L. Donatiello & R. Nelson, Eds., Performance Evaluation of Computer. and Communication Systems, Springer-Verlag, 1993, pp

Size: px
Start display at page:

Download "L. Donatiello & R. Nelson, Eds., Performance Evaluation of Computer. and Communication Systems, Springer-Verlag, 1993, pp"

Transcription

1 L. Donatello & R. Nelson, Eds., Performance Evaluaton of Computer and Communcaton Systems, Sprnger-Verlag, 1993, pp ANALYSIS AND CONTROL OF POLLING SYSTEMS Ur Yechal Department of Statstcs & Operatons Research, School of Mathematcal Scences, Sackler Faculty of Exact Scences, Tel Avv Unversty, Tel Avv 69978, Israel Emal: Abstract. We present methods for analyzng contnuous-tme multchannel queueng systems wth Gated, Exhaustve, or Globally-Gated servce regmes, and wth Cyclc, Hamltonan or Elevator-type pollng mechansms. We dscuss ssues of dynamcally controllng the server's order of vsts to the channels, and derve easly mplementable ndex-type rules that optmze system's performance. Future drectons of research are ndcated. Keywords: Mult-channel queueng systems, pollng, gated, exhaustve, globally-gated, conservaton laws, Hamltonan tours, Elevator pollng, dynamc control. 1 Introducton Queueng systems consstng of N queues (channels) served by a sngle server whch ncurs swtch-over perods when movng from one channel to another have been wdely studed n the lterature and used as a central model for the analyss of a wde varety of applcatons n the areas of computer networks, telecommuncaton systems, multple access protocols, multplexng schemes n ISDNs, reader-head's movements n a computer's hard dsk, exble manufacturng systems, road trac control, repar problems and the lke. Very often such applcatons (e.g. Token Rng networks n whch N statons attempt to transmt ther messages by sharng a sngle transmsson lne) are modeled as a pollng system where the server vsts the channels n a cyclc routne or accordng to an arbtrary pollng table. In many of these applcatons, as well as n most pollng models, t s customary to control the amount of servce gven to each queue durng the server's vst. Common servce polces are the Exhaustve, Gated and Lmted regmes. Under the Exhaustve regme, at each vst the server attends the queue untl t becomes completely empty, and only then s the server allowed to move on. Under the Gated regme, all (and only) customers (packets, obs) present when the server starts vstng (polls) the queue are served durng the vst, whle customers arrvng when the queue s attended wll be served durng the next vst. Under the K -Lmted servce dscplne only a lmted number of obs (at most K ) are served at each server's vst to queue. There s extensve lterature on 630

2 the theory and applcatons of these models. Among the rst works are Cooper & Murray [1969] and Cooper [1970] who studed the cyclc Exhaustve and Gated regmes wth no swtchover tmes. Esenberg [1972] generalzed the results of Cooper & Murray by allowng changeover tmes and by consderng a general pollng table,.e., by allowng a general conguraton of the server's (perodc) sequence of vsts to the channels. Many other authors have nvestgated varous aspects of pollng systems, and for a more detaled descrpton the reader s referred to a book [1986] and an update [1990] by Takag, and to a survey by Levy & Sd [1990]. Recently, Globally-Gated regmes were proposed by Boxma, Levy & Yechal [1992], who provded a thorough analyss of the cyclc Globally-Gated (GG) scheme. Under the Globally-Gated regme the server uses the nstant of cycle begnnng as a reference pont of tme, and serves n each queue only those obs that were present there at the cycle-begnnng. A specal, yet mportant, pollng mechansm s the so-called Elevator (or scan)-type (cf. Shoham & Yechal [1992], Altman, Khamsy & Yechal [1992]): nstead of movng cyclcally through the channels, the server rst vsts the queues n one drecton,.e. n the order 1; 2; : : :; N (`up' cycle) and then reverses ts orentaton and serves the channels n the opposte drecton (`down' cycle). Then t changes drecton agan, and keeps movng n ths manner back and forth. Ths type of servce regme s encountered n many applcatons, e.g. t models a common scheme of addressng a hard dsk for wrtng (or readng) nformaton on (or from) derent tracks. Among ts advantages s that t saves the return walkng tme from channel N to channel 1. All the above models studed open systems wth external arrvals, where obs ext the system after servce completon. Altman & Yechal [1992] studed a closed system n whch the number of obs s xed. They analyzed the Gated, Exhaustve, Mxed and Globally-Gated regmes and derved measures for system's performance. One of the man tools used n the analyss of pollng systems s the dervaton of a set of mult-dmensonal Probablty Generatng Functons (PGS 's) of the number of obs present n the varous channels at a pollng nstant to queue ( = 1; 2; : : :; N). The common method s to derve PGF +1 n terms of PGF and from the set of N (mplct) dependent equatons n the unknown PGF 's one can obtan expressons whch allow for numercal calculaton of the mean queue sze or mean watng tme at each queue. The Globally-Gated regme stands out among the varous dscplnes as t yelds a closed-form analyss and leads to explct expressons for performance measures, such as mean and second moment of watng tme at each queue, as well as the Laplace-Steltes Transform (LST) of the cycle duraton. Most of the work on pollng systems has been concentrated on obtanng equlbrum mean-value or approxmate results for the varous servce dscplnes. Browne & Yechal [1989a], [1989b] were the rst to obtan dynamc control polces for systems under the Exhaustve, Gated or Mxed servce regmes. At the begnnng of each cycle the server decdes on a new Hamltonan tour and 631

3 vsts the channels accordngly. Browne & Yechal showed that f the obectve s to mnmze (or maxmze) cycle-duraton, then an ndex-type rule apples. Such a rule makes t extremely easy for practcal mplementatons. For the Globally- Gated regme Boxma, Levy & Yechal [1992] showed that mnmzng weghted watng costs for each cycle ndvdually, mnmzes the long-run average weghted watng costs of all customers n the system. A surprsng result holds for the Globally-Gated Elevator-type mechansm (Altman, Khamsy & Yechal [1992]): mean watng tmes n all channels are the same. In ths tutoral we present and dscuss () analytcal technques used n studyng pollng systems, and () methods derved and appled for dynamc control of such systems. In sectons 3 and 4 we present the basc tools for analyzng pollng systems wth Gated or wth Exhaustve servce regmes, respectvely. Secton 5 dscusses conservton laws and optmal vst frequences. In secton 6 we address the ssue of dynamc control of pollng systems havng servce regmes wth lnear growth of work. Secon 7 studes the Globally-Gated regme, and n secton 8 the Elevatortype pollng mechansm s analyzed. Future drectons of research are ndcated n secton 9. 2 Models and Notaton A pollng system s composed of N channels (queues), labeled 1; 2; : : :; N, where `customers' (messages, obs) arrve at channel accordng to some arrval process, usually taken as an ndependent Posson process wth rate. There s a sngle server n the system whch moves from channel to channel followng a prescrbed order (`pollng table'), most-commonly cyclc,.e., vstng the queues n the order 1; 2; ; : : :; N? 1; N; 1; 2; : : :. The server stays at a channel for a length of tme determned by the servce dscplne and then moves on to the next channel. Each ob n channel ( = 1; 2; : : :; N) carres an ndependent random servce requrement B, havng dstrbuton functon G (), Laplace-Steltes Transform eb (), mean b, and second moment b (2). The queue dscplne determnes how many obs are to be served n each channel. The dscplnes most often studed are the Exhaustve, Gated and Lmted servce regmes. To llustrate these regmes, assume the server arrves to channel to nd m obs (customers) watng. Under the Exhaustve regme, the server must servce channel untl t s empty before t s allowed to move on. Ths amount of tme s dstrbuted as the sum of m ordnary busy perods n an M=G =1 queue. Under the Gated regme, the server `gates o' those m customers and serves only them before movng on to the next channel. As such, the total servce tme n channel s dstrbuted as the sum of m ordnary servce requrements. Under Lmted servce regmes, the server must serve ether 1 ob, at most K obs, or deplete the queue at channel by 1 (.e., stay one busy perod of M=G =1 type). Accordng to the recently ntroduces Globally-Gated servce regme, at the start of the cycle all channels are `gated o' smultaneously, and only customers gated at that nstant wll be served durng the comng cycle. 632

4 Typcally, the server takes a (random) non-neglgble amount of tme to swtch between channels. Ths tme s called `walkng' or `swtchover' perod. The swtchover duraton from channel to the next s denoted by D, wth LST ed (), mean d, and second moment d (2). In some applcatons (e.g. star con- guraton) the tme to move from channel to channel ( 6= ) s composed of a swtch over tme D, out of channel, plus a swtch-n perod to channel, R. In other applcatons, even for a cyclc pollng procedure, the swtch-n tme R s ncurred only f there s at least one message n queue (see Altman, Blanc, Khamsy & Yechal [1992]), thus savng the swtchng tme nto an empty channel. We wll dscuss here only systems where each channel has an nnte buer capacty, assumng steady state condtons, and we focus on contnuous-tme models where channel s an M=G =1 queue wth Posson arrval rate and servce requrements B. The analyss wll concentrate on three man servce regmes: Gated, Exhaustve and Globally-Gated. 3 Analyss of the Gated Regme Let denote the number of obs present n channel ( = 1; 2; : : :; N) when the server arrves at (polls) channel (= 1; 2; : : :; N). = ( 1; 2 ; : : :; N ) s the state of the system at that nstant. Let A (T ) be the number of Posson arrvals to channel durng a (random) tme nterval of length T. Then, for the Gated servce regme, the evoluton of the state of the system s gven by 8 > < = +1 >: + A P k=1 B k + D ; 6= A P k=1 B k + D ; = where B k are all dstrbuted as B. One of the basc tools of analyss s to derve the multdmensonal Probablty Generatng Functon (PGF ) of the state of the system at the pollng nstant to channel ( = 1; 2; : : :; N). PGF s dened as G (z) = G (z 1 ; z 2 ; : : :; z?1 ; z ; z +1 ; : : :; z N ) = E Then, for the Gated regme, whle usng (1), G +1 (z) = E NY 2 = E 6 4 N Y z +1 6= NY NY P z E z A ( k=1 3 NY Bk) 7 5 E 633 z z A (D) (1) : (2) (3)

5 For a Posson random varable A (T ), and wth e T () denotng the Laplace- Steltes Transform (LST) of T, we have and Therefore G +1 (z) = E 2 6 E[z A(T ) ] = E T [e?(1?z)t ] = T e (1? z ) 4 N Y NY E 6= z z A (T ) eb N = T e (1? z ) : (1? z ) 3 7 N 5 D e Thus, for = 1; 2; : : :; N? 1; N (where we take N + 1 as 1) 1 ; z +1 ; : : :; z N G +1 (z)=g z1 ; z 2 ; : : :; z?1 ; e B (1?z ) (1? z ) : A D e (1?z ) (4) Equatons (4) dene a set of N relatons between the varous PGFs whch are used to derve moments of the varables, as follows. Moments The mean number of messages, f () = E( ), present n channel at a pollng nstant to channel s obtaned by takng dervatves of the PGFs, where f () = E( ) (z) z=1 A set N 2 lnear equatons n f () : ; = 1; 2; : : :; N determnes ther values: f () + f +1 () = b f () + d b f () + d 6= = (6) Indeed, equatons (6) could be obtaned drectly from (1). by P P N Set k = k b k, = N k=1 k, d = 8 < P h?1 k= d k 1? + d k f () = : d 1? k=1 d k. Then, the soluton of (6) s gven 6= = The explanaton of (7) s the followng. It wll be shown shortly that the mean cycle tme s E[C] = d=(1? ). Durng that tme the mean number of arrvals to channel s E[C]. Also, durng a cycle the server renders servce to channel k for an average length of tme k E[C]. Thus, the elapsed tme snce the last gatng nstant of channel ( 6= ) untl the pollng nstant of channel, s 634 (7)

6 P?1 k= k E[C] + d k. Wthn that tme-nterval the mean number of arrvals to channel s f (), as gven by (7). The second moments of the are also derved from the set of PGFs (4). Let f (; k) = E[ k ] k z=1 f (; ) = E (? 1) ] G 2 (; ; k = 1; 2; : : :; N not all equal) z=1 (8)? Clearly, Var[ ] = f (; ) + f ()? f () 2. Takng dervatves, the soluton of (8) s gven (see, Takag [1986]) as a set of N 3 lnear equatons n the N 3 unknowns f (; k). Cycle Tme The mean cycle tme s obtaned from the balance equaton E[C] = E[C] + d. Hence, E[C] = d 1? : The mean soourn tme of the server at channel s f ()b = E[C], and the number of obs served n a cycle s clearly, P N =1 f () =?P N =1 E[C]. The PGF of L and Watng Tmes Consder the probablty generatng functon, Q (z) = E(z L ), of the number of customers, L, left behnd by an arbtrary departng customer from channel n a pollng system wth arbtrary servce regme. As the dstrbutons of the number of customers n the system at epochs of arrval and epochs of departure are dentcal, then by the well known PASTA phenomenon (Posson Arrvals See Tme Averages), Q (z) also stands for the generatng functon of the number of customers at channel n a steady state condton at an arbtrary pont of tme. Let T be the total number of customers served n channel durng a vst of the server to that channel, and let L (n) (n = 1; 2; : : :; T ), be the sequence of random varables denotng the number of customers that the n-th departng customer from channel (countng from the moment that the channel was last polled) leaves behnd t. Then the PGF of L s gven by (see, Takag [1986], p. 78) Q (z) = E?P T n=1 zl(n) E(T ) : (9) As L (n) =? n + A?P n k=1 B k, the evaluaton of the expresson for Q (z) becomes Q (z)= 1 E(T ) E T n=1 z?n+a( P n k=1 Bk) = 1 E(T ) E z T n=1 z?n+a( P n k=1 Bk) 635

7 = 1 E(T ) E z = T n=1 z?n e?( P n k=1 Bk)(1?z) = 1 0 B = 1 E(T ) eb? (1? z) z E(T ) E 1? T z h eb((1?z)) z n=1 T 1? eb((1?z)) z eb? (1?z) n! 1 A C? eb (1? z) E(T ) z? B e? E z?t? z T? eb T ( (1? z)) : (10) (1? z) Let W q denote the queueng tme of an arbtrary message at queue, and let W = W q + B denote the soourn (resdence) tme of a message n the system. As the messages left behnd by a departng message from channel have all arrved durng ts resdence tme W, we have Q (z) = Hence, 1 P number of messages at channel = k z k = 1 z k Z 1 k=0 k=0 0 = W f (1? z) = Wq f (1? z) B e (1? z) fw q (s) = Q (1? s= ) eb (s) For the Gated regme, = T, and therefore Q (z) = eb? (1? z) e?w ( w) k dp (W w) k! E(T ) z? e B ( (1? z)) E[z ]? E ( e B ( (1? z))) (see also Takag [1986], p. 109). As E(T ) = E( ) = E[C], usng (11) and (12) leads to z (11) (12) E? W q = E? ( )2? E( ) 2 E( ) (1 + ) = (1 + )f (; ) 2 2 E[C] (13) By Lttle's law, E[L ] = E(Wq ) + b. 4 Exhaustve Regme To derve the PGF of the state of the system at a pollng nstant to channel + 1 we use the law of moton +1 = 8 < : + A A (D ) ; P k=1 k + D ; 6= = (14) where denotes the length of a regular busy perod n an M/G /1 queue, and k are all dstrbuted as. Then, 636

8 Hence, NY G +1 (z) = E 2 = E 6 = E N Y z +1 6= 4 N Y 6= z z NY E 6= e 6= z A ( P k=1 k) (1? z ) 0 G +1 (z)=g 1 ; z 2 ; : : :; z?1 ; e (1?z ) 6= 3 7 NY 5 A (D) E z D e (1? z ) 1 ; z +1 ; : : :; z N C A D e (1?z ) (15) To get the N 2 values of f () one can derentate (15) or use drectly (14). The result s f () + f +1 () = E( )f () + d d 6= = (16) where E( ) = b =(1? ) s the mean duraton of a regular busy perod at channel. The soluton of (16) s f () = 8 >< >: P?1 d k=+1 k 1? (1? ) d 1? P?1 + k= d k 6= = The nterpretaton of (17) s the followng. The mean cycle tme s agan E[C] = d=(1? ), whch s derved from the same balance equaton as for the Gated regme. The fracton of tme that the server stays at channel s, hence, durng the tme nterval snce the server leaves (an empty) channel untl t arrves there agan, the mean number of accumulated messages at s (1? )E[C]. For channel 6=, the total swtchover tmes P from the moment?1 the server last exted the channel untl t enters channel s k= d k, and the mean tme spent n each of the channels k = + 1; + 2; ; : : :;? 1, s k E[C]. Thus, the expected number of?p obs accumulated P at channel when the server polls channel s gven by?1? k= d?1 k + d k=+1 k 1?. The PGF of the number of messages at channel can be obtaned by usng result (10). For the Exhaustve case, the number of customers served durng a vst to channel s T =?P + A k=1 k, so that E(T ) = f () + f ()E( ) = f ()=(1? ), and by usng (17), E(T ) = E[C]. 637 (17)

9 The PGF of the number of messages at channel at an arbtrary pont of tme s gven by Takag [1986], p. 79: Q (z) = 1 E[C] e B (1? z) z? e B (1? z) E[z ]? 1 (18) The mean number of messages at channel and the mean queueng tmes are derved from (18), E[L ] = + 2 b(2) 2(1? ) + f (; ) 2 (1? )E[C] E[W q ] = b (2) 2(1? ) + f (; ) 2 2 (1? )E[C] Agan, the values of f (; ) have to be calculated numercally by solvng a set of N 3 lnear equatons n the unknowns f (; k) derved (see (8)) by derentatng the PGFs n (15). Remarks on Computatonal Methods Several numercal procedures have been proposed for computng the mean watng tmes n pollng systems wth Gated or wth Exhaustve servce regmes. The procedure mentoned above of determnng the mean delay n varous channels by solvng a set of N 3 lnear equatons s called the Buer Occupancy method. It s of hgh computatonal complexty, but can also be appled to solve models wth swtch-n tmes or wth lmted-servce regmes. A more ecent procedure s known as the Staton Tme method (see Ferguson & Amnetzah [1985]). Ths s an teratve procedure whch has been appled to a number of pollng systems, but cannot be drectly used for closed networks or for open systems wth customers' routng. Sarkar & Zangwll [1989] have developed an algorthm for cyclc (Exhaustve or Gated) systems where the mean watng tmes are obtaned by solvng a set of only N lnear equatons (thus requrng O(N 3 ) computatonal steps). Recently, Konhen, Levy & Srnvasan [1993a] ntroduced a Descedant Set (DS) approach whch s based on countng the number of descedants generated n the system by each customer. The method can be appled to varatons of Exhaustve or Gated pollng systems whch are based on xed order of vsts, and can also be used to derve second and hgher delay moments. It s clamed that the DS s superor to other methods due to ts low computatonal complexty, even though t s based on the buer occupancy varables. In a further eort to develop ecent computatonal methods, the same authors [1993b] ntroduced the Indvdual Staton (IS) technque whch, lke the DS procedure, allows for the determnng of mean watng tme at one or more selected nodes wthout havng to obtan mean watng tmes at all channels smultaneously. The IS s superor to the DS for systems wth hgh utlzaton factor, whle the DS would be preferred for systems wth very large N. 638

10 5 Conservaton Laws and Vst Frequences In an arbtrary sngle-server system (wth sngle or multple queues) when no work s generated or lost wthn the system, the amount of work present does not depend on the order of servce { and hence equals the amount of work n the `correspondng' system wth a sngle queue and FCFS servce dscplne. Ths `prncple' of work conservaton yelds useful expresson whch we now dscuss. Suppose that no swtchng tmes are ncurred n our pollng system, and assume cyclc or any order of the server's vsts. Then t s well known (see, Klenrock [1975]) that the expected amount of work n the system s constant,.e., P N =1 b E[L ] = E(W ) = b (2) W : (19) 2(1? ) =1 =1 When swtchng tmes are ncurred, Boxma & Groenendk [1987] and Boxma [1989] have derved the so called `pseudo-conservaton laws' and showed that for an arbtrary pollng system wth mxed channels =1 E(W ) = W + d(2) 2d + =1 d 2(1? ) E(W ) = W + d(2) 2d + 2? =1 d 2(1? ) 2 + =1 EM (1) (20) where EM (1) s the expected unnshed work at the th queue at an (arbtrary) nstant of departure of the server from that queue. Result (20) holds for any servce regme, and EM (1) depends only on the servce dscplne n channel. For the Exhaustve servce regme EM (1) = 0 for every, so that 2? : (21) For the Gated regme, we use (7) and wrte Hence, for the Gated, =1 EM (1) = (f ()b ) b = 2 E(W ) = W + d(2) 2d + d 1? =1 : 2 d : (22) 2(1? ) =1 It follows that for the same set of parameters, whenever swtchover tmes are ncurred the mean amount of work n the system under the Exhaustve regme s smaller then that under the Gated dscplne. Furthermore, expressons (21) and (22) enabled Boxma, Levy & Weststrate (see, Boxma [1991]) to develop `good' vst frequences of the server to the varous channels so as to construct a pollng table that wll reduce the value of the expected amount of work n the system, 639

11 as expressed n (20). For the Exhaustve and for the Gated regmes the vst frequences v exh, and v gated are gven by p (1 v exh? )=d = p (1? )=d PN v gated = PN p (1 + )=d p (1 + )=d For example, n a 3-channel case for whch the calculated vst frequences are 0.52, 0.32 and 0.16, the approxmate vst frequences are 1/2, 1/3 and 1/6, respectvely, such that a (perodc) pollng table of sze 6 s constructed wth the order of vsts [1,2,1,3,1,2]. Another approach n the attempt to control and otmze the vst frequences of the server to the varous channels s the Cyclc Bernoull Pollng (CBP) ntroduced by Altman & Yechal [1993]. The server moves cyclcally among the N channels where change-over tmes between statons are composed of two parts: walkng tmes requred to `move' from one channel to another and swtch-n tmes that are ncurred only when the server actually enters a staton to render servce. Upon arrval to channel the server swtches n wth probablty p, or moves on to the next channel (wth probablty 1? p ) wthout servng any customer. Altman & Yechal analyzed the Gated and Exhaustve regmes and dened a mathematcal program to nd the optmal values of the swtch-n probabltes fp g N =1 so as to mnmze the expected amount of unnshed work n the system. Any CBP scheme for whch the optmal p 's are not equal to 1 yelds a smaller amount of expected unnshed work n the system than that n the standard cyclc procedure wth equvalent parameters. They showed that even n the case of a sngle queue, t s not always true that p 1 = 1 s the best strategy, and derved condtons under whch t s optmal to have p 1 < 1. 6 Dynamc Control of Server's Vsts: Hamltonan Tours A basc queston that arses when plannng ecent pollng systems concerns the order of vsts performed by the server. For statc order one can thnk of a `good' pollng table that optmzes some measure of eectveness. Steps n ths drecton were taken, as mentoned n secton 5, by varous authors. However, a more reachng goal s to control the system dynamcally, so that the server wll modfy ts order of vsts n response to the stochastc evoluton of the system. In other words, the general control problem facng the server when t exts a specc channel, s \whch of the channels to vst next?". In tryng to solve ths problem Browne & Yechal [1989a], [1989b] developed and formulated sem-markov Decson Processes (SMDP) for the Gated and for the Exhaustve regmes. They derved a set of optmalty equatons where the obectve s to mnmze mean weghted watng costs. However, these equatons are non-tractable, 640

12 so that one should look for alternatve methods. An appealng approach s to look for sem-dynamc control schemes. The dea s to dspatch the server to perform Hamltonan tours, each tour derent from ts prevous one, dependng on the state of the system at the begnnng of the tour, so as to optmze some measure of eectveness. Speccally, suppose that at the begnnng of a cycle the state of the system s (n 1 ; n 2 ; : : :; n N ), where n s the number of obs watng n channel (1 N). Assume for the moment that swtchng tmes between channels are neglgble. The obectve s to choose a path (Hamltonan tour) through the queues so as to mnmze the expected tme of traversng ths path. It was shown by Browne & Yechal [1989a], [1989b] that for both servce dscplnes { the fully Gated and the fully Exhaustve { ths measure of eectveness s mnmzed f the channels are ordered by ncreasng values of the ndex n =. Ths s a surprsng result, as the ndex n = does not nclude the servce tmes at the varous channels. It s surprsng as well that the same ndex-rule holds for both servce regmes (although, obvously, the duraton of a Gated-type cycle that starts wth (n 1 ; n 2 ; : : :; n N ) ders from ts Exhaustve counter-part startng wth the same system-state). The dynamcs of the control are such that at the end of each Hamltonan cycle a new system-state s observed, say (n 0 1; n 0 2; : : :; n 0 N ), and the server follows a new path governed by a new order: ncreasng values of n 0 =, etc. Ths s an extremely smple rule whch can be drectly mplemented. Moreover, suppose that, for one reason or another, there are systems where the obectve s to maxmze the duraton of each cycle. Then, the ndex-rule that determnes the order of vsts to the channels s smply reversed: the server completes a Hamltonan tour determned by a decreasng order of n =. To understand the above surprsng result Browne & Yechal [1990] studes a general schedulng problem wth a lnear growth of work, as follows. Consder a sngle-processor system wth N obs watng to be performed sequentally. Let a be the ntal (expected) processng tme requrement of ob ( = 1; 2; : : :; N), called the `core'. If ob s delayed and s started at tme t, then ts processng requrement grows lnearly wth the delay to Y (t) = a + t where s the growth rate of work requrement by ob. Consder the processng order 0 = (1; 2; : : :; P N), and let Y denote the actual processng length of ob k under 0. Let S k = Y =1 be the completon tme of ob k under 0 (S 0 = 0). Then Y = a + S?1. By addng S?1 to both sdes we obtan a set of derence equatons S? (1 + )S?1 = a ( = 1; 2; : : :; N) (23) The soluton of (23) s S = =1 a Y r=+1 (1 + r ) ( = 1; 2; : : :; N) (24) 641

13 so that the makespan s S N = S N ( 0 ) = P N =1 a Q N r=+1 (1 + r). The obectve s to nd a vst order that mnmzes the makespan S N () over all n! possble permutatons. Consder now the processng sequence 1 = (1; 2; : : :;? 1; + 1; ; + 2; : : :; N), where the order of obs and + 1 s nterchanged. The correspondng makespan s S N ( 1 ). Then, t s easy to show that S N ( 0 ) < S N ( 1 ) a = < a +1 = +1. That s, the makespan s mnmzed (maxmzed) f we process the obs n an ncreasng (decreasng) order of the rato ndex a =,.e., `core' dvded by `growth rate'. Consder agan the Gated regme. If (n 1 ; n 2 ; : : :; n N ) s the state of the system at the start of the Hamltonan tour, then a = n b. The growth rate (.e., the amount of work owng to channel per unt of tme) s. Hence, a = n b b = n :? For the Exhaustve regme, a = n E( ) = n b 1?, whereas = 1? (the duraton of tme that the server has to stay n channel grows lnearly at a rate b of 1? for each new arrval. As the rate of arrvals s, we have = 1? ). Thus, for the Exhaustve case a = n b 1? = n b 1? = n whch s the same ndex as for the Gated regme. We can now rentroduce the swtchover and swtch-n tmes. For llustraton, assume a star-conguraton of the system. Recall that D s the swtchover tme out of and R denotes the swtch-n duraton nto. Then, for the Gated regme, assumng gatng occurs after swtch-n s completed, a = n b + (1 + )r + d = ; so that a = = n b + (1 + )r + d. For the Exhaustve r a = + n b + d 1? 1? = =(1? ) ; so that a = = r + n b + d (1? ). It should be emphaszed that the schedulng prncple a = can be appled to any system wth a mxed set of servce regmes among the channels: Gated, Exhaustve, Bnomal or Bernoull Gated, Bnomal or Bernoull Exhaustve, etc. (see, Yechal [1991]). All that one has to do s to calculate (once) for every channel, and then, at the begnnng of each new Hamltonan tour, to calculate the current `core' a at each channel. Then, performng a vst tour that follows an ncreasng (decreasng) order of a = wll mnmze (maxmze) cycle duraton. 642

14 Browne & Yechal [1991] further employed the above deas to acheve dynamc schedulng n systems wth only a unt buer at each channel. 7 The Globally-Gated Regme A drawback both of the Gated and the Exhaustve regmes s that they are not `far' wth regard to the FCFS prncple. To help resolve ths dchotomy, Boxma, Levy & Yechal [1992] ntroduced a (cyclc) Globally Gated (GG) servce scheme whch uses a tme-stamp mechansm for ts operaton: the server moves cyclcally among the queues, and uses the nstant of cycle-begnnng as a reference pont of tme; when t reaches a queue t serves there all (and only) customers who were present at that queue at the cycle-begnnng. Ths strategy can be mplemented by markng all customers wth a tme-stamp denotng ther arrval tme. In ts nature the GG polcy resembles the regular Gated polcy. However, the GG polcy leads to a mathematcal model whch allows for dervaton of closed-form expressons for the mean delay n the varous queues. As a result, the operaton of the pollng system by the GG polcy s easy to control and optmze. As n earler sectons, the system conssts of N nnte-buer channels, the P rate of N oered load to queue s = b and the total system load-rate s =1. When leavng queue and before startng servce at the next queue, the server ncurs P a random swtchover perod D. The total `walkng' tme n a cycle s N D D =1. (Clearly, other `Global' versons, such as Globally Exhaustve, can be easly magned and analyzed.) Cycle Tme Assume, wthout loss of generalty, that a cycle starts from channel 1. Let ( 1 1 ; 2 1 ; : : :; 1 ; : : :; N 1 ) = ( 1; 2 ; : : :; ; : : :; N ) be the state of the system at the begnnng of the cycle. Then, the cycle duraton s C = D + The LST of C s derved as follows k=1 B k : E? e?wc ( 1 ; 2 ; : : :; N ) = e D(w) N Y? eb (w) : (25) On the other hand, the length of a cycle determnes the ont queue-length dstrbuton at the begnnng of the next cycle. Hence E NY z 2 NY = E C 4 E z C = e C N 3 5 = EC 2 4 exp? (1? z )C 3 5 (1? z ) : (26) 643

15 Combnng (25) and (26) ec(w) = e D(w) e C? 1? e B (w) : (27) The mean cycle tme s derved from (27) E[C] = d + b E[C] : That s, E[C] = d=(1? ), as for the Gated and the Exhaustve regmes. The second moment of C s derved from (27) 2 3 E[C 2 ] = 4 d (2) + 2d + b (2) E[C] 5 (1? 2 ) : (28) Let C P and C R denote, respectvely, the past and resdual duraton of a cycle. It s well known that and E[C P ] = E[C R ] = E[C2 ] 2E[C]. ec P (w) = e CR (w) = 1? e C(w) we[c] Pseudo-Conservaton law To derve a pseudo-conservaton law we use (20) and the observaton that for the cyclc GG regme, E( ) = E[C] and EM (1) " = =1 E( )b + Substtutng (29) n (20) yelds?1 # d =1 =?1 =1 E(W ) = W + d(2) 2d + d 1? 2 + d 1? + d =2?1 =1 d + 2 1? : (29) d : (30) Watng Tmes Consder an arbtrary ob K at channel k. The cycle age at the ob's arrval nstant s C P. The ob's watng tme s composed of () the resdual cycle tme C R, () the servce tmes of all customers who arrve at channels 1 to k? 1 durng the cycle n whch K arrves, () the swtchover tmes of the server through channels 1 to k, and (v) the servce tmes of all customers that arrve at channel k durng the past part of the cycle, C P. Then 644

16 k?1 E(W k ) = E[C R ] + = It readly follows that k?1? k?1 E[CP ] + E[C R ] + k?1 + k E[C R ] + E(W k+1 )? E(W k ) = ( k+1 + k )E[C R ] + d k so that, for the cyclc GG regme, we always have d + k E[C P ] d : (31) E(W 1 ) < E(W 2 ) < : : : < E(W N ) : (32) Boxma, Weststrate & Yechal [1993] extended the cyclc GG model to the case where the server suers perods of breakdown, and appled the results to realworld reparman problems where both preventve and correctve mantenance actons are consdered. Statc Optmzaton Let c k be the cost rate of a watng ob at queue k. Then, the mean weghted watng cost of an arbtrary ob n the system s N k k=1 c k E(W k ) : (33) By substtutng (31) nto (33) and usng an nterchange argument t follows that the cycle whch mnmzed (33) s determned by an ncreasng order of the ndex u = 2E[C R] + d c If d s neglgble, the above ndex reduces to the ndex b =c, whch s the well known \c" rule. Dynamc Control An mportant characterstc of the GG regme s that the order of vsts selected for one cycle does not aect the future stochastc behavour of the system. Moreover, any Hamltonan tour that starts from state (n 1 ; n 2 ; : : :; n M ) yelds the same cycle duraton C(n 1 ; n 2 ; : : :; n N ). Thus, f we consder the costs ncurred durng the cycle by the customers present at ts ntaton and add to t the costs ncurred along that cycle by the new arrvals, then the long-run mmnal cost can be acheved by determnng a new optmal Hamltonan tour for each cycle ndependently. The mean total weghted cost ncurred durng a cycle startng wth (n 1 ; n 2 ; : : :; n N ) s 645

17 k=1 c k n k k?1 (n b + d ) + b k n k?1 =1 (34) + c k k E C(n 1 ; n 2 ; : : :; n N ) 2 2 k=1 where the rst term s the contrbuton to total cost ncurred by the customers present at the cycle begnnng, and the second term s due to the customers arrvng durng that cycle (see, Yechal [1976]). The only term n (34) that depends on the order of vsts s P N k= c kn k P k?1 (n b + d ). It follows (by an nterchange argument) that the optmal order of vsts that mnmzes expected total costs of the comng cycle s determned by an ncreasng order of the (Gttns) ndex n b + d n c : Agan, for neglgble d ths ndex reduces to the \c" rule (.e., b =c ). 8 Elevator-Type Pollng ; 1 nup 2 N In an Elevator-type (scan) pollng mechansm the server alternates between `up' and `down' cycles. In an `up' cycle t vsts the channels n the order 1; 2; : : :; N? 1; N, and n a `down' cycle the order of vsts s reversed to N; N?1; : : :; 2; 1. Ths type of pollng procedure s encountered n many applcatons, e.g., t models a common scheme of addressng a hard dsk for wrtng (readng) nformaton on (from) derent tracks. It s mportant to note that the Elevator-type pollng saves the return walkng tme from channel N to channel 1. A comprehensve analyss of Elevator-type pollng wth four derent servce regmes can be found n Shoham & Yechal [1992]. Here we present the Globally-Gated (GG) regme as dscussed n Altman, Khamsy & Yechal [1992]. Accordng to the Elevator-type pollng wth GG servce regme all channels are gated o at the begnnng of the `up' cycle, where the system-state s (n up ; : : :; nup), and the server resdes n channel for nup regular servce duratons. At the end of the up cycle all channels are gated agan, the system-state s (n down 1 ; n down 2 ; : : :; n down N ), and the server starts ts down cycle, servng ndown customers n channel. We assume that the down walkng tme from channel + 1 to channel has the same dstrbuton as the up walkng tme D from channel to channel + 1. A key observaton s that arbtrary up and down cycles have the same dstrbuton, whch ders from ts cyclc GG counter-part only n that t s smaller by the `saved' walkng tme D N. Hence, the results derved for the cycle tme dstrbuton (27) and for mean watng tmes (31) n a cyclc GG regme are drectly applcable to the Elevator case, wth D N =

18 Watng Tmes Consder an arbtrary ob at channel k. Snce all cycles are dstrbuted alke, the ob arrves durng an up or a down cycle wth equal probabltes, 0.5. Hence, ts mean watng tme s gven by E(W k ) = 0:5E The expresson for E W k W k W k server moves up server moves down server moves down by reversng the order of vsts, we have E = server + 0:5E W k moves up : (35) s gven by (31), wth d N = 0, whereas, =k+1 N?1 + k E[C R ] + d : (36) Combnng (35) wth (31) and (36) yelds the surprsng result =k E(W k ) = (1 + )E[C R ] + 0:5d : (37) That s, expected watng tmes are equal n all channels. Ths s the only-known non-symmetrc pollng system that exhbts such a \farness" phenomenon. An explanaton of result (37) s the followng. An arbtrary arrval has to wat, on the average, E[C R ] unts of tme untl the cycle (up or down) n whch t arrves termnates. Then, t wats untl the server moves back to channel k, whch requres, on the average (takng nto account both drectons), 1 2? E[CR ]+ E[C p ] + d unts of tme. Optmal Arrangement of Channels The nterestng result that E(W k ) s the same for all channels, ndependent of ther locaton, leads to consderng channels' arrangement such that the varaton n watng tmes wll be small. Let a = 2E[C R ] + d ( = 1; 2; : : :; N). Then E W k server moves down server E W k moves up k?1 = E[C R ](1 + k ) + = E[C R ](1 + k ) + =1 =k+1 a a + d k P P Let k P = E(W k down)? E(W P k?1 k up) = a N =1? a =k+1? d k. Now, N 1 =? a N?1 =2? d 1 < 0, N = a =1 > 0 (recall that d N = 0), and k s a monotone ncreasng functon of k. One goal s to arrange the channels such that max 1kN k s as small as possble. Clearly 647

19 max k 1kN = max 1 ; N = max a? 2E[C R ] 1 ; =1 =1 a? 2E[C R ] N (38) It follows from (38) that max 1kN k s mnmzed f channel 1 s the one wth the hghest value of and channel N s the one wth the second hghest value of (or vce versa). 9 Future Drectons of Research We have presented methods of analyss for sngle-server, contnuous-tme, nnte buers pollng systems, and studed several control and optmzaton problems. Dcult problems are nte-capacty models and lmted servce regmes, for whch only partal solutons are gven n the lterature (see, bblography n Takag [1990]). A few authors have studed pollng systems wth multple servers, and recently Browns & Wess [1992] nvestgated dynamc prorty rules for a system wth parallel servers. All the systems mentoned above are open, wth external arrvals, where obs ext the system after servce completon. Closed systems should also be nvestgated, and only recently Altman & Yechal [1992] analyzed such systems wth Gated, Exhaustve or Globally-Gated servce regmes. For other future drectons of research we state a recent `call for papers' on \Dscrete-Tme Models and Analyss Methods": \The past few years have seen an ncreasng nterest n dscrete-tme models and ther soluton technques. One of the drvng forces behnd ths area has been new developments n telecommuncatons, espacally n hgh-speed metropoltan area and wde area networks. Tehcnologcal advances and user demands have shfted the evoluton of telecommuncaton systems towards ntegrated networks where nformaton s transferred n small, ofted xed-sze, packets, slots or cells (e.g., ATM networks, hghspeed LANs and MANs such as DQDB, etc...), opertng n a dscrete-tme envronment. The resultng mathematcal models of such slotted systems, crucal for the evaluaton of desgn alternatves and ther dmensonng, are dscrete-tme models. The complexty of the stochastc processes nvolved (e.g., arrval and departure processes) and of the system operaton mechansms (e.g., servce mechansm, access protocol, etc...) pose an exctng challenge for the development of ecent and tractable methods for dervng the man performance measures of these systems. Papers are solcted on dscrete-tme models and ther analyss methods, n partcular on, but not restrcted to, the followng topcs: Dscrete-tme queueng models (pollng systems, prorty systems, multserver systems, vacaton models, etc...). 648

20 Exact and approxmate soluton methods for dscrete-tme queueng models, wth emphass on the ecency and the numercal tractablty of these methods. Stochastc processes as trac models for performance studes (takng nto account the dversty of tme scales, correlatons between arrvals, etc...) Dscrete-tme markov chans and ther analyss methods". Naturally, we add to the above topcs the nterestng and challengng problems of control and optmzaton of such systems. Bblography 1. Altman, E., Blanc, H., Khamsy, A., Yechal, U.: Gated-type pollng systems wth walkng and swtch-n tmes. Techncal Report, Dept. of Statstcs & OR, Tel Avv Unversty Altman, E., Khamsy, A., Yechal, U.: On elevator pollng wth globally-gated regme. Queueng Systems 11 (1992) Altman, E., Yechal, U.: Pollng n a closed network. Techncal Report SOR-92-14, Dept. of Statstcs & OR, New York Unversty Altman, E., Yechal, U.: Cyclc Bernoull pollng. ZOR-Methods and Models of Operatons Research 38 (1993). 5. Boxma, O.J.: Workloads and watng tmes n sngle-server systems wth multple customer classes. Queueng Systems 5 (1989) Boxma, O.J.: Analyss and optmzaton of pollng systems. In: Cohen, J.W., Pack, C.D. (Eds.) Queueng, Performance and Control n ATM. North-Holland, 1991, pp Boxma, O.J., Groenendk, W.P.: Pseudo conservaton laws n cyclc servce systems. Journal of Appled Probablty 24 (1987) Boxma, O.J., Levy, H., Yechal, U.: Cyclc reservaton schemes for ecent operaton of multple-queue sngle-server Systems. Annals of Operatons Research 35 (1992) Boxma, O.J., Weststrate, J.A., Yechal, U.: A globally gated pollng system wth server nterruptons, and applcatons to the reparman problem. Probablty n the Engneerng and Informatonal Scences 7 (1993). 10. Browne, S., Yechal, U.: Dynamc prorty rules for cyclc-type queues, Advances n Appled Probablty 21 (1989a) Browne, S., Yechal, U.: Dynamc routng n pollng systems. In: M. Bonatt (Ed.) Teletrac Scence for New Cost-Eectve Systems, Networks and Servces. North- Holland, 1989b, pp Browne, S., Yechal, U.: Schedulng deteroratng obs on a sngle processor. Operatons Research 38 (1990) Browne, S., Yechal, U.: Dynamc schedulng n sngle-server multclass servce systems wth unt buers. Naval Research Logstcs 38 (1991) Browne, S., Wess, G.: Dynamc prorty rules when pollng wth multple parallel servers. Operatons Research Letters 12 (1992) Cooper, R.B. Murray, G.: Queues served n cyclc order. Bell System Techncal Journal 48 (1969)

21 16. Cooper, R.B.: Queues served n cyclc order: watng tmes. Bell System Techncal Journal 49 (1970) Esenberg, M.: Queues wth perodc servce and changeover tme. Operatons Research 20 (1972) Ferguson, M.J., Amnetzah, Y.J.: Exact results for nonsymmetrc token rng systems. IEEE Transactons on Communcatons 33 (1985) Klenrock, L.: Queueng Systems, Vol. 1: Theory. John Wley, Konhem, A.G., Levy, H., Srnvasan: Descendant set: an ecent approach for the analyss of pollng systems. IEEE Transactons on Communcatons (to appear 1993a). 21. Konhem, A.G., Levy, H., Srnvasan: The ndvdual staton technque for the analyss of pollng systems. Techncal Report, 1993b. 22. Levy, H., Sd, M.: Pollng systems: applcatons, modelng and optmzaton. IEEE Transactons on Communcatons 8 (1990) Sarkar, D., Zangwll, W.I.: Expected watng tme for nonsymmetrc cyclc queueng systems { exact results and applcatons. Management Scence 35 (1989) Shoham, R., Yechal, U.: Elevator-type pollng systems. Techncal Report, Dept. of Statstcs & OR, Tel Avv Unversty, Takag, H.: Analyss of Pollng Systems. MIT Press, Takag, H.: Queueng analyss of pollng models: an update. In: Takag, H. (ed.) Stochastc Analyss of Computer and Communcatons Systems. North Holland, 1990, pp Yechal, U.: A new dervaton of the Khntchne-Pollaczek formula. In: Haley, K.B. (Ed.) Operatonal Research '75. North Holland, 1976, pp Yechal, U.: Optmal dynamc control of pollng systems. In: Cohen, J.W., Pack, C.D. (Eds.) Queueng, Performance and Control n ATM. North Holland, 1991, pp

Elevator-Type Polling Systems

Elevator-Type Polling Systems Elevator-Type Pollng Systems Ruth Shoham Ur Yechal Department of Statstcs, School of Mathematcal Scence Raymond and Beverly Sackler Faculty of Exact Scences Tel Avv Unversty, Tel Avv 69978, Israel November

More information

Analysis of Discrete Time Queues (Section 4.6)

Analysis of Discrete Time Queues (Section 4.6) Analyss of Dscrete Tme Queues (Secton 4.6) Copyrght 2002, Sanjay K. Bose Tme axs dvded nto slots slot slot boundares Arrvals can only occur at slot boundares Servce to a job can only start at a slot boundary

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

Adaptive Dynamical Polling in Wireless Networks

Adaptive Dynamical Polling in Wireless Networks BULGARIA ACADEMY OF SCIECES CYBERETICS AD IFORMATIO TECHOLOGIES Volume 8, o Sofa 28 Adaptve Dynamcal Pollng n Wreless etworks Vladmr Vshnevsky, Olga Semenova Insttute for Informaton Transmsson Problems

More information

Equilibrium Analysis of the M/G/1 Queue

Equilibrium Analysis of the M/G/1 Queue Eulbrum nalyss of the M/G/ Queue Copyrght, Sanay K. ose. Mean nalyss usng Resdual Lfe rguments Secton 3.. nalyss usng an Imbedded Marov Chan pproach Secton 3. 3. Method of Supplementary Varables done later!

More information

COMPLETE BUFFER SHARING IN ATM NETWORKS UNDER BURSTY ARRIVALS

COMPLETE BUFFER SHARING IN ATM NETWORKS UNDER BURSTY ARRIVALS COMPLETE BUFFER SHARING WITH PUSHOUT THRESHOLDS IN ATM NETWORKS UNDER BURSTY ARRIVALS Ozgur Aras and Tugrul Dayar Abstract. Broadband Integrated Servces Dgtal Networks (B{ISDNs) are to support multple

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

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

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

arxiv: v1 [math.pr] 1 Aug 2014

arxiv: v1 [math.pr] 1 Aug 2014 A Pollng Model wth Smart Customers M.A.A. Boon marko@wn.tue.nl A.C.C. van Wk a.c.c.v.wk@tue.nl O.J. Boxma boxma@wn.tue.nl I.J.B.F. Adan adan@wn.tue.nl arxv:408.029v [math.pr] Aug 204 September, 200 Abstract

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

M/G/1 POLLING SYSTEMS WITH RANDOM VISIT TIMES

M/G/1 POLLING SYSTEMS WITH RANDOM VISIT TIMES Probablty n the Engneerng and Informatonal Scences, 22, 2008, 81 105. Prnted n the U.S.A. DOI: 10.1017/S0269964808000065 M/G/1 POLLING SYSTEMS WITH RANDOM VISIT TIMES M. VLASIOU Georga Insttute of Technology

More information

Minimisation of the Average Response Time in a Cluster of Servers

Minimisation of the Average Response Time in a Cluster of Servers Mnmsaton of the Average Response Tme n a Cluster of Servers Valery Naumov Abstract: In ths paper, we consder task assgnment problem n a cluster of servers. We show that optmal statc task assgnment s tantamount

More information

TCOM 501: Networking Theory & Fundamentals. Lecture 7 February 25, 2003 Prof. Yannis A. Korilis

TCOM 501: Networking Theory & Fundamentals. Lecture 7 February 25, 2003 Prof. Yannis A. Korilis TCOM 501: Networkng Theory & Fundamentals Lecture 7 February 25, 2003 Prof. Yanns A. Korls 1 7-2 Topcs Open Jackson Networks Network Flows State-Dependent Servce Rates Networks of Transmsson Lnes Klenrock

More information

EURANDOM PREPRINT SERIES October, Performance analysis of polling systems with retrials and glue periods

EURANDOM PREPRINT SERIES October, Performance analysis of polling systems with retrials and glue periods EURADOM PREPRIT SERIES 2016-012 October, 2016 Performance analyss of pollng systems wth retrals and glue perods M. Abdn, O. Boxma, B. Km, J. Km, J. Resng ISS 1389-2355 Performance analyss of pollng systems

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

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

ECE559VV Project Report

ECE559VV Project Report ECE559VV Project Report (Supplementary Notes Loc Xuan Bu I. MAX SUM-RATE SCHEDULING: THE UPLINK CASE We have seen (n the presentaton that, for downlnk (broadcast channels, the strategy maxmzng the sum-rate

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

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

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

Introduction to Continuous-Time Markov Chains and Queueing Theory

Introduction to Continuous-Time Markov Chains and Queueing Theory Introducton to Contnuous-Tme Markov Chans and Queueng Theory From DTMC to CTMC p p 1 p 12 1 2 k-1 k p k-1,k p k-1,k k+1 p 1 p 21 p k,k-1 p k,k-1 DTMC 1. Transtons at dscrete tme steps n=,1,2, 2. Past doesn

More information

Global Sensitivity. Tuesday 20 th February, 2018

Global Sensitivity. Tuesday 20 th February, 2018 Global Senstvty Tuesday 2 th February, 28 ) Local Senstvty Most senstvty analyses [] are based on local estmates of senstvty, typcally by expandng the response n a Taylor seres about some specfc values

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

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

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

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

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

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

Note 10. Modeling and Simulation of Dynamic Systems

Note 10. Modeling and Simulation of Dynamic Systems Lecture Notes of ME 475: Introducton to Mechatroncs Note 0 Modelng and Smulaton of Dynamc Systems Department of Mechancal Engneerng, Unversty Of Saskatchewan, 57 Campus Drve, Saskatoon, SK S7N 5A9, Canada

More information

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U) Econ 413 Exam 13 H ANSWERS Settet er nndelt 9 deloppgaver, A,B,C, som alle anbefales å telle lkt for å gøre det ltt lettere å stå. Svar er gtt . Unfortunately, there s a prntng error n the hnt of

More information

arxiv: v1 [math.pr] 19 Oct 2016

arxiv: v1 [math.pr] 19 Oct 2016 Performance analyss of pollng systems wth retrals and glue perods Murtuza Al Abdn, Onno Boxma, Bara Km, Jeongsm Km, Jacques Resng EURADOM and Department of Mathematcs and Computer Scence arxv:1610.05960v1

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

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

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

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

EEE 241: Linear Systems

EEE 241: Linear Systems EEE : Lnear Systems Summary #: Backpropagaton BACKPROPAGATION The perceptron rule as well as the Wdrow Hoff learnng were desgned to tran sngle layer networks. They suffer from the same dsadvantage: they

More information

Polling Models with Two-Stage Gated Service: Fairness versus Efficiency

Polling Models with Two-Stage Gated Service: Fairness versus Efficiency Pollng Models wth Two-Stage Gated Servce: Farness versus Effcency R.D. van der Me a,b and J.A.C. Resng c a CWI, Advanced Communcaton Networks P.O. Box 94079, 1090 GB Amsterdam, The Netherlands b Vrje Unverstet,

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

Hidden Markov Models & The Multivariate Gaussian (10/26/04)

Hidden Markov Models & The Multivariate Gaussian (10/26/04) CS281A/Stat241A: Statstcal Learnng Theory Hdden Markov Models & The Multvarate Gaussan (10/26/04) Lecturer: Mchael I. Jordan Scrbes: Jonathan W. Hu 1 Hdden Markov Models As a bref revew, hdden Markov models

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

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

36.1 Why is it important to be able to find roots to systems of equations? Up to this point, we have discussed how to find the solution to

36.1 Why is it important to be able to find roots to systems of equations? Up to this point, we have discussed how to find the solution to ChE Lecture Notes - D. Keer, 5/9/98 Lecture 6,7,8 - Rootndng n systems o equatons (A) Theory (B) Problems (C) MATLAB Applcatons Tet: Supplementary notes rom Instructor 6. Why s t mportant to be able to

More information

Dynamic Programming. Lecture 13 (5/31/2017)

Dynamic Programming. Lecture 13 (5/31/2017) Dynamc Programmng Lecture 13 (5/31/2017) - A Forest Thnnng Example - Projected yeld (m3/ha) at age 20 as functon of acton taken at age 10 Age 10 Begnnng Volume Resdual Ten-year Volume volume thnned volume

More information

MATH 829: Introduction to Data Mining and Analysis The EM algorithm (part 2)

MATH 829: Introduction to Data Mining and Analysis The EM algorithm (part 2) 1/16 MATH 829: Introducton to Data Mnng and Analyss The EM algorthm (part 2) Domnque Gullot Departments of Mathematcal Scences Unversty of Delaware Aprl 20, 2016 Recall 2/16 We are gven ndependent observatons

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

Research Article Green s Theorem for Sign Data

Research Article Green s Theorem for Sign Data Internatonal Scholarly Research Network ISRN Appled Mathematcs Volume 2012, Artcle ID 539359, 10 pages do:10.5402/2012/539359 Research Artcle Green s Theorem for Sgn Data Lous M. Houston The Unversty of

More information

12. The Hamilton-Jacobi Equation Michael Fowler

12. The Hamilton-Jacobi Equation Michael Fowler 1. The Hamlton-Jacob Equaton Mchael Fowler Back to Confguraton Space We ve establshed that the acton, regarded as a functon of ts coordnate endponts and tme, satsfes ( ) ( ) S q, t / t+ H qpt,, = 0, and

More information

Markov chains. Definition of a CTMC: [2, page 381] is a continuous time, discrete value random process such that for an infinitesimal

Markov chains. Definition of a CTMC: [2, page 381] is a continuous time, discrete value random process such that for an infinitesimal Markov chans M. Veeraraghavan; March 17, 2004 [Tp: Study the MC, QT, and Lttle s law lectures together: CTMC (MC lecture), M/M/1 queue (QT lecture), Lttle s law lecture (when dervng the mean response tme

More information

RELIABILITY ASSESSMENT

RELIABILITY ASSESSMENT CHAPTER Rsk Analyss n Engneerng and Economcs RELIABILITY ASSESSMENT A. J. Clark School of Engneerng Department of Cvl and Envronmental Engneerng 4a CHAPMAN HALL/CRC Rsk Analyss for Engneerng Department

More information

Scheduling in polling systems

Scheduling in polling systems Schedulng n pollng systems Adam Werman, Erk M.M. Wnands, and Onno J. Boxma Abstract: We present a smple mean value analyss MVA framework for analyzng the effect of schedulng wthn queues n classcal asymmetrc

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

Distributions /06. G.Serazzi 05/06 Dimensionamento degli Impianti Informatici distrib - 1

Distributions /06. G.Serazzi 05/06 Dimensionamento degli Impianti Informatici distrib - 1 Dstrbutons 8/03/06 /06 G.Serazz 05/06 Dmensonamento degl Impant Informatc dstrb - outlne densty, dstrbuton, moments unform dstrbuton Posson process, eponental dstrbuton Pareto functon densty and dstrbuton

More information

Canonical transformations

Canonical transformations Canoncal transformatons November 23, 2014 Recall that we have defned a symplectc transformaton to be any lnear transformaton M A B leavng the symplectc form nvarant, Ω AB M A CM B DΩ CD Coordnate transformatons,

More information

O-line Temporary Tasks Assignment. Abstract. In this paper we consider the temporary tasks assignment

O-line Temporary Tasks Assignment. Abstract. In this paper we consider the temporary tasks assignment O-lne Temporary Tasks Assgnment Yoss Azar and Oded Regev Dept. of Computer Scence, Tel-Avv Unversty, Tel-Avv, 69978, Israel. azar@math.tau.ac.l??? Dept. of Computer Scence, Tel-Avv Unversty, Tel-Avv, 69978,

More information

Linear Approximation with Regularization and Moving Least Squares

Linear Approximation with Regularization and Moving Least Squares Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...

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

Temperature. Chapter Heat Engine

Temperature. Chapter Heat Engine Chapter 3 Temperature In prevous chapters of these notes we ntroduced the Prncple of Maxmum ntropy as a technque for estmatng probablty dstrbutons consstent wth constrants. In Chapter 9 we dscussed the

More information

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method Appled Mathematcal Scences, Vol. 7, 0, no. 47, 07-0 HIARI Ltd, www.m-hkar.com Comparson of the Populaton Varance Estmators of -Parameter Exponental Dstrbuton Based on Multple Crtera Decson Makng Method

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

A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS

A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS HCMC Unversty of Pedagogy Thong Nguyen Huu et al. A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS Thong Nguyen Huu and Hao Tran Van Department of mathematcs-nformaton,

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

FUZZY GOAL PROGRAMMING VS ORDINARY FUZZY PROGRAMMING APPROACH FOR MULTI OBJECTIVE PROGRAMMING PROBLEM

FUZZY GOAL PROGRAMMING VS ORDINARY FUZZY PROGRAMMING APPROACH FOR MULTI OBJECTIVE PROGRAMMING PROBLEM Internatonal Conference on Ceramcs, Bkaner, Inda Internatonal Journal of Modern Physcs: Conference Seres Vol. 22 (2013) 757 761 World Scentfc Publshng Company DOI: 10.1142/S2010194513010982 FUZZY GOAL

More information

= z 20 z n. (k 20) + 4 z k = 4

= z 20 z n. (k 20) + 4 z k = 4 Problem Set #7 solutons 7.2.. (a Fnd the coeffcent of z k n (z + z 5 + z 6 + z 7 + 5, k 20. We use the known seres expanson ( n+l ( z l l z n below: (z + z 5 + z 6 + z 7 + 5 (z 5 ( + z + z 2 + z + 5 5

More information

Lecture Randomized Load Balancing strategies and their analysis. Probability concepts include, counting, the union bound, and Chernoff bounds.

Lecture Randomized Load Balancing strategies and their analysis. Probability concepts include, counting, the union bound, and Chernoff bounds. U.C. Berkeley CS273: Parallel and Dstrbuted Theory Lecture 1 Professor Satsh Rao August 26, 2010 Lecturer: Satsh Rao Last revsed September 2, 2010 Lecture 1 1 Course Outlne We wll cover a samplng of the

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

CS : Algorithms and Uncertainty Lecture 17 Date: October 26, 2016

CS : Algorithms and Uncertainty Lecture 17 Date: October 26, 2016 CS 29-128: Algorthms and Uncertanty Lecture 17 Date: October 26, 2016 Instructor: Nkhl Bansal Scrbe: Mchael Denns 1 Introducton In ths lecture we wll be lookng nto the secretary problem, and an nterestng

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

CHAPTER 14 GENERAL PERTURBATION THEORY

CHAPTER 14 GENERAL PERTURBATION THEORY CHAPTER 4 GENERAL PERTURBATION THEORY 4 Introducton A partcle n orbt around a pont mass or a sphercally symmetrc mass dstrbuton s movng n a gravtatonal potental of the form GM / r In ths potental t moves

More information

COS 521: Advanced Algorithms Game Theory and Linear Programming

COS 521: Advanced Algorithms Game Theory and Linear Programming COS 521: Advanced Algorthms Game Theory and Lnear Programmng Moses Charkar February 27, 2013 In these notes, we ntroduce some basc concepts n game theory and lnear programmng (LP). We show a connecton

More information

Absorbing Markov Chain Models to Determine Optimum Process Target Levels in Production Systems with Rework and Scrapping

Absorbing Markov Chain Models to Determine Optimum Process Target Levels in Production Systems with Rework and Scrapping Archve o SID Journal o Industral Engneerng 6(00) -6 Absorbng Markov Chan Models to Determne Optmum Process Target evels n Producton Systems wth Rework and Scrappng Mohammad Saber Fallah Nezhad a, Seyed

More information

PHYS 705: Classical Mechanics. Canonical Transformation II

PHYS 705: Classical Mechanics. Canonical Transformation II 1 PHYS 705: Classcal Mechancs Canoncal Transformaton II Example: Harmonc Oscllator f ( x) x m 0 x U( x) x mx x LT U m Defne or L p p mx x x m mx x H px L px p m p x m m H p 1 x m p m 1 m H x p m x m m

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

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

Application of Queuing Theory to Waiting Time of Out-Patients in Hospitals.

Application of Queuing Theory to Waiting Time of Out-Patients in Hospitals. Applcaton of Queung Theory to Watng Tme of Out-Patents n Hosptals. R.A. Adeleke *, O.D. Ogunwale, and O.Y. Hald. Department of Mathematcal Scences, Unversty of Ado-Ekt, Ado-Ekt, Ekt State, Ngera. E-mal:

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

Computation of Higher Order Moments from Two Multinomial Overdispersion Likelihood Models

Computation of Higher Order Moments from Two Multinomial Overdispersion Likelihood Models Computaton of Hgher Order Moments from Two Multnomal Overdsperson Lkelhood Models BY J. T. NEWCOMER, N. K. NEERCHAL Department of Mathematcs and Statstcs, Unversty of Maryland, Baltmore County, Baltmore,

More information

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur Analyss of Varance and Desgn of Exerments-I MODULE III LECTURE - 2 EXPERIMENTAL DESIGN MODELS Dr. Shalabh Deartment of Mathematcs and Statstcs Indan Insttute of Technology Kanur 2 We consder the models

More information

10. Canonical Transformations Michael Fowler

10. Canonical Transformations Michael Fowler 10. Canoncal Transformatons Mchael Fowler Pont Transformatons It s clear that Lagrange s equatons are correct for any reasonable choce of parameters labelng the system confguraton. Let s call our frst

More information

Suggested solutions for the exam in SF2863 Systems Engineering. June 12,

Suggested solutions for the exam in SF2863 Systems Engineering. June 12, Suggested solutons for the exam n SF2863 Systems Engneerng. June 12, 2012 14.00 19.00 Examner: Per Enqvst, phone: 790 62 98 1. We can thnk of the farm as a Jackson network. The strawberry feld s modelled

More information

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4) I. Classcal Assumptons Econ7 Appled Econometrcs Topc 3: Classcal Model (Studenmund, Chapter 4) We have defned OLS and studed some algebrac propertes of OLS. In ths topc we wll study statstcal propertes

More information

Transient results for M/M/1/c queues via path counting. M. Hlynka*, L.M. Hurajt and M. Cylwa

Transient results for M/M/1/c queues via path counting. M. Hlynka*, L.M. Hurajt and M. Cylwa Int. J. Mathematcs n Operatonal Research, Vol. X, No. X, xxxx 1 Transent results for M/M/1/c queues va path countng M. Hlynka*, L.M. Hurajt and M. Cylwa Department of Mathematcs and Statstcs, Unversty

More information

Buffer Dumping Management for High Speed Routers

Buffer Dumping Management for High Speed Routers Buffer Dumpng Management for Hgh Speed Routers Carolne Fayet, André-Luc Beylot IT, 9 rue Charles Fourer, 90 Evry Cedex, France carolne.fayet@nt-evry.fr, Groupe des Ecoles des Télécommuncatons ESEEIHT,,

More information

Lecture 3 Stat102, Spring 2007

Lecture 3 Stat102, Spring 2007 Lecture 3 Stat0, Sprng 007 Chapter 3. 3.: Introducton to regresson analyss Lnear regresson as a descrptve technque The least-squares equatons Chapter 3.3 Samplng dstrbuton of b 0, b. Contnued n net lecture

More information

Chapter 3 Differentiation and Integration

Chapter 3 Differentiation and Integration MEE07 Computer Modelng Technques n Engneerng Chapter Derentaton and Integraton Reerence: An Introducton to Numercal Computatons, nd edton, S. yakowtz and F. zdarovsky, Mawell/Macmllan, 990. Derentaton

More information

A Hybrid Variational Iteration Method for Blasius Equation

A Hybrid Variational Iteration Method for Blasius Equation Avalable at http://pvamu.edu/aam Appl. Appl. Math. ISSN: 1932-9466 Vol. 10, Issue 1 (June 2015), pp. 223-229 Applcatons and Appled Mathematcs: An Internatonal Journal (AAM) A Hybrd Varatonal Iteraton Method

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

Meenu Gupta, Man Singh & Deepak Gupta

Meenu Gupta, Man Singh & Deepak Gupta IJS, Vol., o. 3-4, (July-December 0, pp. 489-497 Serals Publcatons ISS: 097-754X THE STEADY-STATE SOLUTIOS OF ULTIPLE PARALLEL CHAELS I SERIES AD O-SERIAL ULTIPLE PARALLEL CHAELS BOTH WITH BALKIG & REEGIG

More information

On the Throughput of Clustered Photolithography Tools:

On the Throughput of Clustered Photolithography Tools: On the hroughput of lustered Photolthography ools: Wafer Advancement and Intrnsc Equpment Loss Maruth Kumar Mutnur James R. Morrson, Ph.D. September 23, 2007 Presentaton Outlne Motvaton Model : Synchronous

More information

Iterative approximation of k-limited polling systems

Iterative approximation of k-limited polling systems Queueng Syst (2007) 55: 6 78 DOI 0.007/s34-007-900-4 Iteratve approxmaton of k-lmted pollng systems M. van Vuuren E.M.M. Wnands Receved: May 2006 / Revsed: 5 December 2006 / Publshed onlne: 2 March 2007

More information

Indeterminate pin-jointed frames (trusses)

Indeterminate pin-jointed frames (trusses) Indetermnate pn-jonted frames (trusses) Calculaton of member forces usng force method I. Statcal determnacy. The degree of freedom of any truss can be derved as: w= k d a =, where k s the number of all

More information

Formulas for the Determinant

Formulas for the Determinant page 224 224 CHAPTER 3 Determnants e t te t e 2t 38 A = e t 2te t e 2t e t te t 2e 2t 39 If 123 A = 345, 456 compute the matrx product A adj(a) What can you conclude about det(a)? For Problems 40 43, use

More information

J. Parallel Distrib. Comput.

J. Parallel Distrib. Comput. J. Parallel Dstrb. Comput. 7 (20) 537 555 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. ournal homepage: www.elsever.com/locate/pdc Game-theoretc statc load balancng for dstrbuted systems

More information

Case A. P k = Ni ( 2L i k 1 ) + (# big cells) 10d 2 P k.

Case A. P k = Ni ( 2L i k 1 ) + (# big cells) 10d 2 P k. THE CELLULAR METHOD In ths lecture, we ntroduce the cellular method as an approach to ncdence geometry theorems lke the Szemeréd-Trotter theorem. The method was ntroduced n the paper Combnatoral complexty

More information

Section 8.3 Polar Form of Complex Numbers

Section 8.3 Polar Form of Complex Numbers 80 Chapter 8 Secton 8 Polar Form of Complex Numbers From prevous classes, you may have encountered magnary numbers the square roots of negatve numbers and, more generally, complex numbers whch are the

More information

Applied Stochastic Processes

Applied Stochastic Processes STAT455/855 Fall 23 Appled Stochastc Processes Fnal Exam, Bref Solutons 1. (15 marks) (a) (7 marks) The dstrbuton of Y s gven by ( ) ( ) y 2 1 5 P (Y y) for y 2, 3,... The above follows because each of

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

ONE-DIMENSIONAL COLLISIONS

ONE-DIMENSIONAL COLLISIONS Purpose Theory ONE-DIMENSIONAL COLLISIONS a. To very the law o conservaton o lnear momentum n one-dmensonal collsons. b. To study conservaton o energy and lnear momentum n both elastc and nelastc onedmensonal

More information

COMPOSITE BEAM WITH WEAK SHEAR CONNECTION SUBJECTED TO THERMAL LOAD

COMPOSITE BEAM WITH WEAK SHEAR CONNECTION SUBJECTED TO THERMAL LOAD COMPOSITE BEAM WITH WEAK SHEAR CONNECTION SUBJECTED TO THERMAL LOAD Ákos Jósef Lengyel, István Ecsed Assstant Lecturer, Professor of Mechancs, Insttute of Appled Mechancs, Unversty of Mskolc, Mskolc-Egyetemváros,

More information

Towards a unifying theory on branching-type polling systems in heavy traffic

Towards a unifying theory on branching-type polling systems in heavy traffic Queueng Syst 2007 57: 29 46 DOI 0007/s34-007-9044-7 Towards a unfyng theory on branchng-type pollng systems n heavy traffc RD van der Me Receved: 2 August 2006 / Revsed: 2 May 2007 / Publshed onlne: 24

More information