to as A nv (t). Ths s formalzed n Equaton 6 (t s referred to as servce tme bound functon n []). R = earlest t: A_ (t) = H (t) () R = earlest t: t = A nv (H (t)) (5) Formally, nvertng A_ (t) requres specfyng the value of the nverse correspondng to the flat segments of the orgnal functon. We consder the lowest values of such ntervals, as ndcated n Fgure, whch result n the lowest avalablty. A nv (u) = + m * T + (u m*c ), m = u/c A nv (u) = (β + u/c ) * (T C ) + u (6) Equaton 5 can be solved teratvely wth R R =H () and R n+ = A nv (H (R n )) that ether converges to R = R n+ = R n or grows beyond the deadlne D, wthn fnte teratons. Ths can be easly demonstrated by the fact that, from teraton to teraton, the ncrement n H (t) s lower bounded by mn =.. (C ). If the whole CPU s avalable to execute the applcaton, then A_ (t) = A nv (t) = t and Equatons and 5 reduce to the usual response tme analyss for fxed prortes schedulng for a smlar task model []. A smpler but less tght upper bound R for the response tme of each task can be obtaned consderng a lnear lower bound to the avalablty functon, also proposed n [] and [], heren referred to as lnear avalablty functon A_ (t) (Equaton 7). Ths functon s depcted n Fgure 5, wth an ntal latency of, such as A_ (t), and then grows lnearly wth slope α = C / T,.e. the server bandwdth. A_ (t) = (t ) * α, for t > and otherwse (7) The response tme upper bounds can be obtaned rewrtng Equaton 5 usng the nverse of A_ (t) (Equaton 8). R = earlest t: t = + H (t)/α (8) Ths allows us to state Theorem, whch wll be partcularly helpful n the followng secton. Theorem. The task set Γ Ω, executed wthn a server wth ntal latency and bandwdth α, s schedulable f τ Γ Ω R D. Proof: Ths can be proved notcng that, for every task τ the ntersecton between H (t) and A_ (t) (.e. R ), s always later than or concdent wth that between H (t) and A_ (t) (.e. R ). Thus, τ Γ Ω R R, provng the theorem. 5. ERVER DEIGN In ths secton we tackle the opposte problem of the prevous one,.e. gven an applcaton Ω, whch parameters (C,T ), or equvalently (α, ), should the server Ω have so that t requres the least system resources and stll meets the applcaton tme constrants? In order to address ths problem we wll start by notcng that such a server should be as tght as possble concernng fulfllng the applcaton tme constrants, otherwse t would over consume system resources. Therefore, usng the approach presented n the prevous secton for Theorem, we wll consder that, for each task τ the respectve R s ust on the deadlne D. Ths means that the avalablty functon A_ (t) should be such that ntersects H (t) exactly at t=d. To acheve ths, we start by defnng the set of deadlne ponts DP Ω {DP (D,H (D )), =..N Ω } (Fgure 5), whch represent the lowest avalablty requred for meetng all applcaton deadlnes. Our purpose, then, s to defne A_ (t) so that t s hgher than but as close as possble to the set of such ponts. In [] (theorem 5), gven a fxed server perod T, the authors suggest performng an extensve search for the mnmum server capacty C that stll allows meetng the tasks deadlnes. However, the search method s not specfed. We suggest usng Bnary earch n the nterval [, T ], whch s relatvely effcent. For example, f a resoluton of /56 of T s enough, only 8 teratons are needed. Correspondngly, we can use the approach suggested n Theorem 98

5 to fnd the lowest lnear bound A_ (t) that passes above or through all deadlne ponts DP. Ths s also proposed n [] (theorem 6), where C s determned as a functon of the perod T so that A_ (t) touches at least one deadlne pont. Notce that all task deadlnes,.e. all deadlne ponts, are tested aganst the lnear lower bound. 8 deadlne ponts external ponts A_ (t) A_' (t) N Ω and N Ω. Also, n general, the number of external ponts N E s substantally smaller than the number of deadlne ponts N Ω (.e. the number of tasks), whch contrbutes to decrease the complexty of the remanng part of the process. Fgure 6 plots the number of external ponts aganst the number of tasks for 5 randomly generated sets, each constraned to a total bandwdth smaller than but close to 5% and wth up to tasks. It s surprsng to see how small the number of external ponts normally s, between and 5, and that there s stll a tendency for reducton as the number of tasks grows (for large task sets, the most frequent number of external ponts s!). H (D ) 6 slope α 5 5 D Fgure 5. Determnng the avalablty functon A_ (t). However, we are not nterested n determnng ust C as a functon of T but rather n analyzng the space of possble solutons (C, T ) to choose the one that mnmzes the use of system resources. For ths purpose we determne the subset of external ponts, E Ω {E (x,y ), =..N E } DP Ω (Fgure 5), that s the subset of deadlne ponts through whch a lnear bound (α, ) can be drawn that fulflls Theorem. Necessarly, such lnear bound must also respect ts boundary constrants namely α (the server cannot use more than the total CPU bandwdth) and, for each E, α y /x (otherwse, < and, snce α >, t would lead to T < and C <). Ths drectly leads to Theorem. Theorem. The task set Γ Ω, executed wthn a server wth ntal latency and bandwdth α, s schedulable f τ : DP E Ω, R D. Proof: Informally, Theorem says that t s suffcent to test the schedulablty of the external ponts to guarantee that the task set s schedulable (Fgure 5). It can be proved ust by observng the defnton of external ponts. If these ponts are below a gven lnear avalablty functon, then all other deadlne ponts are and thus the whole set s schedulable. The determnaton of the E Ω subset s carred out usng a smple algorthm that goes through all deadlne ponts n ascendng deadlne order, startng wth the assumpton that the frst deadlne pont s an external pont. Notce that the slopes of the segments that on every two consecutve external ponts must be monotoncally decreasng from the maxmum α = to the mnmum α = y NE /x NE. Therefore, the algorthm removes from the set of deadlne ponts all those that would volate ths decreasng slope pattern and the referred lmts. At the end, the ponts that reman are the external ponts. The tme complexty of the algorthm vares wth the characterstcs of the task set between Fgure 6. Number of external ponts n random sets. Hence, the use of external ponts may accelerate substantally the executon of repettve response tme based schedulablty tests, for example n optmzaton processes, n whch the task set remans unchanged and several server parameters are tested. Ths s depcted n Fgure 7 for several mplementatons of the test correspondng to theorem 6 n [], to determne C for dfferent values of T, and usng random task sets of dfferent szes (each pont corresponds to the average of sets wth the same sze). We used a drect mplementaton n Matlab ([](th.6)) and an mproved verson ([] (th.6)_mp), as well as both cases usng external ponts, (wth ext.pt.) and (wth ext.pt._mp), respectvely. The benefts of usng external ponts are clear for task sets wth more than 8 tasks. The set of external ponts defnes a soluton space for the server desgn problem. In fact, all lnear avalabltes A_ (t) that touch at least one external pont wth a vald slope are possble solutons. Vald slopes are those n between the slopes of the segments that on each pont wth the prevous and wth the next, consderng the absolute maxmum and mnmum for the ponts n the extremes. Ths can be represented by the unon of the followng N E subntervals: α {[ α(e,e )] [α(e,e ) α(e,e )] [α(e N E-,E N E) y N E /x N E]} 99

6 Executon tm e n seconds wth ext.pt._m p [](th.6)_mp wth ext.pt. [](th.6) 6 8 Number of task per taskset Fgure 7. peed up by usng external ponts. Each of these subntervals corresponds to A_ (t) lnes that pass through one external pont E (x,y ), =..N E, whch can be characterzed by the straght-lne Equaton 9 (left). y=α ( x- x ) + y and = x - y /α (9) Equaton 9(rght) shows a hyperbolc relatonshp between α and around each pont E and also allows determnng the respectve server perod T usng Equaton. Ts = /((+β)(-α )) () The complete soluton space n terms of the server parameters (α, ) can be obtaned by calculatng the followng ntersecton for all N E external ponts: =..NE x - y /α and α () Ths soluton space can equvalently be expressed n terms of the server parameters (C,T ) by nsertng from Equaton 9 (rght) nto and replacng α wth Cs /Ts. The result s the followng ntersecton for all N E external ponts: Cs =..NE Cs Ts and ( x ( + β ) Ts ) + x ( + β ) Ts ( + β ) ( ) + ( + β ) y Ts () Condton s equvalent to the result presented n [] (theorem 6). Recall that x = D and y =H (D ). Fnally, n order to fnd one specfc soluton, we use the cost functon proposed n [], whch consders both the bandwdth drectly requested by the server, α=c /T, as well as the overhead bandwdth mplctly used by the server n context swtchng between applcatons at the system level. Ths latter factor can be roughly computed as C O /T where C O s a system parameter representng the context swtchng tme. The cost functon can thus be expressed as F=α+C O /T. Mnmzng F corresponds to fndng the balance between mnmzng α and maxmzng T. Agan nsertng from Equaton 9 (rght) nto Equaton we obtan Ts =Ts (α ), for each α subnterval,.e. each external pont. Then, usng Ts (α ) wthn the cost functon results n F=F(α ). The α of mnmum cost for each subnterval (referred to as α,mn ) can be easly determned wth a closed formula (Equaton ) obtaned by dfferentatng F(α ) wth respect to α and calculatng the respectve root. Whenever α,mn les outsde the respectve subnterval, the closest extreme s consdered for mnmum. After havng determned α,mn for all subntervals (=..N E ) t s then ust a matter of selectng the one that generates the absolute mnmum cost (α mn ). It s also necessary to dentfy to whch subnterval the α mn value belongs to, n order to determne (Equaton 9 rght) and then T (Equaton ). ( y ( + β)* CO ) /( x ( + β )* CO ) α = + y / x * (),mn y / x The server parameters generated ths way are not optmal due to several factors such as the pessmsm ncluded n the server ntal latency, the successve approxmatons of the effectve server avalablty functon A (t) and the use of the deadlne ponts that may lead to worse than necessary bandwdth requrements. However, concernng the A_ (t) approxmaton by A_ (t),.e. the lnear bound, t s possble to carry out a smple fnal mprovement step. The fact that A_ (t) touches at least one deadlne (external) pont accordng to the desgn method presented above, t does not mply that the correspondng A_ (t) functon also touches one, because t A_ (t) A_ (t). Hence, we may allow A_ (t) to actually enter the area bounded by A_ (t) untl t also touches one deadlne pont (ths assures that Theorem s stll met). Ths can be done decreasng C, ncreasng T or a combnaton of both. For smplcty and effcency accordng to tests wth random sets, we propose ncreasng T, accordng to Equaton, whch corresponds to mantanng the heght of the steps n A_ (t) whle expandng the functon to the rght. Ths also leads to a further reducton n α. The amount of beneft, however, depends on the task set. Usng random task sets wth unformly dstrbuted perods, we acheved a best mprovement of 8% ncrease n T and 7.% reducton n α. However, on average, the benefts were lower, wth.8% ncrease n T and.7% reducton n α. nv D A ( H ( D )) T = + mp T mn = N + + (),.. Ω ( D ( β ) * C ) / T Ths mprovement can be carred out over any non-optmal result obtaned by any perodc server based method. 6. EXAMPLE To llustrate the desgn methodology presented above we make use of the example suggested n []. Ths example conssts of an applcaton Ω, wth the task set Γ Ω ={(C,T )= (,), (,), (,5) and (D =T, J =, P =/)}. Frstly, we obtan the sets of deadlne ponts DP Ω ={(,), (,), (5,)} and external ponts E Ω ={(,), (5,)}. Then, usng the external ponts we can derve the α subntervals and usng expressons and we can establsh the (α, ) and (C,T ) soluton spaces, respectvely (Fgure 8). In the (α, ) fgure we can see that our soluton space s contaned n the one derved n [] and thus t s worse, although for a small dfference (less than % n the low values of α). On the other

7 hand, the worst-case tme complexty of the method n [] may reach NΩ -, contrastng wth the smplcty of our method wth a worst-case tme complexty of N Ω. Executng the fnal mprovement step wth Equaton consderng C O =.6 generates an operatng pont (Fgure 8) that s slghtly better than both soluton spaces, wth (α, ) = (.5,.8) or equvalently (C,T ) = (.,.9). Just for comparson, the corner n the sold lne of Fgure 8-a) corresponds to (α, ) = (.55,.8). Fgure 9 shows the obtaned avalablty functons (top) and varaton of the cost functon F wth respect to T (bottom). 5 5 A_s(t) A_'s(t) o deadlne ponts + external ponts.5 Our soluton space oluton space n [] 5 t 5 5 a) Avalablty functons plus deadlne and external ponts delta.5.5 Operatng pont wth mproved Ts alfa.8.6 F.. alfa Co overhead Total bandwdth Cs 5 a) (α, ) soluton space Cs Cs Ts max(cs) (Cs,Ts) soluton space Operatonal pont wth mproved Ts 5 Ts b) (C s,t s) soluton space Fgure 8. oluton spaces for the gven example 7. YNCHRONIZATION AMONG TAK In many practcal applcatons there s the need for synchronzaton n the access to shared resources that consttute crtcal sectons. There are several protocols specfcally developed for real-tme systems that allow boundng prorty nversons and blockng ntervals as well as preventng deadlocks. Our ssue here s to dscuss whether and how can such protocols be used wthn the temporally parttoned framework consdered n ths paper, wth fxed prortes local schedulers. 6 8 Ts b) cost functon wth respect to Ts Fgure 9. Avalablty and cost functons for the example In ths paper, we consder ntra-server synchronzaton, only,.e. synchronzaton among tasks of the same applcaton (Fgure ), whch execute wthn the same partton and requre access to resources that are exclusve to that applcaton (applcaton resources). These tasks eventually cause blockng to one another, the duraton of whch depends on the specfc synchronzaton protocol used. uch blockng (applcaton level blockng - B a ) conssts on the executon of crtcal sectons of lower prorty tasks,.e. the blockng tasks, whle there are hgher prorty ones pendng,.e. the blocked tasks. Therefore, the applcaton level blockng B a causes nterference to the blocked tasks, delayng them, and consequently must be consdered n the respectve level submtted load H (t), whch must be updated accordngly (Equaton 5). server- τ τ ( t + J ) / T a H ( t) = B + * C (5) shared resource = blockng of τ server suspenson Fgure. Intra-server blockng. server- tme

8 Moreover, we can state Lemma wth respect to the propertes of the synchronzaton protocols. Lemma. All propertes of any synchronzaton protocol appled to shared applcaton resources wthn a server, reman the same as establshed n a non-parttoned processng system, namely the number and duraton of prorty nversons and deadlock avodance. Proof: To prove Lemma notce that semaphore lockng and unlockng and the related blockng take place wthn the applcaton, only, durng the executon of the respectve server. When the server s suspended untl the next perodc nstance, the status of all semaphores stays unaltered. Therefore, from the pont of vew of the synchronzaton protocol, whchever t s, these perods of tme have no mpact on the protocol propertes, except on the nflaton of the blockng that, obvously, can now extend across consecutve server nstances. However, ths extra delay due to server suspenson s the same that the task would suffer ust because of executng wthn a partton and that s already accounted for n the avalablty functon of the server and should not be consdered blockng. Thus, excludng ths suspenson delay, the applcaton level blockng (B a ) that a task can suffer s the same as t would be f the respectve applcaton was executed n a dedcated system,.e. non-parttoned. Lemma plus the updated load functon lead to Theorem : Theorem. The response tme analyss and the server desgn method proposed prevously n ths paper apply equally when the task model consders applcaton level blockng as long as: the updated load functon s used (Equaton 5); the applcaton tasks can always be suspended at any pont of ther executon when the server capacty s exhausted or the server s preempted as determned by the system scheduler. Proof: The proof of ths theorem can be easly establshed by notng that, from a response tme analyss pont of vew, and gven Lemma, the blockng ust represents extra load that s accounted for n the updated load functon. Thus, all the reasonng behnd the prevously shown analyss that s based on the matchng between server avalablty and submtted load stll apples. The second requrement guarantees that there s no couplng between the applcaton level synchronzaton protocol and the system level schedulng. Ths assures that a server does not execute more than ts capacty n any nstance, not causng/sufferng extra nterference on/from other servers runnng n the system, thus behavng accordng to our perodc model. The second requrement of Theorem may conflct wth synchronzaton protocols based on non-preempton. Thus, ether preemptve protocols are used, e.g. Prorty Inhertance (PIP), Prorty Celng (PCP) and tack Resource Protocols (RP), or the executon platform must be able to separate preempton wthn the applcaton from preempton at the system level and assurng that the latter s always possble, even when the server s executng a local non-preemptve applcaton secton. 8. CONCLUION Ths paper consdered the case n whch an applcaton composed by several tasks executes wthn a perodc server wth a fxed prortes local schedulng polcy. Two man results are presented, the response tme analyss for such tasks and the desgn of the server to allow fulfllng the applcaton tme constrants usng the least system resources. The former contrbuton s a generalzaton of the well-known worst-case response tme analyss for fxed-prorty systems [] that copes wth the lmted processor avalablty delvered by a server and t s based on the analyss prevously developed for traffc schedulng wthn the asynchronous messagng system of FTT-CAN []. It s also equvalent to the one n [] but ncludes a more complete task model wth release tter, deadlnes earler than perods and synchronzaton blockng. The latter contrbuton goes n the same drecton as that of [] but presents a dfferent method that leads to a more favorable compromse between tghtness of the soluton (slghtly lower) and complexty of the process (substantally lower). Agan, the presented method also bears smlartes wth the one n [] but, not only t ncludes a more complete task model as referred above, as t also presents an heurstc to deduce the server parameters that mnmze resource utlzaton takng nto account the context swtch overhead at the system level. Moreover, we present a fnal optmzaton step that s applcable to our method as well as to [] and [], whch reduces the requred server utlzaton when a lnear bound to the server servce s used. A spn-off result that seems to have potental for more generalzed use s the utlzaton of external ponts n the response tme analyss. Further work wll address the case of nter-server synchronzaton blockng as well as non-preempton at the applcaton level and ts mpact at the system level. 9. ACKNOWLEDGMENT The authors would lke to thank E. Bn and G. Lpar for the frutful dscussons concernng the server desgn problem that helped n mprovng the respectve part of ths paper.. REFERENCE [] Almeda L., P. Pedreras, J. A. Fonseca, The FTT-CAN Protocol: Why and How, IEEE Transactons on Industral Electroncs, 9(6), December. [] Almeda L., J. Fonseca. Analyss of a mple Model for Non- Preemptve Blockng-Free chedulng. Proc. of ECRT (EUROMICRO Conf. on Real-Tme ystems). Delft, Holland. June. [] Audsley, N., A. Burns, M. Rchardson, K. Tndell and A. Wellngs. Applyng New chedulng Theory to tatc Prorty Pre-Emptve chedulng. oftware Engneerng Journal, 8(5): 85-9, 99. [] Lpar G. and E. Bn. Resource Parttonng among Real- Tme Applcatons. Proc. of ECRT (EUROMICRO Conf. on Real-Tme ystems). Porto, Portugal. July. [5] Xu, J., D.L. Parnas. chedulng processes wth release tmes, deadlnes, precedence and excluson relatons. IEEE Trans. on oftware Engneerng, 6:6-69 March 99.

9 [6] aewong,., R. Rakumar, J.P. Lehoczky, M.H. Klen. Analyss of herarchcal fxed prorty schedulng. Proc. of ECRT (EUROMICRO Conf. on Real-Tme ystems). Venna, Austra. June. [7] Mok, A., X. Feng. A model of herarchcal real-tme vrtual resources. Proc. of RT (IEEE Real-Tme ystems ymposum). Austn, UA. December. [8] Rushby, J., A Comparson of Bus Archtectures for afety- Crtcal Embedded ystems, CL Techncal Report, RI Internatonal, eptember. [9] Howell, R. and M. Venkatrao. On Non-Preemptve chedulng of Recurrng Tasks Usng Inserted Idle Tmes. Informaton and Computaton, 7, 995. [] Almeda, L. Response-Tme Analyss and erver Desgn for Herarchcal chedulng. Proc. of the Work-n-Progress sesson of RT (IEEE Real-Tme ystems ymposum). Cancun, Mexco. December. [] hn, I. and I. Lee. Perodc Resource Model for Compostonal Real-Tme Guarantees. Proc. of RT (IEEE Real-Tme ystems ymposum). Cancun, Mexco. Dec. [] Harbour M. and J. Palenca. Response-Tme Analyss for Tasks cheduled under EDF wthn Fxed Prortes. Proc. of RT (IEEE Real-Tme ystems ymposum). Cancun, Mexco. December. [] Pedreras, P. and L. Almeda. Combnng Tme and Eventtrggered Traffc n FTT-CAN. Proc. of WFC (IEEE Work. Factory Communcaton ystems). Porto, Portugal. ept. [] Mok, A., X. Feng. Real-tme vrtual resource: A Tmely Abstracton for Embedded ystems. Proc. of Emoft ( nd Int. Conf. on Embedded oftware). Grenoble, France. October.

Real-Time Systems. Multiprocessor scheduling. Multiprocessor scheduling. Multiprocessor scheduling

Real-Time Systems. Multiprocessor scheduling. Multiprocessor scheduling. Multiprocessor scheduling Real-Tme Systems Multprocessor schedulng Specfcaton Implementaton Verfcaton Multprocessor schedulng -- -- Global schedulng How are tasks assgned to processors? Statc assgnment The processor(s) used for

More information

Embedded Systems. 4. Aperiodic and Periodic Tasks

Embedded Systems. 4. Aperiodic and Periodic Tasks Embedded Systems 4. Aperodc and Perodc Tasks Lothar Thele 4-1 Contents of Course 1. Embedded Systems Introducton 2. Software Introducton 7. System Components 10. Models 3. Real-Tme Models 4. Perodc/Aperodc

More information

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud Resource Allocaton wth a Budget Constrant for Computng Independent Tasks n the Cloud Wemng Sh and Bo Hong School of Electrcal and Computer Engneerng Georga Insttute of Technology, USA 2nd IEEE Internatonal

More information

Two Methods to Release a New Real-time Task

Two Methods to Release a New Real-time Task Two Methods to Release a New Real-tme Task Abstract Guangmng Qan 1, Xanghua Chen 2 College of Mathematcs and Computer Scence Hunan Normal Unversty Changsha, 410081, Chna qqyy@hunnu.edu.cn Gang Yao 3 Sebel

More information

Problem Set 9 Solutions

Problem Set 9 Solutions Desgn and Analyss of Algorthms May 4, 2015 Massachusetts Insttute of Technology 6.046J/18.410J Profs. Erk Demane, Srn Devadas, and Nancy Lynch Problem Set 9 Solutons Problem Set 9 Solutons Ths problem

More information

Last Time. Priority-based scheduling. Schedulable utilization Rate monotonic rule: Keep utilization below 69% Static priorities Dynamic priorities

Last Time. Priority-based scheduling. Schedulable utilization Rate monotonic rule: Keep utilization below 69% Static priorities Dynamic priorities Last Tme Prorty-based schedulng Statc prortes Dynamc prortes Schedulable utlzaton Rate monotonc rule: Keep utlzaton below 69% Today Response tme analyss Blockng terms Prorty nverson And solutons Release

More information

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE Analytcal soluton s usually not possble when exctaton vares arbtrarly wth tme or f the system s nonlnear. Such problems can be solved by numercal tmesteppng

More information

Lecture Notes on Linear Regression

Lecture Notes on Linear Regression Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume

More information

Module 9. Lecture 6. Duality in Assignment Problems

Module 9. Lecture 6. Duality in Assignment Problems Module 9 1 Lecture 6 Dualty n Assgnment Problems In ths lecture we attempt to answer few other mportant questons posed n earler lecture for (AP) and see how some of them can be explaned through the concept

More information

Structure and Drive Paul A. Jensen Copyright July 20, 2003

Structure and Drive Paul A. Jensen Copyright July 20, 2003 Structure and Drve Paul A. Jensen Copyrght July 20, 2003 A system s made up of several operatons wth flow passng between them. The structure of the system descrbes the flow paths from nputs to outputs.

More information

Improved Worst-Case Response-Time Calculations by Upper-Bound Conditions

Improved Worst-Case Response-Time Calculations by Upper-Bound Conditions Improved Worst-Case Response-Tme Calculatons by Upper-Bound Condtons Vctor Pollex, Steffen Kollmann, Karsten Albers and Frank Slomka Ulm Unversty Insttute of Embedded Systems/Real-Tme Systems {frstname.lastname}@un-ulm.de

More information

Kernel Methods and SVMs Extension

Kernel Methods and SVMs Extension Kernel Methods and SVMs Extenson The purpose of ths document s to revew materal covered n Machne Learnng 1 Supervsed Learnng regardng support vector machnes (SVMs). Ths document also provdes a general

More information

Chapter - 2. Distribution System Power Flow Analysis

Chapter - 2. Distribution System Power Flow Analysis Chapter - 2 Dstrbuton System Power Flow Analyss CHAPTER - 2 Radal Dstrbuton System Load Flow 2.1 Introducton Load flow s an mportant tool [66] for analyzng electrcal power system network performance. Load

More information

Difference Equations

Difference Equations Dfference Equatons c Jan Vrbk 1 Bascs Suppose a sequence of numbers, say a 0,a 1,a,a 3,... s defned by a certan general relatonshp between, say, three consecutve values of the sequence, e.g. a + +3a +1

More information

AN EXTENDIBLE APPROACH FOR ANALYSING FIXED PRIORITY HARD REAL-TIME TASKS

AN EXTENDIBLE APPROACH FOR ANALYSING FIXED PRIORITY HARD REAL-TIME TASKS AN EXENDIBLE APPROACH FOR ANALYSING FIXED PRIORIY HARD REAL-IME ASKS K. W. ndell 1 Department of Computer Scence, Unversty of York, England YO1 5DD ABSRAC As the real-tme computng ndustry moves away from

More information

NUMERICAL DIFFERENTIATION

NUMERICAL DIFFERENTIATION NUMERICAL DIFFERENTIATION 1 Introducton Dfferentaton s a method to compute the rate at whch a dependent output y changes wth respect to the change n the ndependent nput x. Ths rate of change s called the

More information

Computer Control: Task Synchronisation in Dynamic Priority Scheduling

Computer Control: Task Synchronisation in Dynamic Priority Scheduling Computer Control: Task Synchronsaton n Dynamc Prorty Schedulng Sérgo Adrano Fernandes Lopes Department of Industral Electroncs Engneerng School Unversty of Mnho Campus de Azurém 4800 Gumarães - PORTUGAL

More information

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009 College of Computer & Informaton Scence Fall 2009 Northeastern Unversty 20 October 2009 CS7880: Algorthmc Power Tools Scrbe: Jan Wen and Laura Poplawsk Lecture Outlne: Prmal-dual schema Network Desgn:

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

Improving the Sensitivity of Deadlines with a Specific Asynchronous Scenario for Harmonic Periodic Tasks scheduled by FP

Improving the Sensitivity of Deadlines with a Specific Asynchronous Scenario for Harmonic Periodic Tasks scheduled by FP Improvng the Senstvty of Deadlnes wth a Specfc Asynchronous Scenaro for Harmonc Perodc Tasks scheduled by FP P. Meumeu Yoms, Y. Sorel, D. de Rauglaudre AOSTE Project-team INRIA Pars-Rocquencourt Le Chesnay,

More information

EEL 6266 Power System Operation and Control. Chapter 3 Economic Dispatch Using Dynamic Programming

EEL 6266 Power System Operation and Control. Chapter 3 Economic Dispatch Using Dynamic Programming EEL 6266 Power System Operaton and Control Chapter 3 Economc Dspatch Usng Dynamc Programmng Pecewse Lnear Cost Functons Common practce many utltes prefer to represent ther generator cost functons as sngle-

More information

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification E395 - Pattern Recognton Solutons to Introducton to Pattern Recognton, Chapter : Bayesan pattern classfcaton Preface Ths document s a soluton manual for selected exercses from Introducton to Pattern Recognton

More information

Simultaneous Optimization of Berth Allocation, Quay Crane Assignment and Quay Crane Scheduling Problems in Container Terminals

Simultaneous Optimization of Berth Allocation, Quay Crane Assignment and Quay Crane Scheduling Problems in Container Terminals Smultaneous Optmzaton of Berth Allocaton, Quay Crane Assgnment and Quay Crane Schedulng Problems n Contaner Termnals Necat Aras, Yavuz Türkoğulları, Z. Caner Taşkın, Kuban Altınel Abstract In ths work,

More information

Annexes. EC.1. Cycle-base move illustration. EC.2. Problem Instances

Annexes. EC.1. Cycle-base move illustration. EC.2. Problem Instances ec Annexes Ths Annex frst llustrates a cycle-based move n the dynamc-block generaton tabu search. It then dsplays the characterstcs of the nstance sets, followed by detaled results of the parametercalbraton

More information

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X Statstcs 1: Probablty Theory II 37 3 EPECTATION OF SEVERAL RANDOM VARIABLES As n Probablty Theory I, the nterest n most stuatons les not on the actual dstrbuton of a random vector, but rather on a number

More information

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS Avalable onlne at http://sck.org J. Math. Comput. Sc. 3 (3), No., 6-3 ISSN: 97-537 COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

More information

Physics 5153 Classical Mechanics. Principle of Virtual Work-1

Physics 5153 Classical Mechanics. Principle of Virtual Work-1 P. Guterrez 1 Introducton Physcs 5153 Classcal Mechancs Prncple of Vrtual Work The frst varatonal prncple we encounter n mechancs s the prncple of vrtual work. It establshes the equlbrum condton of a mechancal

More information

Resource Sharing. CSCE 990: Real-Time Systems. Steve Goddard. Resources & Resource Access Control (Chapter 8 of Liu)

Resource Sharing. CSCE 990: Real-Time Systems. Steve Goddard. Resources & Resource Access Control (Chapter 8 of Liu) CSCE 990: Real-Tme Systems Resource Sharng Steve Goddard goddard@cse.unl.edu http://www.cse.unl.edu/~goddard/courses/realtmesystems Resources & Resource Access Control (Chapter 8 of Lu) Real-Tme Systems

More information

NP-Completeness : Proofs

NP-Completeness : Proofs NP-Completeness : Proofs Proof Methods A method to show a decson problem Π NP-complete s as follows. (1) Show Π NP. (2) Choose an NP-complete problem Π. (3) Show Π Π. A method to show an optmzaton problem

More information

TOPICS MULTIPLIERLESS FILTER DESIGN ELEMENTARY SCHOOL ALGORITHM MULTIPLICATION

TOPICS MULTIPLIERLESS FILTER DESIGN ELEMENTARY SCHOOL ALGORITHM MULTIPLICATION 1 2 MULTIPLIERLESS FILTER DESIGN Realzaton of flters wthout full-fledged multplers Some sldes based on support materal by W. Wolf for hs book Modern VLSI Desgn, 3 rd edton. Partly based on followng papers:

More information

Calculation of time complexity (3%)

Calculation of time complexity (3%) Problem 1. (30%) Calculaton of tme complexty (3%) Gven n ctes, usng exhaust search to see every result takes O(n!). Calculaton of tme needed to solve the problem (2%) 40 ctes:40! dfferent tours 40 add

More information

Design and Optimization of Fuzzy Controller for Inverse Pendulum System Using Genetic Algorithm

Design and Optimization of Fuzzy Controller for Inverse Pendulum System Using Genetic Algorithm Desgn and Optmzaton of Fuzzy Controller for Inverse Pendulum System Usng Genetc Algorthm H. Mehraban A. Ashoor Unversty of Tehran Unversty of Tehran h.mehraban@ece.ut.ac.r a.ashoor@ece.ut.ac.r Abstract:

More information

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers Psychology 282 Lecture #24 Outlne Regresson Dagnostcs: Outlers In an earler lecture we studed the statstcal assumptons underlyng the regresson model, ncludng the followng ponts: Formal statement of assumptons.

More information

Lecture 4: November 17, Part 1 Single Buffer Management

Lecture 4: November 17, Part 1 Single Buffer Management Lecturer: Ad Rosén Algorthms for the anagement of Networs Fall 2003-2004 Lecture 4: November 7, 2003 Scrbe: Guy Grebla Part Sngle Buffer anagement In the prevous lecture we taled about the Combned Input

More information

Foundations of Arithmetic

Foundations of Arithmetic Foundatons of Arthmetc Notaton We shall denote the sum and product of numbers n the usual notaton as a 2 + a 2 + a 3 + + a = a, a 1 a 2 a 3 a = a The notaton a b means a dvdes b,.e. ac = b where c s an

More information

Fixed-Priority Multiprocessor Scheduling with Liu & Layland s Utilization Bound

Fixed-Priority Multiprocessor Scheduling with Liu & Layland s Utilization Bound Fxed-Prorty Multprocessor Schedulng wth Lu & Layland s Utlzaton Bound Nan Guan, Martn Stgge, Wang Y and Ge Yu Department of Informaton Technology, Uppsala Unversty, Sweden Department of Computer Scence

More information

The optimal delay of the second test is therefore approximately 210 hours earlier than =2.

The optimal delay of the second test is therefore approximately 210 hours earlier than =2. THE IEC 61508 FORMULAS 223 The optmal delay of the second test s therefore approxmately 210 hours earler than =2. 8.4 The IEC 61508 Formulas IEC 61508-6 provdes approxmaton formulas for the PF for smple

More information

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan Wnter 2008 CS567 Stochastc Lnear/Integer Programmng Guest Lecturer: Xu, Huan Class 2: More Modelng Examples 1 Capacty Expanson Capacty expanson models optmal choces of the tmng and levels of nvestments

More information

An Interactive Optimisation Tool for Allocation Problems

An Interactive Optimisation Tool for Allocation Problems An Interactve Optmsaton ool for Allocaton Problems Fredr Bonäs, Joam Westerlund and apo Westerlund Process Desgn Laboratory, Faculty of echnology, Åbo Aadem Unversty, uru 20500, Fnland hs paper presents

More information

Overhead-Aware Compositional Analysis of Real-Time Systems

Overhead-Aware Compositional Analysis of Real-Time Systems Overhead-Aware ompostonal Analyss of Real-Tme Systems Lnh T.X. Phan, Meng Xu, Jaewoo Lee, nsup Lee, Oleg Sokolsky PRESE enter Department of omputer and nformaton Scence Unversty of Pennsylvana ompostonal

More information

Worst-case response time analysis of real-time tasks under fixed-priority scheduling with deferred preemption

Worst-case response time analysis of real-time tasks under fixed-priority scheduling with deferred preemption Real-Tme Syst (2009) 42: 63 119 DOI 10.1007/s11241-009-9071-z Worst-case response tme analyss of real-tme tasks under fxed-prorty schedulng wth deferred preempton Render J. Brl Johan J. Lukken Wm F.J.

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have

More information

Improving the Quality of Control of Periodic Tasks Scheduled by FP with an Asynchronous Approach

Improving the Quality of Control of Periodic Tasks Scheduled by FP with an Asynchronous Approach Improvng the Qualty of Control of Perodc Tasks Scheduled by FP wth an Asynchronous Approach P. Meumeu Yoms, L. George, Y. Sorel, D. de Rauglaudre AOSTE Project-team INRIA Pars-Rocquencourt Le Chesnay,

More information

The Minimum Universal Cost Flow in an Infeasible Flow Network

The Minimum Universal Cost Flow in an Infeasible Flow Network Journal of Scences, Islamc Republc of Iran 17(2): 175-180 (2006) Unversty of Tehran, ISSN 1016-1104 http://jscencesutacr The Mnmum Unversal Cost Flow n an Infeasble Flow Network H Saleh Fathabad * M Bagheran

More information

A 2D Bounded Linear Program (H,c) 2D Linear Programming

A 2D Bounded Linear Program (H,c) 2D Linear Programming A 2D Bounded Lnear Program (H,c) h 3 v h 8 h 5 c h 4 h h 6 h 7 h 2 2D Lnear Programmng C s a polygonal regon, the ntersecton of n halfplanes. (H, c) s nfeasble, as C s empty. Feasble regon C s unbounded

More information

9 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations

9 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations Physcs 171/271 - Chapter 9R -Davd Klenfeld - Fall 2005 9 Dervaton of Rate Equatons from Sngle-Cell Conductance (Hodgkn-Huxley-lke) Equatons We consder a network of many neurons, each of whch obeys a set

More information

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems Numercal Analyss by Dr. Anta Pal Assstant Professor Department of Mathematcs Natonal Insttute of Technology Durgapur Durgapur-713209 emal: anta.bue@gmal.com 1 . Chapter 5 Soluton of System of Lnear Equatons

More information

1 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations

1 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations Physcs 171/271 -Davd Klenfeld - Fall 2005 (revsed Wnter 2011) 1 Dervaton of Rate Equatons from Sngle-Cell Conductance (Hodgkn-Huxley-lke) Equatons We consder a network of many neurons, each of whch obeys

More information

Quantifying the Sub-optimality of Uniprocessor Fixed Priority Non-Pre-emptive Scheduling

Quantifying the Sub-optimality of Uniprocessor Fixed Priority Non-Pre-emptive Scheduling Quantfyng the Sub-optmalty of Unprocessor Fxed Prorty Non-Pre-emptve Schedulng Robert I Davs Real-Tme Systems Research Group, Department of Computer Scence, Unversty of York, York, UK robdavs@csyorkacuk

More information

Lecture 4. Instructor: Haipeng Luo

Lecture 4. Instructor: Haipeng Luo Lecture 4 Instructor: Hapeng Luo In the followng lectures, we focus on the expert problem and study more adaptve algorthms. Although Hedge s proven to be worst-case optmal, one may wonder how well t would

More information

Limited Preemptive Scheduling for Real-Time Systems: a Survey

Limited Preemptive Scheduling for Real-Time Systems: a Survey Lmted Preemptve Schedulng for Real-Tme Systems: a Survey Gorgo C. Buttazzo, Fellow Member, IEEE, Marko Bertogna, Senor Member, IEEE, and Gang Yao Abstract The queston whether preemptve algorthms are better

More information

On the correction of the h-index for career length

On the correction of the h-index for career length 1 On the correcton of the h-ndex for career length by L. Egghe Unverstet Hasselt (UHasselt), Campus Depenbeek, Agoralaan, B-3590 Depenbeek, Belgum 1 and Unverstet Antwerpen (UA), IBW, Stadscampus, Venusstraat

More information

On the Multicriteria Integer Network Flow Problem

On the Multicriteria Integer Network Flow Problem BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 5, No 2 Sofa 2005 On the Multcrtera Integer Network Flow Problem Vassl Vasslev, Marana Nkolova, Maryana Vassleva Insttute of

More information

4DVAR, according to the name, is a four-dimensional variational method.

4DVAR, according to the name, is a four-dimensional variational method. 4D-Varatonal Data Assmlaton (4D-Var) 4DVAR, accordng to the name, s a four-dmensonal varatonal method. 4D-Var s actually a drect generalzaton of 3D-Var to handle observatons that are dstrbuted n tme. The

More information

MMA and GCMMA two methods for nonlinear optimization

MMA and GCMMA two methods for nonlinear optimization MMA and GCMMA two methods for nonlnear optmzaton Krster Svanberg Optmzaton and Systems Theory, KTH, Stockholm, Sweden. krlle@math.kth.se Ths note descrbes the algorthms used n the author s 2007 mplementatons

More information

2016 Wiley. Study Session 2: Ethical and Professional Standards Application

2016 Wiley. Study Session 2: Ethical and Professional Standards Application 6 Wley Study Sesson : Ethcal and Professonal Standards Applcaton LESSON : CORRECTION ANALYSIS Readng 9: Correlaton and Regresson LOS 9a: Calculate and nterpret a sample covarance and a sample correlaton

More information

Lecture 10 Support Vector Machines II

Lecture 10 Support Vector Machines II Lecture 10 Support Vector Machnes II 22 February 2016 Taylor B. Arnold Yale Statstcs STAT 365/665 1/28 Notes: Problem 3 s posted and due ths upcomng Frday There was an early bug n the fake-test data; fxed

More information

Lecture Note 3. Eshelby s Inclusion II

Lecture Note 3. Eshelby s Inclusion II ME340B Elastcty of Mcroscopc Structures Stanford Unversty Wnter 004 Lecture Note 3. Eshelby s Incluson II Chrs Wenberger and We Ca c All rghts reserved January 6, 004 Contents 1 Incluson energy n an nfnte

More information

8 Derivation of Network Rate Equations from Single- Cell Conductance Equations

8 Derivation of Network Rate Equations from Single- Cell Conductance Equations Physcs 178/278 - Davd Klenfeld - Wnter 2015 8 Dervaton of Network Rate Equatons from Sngle- Cell Conductance Equatons We consder a network of many neurons, each of whch obeys a set of conductancebased,

More information

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg prnceton unv. F 17 cos 521: Advanced Algorthm Desgn Lecture 7: LP Dualty Lecturer: Matt Wenberg Scrbe: LP Dualty s an extremely useful tool for analyzng structural propertes of lnear programs. Whle there

More information

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011 Stanford Unversty CS359G: Graph Parttonng and Expanders Handout 4 Luca Trevsan January 3, 0 Lecture 4 In whch we prove the dffcult drecton of Cheeger s nequalty. As n the past lectures, consder an undrected

More information

Quantifying the Sub-optimality of Uniprocessor Fixed Priority Pre-emptive Scheduling for Sporadic Tasksets with Arbitrary Deadlines

Quantifying the Sub-optimality of Uniprocessor Fixed Priority Pre-emptive Scheduling for Sporadic Tasksets with Arbitrary Deadlines Quantfyng the Sub-optmalty of Unprocessor Fxed Prorty Pre-emptve Schedulng for Sporadc Tasksets wth Arbtrary Deadlnes Robert Davs, Sanjoy Baruah, Thomas Rothvoss, Alan Burns To cte ths verson: Robert Davs,

More information

Prof. Dr. I. Nasser Phys 630, T Aug-15 One_dimensional_Ising_Model

Prof. Dr. I. Nasser Phys 630, T Aug-15 One_dimensional_Ising_Model EXACT OE-DIMESIOAL ISIG MODEL The one-dmensonal Isng model conssts of a chan of spns, each spn nteractng only wth ts two nearest neghbors. The smple Isng problem n one dmenson can be solved drectly n several

More information

Feature Selection: Part 1

Feature Selection: Part 1 CSE 546: Machne Learnng Lecture 5 Feature Selecton: Part 1 Instructor: Sham Kakade 1 Regresson n the hgh dmensonal settng How do we learn when the number of features d s greater than the sample sze n?

More information

The Study of Teaching-learning-based Optimization Algorithm

The Study of Teaching-learning-based Optimization Algorithm Advanced Scence and Technology Letters Vol. (AST 06), pp.05- http://dx.do.org/0.57/astl.06. The Study of Teachng-learnng-based Optmzaton Algorthm u Sun, Yan fu, Lele Kong, Haolang Q,, Helongang Insttute

More information

Supplement: Proofs and Technical Details for The Solution Path of the Generalized Lasso

Supplement: Proofs and Technical Details for The Solution Path of the Generalized Lasso Supplement: Proofs and Techncal Detals for The Soluton Path of the Generalzed Lasso Ryan J. Tbshran Jonathan Taylor In ths document we gve supplementary detals to the paper The Soluton Path of the Generalzed

More information

Global EDF Scheduling for Parallel Real-Time Tasks

Global EDF Scheduling for Parallel Real-Time Tasks Washngton Unversty n St. Lous Washngton Unversty Open Scholarshp Engneerng and Appled Scence Theses & Dssertatons Engneerng and Appled Scence Sprng 5-15-2014 Global EDF Schedulng for Parallel Real-Tme

More information

Appendix B: Resampling Algorithms

Appendix B: Resampling Algorithms 407 Appendx B: Resamplng Algorthms A common problem of all partcle flters s the degeneracy of weghts, whch conssts of the unbounded ncrease of the varance of the mportance weghts ω [ ] of the partcles

More information

More metrics on cartesian products

More metrics on cartesian products More metrcs on cartesan products If (X, d ) are metrc spaces for 1 n, then n Secton II4 of the lecture notes we defned three metrcs on X whose underlyng topologes are the product topology The purpose of

More information

Errors for Linear Systems

Errors for Linear Systems Errors for Lnear Systems When we solve a lnear system Ax b we often do not know A and b exactly, but have only approxmatons  and ˆb avalable. Then the best thng we can do s to solve ˆx ˆb exactly whch

More information

Some modelling aspects for the Matlab implementation of MMA

Some modelling aspects for the Matlab implementation of MMA Some modellng aspects for the Matlab mplementaton of MMA Krster Svanberg krlle@math.kth.se Optmzaton and Systems Theory Department of Mathematcs KTH, SE 10044 Stockholm September 2004 1. Consdered optmzaton

More information

Chapter 11: Simple Linear Regression and Correlation

Chapter 11: Simple Linear Regression and Correlation Chapter 11: Smple Lnear Regresson and Correlaton 11-1 Emprcal Models 11-2 Smple Lnear Regresson 11-3 Propertes of the Least Squares Estmators 11-4 Hypothess Test n Smple Lnear Regresson 11-4.1 Use of t-tests

More information

Chapter 13: Multiple Regression

Chapter 13: Multiple Regression Chapter 13: Multple Regresson 13.1 Developng the multple-regresson Model The general model can be descrbed as: It smplfes for two ndependent varables: The sample ft parameter b 0, b 1, and b are used to

More information

Single-Facility Scheduling over Long Time Horizons by Logic-based Benders Decomposition

Single-Facility Scheduling over Long Time Horizons by Logic-based Benders Decomposition Sngle-Faclty Schedulng over Long Tme Horzons by Logc-based Benders Decomposton Elvn Coban and J. N. Hooker Tepper School of Busness, Carnege Mellon Unversty ecoban@andrew.cmu.edu, john@hooker.tepper.cmu.edu

More information

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results.

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results. Neural Networks : Dervaton compled by Alvn Wan from Professor Jtendra Malk s lecture Ths type of computaton s called deep learnng and s the most popular method for many problems, such as computer vson

More information

Lecture 12: Discrete Laplacian

Lecture 12: Discrete Laplacian Lecture 12: Dscrete Laplacan Scrbe: Tanye Lu Our goal s to come up wth a dscrete verson of Laplacan operator for trangulated surfaces, so that we can use t n practce to solve related problems We are mostly

More information

Spectral Graph Theory and its Applications September 16, Lecture 5

Spectral Graph Theory and its Applications September 16, Lecture 5 Spectral Graph Theory and ts Applcatons September 16, 2004 Lecturer: Danel A. Spelman Lecture 5 5.1 Introducton In ths lecture, we wll prove the followng theorem: Theorem 5.1.1. Let G be a planar graph

More information

Assortment Optimization under MNL

Assortment Optimization under MNL Assortment Optmzaton under MNL Haotan Song Aprl 30, 2017 1 Introducton The assortment optmzaton problem ams to fnd the revenue-maxmzng assortment of products to offer when the prces of products are fxed.

More information

Critical sections. Using semaphores. Using semaphores. Using semaphores. How long is blocking time? 17/10/2016. Problems caused by mutual exclusion

Critical sections. Using semaphores. Using semaphores. Using semaphores. How long is blocking time? 17/10/2016. Problems caused by mutual exclusion rtcal sectons Problems caused by mutual excluson crtcal secton wat(s) x = ; y = 5; sgnal(s) wrte global memory buffer nt x; nt y; read wat(s) a = x+; b = y+; c = x+y; crtcal secton sgnal(s) Usng semaphores

More information

Outline. Communication. Bellman Ford Algorithm. Bellman Ford Example. Bellman Ford Shortest Path [1]

Outline. Communication. Bellman Ford Algorithm. Bellman Ford Example. Bellman Ford Shortest Path [1] DYNAMIC SHORTEST PATH SEARCH AND SYNCHRONIZED TASK SWITCHING Jay Wagenpfel, Adran Trachte 2 Outlne Shortest Communcaton Path Searchng Bellmann Ford algorthm Algorthm for dynamc case Modfcatons to our algorthm

More information

find (x): given element x, return the canonical element of the set containing x;

find (x): given element x, return the canonical element of the set containing x; COS 43 Sprng, 009 Dsjont Set Unon Problem: Mantan a collecton of dsjont sets. Two operatons: fnd the set contanng a gven element; unte two sets nto one (destructvely). Approach: Canoncal element method:

More information

EDF Scheduling for Identical Multiprocessor Systems

EDF Scheduling for Identical Multiprocessor Systems EDF Schedulng for dentcal Multprocessor Systems Maro Bertogna Unversty of Modena, taly As Moore s law goes on Number of transstor/chp doubles every 18 to 24 mm heatng becomes a problem Power densty (W/cm

More information

U.C. Berkeley CS294: Spectral Methods and Expanders Handout 8 Luca Trevisan February 17, 2016

U.C. Berkeley CS294: Spectral Methods and Expanders Handout 8 Luca Trevisan February 17, 2016 U.C. Berkeley CS94: Spectral Methods and Expanders Handout 8 Luca Trevsan February 7, 06 Lecture 8: Spectral Algorthms Wrap-up In whch we talk about even more generalzatons of Cheeger s nequaltes, and

More information

DUE: WEDS FEB 21ST 2018

DUE: WEDS FEB 21ST 2018 HOMEWORK # 1: FINITE DIFFERENCES IN ONE DIMENSION DUE: WEDS FEB 21ST 2018 1. Theory Beam bendng s a classcal engneerng analyss. The tradtonal soluton technque makes smplfyng assumptons such as a constant

More information

Analysis of Queuing Delay in Multimedia Gateway Call Routing

Analysis of Queuing Delay in Multimedia Gateway Call Routing Analyss of Queung Delay n Multmeda ateway Call Routng Qwe Huang UTtarcom Inc, 33 Wood Ave. outh Iseln, NJ 08830, U..A Errol Lloyd Computer Informaton cences Department, Unv. of Delaware, Newark, DE 976,

More information

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity Week3, Chapter 4 Moton n Two Dmensons Lecture Quz A partcle confned to moton along the x axs moves wth constant acceleraton from x =.0 m to x = 8.0 m durng a 1-s tme nterval. The velocty of the partcle

More information

VQ widely used in coding speech, image, and video

VQ widely used in coding speech, image, and video at Scalar quantzers are specal cases of vector quantzers (VQ): they are constraned to look at one sample at a tme (memoryless) VQ does not have such constrant better RD perfomance expected Source codng

More information

Amiri s Supply Chain Model. System Engineering b Department of Mathematics and Statistics c Odette School of Business

Amiri s Supply Chain Model. System Engineering b Department of Mathematics and Statistics c Odette School of Business Amr s Supply Chan Model by S. Ashtab a,, R.J. Caron b E. Selvarajah c a Department of Industral Manufacturng System Engneerng b Department of Mathematcs Statstcs c Odette School of Busness Unversty of

More information

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix Lectures - Week 4 Matrx norms, Condtonng, Vector Spaces, Lnear Independence, Spannng sets and Bass, Null space and Range of a Matrx Matrx Norms Now we turn to assocatng a number to each matrx. We could

More information

Module 2. Random Processes. Version 2 ECE IIT, Kharagpur

Module 2. Random Processes. Version 2 ECE IIT, Kharagpur Module Random Processes Lesson 6 Functons of Random Varables After readng ths lesson, ou wll learn about cdf of functon of a random varable. Formula for determnng the pdf of a random varable. Let, X be

More information

Tornado and Luby Transform Codes. Ashish Khisti Presentation October 22, 2003

Tornado and Luby Transform Codes. Ashish Khisti Presentation October 22, 2003 Tornado and Luby Transform Codes Ashsh Khst 6.454 Presentaton October 22, 2003 Background: Erasure Channel Elas[956] studed the Erasure Channel β x x β β x 2 m x 2 k? Capacty of Noseless Erasure Channel

More information

Minimizing Energy Consumption of MPI Programs in Realistic Environment

Minimizing Energy Consumption of MPI Programs in Realistic Environment Mnmzng Energy Consumpton of MPI Programs n Realstc Envronment Amna Guermouche, Ncolas Trquenaux, Benoît Pradelle and Wllam Jalby Unversté de Versalles Sant-Quentn-en-Yvelnes arxv:1502.06733v2 [cs.dc] 25

More information

Solution Thermodynamics

Solution Thermodynamics Soluton hermodynamcs usng Wagner Notaton by Stanley. Howard Department of aterals and etallurgcal Engneerng South Dakota School of nes and echnology Rapd Cty, SD 57701 January 7, 001 Soluton hermodynamcs

More information

Scheduling Motivation

Scheduling Motivation 76 eal-me & Embedded Systems 7 Uwe. Zmmer - he Australan Natonal Unversty 78 Motvaton n eal-me Systems Concurrency may lead to non-determnsm. Non-determnsm may make t harder to predct the tmng behavour.

More information

Second Order Analysis

Second Order Analysis Second Order Analyss In the prevous classes we looked at a method that determnes the load correspondng to a state of bfurcaton equlbrum of a perfect frame by egenvalye analyss The system was assumed to

More information

The Second Eigenvalue of Planar Graphs

The Second Eigenvalue of Planar Graphs Spectral Graph Theory Lecture 20 The Second Egenvalue of Planar Graphs Danel A. Spelman November 11, 2015 Dsclamer These notes are not necessarly an accurate representaton of what happened n class. The

More information

Queueing Networks II Network Performance

Queueing Networks II Network Performance Queueng Networks II Network Performance Davd Tpper Assocate Professor Graduate Telecommuncatons and Networkng Program Unversty of Pttsburgh Sldes 6 Networks of Queues Many communcaton systems must be modeled

More information

PHYS 705: Classical Mechanics. Calculus of Variations II

PHYS 705: Classical Mechanics. Calculus of Variations II 1 PHYS 705: Classcal Mechancs Calculus of Varatons II 2 Calculus of Varatons: Generalzaton (no constrant yet) Suppose now that F depends on several dependent varables : We need to fnd such that has a statonary

More information

This column is a continuation of our previous column

This column is a continuation of our previous column Comparson of Goodness of Ft Statstcs for Lnear Regresson, Part II The authors contnue ther dscusson of the correlaton coeffcent n developng a calbraton for quanttatve analyss. Jerome Workman Jr. and Howard

More information

Numerical Heat and Mass Transfer

Numerical Heat and Mass Transfer Master degree n Mechancal Engneerng Numercal Heat and Mass Transfer 06-Fnte-Dfference Method (One-dmensonal, steady state heat conducton) Fausto Arpno f.arpno@uncas.t Introducton Why we use models and

More information