A DISCRETE-TIME MODEL OF TCP WITH ACTIVE QUEUE MANAGEMENT

Size: px
Start display at page:

Download "A DISCRETE-TIME MODEL OF TCP WITH ACTIVE QUEUE MANAGEMENT"

Transcription

1 A DISCRETE-TIME MODEL OF TCP WITH ACTIVE QUEUE MANAGEMENT by Hui Zhang B.Sc., Insiuion of Informaion and Communicaion, Dong Hua Universiy, P. R. China, 1 THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF Maser of Applied Science In he School of Engineering Science Hui Zhang 4 SIMON FRASER UNIVERSITY Augus 4 All righs reserved. This wor may no be reproduced in whole or in par, by phoocopy or oher means, wihou permission of he auhor.

2 APPROVAL Name: Degree: Tile of Thesis: Hui Zhang Maser of Applied Science A Discree-Time Model of TCP wih Acive Queue Managemen Examining Commiee: Chair: Dr. Glenn H. Chapman Professor, School of Engineering Science Dr. Ljiljana Trajović Senior Supervisor Professor, School of Engineering Science Dr. Uwe Gläesser Supervisor Associae Professor, School of Compuing Science Dr. Qianping Gu Examiner Professor, School of Compuing Science Dae Defended/Approved: ii

3 ABSTRACT The ineracion beween Transmission Conrol Proocol TCP) and Acive Queue Managemen AQM) can be capured using dynamical models. The communicaion newor is viewed as a discree-ime feedbac conrol sysem, where TCP adjuss is window size based on pace loss deecion during he previous Round Trip Time RTT) inerval. In his research, a discree-ime dynamical model is proposed for he ineracion beween TCP and Random Early Deecion RED), one of he mos widely used AQM schemes in oday s Inerne. The model, consruced using an ieraive map, capures he deailed dynamical behaviour of TCP/RED, including he slow sar and fas reransmi phases, as well as he imeou even in TCP. Model performance for various RED parameers is evaluaed using he ns- newor simulaor. The resuls are hen compared wih oher exising models. Based on evaluaions and comparisons of our resuls, we can conclude ha he proposed model capures he dynamical behaviour of he TCP/RED sysem. In addiion, he model provides reasonable predicions for sysem parameers such as RTT, pace sending rae, and pace drop rae. An exension of he proposed model is also presened, which is a dynamical model of TCP wih Adapive RED ARED), a modified AQM algorihm based on RED. iii

4 ACKNOWLEDGEMENTS I would lie o han my senior supervisor, Professor Lijljana Trajović, for her suppor, guidance, and nowledge. I grealy appreciae ha she gave me his precious chance o coninue my sudies and become a member of he Communicaion Newors Lab. In addiion, I han her for encouragemen, grea help and sincere care. A he same ime, I would lie o acnowledge he suppors of all of he members in our lab, especially Judy Liu, Kenny Shao, Niola Cacov, Vladimir Vuadinović, and Wan Zeng Gang. My special hans o Dr. Qianping Gu and Dr. Uwe Glaesser for heir valuable commens and for serving on my supervisory commiee. iv

5 TABLE OF CONTENTS Approval...ii Absrac...iii Acnowledgemens... iv Table of Conens... v Lis of Figures...vii Lis of Tables... ix 1 Chaper One: Moivaion... 1 Chaper Two: Bacground TCP algorihm TCP congesion conrol TCP implemenaions RED algorihm and evaluaion of is parameers Acive Queue Managemen RED algorihms Evaluaion of RED parameers Recen sudies on AQM Simulaion ool: ns Chaper Three: A survey of exising models Overview of modeling TCP Reno Overview of modeling TCP Reno wih RED... 4 Chaper Four: A discree approach for modeling TCP RENO wih Acive Queue Managemen TCP/RED model Case 1: No loss Case : One pace loss Case 3: A leas wo pace losses Properies of he S-RED Model Model validaion Validaion wih defaul RED parameers Validaion for various ueue weighs w Validaion for various drop probabiliies p max Validaion wih various hresholds min and max Validaion summary Comparison of S-RED model and M-model Comparison for defaul RED parameers v

6 4.4. Comparison for various ueue weighs w Comparison for various drop probabiliies p max Comparison for various hresholds min and max Model modificaion Model exension: TCP wih Adapive RED S-ARED model) S-ARED model validaion Model evaluaion Chaper Five: Conclusions and fuure wor... 7 Reference Lis...74 vi

7 LIST OF FIGURES Figure 1. Evoluion of he window size in TCP Reno, consising of slow sar, congesion avoidance, fas reransmi, and fas recovery phases Figure. RED drop probabiliy as a funcion of he average ueue size... 1 Figure 3. RED drop probabiliy wih p max =... 1 Figure 4. RED drop probabiliy wih p max = Figure 5. TCP window size evoluion Figure 6. Topology of he modeled and simulaed newor Figure 7. Evoluion of he window size in he S-RED model. The fas recovery phase has been simplified Figure 8. Topology of he simulaed newor Figure 9. Evoluion of he window size wih defaul RED parameers: a) S-RED model, and b) zoom-in, c) ns- simulaion resuls and d) zoom-in. Waveforms show good mach beween he S-RED model and he ns- simulaion resuls Figure 1. Evoluion of he average ueue size wih defaul RED parameers: a) S-RED model and b) zoom-in, c) ns- simulaion resul and d) zoomin. The average ueue size obained using he S-RED model is higher han he average ueue size obained using ns- simulaions Figure 11. S-RED model: evoluion of he window size for w = Figure 1. ns-: evoluion of he window size for w = Figure 13. Comparison of he average ueue size for various w Figure 14. S-RED model: evoluion of he window size for p max = Figure 15. ns-: evoluion of he window size for p max = Figure 16. Comparison of he average ueue size for various p max Figure 17. S-RED model: evoluion of he window size for min = 1 paces Figure 18. ns-: evoluion of he window size for min = 1 paces Figure 19. Comparison of he average ueue size for various min and max... 5 Figure. M-model: evoluion of he average ueue size wih defaul RED parameers Figure 1. Comparison of he average ueue size for various w vii

8 Figure. Comparison of he average ueue size for various p max Figure 3. Comparison of he average ueue size for various min and max Figure 4. S-RED model wih modificaion: a) evoluion of he window size and b) he average ueue size Figure 5. Comparison of he average ueue size for various w... 6 Figure 6. Comparison of he average ueue size for various p max Figure 7. Comparison of average ueue size for various min and max Figure 8. Evoluion of he window size: a) S-ARED and b) ns- simulaion resuls Figure 9. Evoluion of he average ueue size: a) S-ARED and b) ns- simulaion resuls viii

9 LIST OF TABLES Table 1. Defaul RED parameers in ns Table. Saisic values of sae variables for S-RED model and ns Table 3. Saisical analysis of window size for various w Table 4. Saisical analysis of he average ueue size for various w Table 5. Sysem variable for various w Table 6. Saisical analysis of he window size for various p max Table 7. Saisic analysis of he average ueue size for various p max Table 8. Sysem parameers for various p max Table 9. Saisical analysis of he window size for various max and min Table 1. Saisical analysis of he average ueue size for various max and min Table 11. Saisic analysis of he average ueue size for various w Table 1. Sysem parameers Table 13. Sysem variables for various w Table 14. Sysem variables for various p max Table 15. Sysem parameers wih various min and max Table 16. Defaul ARED parameers in ns Table 17. Average values of he window size, average ueue size, and sysem parameers ix

10 1 CHAPTER ONE: MOTIVATION Today s Inerne applicaions, such as he World Wide Web, file ransfer, Usene news, and remoe login, are delivered via he Transmission Conrol Proocol TCP). Wih an increasing number and variey of Inerne applicaions, congesion conrol becomes a ey issue. Acive Queue Managemen AQM) ineracs wih TCP congesion conrol mechanisms, and plays an imporan role in meeing oday s increasing demand for ualiy of service QoS). Random Early Deecion RED), a widely employed AQM algorihm, is a gaeway-based congesion conrol mechanism. An accurae model of TCP wih RED can aid in he undersanding and predicion of he dynamical behaviour of he newor. In addiion, he model may help in analyzing he sysem s sabiliy margins, and providing he design guidelines for selecing newor parameers. These design guidelines are imporan for newor designers whose aim is o improve newor robusness. Therefore, modeling TCP wih RED is an imporan sep owards improving he newor efficiency and he service provided o Inerne users. Modeling TCP performance has received increasing aenion during he las few years due o he benefis i offers o he neworing communiy. Analyical TCP models enable researchers o closely examine he exising congesion conrol algorihms, address heir shorcomings, and propose mehods for heir improvemen. They may also be used o compare various TCP schemes and implemenaions, and o deermine heir performance under various operaing condiions. Moreover, hese models help examine he ineracions beween TCP and he ueuing algorihms implemened in rouers. Hence, hey aid in he improvemen of exising algorihms and in designing beer algorihms, such as AQM echniues. Finally, such models offer he possibiliy of defining TCP-friendly behaviour in erms of hroughpu for non-tcp flows ha coexis wih TCP connecions in he same newor. 1

11 In his hesis, our goal is o model TCP and o invesigae he nonlinear phenomena in a TCP/RED sysem. We are paricular ineresed in he bifurcaion and chaos phenomena ha are recenly observed in a TCP/IP newor [4]. We use an ieraive map o model he sysem. A second-order discree model is derived in order o capure he ineracions of he TCP congesion conrol algorihm wih he RED ueuing mechanism. We use he conceps proposed in [34] and consruc a nonlinear dynamical model of TCP/RED ha employs wo sae variables: he window size and he average ueue size. The change of he window size reflecs he dynamics of TCP congesion conrol, while he average ueue size capures he ueue dynamics in he RED gaeway. The novely of he proposed model is in capuring he deailed dynamical behaviour of TCP/RED. The proposed model, called S-RED model, considers he slow sar, he congesion avoidance, he fas reransmi, and pars of fas recovery phases. I also aes ino accoun imeou evens. In addiion, he ineracion of TCP wih Adapive RED ARED), a revised RED algorihm, is discussed. We also derive a discree-ime model for TCP wih ARED, named S-ARED model, based on S-RED model.

12 CHAPTER TWO: BACKGROUND.1 TCP algorihm TCP, or Transmission Conrol Proocol, is designed o provide a connecion-oriened, reliable, and bye-sream service [4]. Before communicaion can begin beween wo hoss, a connecion mus be esablished by employing a hree-way handshae algorihm. This algorihm also allows he sender and receiver o iniialize and synchronize imporan parameers, seuence numbers, adverised window size, and maximum segmen size. The basic concep of he TCP congesion conrol algorihm is window based flow conrol. The window size deermines he amoun of daa o be sen by he source. Is value will increase when all of he acnowledgemens are properly received; oherwise, he window size may decrease. There are numerous versions of TCP implemenaions. In his research, he primary focus is on TCP Reno [1], [], [1], as i is one of he mos widely used versions in oday s newor. Oher versions of TCP implemenaions are briefly overviewed in Secion TCP congesion conrol To adjus he window size, he TCP Reno congesion conrol mechanism employs four algorihms: slow sar, congesion avoidance, fas reransmi, and fas recovery, as shown in Figure 1. They were inroduced by Jacobson [], [1] and are described in RFCs 11 [4] and 581 [1]. 3

13 Figure 1. Evoluion of he window size in TCP Reno, consising of slow sar, congesion avoidance, fas reransmi, and fas recovery phases. In order o avoid congesing he newor wih large daa burss, an esablished TCP connecion firs employs he slow sar algorihm o deec he available bandwidh in he newor. Typically, a TCP sender iniializes is congesion window cwnd) o one or wo segmens, depending on he TCP implemenaion. Upon receip of each acnowledgemen ACK) for new daa, TCP incremens cwnd by one segmen size. When cwnd exceeds a hreshold sshresh), he sender swiches from slow sar o he congesion avoidance phase. During congesion avoidance, cwnd is incremened by one segmen size per Round Trip Time RTT). Each ime a pace is sen by he sender, a imer is se. Pace loss is deeced by his imeou mechanism if he imer expires before receip of he pace has been acnowledged. When muliple paces are dropped from he same window or he TCP sender does no receive enough paces o rigger fas reransmi, he TCP sender may need o wai for a imeou o occur before he los paces ge reransmied. If a pace loss is deeced by he imeou mechanism, he TCP sender adjuss is sshresh and swiches bac o slow sar. The duraion of he imeou TO) is usually updaed wih he RTT. If anoher imeou occurrs before he los pace is successfully reransmied, he period of imeou TO) doubles o wice he value of he previous duraion TO). This duraion ime coninues o 4

14 increase exponenially for each unsuccessful reransmission unil he imeou period reaches 64 imes he firs imeou period 64TO). The imeou period hen remains consan a 64TO. The fas reransmi algorihm is used for recovery from losses deeced by riple duplicae ACKs; i.e. four consecuive ACKs acnowledging he same pace. When an ou-of-order segmen is received by a TCP receiver, a duplicae ACK is immediaely sen ou o inform he sender of he seuence number for he pace ha he receiver is expecing. The receip of riple duplicae ACKs is used as an indicaion of pace loss. The TCP sender reacs o he pace loss by reducing cwnd by half and by re-ransmiing he los pace wihou waiing for he reransmission imer o expire. The fas recovery algorihm is used o conrol daa ransmission afer fas reransmission of he los pace. During his phase, he TCP sender firs ses is sshresh o half of he curren cwnd. Then i increases is cwnd for each duplicae ACK received. The fas recovery algorihm recognizes each duplicae ACK as an indicaion ha one pace has lef he channel and has reached he desinaion. Since he number of ousanding paces has decreased by one, he TCP sender is allowed o incremen is cwnd. When a non-duplicae ACK is received, TCP sender ses cwnd eual o sshresh and swiches from fas recovery o he congesion avoidance phase..1. TCP implemenaions Older TCP implemenaions, released in he early 198s, employed a simple windowbased congesion conrol mechanism as specified in RFC 793 [38]. TCP Tahoe, released in he lae 198s, employed he slow sar, congesion avoidance, and fas reransmi algorihms. When pace loss is deeced, he sender reransmis he los pace, reduces he congesion window o one and eners slow sar. When he sender receives an ACK for he reransmied pace, i will coninue o send paces from he seuence number indicaed in he ACK even if some of hem are already sen. Thus, if any pace is los in he 5

15 meanime, i will be auomaically reransmied via his go-bac procedure. In oher words, TCP Tahoe will no have o wai for a imeou in he case where muliple paces are los from he same window. However, freuen losses will force he TCP sender o operae in slow sar phase for mos of he ime. TCP Reno, inroduced in he early 199s, added he fas recovery algorihm. This algorihm does no reuire slow sar for every pace loss. Once he hree duplicae ACKs are received, he TCP sender reransmis he los pace, reduces is congesion window by half, and eners he fas recovery phase insead of he slow sar algorihm. When he firs non-duplicae ACK is received, he sender exis fas recovery mode and eners he congesion avoidance phase. TCP Reno's fas recovery algorihm is opimized for he case when one pace is los from a window. When muliple paces are los from he same window, TCP Reno generally mus wai for imeou. TCP Reno s congesion conrol algorihm wors well when dealing wih congesioninduced loss. However i resuls in reduced hroughpu wihou providing any benefis. I is also inappropriae when faced wih loss due o corrupion raher han congesion, such as saellie newors. In addiion, using ns- simulaions, Fall and Floyd [11] demonsraed ha TCP Reno exhibis poor performance in erms of lin uilizaion whenever muliple paces are dropped from a single window of daa. To alleviae his problem, hey inroduced wo modificaions o TCP Reno: TCP New-Reno and TCP SACK [5]. A large number of Inerne Web servers sill use TCP Reno and is varians [35]. For example, in he TCP SACK implemenaion, TCP receivers can inform senders exacly which segmens have arrived, raher han replying on TCP s cumulaive acnowledgemens. This allows a TCP sender o efficienly recover from muliple los segmens wihou revering o using a cosly reransmission imeou o deermine which segmens need o be reransmied. I also hasens recovery and prevens he sender from becoming window limied, hus allowing he pipe o drain while waiing o learn abou los segmens. This abiliy is 6

16 of paricular benefi in eeping he pipe full and allowing ransmission o coninue while recovering from losses. Oher TCP implemenaions, such as TCP Vegas [5] and TCP Weswood [7], use various echniues o avoid congesion. They adjus he congesion window size based on esimaes of he hroughpu a he bolenec. For insance, TCP Vegas conrols he congesion window size by esimaing he RTT and by calculaing he difference beween he expeced flow rae and he acual flow rae. I linearly adjuss TCP s congesion window size upwards or downwards, so as o consume a consan amoun of buffer space in he newor. I deecs pace losses earlier han Reno and uses a slower muliple decrease han Reno. TCP Vegas eliminaes he need o une he receive window o serve as an upper bound on he size of he congesion window. I can avoid newor congesion wihou overdriving he lin o find he upper bound, even when operaing wih large windows. TCP Vegas increases is congesion window size more slowly han Reno by measuring he achieved hroughpu gain afer each increase o deec he available capaciy wihou incurring loss.. RED algorihm and evaluaion of is parameers..1 Acive Queue Managemen A radiional DropTail ueue managemen mechanism drops he paces ha arrive when he buffer is full. However, his mehod has wo drawbacs. Firs, i may allow a few connecions o monopolize he ueue space so ha oher flows are sarved. Second, DropTail allows ueues o be full for a long period of ime. During ha period, incoming paces are dropped in burss. This causes a severe reducion in hroughpu of he TCP flows. One soluion, recommended in RFC 39 [], is o employ acive ueue managemen AQM) algorihms. The purpose of AQM is o reac o incipien congesion before he buffer overflows. AQM allows responsive flows, such as 7

17 TCP flows, o reac imely and reduce heir sending raes in order o preven congesion and severe pace losses... RED algorihms The mos popular AQM algorihm is Random Early Deecion RED), proposed by Floyd and Jacobson [16]. The RED mechanism conains wo ey algorihms. One is used o calculae he exponenially weighed moving average of he ueue size, so as o deermine he bursiness ha is allowed in he gaeway ueue and o deec possible congesion. The second algorihm is for compuing he drop or maring probabiliy, which deermines how freuenly he gaeway drops or mars arrival paces. This algorihm can avoid global synchronizaion by dropping or maring paces a fairly evenly-spaced inervals. Furhermore, sufficienly dropping or maring paces, his algorihm can mainain a reasonable bound of he average delay, if he average ueue lengh is under conrol. The RED algorihm is given in Algorihm 1, where is he average ueue size and is he insananeous ueue size. Le w be he weigh facor and +1 be he new insananeous ueue size. A every pace arrival, he RED gaeway updaes he average ueue size as. 1) + 1 = 1 w ) + w + 1 During he period when he RED gaeway ueue is empy, he sysem will esimae he number of paces m ha migh have been ransmied by he rouer. Hence, he average ueue size is updaed as m + 1 = 1 w ), ) m = idle _ ime ransmission _ ime, 3) where idle_ime is he period ha he ueue is empy and ransmission_ime is he ypical ime ha a pace aes o be ransmied. 8

18 Algorihm 1: The pseudo-code of RED algorihm. Iniializaion: = ; coun = -1; for each pace arrival { calculae } if ueue empy) { m = idle_ime / ransmission_ime = 1- w ) m ; } else { = 1- w ) + w ; } if min < < max ) { coun = coun +1; calculae Pa: Pb = p max min ) / max min ); Pa = Pb / 1- coun Pb); Drop arriving pace wih Pa; if arrival pace is dropped, coun = ) } else if > max ) or = buffersize)) { } Drop arriving pace; coun = ; 9

19 The average ueue size is compared o wo parameers: he minimum ueue hreshold min, and he maximum ueue hreshold max. If he average ueue size is smaller han min, he pace is admied o he ueue. If i exceeds max, he pace is mared or dropped. If he average ueue size is beween min and max, he pace is dropped wih a drop probabiliy p b ha is a funcion of he average ueue size p b + 1 = max min min p max if if oherwise min max, 4) where p max is he maximum pace drop probabiliy. The relaionship beween he drop probabiliy and average ueue size is shown in Figure. Figure. RED drop probabiliy as a funcion of he average ueue size. p b 1 p max min max The final drop probabiliy p a is given by E. 5). I is inroduced in order o avoid he severe increase of he drop rae shown in Figure. p p = b a coun p. 5) 1 b 1

20 Here, coun is he cumulaive number of he paces ha are no mared or dropped since he las mared or dropped pace. I is increased by one if he incoming pace is no mared or dropped. Therefore, as coun increases, he drop probabiliy increases. However, if he incoming pace is mared or dropped, coun is rese o...3 Evaluaion of RED parameers Unlie DropTail, he performance of he RED algorihm is no deermined only by he buffer size. Oher parameers, such as ueue weigh w ), maximum drop rae p max ), and he wo ueue hresholds min and max ), are also of grea imporance. The newor designer should selec hese appropriaely, so ha he sysem can achieve beer performance. w is an exponenially weighed average facor, which deermines he ime consan for he averaging of he ueue size. If w is oo low, hen he esimaed average ueue size is probably responding oo slowly o ransien congesion. If w is oo high, hen he esimaed average ueue size will rac he insananeous ueue size oo closely, and may resul in an unsable sysem. In [17], uaniaive guidelines are given for seing w in erms of he ransien burs size ha he ueue can accommodae wihou dropping any paces. The opimal seing for w is beween.1 and.5. The pace drop probabiliy is calculaed as a linear funcion of he average ueue size if he average ueue size is beween min and max. However, he maximum drop probabiliy deermines he oscillaion of he drop rae. If p max is se oo small, hen he newor behaves similar o he DropTail shown in Figure 3. However, if his value is se oo high Figure 4), he pace drop probabiliy increases, hus enforcing he sysem o oscillae severely and experience a decrease in hroughpu. I has been shown ha seady-sae pace drop raes of 5% o 1% would no be unusual a a rouer. There is, herefore, no need o se p max higher han.1 for real newor implemenaions [17]. 11

21 Figure 3. RED drop probabiliy wih p max =. p b 1 p max = min max Figure 4. RED drop probabiliy wih p max = 1. pb p = 1 max min max The opimal values for min and max depend on he desired average ueue size. If he ypical raffic paerns are fairly bursy, hen min mus be correspondingly large o allow he lin uilizaion o be mainained a an accepably high level. This value should also be associaed wih he buffer size of he newor. If min is se oo small, i leads o low bandwidh uilizaion. Conversely, if min is se oo high, i may resul in unfair compeiion for bandwidh among muliple lins, hereby cancelling ou he benefis of he RED algorihm. The opimal value for max depends in par on he maximum average delay ha can be allowed by he gaeway. The RED 1

22 gaeway funcions mos effecively when he difference beween max and min is larger han he ypical increase in he calculaed average ueue size, in one roundrip ime [16]. A useful rule of humb is o se max o a leas wice he size of min. If he difference beween max and min is se oo small, hen he compued average ueue size can regularly oscillae up o max. This behaviour is similar o he oscillaions of he ueue size up o he maximum ueue size experienced wih DropTail gaeways...4 Recen sudies on AQM Alhough RED is one of he mos prominen acive ueue managemen schemes, a number of research effors have focused on possible shorcomings of he original RED algorihm and have proposed modificaions and alernaives, such as Adapive RED ARED) [15], CHOKe [36], BLUE [13], Sabilized RED, Flow RED, and Balanced RED. For example, ARED, proposed in 1, aims o improve he robusness of he original RED wih only minimal changes. I aemps o achieve a sable average ueue size around a pre-se arge value. Is defaul seing is half way beween min and max. I has been shown [15] hrough various simulaion resuls ha ARED ends o be more sable and performs beer han he original RED. The ey modificaion of ARED is updaing he maximum drop rae for every inerval period so ha he pace drop probabiliy changes more slowly han in he original RED algorihm. The calculaion of he modified p max is given in Algorihm. Oher algorihms are aimed a solving problems ha he original RED could no address. For insance, CHOKe [36] is designed o achieve beer performance on fairness, while he BLUE algorihm performs ueue managemen direcly based on pace loss and lin uilizaion, raher han on he insananeous or average ueue lenghs. In addiion, he BLUE algorihm can reac faser o subsanial increases of raffic load han he RED algorihm. 13

23 Algorihm : The pseudo-code of ARED algorihm. For every inerval usually he inerval is se o.5 seconds): { If > arge ) && p max <=.5)) } where { p max = p max + α ; } Else if < arge) && p max >=.1)) { } p max = p max * β ; arge: [ min +.4 * max min ), min +.6 * max min )]; α : min.1, p max /4); β :.9..3 Simulaion ool: ns- Ns- [33] is a discree even newor simulaor argeed a neworing research, which is currenly suppored by he Defence Advanced Research Projec Agency DARPA) and he Naional Science Foundaion NSF). I provides subsanial suppor for simulaion of TCP, rouing, and mulicas proocols over wired and wireless local and saellie) newors. In his research, ns- was used o perform he newor simulaion and he simulaion resuls were used o evaluae he proposed S-RED model. 14

24 3 CHAPTER THREE: A SURVEY OF EXISTING MODELS Many papers have been published on differen models for TCP. In his chaper, we describe he differen models and implemenaions of TCP. From he viewpoin of flow characerisics, analyical TCP models can be classified in hree caegories based on he duraion of he TCP flows. This, in urn, deermines he TCP congesion conrol algorihms o be modeled and he aspecs of TCP performance ha can be capured by he model []. The firs caegory models shor-lived flows where TCP performance is srongly affeced by he connecion esablishmen and slow sar phases [3]. These models ypically approximae average laency or compleion ime, i.e., he ime i aes o ransfer a cerain amoun of daa. The second caegory models long-lived flows, which characerize he seady-sae performance of bul TCP ransfers during he congesion avoidance phase [8], [3], [34], [41]. These models approximae aspecs such as he window size, average hroughpu, sending rae, and goodpu goodpu is used o describe he daa ha has been properly received). The final caegory includes models for flows of arbirary duraion, i.e., hose ha can accommodae boh shor and long-lived flows [6], [8]. All hese models provide funcions o compue he average values of hroughpu, sending rae, goodpu, or laency of TCP flows. From he conrol heoreic poin of view, he developed models of TCP and TCP/RED [18], [19], [3], [4], [31], [39], [4], can be classified ino wo ypes: averaged models and ieraive map models. An averaged model is described by a se of coninuous differenial euaions. I neglecs he deailed dynamics and only capures he low freuency characerisics of he sysem. I can be used o analyze he seady-sae performance and o predic low freuency slow-scale bifurcaion behaviour, such as Hopf bifurcaions. Examples of such models are given in [18], [19], [3]. In conras, an ieraive map model has a discree-ime form and employs a se 15

25 of difference euaions. I provides relaively complee dynamical informaion. The discree-ime map is adeuae o explore nonlinear phenomena such as period-doubling and saddle-node bifurcaions, which may appear across a wide specrum of freuencies and cause he exisence of soluions in he high freuency range. Examples of ieraive maps are given in [4], [39], [4]. 3.1 Overview of modeling TCP Reno Several models have been proposed in order o analyze he performance of pace newors. Prior o 1997, research was primarily focused on modeling he TCP congesion avoidance phase and ignored slow sar, fas recovery, and imeou evens. In [7], he auhors assume random pace loss wih a consan probabiliy p, which means ha approximaely 1 / p paces are o be sen followed by one pace loss. Hence, he evoluion of he cwnd resuls a periodic sawooh shown in Figure 5. Figure 5. TCP window size evoluion. 16

26 The cwnd is halved when a pace is dropped. During his period, here are 1 W W 3W + W ) = paces o be delivered, where W is he maximum cwnd. Since, 8 1 3W p = 8, i follows ha cwnd is W = 8 3p. 6) As cwnd increases by one pace size every RTT, he ime reuired o send 3W 8 paces is W RTT. Hence, he sending rae becomes BW 3 M W = 8 W RTT M C = RTT p, 7) where consan C = 3 /. By selecing an appropriae value for he consan C, he model is able o esimae he TCP performance over a range of loss condiions and in environmens wih boh DropTail and RED ueuing [8]. However, he predicion of he model is no very accurae, because i simplifies he TCP congesion conrol algorihm. A more precise seady-sae model of TCP Reno, inroduced in [34], also employs he seady-sae sending rae as a funcion of loss rae and RTT of a bul daa ransfer TCP flow. The model provides a deailed explanaion of he TCP algorihm. In addiion o capuring he essence of TCP s fas reransmi and congesion avoidance phases, i also accouns for he effec of he 17

27 imeou mechanism, which is imporan from a modeling perspecive. The model also aes ino accoun delayed ACKs, which are no discussed in his research. However, his model ignores he influence of ueue managemen schemes. In [34], auhors ae several seps o analyze he sending rae by separaing he congesions caused by hree-duplicae ACKs and imeous. They also consider oher upper bound parameers, such as adverised window size. The model describes he sysem in erms of duraion of he RTT, called rounds. The model is buil based on he following assumpions: RTT is a consan and i is considered o be independen of he window size; he maximum window size is also consan; only long lived connecions are sudied, which implies ha he source buffer always has enough daa o send; any pace loss is considered as lin congesion; he pace loss is correlaed wihin one round, i.e., if a pace is los, hen he remaining paces ransmied in ha round are also los. However, he pace loss is independen in differen rounds. Slow sar and fas recovery phases are no considered in his model. The following wo siuaions describe he cases sudied in [34]. Case one: The imeou is no considered and congesion is only deeced by hree duplicae ACKs. The expecaion of he oal number of paces Y ha are o be ransmied before a pace loss, is given by 1 p E[ Y ] = + E[ W ] p p = + p 3p, 8) where p is he drop probabiliy and W is he value of congesion window size. The ime reuired o send Y paces is noed by A. The expecaion of A is 18

28 RTT 8 5 p E[ A] = 3 + ). 9) 3p From Es 8) and 9), he sending rae BW can be derived as BW = = E[ Y ] E[ A] 1 8 5p + p 3p. 1) RTT 8 5p 3 + ) 3p Euaion 1) can be simplified o E. 11) for small values of p as BW 1 3 =. 11) RTT p Case Two: The loss indicaion is deeced by hree duplicae ACKs and imeous: The sending rae in E. 1) is modified o BW E[ Y ] + Q E[ R] =, 1) TO E[ A] + Q E[ Z ] where Q is he probabiliy ha a pace loss is deeced by a imeou, R is he oal number of paces ha are o be sen during he imeou inerval, and Z TO is he duraion of he imeou. The final simplified euaion for small values of p) for he sending rae is BW 1, 13) p 3p RTT + T min1,3 ) p1 + 3 p ) 3 8 where T is he iniial value of a imeou. Measuremens demonsraed ha he model was adeuae over a range of loss raes [34]. 19

29 The model developed in [6] is a furher exension of he model developed in [34], which aims for a more precise model of TCP laency. The model aes ino accoun he ime used o se up he connecion, which is nown as he hree-way handshae algorihm, and he slow sar phase. This informaion is very imporan for modeling shor-lived connecions. In addiion, he model considers deailed informaion abou daa ransmission including congesion avoidance, fas reransmi, fas recovery, and imeou evens. The oal laency of he newor is E T ] = E[ T ] + E[ T ] + E[ T ] + E[ T ], 14) [ ss loss ca delac where T ss is he duraion of daa ransfer wihin he slow sar and congesion avoidance phases, T loss is he cos for imeou or fas recovery, T ca is he ime reuired o ransmi he remaining paces before loss recovery, and T delac is he cos of he delayed ACKs. 3. Overview of modeling TCP Reno wih RED A simplified firs-order discree nonlinear dynamical model was developed for simplified TCP wih RED conrol in [4], [39], [4]. The exponenially weighed average ueue size was used as he sae variable. The model describes he sysem dynamics over large parameer variaions and samples he buffer occupancy a cerain ime insances. This dynamical model was used o invesigae he sabiliy, bifurcaion, and roues o chaos of a newor for various sysem parameers. Based on he developed model, he auhors demonsraed ha nonlinear phenomena, such as bifurcaion and chaos, migh occur if he sysem parameers were no properly seleced. However, his discree model neglecs he dynamics of TCP. The derived map is given by ave w) = 1 w) 1 w) ave ave ave + w B N K + w p C d ) M if if ave ave oherwise ave u ave l, 15)

30 where ave 1 + : average ueue size in round +1 ave : average ueue size in round ave u : upper bound of he average ueue size ave l : lower bound of he average ueue size w : ueue weigh in RED algorihm B : buffer size N : number of TCP connecions K : consan [1, 8 / 3 ] p : drop probabiliy in round C : lin capaciy d : round-rip propagaion delay M : pace size. In [31], a second-order nonlinear dynamical model was developed o examine he ineracions of a se of TCP flows wih RED. The model employs he fluid-flow and sochasic differenial euaions. Window size and average ueue lengh are used as sae variables. From a se of coupled ordinary differenial euaions, he auhors develop a numerical algorihm o obain he ransien average behaviour of ueue lengh, round rip ime, and hroughpu of TCP flows. This model is described by nonlinear differenial euaions ha employ he average values of ey newor variables, given by 1

31 C W R N R p R R R W W R W = = ) ) ) ) )) )) )) ) ) 1 ) D D, 16) where ) W : expeced TCP window size ) : expeced ueue lengh ) R : round rip ime ) N : load facor number of TCP sessions) ) p : probabiliy of pace mar/drop C : lin capaciy. A hird-order dynamical model ha describes he ineracion of TCP flows wih an RED conrolled ueue was developed in [3]. The sae variables of he model are average ueue lengh, insananeous ueue lengh, and hroughpu. The TCP sources are idealized o operae only during he congesion avoidance phase, when he congesion window size follows he rule of linear increase and muliplicaive decrease. This dynamical model is used o explore various parameer seings and observe ransien and euilibrium behaviour of he sysem. The validiy of he model is verified by comparison wih simulaion resuls. The ineracion beween TCP and RED is modeled as )) )) )) )) ) ) ) )) / 1 ) ) ) / ) ))) 1 ))) 1 ))) 1 ) / ) )) ) ) / ) s p s p P R R m R P R m R P d d s p R c d s R s d d K K L L L K π π λ λ λ λ λ π µ π λ β λ + = + = = =, 17)

32 where s ) : expecaion of he exponenially averaged ueue lengh ) : expecaion of he insananeous ueue lengh λ ) : expecaion of TCP hroughpu a he source P L ) : loss probabiliy in he ueue a ime π )) : seady-sae probabiliies for he ueue o be full K π )) : seady-sae probabiliies for he ueue o be empy p : drop probabiliy R : round rip ime β : ueue weigh in RED algorihm m : number of idenical TCP sources. In [18], [19], he auhors describe a second-order linear model for TCP and RED, by linearizing he fluid-based nonlinear TCP model. Window size and average ueue lengh are used as sae variables of he sysem. The auhors performed analysis of TCP ineracions wih RED from a conrol heoreic viewpoin. They presened design guidelines for choosing RED parameers ha led o local sabiliy of he sysem. In addiion, hey proposed wo alernaive conrollers o improve he ransien behaviour and sabiliy of he sysem: a proporional P) conroller ha possesses a good ransien response, bu suffers from seady-sae errors in ueue regulaion, and a classical proporional-inegral PI) conroller ha exhibis zero seady-sae regulaion error and accepable ransien behaviour. One imporan conribuion of his wor is providing a good example of how o use classical conrol heory o solve problems in complex communicaion sysems. The linearized model around he operaing poin is described as 3

33 4 ) 1 ) ) ) )) ) 1 )) ) ) R W R N C p N C R R C R R W W C R N W δ δ δ δ δ δ δ δ δ = + =, 18) where W W W = δ = δ p p p = δ ),, p W : he se of operaing poins W : expecaion of he TCP window size : expecaion of he ueue lengh R : round rip ime C : lin capaciy p T : propagaion delay N : load facor number of TCP sessions) p : probabiliy of pace mar/drop. A muli-lin muli-source model, developed in [7], was used o sudy he sabiliy of a general TCP/AQM sysem. A local sabiliy condiion in he case of a single lin wih heerogeneous sources and he sabiliy region of TCP/RED were derived. The sae variables of his model are window size, insananeous ueue lengh, and average ueue lengh. The auhors demonsraed ha TCP/RED becomes unsable when he lin capaciy increases. Finally, hey devised a new disribued congesion conrol algorihm ha mainains local sabiliy for arbirary

34 delay, capaciy, and raffic load. Preliminary simulaion resuls were provided in order o illusrae he model s behaviour. 5

35 4 CHAPTER FOUR: A DISCRETE APPROACH FOR MODELING TCP RENO WITH ACTIVE QUEUE MANAGEMENT 4.1 TCP/RED model The basic idea behind RED is o sense impending congesion before i happens and o ry o provide feedbac o senders by eiher dropping or maring paces. Hence, from he conrol heoreic poin of view, he newor may be considered as a complex feedbac conrol sysem. TCP adjuss is sending rae depending on wheher or no i has deeced a pace drop in he previous RTT inerval. The drop probabiliy of RED can be considered as he conrol law of he newor sysem. Is disconinuiy is he main reason for oscillaions and chaos in he sysem. Hence, i is naural o model he newor sysem as a discree model. In his hesis, we model he TCP/RED sysem using a sroboscopic map, which is he mos widely used ype of discree-ime maps for modeling power converers [3], [9], [1], [43]. This map is obained by periodically sampling he sysem sae. In our sudy, he sampling period is one RTT. Since window size and ueue size behave as sep funcions of RTT, one RTT is he sampling period ha capures heir changes [14]. Higher sampling raes would no significanly improve he accuracy of he model. Conversely, lower sampling raes would ignore he changes and affec he model accuracy. The sae variables employed in he S-RED model are window size and average ueue size. These sae variables capure he deailed behaviour of TCP/RED. Variaions of he window size reflec he dynamics of TCP congesion conrol. The window size increases exponenially and linearly during he slow sar and he congesion avoidance phases, respecively. I muliplicaively decreases when loss occurs. The average ueue size capures he ueue dynamics 6

36 in RED as i is updaed upon every pace arrival. In our sudy, we did no consider insananeous ueue size as an independen sae variable. The S-RED model is suied for invesigaing he nonlinear behavior of TCP/RED and analyzing he sysem s ransiions ino chaos. Hence, in his hesis, we consider a simple newor opology as shown in Figure 6. I consiss of one TCP source, wo rouers, and a desinaion. RED is employed in he firs rouer. A TCP Reno connecion is esablished beween he source and he desinaion. Daa paces are sen from he source o he desinaion, while raffic in he opposie direcion consiss of ACK paces only. We made several assumpions in order o consruc an approximae model. We assume ha ACK paces are never los. The connecion is long-lived and he source always has sufficien daa o send. Round rip propagaion delay d beween he source and desinaion is consan. Daa pace size M is consan. The lin ha connecs he wo rouers is he only bolenec in he newor. We also assume ha he imeou is only caused by pace loss and ha he duraion of he imeou period is 5 RTTs [14]. The sae variables of he sysem are sampled a he end of every RTT period. We assume ha he ueue size is consan during each sampling period. The model includes hree cases, depending on he number of paces los in he previous RTT period: no loss, single loss, and muliple pace losses. Figure 6. Topology of he modeled and simulaed newor. Bolenec Source Rouer 1 Rouer Desinaion 7

37 4.1.1 Case 1: No loss Le W,, and be he window size, insananeous ueue size, and average ueue size, respecively, a he end of he sampling period. If no pace is dropped during he las RTT period, TCP Reno increases is window size. The window size is increased exponenially in he slow sar phase and linearly in he congesion avoidance phase: W + 1 min W, sshresh) if W < sshresh =, 19) min W + 1, rwnd ) if W sshresh where rwnd is he receiver s adverised window size; i.e., he larges window size ha he receiver can accep in one round. Usually, rwnd is greaer han he window size. In his case, rwnd does no affec he variaions of he window size. In he even ha he window size increases linearly and reaches he value rwnd, he window size is ep a rwnd unil a loss occurs in he newor. The round rip ime in +1 round is calculaed by M RTT +1 = d +, ) C where RTT + 1 : round rip ime in round + 1 d : round rip propagaion delay C : lin capaciy M : pace size. In order o calculae he average ueue size given by E. 1), we need o now he ueue size a he sampling period +1. The insananeous ueue size depends on he ueue size in he previous period, he curren window size, and he number of paces ha have lef he ueue during he previous sampling period. Therefore, he insananeous ueue size is given by 8

38 9 M d C W C M d M C W M RTT C W = + + = + = ). 1) where 1 + : insananeous ueue size in round + 1 : insananeous ueue size in round +1 W : curren TCP window size in round + 1 Subsiuing +1 in E. 1) yields he average ueue size:,) max ) M d C W w w + = + +. ) Because RED updaes he average ueue size a every pace arrival, is updaed W +1 imes during he curren sampling period. For he firs pace arrival, he average ueue size is updaed according o,) max ) 1 M d C W w w + =. 3) For he second pace arrival, he average ueue size is given by [ ] ), max ), max ) 1 ) 1 M d C W w M d C W w w w + + =. 4) Since we assume ha ueue size is consan during each period, and ha here are W +1 pace arrivals during his ime inerval, is updaed W +1 imes:,) max ) ) 1 1 ) M d C W w w W W + = )

39 In summary, if p W <.5, which implies ha no pace loss occurred in he previous sampling period, he sae variables of he model are W minw, sshresh) if = min W + 1, rwnd) if = 1 w ) W w W ) W W + 1 < sshresh sshresh ) max W ) C d,) M 4.1. Case : One pace loss If.5 p W < 1.5, which implies ha one pace loss occurred in he previous RTT period, he congesion conrol mechanism of TCP Reno halves he window size in he curren sampling period: 1 W = W ) The average ueue size is updaed in he same manner as in Case 1: W = W = 1 w ) W w ) W + 1 ) max W ) C d,) M Case 3: A leas wo pace losses In his case p W 1.5, which implies ha a leas wo paces are los in he previous RTT period. When muliple paces are los from he same window, TCP Reno may no be able o send a sufficien number of new paces in order o receive hree duplicae ACKs for each los pace. The TCP source will probably have o wai for a imeou o occur before reransmiing he los pace [11]. During he imeou period, he source does no send paces ino he newor. In our model, window size is euivalen o he number of paces ha are sen by he source during one RTT period. Hence, we assume ha he window size is zero during he imeou period. 3

40 The RED mechanism updaes he average ueue size for each pace arrival. However, when he ueue is empy, i calculaes he duraion of his period named idle ime. Then, he average ueue size is updaed based on esimaing he number of paces ha could have been sen during his ime. Our model does no ae ino accoun he idle ime. Therefore, during he imeou period, here are no pace arrivals. The average ueue size is no updaed and has he same value as in he previous RTT period. The TCP/RED sysem during he imeou period is described a W =. 9) = The pseudo-code describing he S-RED model is given in Algorihm 3. 31

41 Algorihm 3: The pseudo-code of he S-RED model. Iniializaion: ; ; p ; for every round { calculae he produc of if p W <. 5 p W { compare W wih sshresh if W < sshresh { } else { } W + 1 minw, sshresh) W + 1 W w ) w ) ) max W = + W 1 min W 1, rwnd) + + W + 1 W w ) w ) ) max W = + calculae he drop probabiliy using E. 4) } elseif.5 p W < 1. 5 { W 1 +1 W W + 1 W w ) w ) ) max W = + calculae he drop probabiliy using E. 4) C d,) M C d,) M C d,) M 3

42 } else { W calculae he drop probabiliy using E. 4) } } 4. Properies of he S-RED Model The proposed second-order discree-ime model capures he ineracions beween he TCP Reno congesion conrol algorihm and he RED mechanism. I models he dynamics of he TCP/RED sysem. Unlie pas models [3], [4], [34], [39], [4], he S-RED model includes he slow sar phase. I also aes ino accoun imeou, common in TCP [34], ha he models in [3], [4], [39], [4] ignore. However, our model does no capure all he deails of he fas recovery phase. Congesion window size in he S-RED model is no increased for each duplicae ACK received afer reransmiing he los pace. Insead of inflaing he window size, we assume ha he TCP sender swiches o he congesion avoidance phase wihou performing slow sar. Evoluion of he window size in he S-RED model is shown in Figure 7. The S-RED model capures he mos imporan characerisics of he RED algorihm. The average ueue size is updaed afer each pace arrival; i.e., W +1 imes in he sampling period +1. In conras, models presened in [4], [39], [4] updae he average ueue size only once every RTT period. However, in deriving he new model, we have also made simplificaions o he RED algorihm. RED uses coun o modify he drop probabiliy according o E. 5), while we ignored he coun ha eeps rac of he number of pace arrivals since he las pace drop. We 33

43 have also ignored he idle ime period, since i has no significan impac on he dynamics of he sysem. Figure 7. Evoluion of he window size in he S-RED model. The fas recovery phase has been simplified. 4.3 Model validaion In order o evaluae he accuracy of he S-RED model, we compare is performance wih he ns- simulaion resuls. The opology of he simulaed newor is shown in Figure 8. I consiss of one source, one sin, and wo rouers: R1 and R. RED is employed in rouer R1. The lin beween R1 and R is he only bolenec in he newor. Is capaciy is 1.54 Mbps and is propagaion delay is 1 ms. The capaciy of he lins beween he source and R1, and beween R and he sin, is 1 Mbps. This is sufficien o guaranee no congesion in hese lins. Their propagaion delay is ms. We evaluaed he window size and he average ueue size in he S- RED model using MATLAB [9] and compared he resuls wih ns- simulaion resuls. 34

44 Figure 8. Topology of he simulaed newor. The model validaion is divided ino four sages. Firs, we used ns- defaul RED parameers. Second, we chose various ueue weighs w, while eeping oher sysem parameers consan. In he hird scenario, we varied he maximum drop probabiliy p max. Finally, we varied he minimum and maximum ueue hresholds min and max simulaneously, while eeping heir raio max / min = 3. In each simulaion scenario, we observed sysem behaviour and capured average RTT, sending rae, and drop probabiliy Validaion wih defaul RED parameers In order o validae ha he S-RED model can capure he deails of he sysem behaviour, we evaluaed he ime waveforms of he wo sae variables: window size and average ueue size. The ns- simulaion resuls are obained using defaul RED parameers lised in Table 1. Table 1. Defaul RED parameers in ns-. Pace size M byes) 5 Maximum drop probabiliy p max.1 Minimum ueue hreshold min paces) 5 Maximum ueue hreshold max paces) 15 Queue weigh w. The waveforms of he window size for various ime scales are shown in Figure 9. The proposed S-RED model and ns- simulaion resuls are uie similar, especially during he seadysae. 35

45 Figure 1 shows waveforms of he average ueue size for various ime scales. The average ueue size in he S-RED model is approximaely one pace larger han he average ueue size resuling from he ns- simulaion resuls. This difference is due o he simplificaions ha we inroduced in he S-RED model. Our S-RED model employs pace-maring/drop probabiliy calculaed by E. 4), while he RED algorihm adops a smooh drop probabiliy p a ha increases slowly as he coun increases: p a p = b, 3) 1 coun p b where p b is he drop probabiliy given by E. 4) and coun measures he number of paces ha have arrived since he las dropped pace. In he S-RED model, p b is used as he final drop probabiliy and coun is ignored. Since p b < p a, he average ueue size of he S-RED model is larger han ha obained via ns- simulaions. Figure 9. Evoluion of he window size wih defaul RED parameers: a) S-RED model, and b) zoom-in, c) ns- simulaion resuls and d) zoom-in. Waveforms show good mach beween he S-RED model and he ns- simulaion resuls. a) 35 Window size paces) Time sec) 36

Trends in TCP/IP Retransmissions and Resets

Trends in TCP/IP Retransmissions and Resets Trends in TCP/IP Reransmissions and Reses Absrac Concordia Chen, Mrunal Mangrulkar, Naomi Ramos, and Mahaswea Sarkar {cychen, mkulkarn, msarkar,naramos}@cs.ucsd.edu As he Inerne grows larger, measuring

More information

Improvement of a TCP Incast Avoidance Method for Data Center Networks

Improvement of a TCP Incast Avoidance Method for Data Center Networks Improvemen of a Incas Avoidance Mehod for Daa Cener Neworks Kazuoshi Kajia, Shigeyuki Osada, Yukinobu Fukushima and Tokumi Yokohira The Graduae School of Naural Science and Technology, Okayama Universiy

More information

TSG-RAN Working Group 1 (Radio Layer 1) meeting #3 Nynashamn, Sweden 22 nd 26 th March 1999

TSG-RAN Working Group 1 (Radio Layer 1) meeting #3 Nynashamn, Sweden 22 nd 26 th March 1999 TSG-RAN Working Group 1 (Radio Layer 1) meeing #3 Nynashamn, Sweden 22 nd 26 h March 1999 RAN TSGW1#3(99)196 Agenda Iem: 9.1 Source: Tile: Documen for: Moorola Macro-diversiy for he PRACH Discussion/Decision

More information

PROFIT TEST MODELLING IN LIFE ASSURANCE USING SPREADSHEETS PART ONE

PROFIT TEST MODELLING IN LIFE ASSURANCE USING SPREADSHEETS PART ONE Profi Tes Modelling in Life Assurance Using Spreadshees PROFIT TEST MODELLING IN LIFE ASSURANCE USING SPREADSHEETS PART ONE Erik Alm Peer Millingon 2004 Profi Tes Modelling in Life Assurance Using Spreadshees

More information

Task is a schedulable entity, i.e., a thread

Task is a schedulable entity, i.e., a thread Real-Time Scheduling Sysem Model Task is a schedulable eniy, i.e., a hread Time consrains of periodic ask T: - s: saring poin - e: processing ime of T - d: deadline of T - p: period of T Periodic ask T

More information

Morningstar Investor Return

Morningstar Investor Return Morningsar Invesor Reurn Morningsar Mehodology Paper Augus 31, 2010 2010 Morningsar, Inc. All righs reserved. The informaion in his documen is he propery of Morningsar, Inc. Reproducion or ranscripion

More information

Chapter 8: Regression with Lagged Explanatory Variables

Chapter 8: Regression with Lagged Explanatory Variables Chaper 8: Regression wih Lagged Explanaory Variables Time series daa: Y for =1,..,T End goal: Regression model relaing a dependen variable o explanaory variables. Wih ime series new issues arise: 1. One

More information

The Application of Multi Shifts and Break Windows in Employees Scheduling

The Application of Multi Shifts and Break Windows in Employees Scheduling The Applicaion of Muli Shifs and Brea Windows in Employees Scheduling Evy Herowai Indusrial Engineering Deparmen, Universiy of Surabaya, Indonesia Absrac. One mehod for increasing company s performance

More information

Duration and Convexity ( ) 20 = Bond B has a maturity of 5 years and also has a required rate of return of 10%. Its price is $613.

Duration and Convexity ( ) 20 = Bond B has a maturity of 5 years and also has a required rate of return of 10%. Its price is $613. Graduae School of Business Adminisraion Universiy of Virginia UVA-F-38 Duraion and Convexiy he price of a bond is a funcion of he promised paymens and he marke required rae of reurn. Since he promised

More information

11/6/2013. Chapter 14: Dynamic AD-AS. Introduction. Introduction. Keeping track of time. The model s elements

11/6/2013. Chapter 14: Dynamic AD-AS. Introduction. Introduction. Keeping track of time. The model s elements Inroducion Chaper 14: Dynamic D-S dynamic model of aggregae and aggregae supply gives us more insigh ino how he economy works in he shor run. I is a simplified version of a DSGE model, used in cuing-edge

More information

The Grantor Retained Annuity Trust (GRAT)

The Grantor Retained Annuity Trust (GRAT) WEALTH ADVISORY Esae Planning Sraegies for closely-held, family businesses The Granor Reained Annuiy Trus (GRAT) An efficien wealh ransfer sraegy, paricularly in a low ineres rae environmen Family business

More information

A Scalable and Lightweight QoS Monitoring Technique Combining Passive and Active Approaches

A Scalable and Lightweight QoS Monitoring Technique Combining Passive and Active Approaches A Scalable and Lighweigh QoS Monioring Technique Combining Passive and Acive Approaches On he Mahemaical Formulaion of CoMPACT Monior Masai Aida, Naoo Miyoshi and Keisue Ishibashi NTT Informaion Sharing

More information

Automatic measurement and detection of GSM interferences

Automatic measurement and detection of GSM interferences Auomaic measuremen and deecion of GSM inerferences Poor speech qualiy and dropped calls in GSM neworks may be caused by inerferences as a resul of high raffic load. The radio nework analyzers from Rohde

More information

II.1. Debt reduction and fiscal multipliers. dbt da dpbal da dg. bal

II.1. Debt reduction and fiscal multipliers. dbt da dpbal da dg. bal Quarerly Repor on he Euro Area 3/202 II.. Deb reducion and fiscal mulipliers The deerioraion of public finances in he firs years of he crisis has led mos Member Saes o adop sizeable consolidaion packages.

More information

MACROECONOMIC FORECASTS AT THE MOF A LOOK INTO THE REAR VIEW MIRROR

MACROECONOMIC FORECASTS AT THE MOF A LOOK INTO THE REAR VIEW MIRROR MACROECONOMIC FORECASTS AT THE MOF A LOOK INTO THE REAR VIEW MIRROR The firs experimenal publicaion, which summarised pas and expeced fuure developmen of basic economic indicaors, was published by he Minisry

More information

CALCULATION OF OMX TALLINN

CALCULATION OF OMX TALLINN CALCULATION OF OMX TALLINN CALCULATION OF OMX TALLINN 1. OMX Tallinn index...3 2. Terms in use...3 3. Comuaion rules of OMX Tallinn...3 3.1. Oening, real-ime and closing value of he Index...3 3.2. Index

More information

USE OF EDUCATION TECHNOLOGY IN ENGLISH CLASSES

USE OF EDUCATION TECHNOLOGY IN ENGLISH CLASSES USE OF EDUCATION TECHNOLOGY IN ENGLISH CLASSES Mehme Nuri GÖMLEKSİZ Absrac Using educaion echnology in classes helps eachers realize a beer and more effecive learning. In his sudy 150 English eachers were

More information

Multiprocessor Systems-on-Chips

Multiprocessor Systems-on-Chips Par of: Muliprocessor Sysems-on-Chips Edied by: Ahmed Amine Jerraya and Wayne Wolf Morgan Kaufmann Publishers, 2005 2 Modeling Shared Resources Conex swiching implies overhead. On a processing elemen,

More information

Making a Faster Cryptanalytic Time-Memory Trade-Off

Making a Faster Cryptanalytic Time-Memory Trade-Off Making a Faser Crypanalyic Time-Memory Trade-Off Philippe Oechslin Laboraoire de Securié e de Crypographie (LASEC) Ecole Polyechnique Fédérale de Lausanne Faculé I&C, 1015 Lausanne, Swizerland philippe.oechslin@epfl.ch

More information

DYNAMIC MODELS FOR VALUATION OF WRONGFUL DEATH PAYMENTS

DYNAMIC MODELS FOR VALUATION OF WRONGFUL DEATH PAYMENTS DYNAMIC MODELS FOR VALUATION OF WRONGFUL DEATH PAYMENTS Hong Mao, Shanghai Second Polyechnic Universiy Krzyszof M. Osaszewski, Illinois Sae Universiy Youyu Zhang, Fudan Universiy ABSTRACT Liigaion, exper

More information

Working Paper No. 482. Net Intergenerational Transfers from an Increase in Social Security Benefits

Working Paper No. 482. Net Intergenerational Transfers from an Increase in Social Security Benefits Working Paper No. 482 Ne Inergeneraional Transfers from an Increase in Social Securiy Benefis By Li Gan Texas A&M and NBER Guan Gong Shanghai Universiy of Finance and Economics Michael Hurd RAND Corporaion

More information

Model-Based Monitoring in Large-Scale Distributed Systems

Model-Based Monitoring in Large-Scale Distributed Systems Model-Based Monioring in Large-Scale Disribued Sysems Diploma Thesis Carsen Reimann Chemniz Universiy of Technology Faculy of Compuer Science Operaing Sysem Group Advisors: Prof. Dr. Winfried Kalfa Dr.

More information

Individual Health Insurance April 30, 2008 Pages 167-170

Individual Health Insurance April 30, 2008 Pages 167-170 Individual Healh Insurance April 30, 2008 Pages 167-170 We have received feedback ha his secion of he e is confusing because some of he defined noaion is inconsisen wih comparable life insurance reserve

More information

Constant Data Length Retrieval for Video Servers with Variable Bit Rate Streams

Constant Data Length Retrieval for Video Servers with Variable Bit Rate Streams IEEE Inernaional Conference on Mulimedia Compuing & Sysems, June 17-3, 1996, in Hiroshima, Japan, p. 151-155 Consan Lengh Rerieval for Video Servers wih Variable Bi Rae Sreams Erns Biersack, Frédéric Thiesse,

More information

Chapter 2 Problems. 3600s = 25m / s d = s t = 25m / s 0.5s = 12.5m. Δx = x(4) x(0) =12m 0m =12m

Chapter 2 Problems. 3600s = 25m / s d = s t = 25m / s 0.5s = 12.5m. Δx = x(4) x(0) =12m 0m =12m Chaper 2 Problems 2.1 During a hard sneeze, your eyes migh shu for 0.5s. If you are driving a car a 90km/h during such a sneeze, how far does he car move during ha ime s = 90km 1000m h 1km 1h 3600s = 25m

More information

Real-time Particle Filters

Real-time Particle Filters Real-ime Paricle Filers Cody Kwok Dieer Fox Marina Meilă Dep. of Compuer Science & Engineering, Dep. of Saisics Universiy of Washingon Seale, WA 9895 ckwok,fox @cs.washingon.edu, mmp@sa.washingon.edu Absrac

More information

cooking trajectory boiling water B (t) microwave 0 2 4 6 8 101214161820 time t (mins)

cooking trajectory boiling water B (t) microwave 0 2 4 6 8 101214161820 time t (mins) Alligaor egg wih calculus We have a large alligaor egg jus ou of he fridge (1 ) which we need o hea o 9. Now here are wo accepable mehods for heaing alligaor eggs, one is o immerse hem in boiling waer

More information

SELF-EVALUATION FOR VIDEO TRACKING SYSTEMS

SELF-EVALUATION FOR VIDEO TRACKING SYSTEMS SELF-EVALUATION FOR VIDEO TRACKING SYSTEMS Hao Wu and Qinfen Zheng Cenre for Auomaion Research Dep. of Elecrical and Compuer Engineering Universiy of Maryland, College Park, MD-20742 {wh2003, qinfen}@cfar.umd.edu

More information

TEMPORAL PATTERN IDENTIFICATION OF TIME SERIES DATA USING PATTERN WAVELETS AND GENETIC ALGORITHMS

TEMPORAL PATTERN IDENTIFICATION OF TIME SERIES DATA USING PATTERN WAVELETS AND GENETIC ALGORITHMS TEMPORAL PATTERN IDENTIFICATION OF TIME SERIES DATA USING PATTERN WAVELETS AND GENETIC ALGORITHMS RICHARD J. POVINELLI AND XIN FENG Deparmen of Elecrical and Compuer Engineering Marquee Universiy, P.O.

More information

Chapter 7. Response of First-Order RL and RC Circuits

Chapter 7. Response of First-Order RL and RC Circuits Chaper 7. esponse of Firs-Order L and C Circuis 7.1. The Naural esponse of an L Circui 7.2. The Naural esponse of an C Circui 7.3. The ep esponse of L and C Circuis 7.4. A General oluion for ep and Naural

More information

Distributing Human Resources among Software Development Projects 1

Distributing Human Resources among Software Development Projects 1 Disribuing Human Resources among Sofware Developmen Proecs Macario Polo, María Dolores Maeos, Mario Piaini and rancisco Ruiz Summary This paper presens a mehod for esimaing he disribuion of human resources

More information

CHARGE AND DISCHARGE OF A CAPACITOR

CHARGE AND DISCHARGE OF A CAPACITOR REFERENCES RC Circuis: Elecrical Insrumens: Mos Inroducory Physics exs (e.g. A. Halliday and Resnick, Physics ; M. Sernheim and J. Kane, General Physics.) This Laboraory Manual: Commonly Used Insrumens:

More information

Measuring macroeconomic volatility Applications to export revenue data, 1970-2005

Measuring macroeconomic volatility Applications to export revenue data, 1970-2005 FONDATION POUR LES ETUDES ET RERS LE DEVELOPPEMENT INTERNATIONAL Measuring macroeconomic volailiy Applicaions o expor revenue daa, 1970-005 by Joël Cariolle Policy brief no. 47 March 01 The FERDI is a

More information

The Transport Equation

The Transport Equation The Transpor Equaion Consider a fluid, flowing wih velociy, V, in a hin sraigh ube whose cross secion will be denoed by A. Suppose he fluid conains a conaminan whose concenraion a posiion a ime will be

More information

Chapter 1.6 Financial Management

Chapter 1.6 Financial Management Chaper 1.6 Financial Managemen Par I: Objecive ype quesions and answers 1. Simple pay back period is equal o: a) Raio of Firs cos/ne yearly savings b) Raio of Annual gross cash flow/capial cos n c) = (1

More information

Option Put-Call Parity Relations When the Underlying Security Pays Dividends

Option Put-Call Parity Relations When the Underlying Security Pays Dividends Inernaional Journal of Business and conomics, 26, Vol. 5, No. 3, 225-23 Opion Pu-all Pariy Relaions When he Underlying Securiy Pays Dividends Weiyu Guo Deparmen of Finance, Universiy of Nebraska Omaha,

More information

The naive method discussed in Lecture 1 uses the most recent observations to forecast future values. That is, Y ˆ t + 1

The naive method discussed in Lecture 1 uses the most recent observations to forecast future values. That is, Y ˆ t + 1 Business Condiions & Forecasing Exponenial Smoohing LECTURE 2 MOVING AVERAGES AND EXPONENTIAL SMOOTHING OVERVIEW This lecure inroduces ime-series smoohing forecasing mehods. Various models are discussed,

More information

µ r of the ferrite amounts to 1000...4000. It should be noted that the magnetic length of the + δ

µ r of the ferrite amounts to 1000...4000. It should be noted that the magnetic length of the + δ Page 9 Design of Inducors and High Frequency Transformers Inducors sore energy, ransformers ransfer energy. This is he prime difference. The magneic cores are significanly differen for inducors and high

More information

Mathematics in Pharmacokinetics What and Why (A second attempt to make it clearer)

Mathematics in Pharmacokinetics What and Why (A second attempt to make it clearer) Mahemaics in Pharmacokineics Wha and Why (A second aemp o make i clearer) We have used equaions for concenraion () as a funcion of ime (). We will coninue o use hese equaions since he plasma concenraions

More information

The Real Business Cycle paradigm. The RBC model emphasizes supply (technology) disturbances as the main source of

The Real Business Cycle paradigm. The RBC model emphasizes supply (technology) disturbances as the main source of Prof. Harris Dellas Advanced Macroeconomics Winer 2001/01 The Real Business Cycle paradigm The RBC model emphasizes supply (echnology) disurbances as he main source of macroeconomic flucuaions in a world

More information

Analysis of Pricing and Efficiency Control Strategy between Internet Retailer and Conventional Retailer

Analysis of Pricing and Efficiency Control Strategy between Internet Retailer and Conventional Retailer Recen Advances in Business Managemen and Markeing Analysis of Pricing and Efficiency Conrol Sraegy beween Inerne Reailer and Convenional Reailer HYUG RAE CHO 1, SUG MOO BAE and JOG HU PARK 3 Deparmen of

More information

How To Calculate Price Elasiciy Per Capia Per Capi

How To Calculate Price Elasiciy Per Capia Per Capi Price elasiciy of demand for crude oil: esimaes for 23 counries John C.B. Cooper Absrac This paper uses a muliple regression model derived from an adapaion of Nerlove s parial adjusmen model o esimae boh

More information

Detection of DDoS Attack in SIP Environment with Non-parametric CUSUM Sensor

Detection of DDoS Attack in SIP Environment with Non-parametric CUSUM Sensor Deecion of DDoS Aac in SIP Environmen wih Non-parameric CUSUM Sensor Luigi Alcuri Universiy of Palermo Deparmen of Elecrical, Elecronic and Telecommunicaion Engineering luigi.alcuri@i.unipa.i Piero Cassarà

More information

Forecasting, Ordering and Stock- Holding for Erratic Demand

Forecasting, Ordering and Stock- Holding for Erratic Demand ISF 2002 23 rd o 26 h June 2002 Forecasing, Ordering and Sock- Holding for Erraic Demand Andrew Eaves Lancaser Universiy / Andalus Soluions Limied Inroducion Erraic and slow-moving demand Demand classificaion

More information

Task-Execution Scheduling Schemes for Network Measurement and Monitoring

Task-Execution Scheduling Schemes for Network Measurement and Monitoring Task-Execuion Scheduling Schemes for Nework Measuremen and Monioring Zhen Qin, Robero Rojas-Cessa, and Nirwan Ansari Deparmen of Elecrical and Compuer Engineering New Jersey Insiue of Technology Universiy

More information

Price Controls and Banking in Emissions Trading: An Experimental Evaluation

Price Controls and Banking in Emissions Trading: An Experimental Evaluation This version: March 2014 Price Conrols and Banking in Emissions Trading: An Experimenal Evaluaion John K. Sranlund Deparmen of Resource Economics Universiy of Massachuses-Amhers James J. Murphy Deparmen

More information

Market Liquidity and the Impacts of the Computerized Trading System: Evidence from the Stock Exchange of Thailand

Market Liquidity and the Impacts of the Computerized Trading System: Evidence from the Stock Exchange of Thailand 36 Invesmen Managemen and Financial Innovaions, 4/4 Marke Liquidiy and he Impacs of he Compuerized Trading Sysem: Evidence from he Sock Exchange of Thailand Sorasar Sukcharoensin 1, Pariyada Srisopisawa,

More information

STUDY ON THE GRAVIMETRIC MEASUREMENT OF THE SWELLING BEHAVIORS OF POLYMER FILMS

STUDY ON THE GRAVIMETRIC MEASUREMENT OF THE SWELLING BEHAVIORS OF POLYMER FILMS 452 Rev. Adv. Maer. Sci. 33 (2013) 452-458 J. Liu, X.J. Zheng and K.Y. Tang STUDY ON THE GRAVIMETRIC MEASUREMENT OF THE SWELLING BEHAVIORS OF POLYMER FILMS J. Liu, X. J. Zheng and K. Y. Tang College of

More information

ANALYSIS AND COMPARISONS OF SOME SOLUTION CONCEPTS FOR STOCHASTIC PROGRAMMING PROBLEMS

ANALYSIS AND COMPARISONS OF SOME SOLUTION CONCEPTS FOR STOCHASTIC PROGRAMMING PROBLEMS ANALYSIS AND COMPARISONS OF SOME SOLUTION CONCEPTS FOR STOCHASTIC PROGRAMMING PROBLEMS R. Caballero, E. Cerdá, M. M. Muñoz and L. Rey () Deparmen of Applied Economics (Mahemaics), Universiy of Málaga,

More information

OPERATION MANUAL. Indoor unit for air to water heat pump system and options EKHBRD011ABV1 EKHBRD014ABV1 EKHBRD016ABV1

OPERATION MANUAL. Indoor unit for air to water heat pump system and options EKHBRD011ABV1 EKHBRD014ABV1 EKHBRD016ABV1 OPERAION MANUAL Indoor uni for air o waer hea pump sysem and opions EKHBRD011ABV1 EKHBRD014ABV1 EKHBRD016ABV1 EKHBRD011ABY1 EKHBRD014ABY1 EKHBRD016ABY1 EKHBRD011ACV1 EKHBRD014ACV1 EKHBRD016ACV1 EKHBRD011ACY1

More information

The Interaction of Guarantees, Surplus Distribution, and Asset Allocation in With Profit Life Insurance Policies

The Interaction of Guarantees, Surplus Distribution, and Asset Allocation in With Profit Life Insurance Policies 1 The Ineracion of Guaranees, Surplus Disribuion, and Asse Allocaion in Wih Profi Life Insurance Policies Alexander Kling * Insiu für Finanz- und Akuarwissenschafen, Helmholzsr. 22, 89081 Ulm, Germany

More information

UNDERSTANDING THE DEATH BENEFIT SWITCH OPTION IN UNIVERSAL LIFE POLICIES. Nadine Gatzert

UNDERSTANDING THE DEATH BENEFIT SWITCH OPTION IN UNIVERSAL LIFE POLICIES. Nadine Gatzert UNDERSTANDING THE DEATH BENEFIT SWITCH OPTION IN UNIVERSAL LIFE POLICIES Nadine Gazer Conac (has changed since iniial submission): Chair for Insurance Managemen Universiy of Erlangen-Nuremberg Lange Gasse

More information

Supplementary Appendix for Depression Babies: Do Macroeconomic Experiences Affect Risk-Taking?

Supplementary Appendix for Depression Babies: Do Macroeconomic Experiences Affect Risk-Taking? Supplemenary Appendix for Depression Babies: Do Macroeconomic Experiences Affec Risk-Taking? Ulrike Malmendier UC Berkeley and NBER Sefan Nagel Sanford Universiy and NBER Sepember 2009 A. Deails on SCF

More information

Packet-Oriented Communication Protocols for Smart Grid Services over Low-Speed PLC

Packet-Oriented Communication Protocols for Smart Grid Services over Low-Speed PLC Packe-Oriened Communicaion Proocols for Smar Grid Services over Low-Speed PLC Michael Bauer, Wolfgang Plapper, Chong Wang, Klaus Doser Insiue of Indusrial Informaion Technology Universiä Karlsruhe (TH)

More information

Single-machine Scheduling with Periodic Maintenance and both Preemptive and. Non-preemptive jobs in Remanufacturing System 1

Single-machine Scheduling with Periodic Maintenance and both Preemptive and. Non-preemptive jobs in Remanufacturing System 1 Absrac number: 05-0407 Single-machine Scheduling wih Periodic Mainenance and boh Preempive and Non-preempive jobs in Remanufacuring Sysem Liu Biyu hen Weida (School of Economics and Managemen Souheas Universiy

More information

Risk Modelling of Collateralised Lending

Risk Modelling of Collateralised Lending Risk Modelling of Collaeralised Lending Dae: 4-11-2008 Number: 8/18 Inroducion This noe explains how i is possible o handle collaeralised lending wihin Risk Conroller. The approach draws on he faciliies

More information

Journal Of Business & Economics Research September 2005 Volume 3, Number 9

Journal Of Business & Economics Research September 2005 Volume 3, Number 9 Opion Pricing And Mone Carlo Simulaions George M. Jabbour, (Email: jabbour@gwu.edu), George Washingon Universiy Yi-Kang Liu, (yikang@gwu.edu), George Washingon Universiy ABSTRACT The advanage of Mone Carlo

More information

Performance Center Overview. Performance Center Overview 1

Performance Center Overview. Performance Center Overview 1 Performance Cener Overview Performance Cener Overview 1 ODJFS Performance Cener ce Cener New Performance Cener Model Performance Cener Projec Meeings Performance Cener Execuive Meeings Performance Cener

More information

Research on Inventory Sharing and Pricing Strategy of Multichannel Retailer with Channel Preference in Internet Environment

Research on Inventory Sharing and Pricing Strategy of Multichannel Retailer with Channel Preference in Internet Environment Vol. 7, No. 6 (04), pp. 365-374 hp://dx.doi.org/0.457/ijhi.04.7.6.3 Research on Invenory Sharing and Pricing Sraegy of Mulichannel Reailer wih Channel Preference in Inerne Environmen Hanzong Li College

More information

SPEC model selection algorithm for ARCH models: an options pricing evaluation framework

SPEC model selection algorithm for ARCH models: an options pricing evaluation framework Applied Financial Economics Leers, 2008, 4, 419 423 SEC model selecion algorihm for ARCH models: an opions pricing evaluaion framework Savros Degiannakis a, * and Evdokia Xekalaki a,b a Deparmen of Saisics,

More information

Segmentation, Probability of Default and Basel II Capital Measures. for Credit Card Portfolios

Segmentation, Probability of Default and Basel II Capital Measures. for Credit Card Portfolios Segmenaion, Probabiliy of Defaul and Basel II Capial Measures for Credi Card Porfolios Draf: Aug 3, 2007 *Work compleed while a Federal Reserve Bank of Philadelphia Dennis Ash Federal Reserve Bank of Philadelphia

More information

BALANCE OF PAYMENTS. First quarter 2008. Balance of payments

BALANCE OF PAYMENTS. First quarter 2008. Balance of payments BALANCE OF PAYMENTS DATE: 2008-05-30 PUBLISHER: Balance of Paymens and Financial Markes (BFM) Lena Finn + 46 8 506 944 09, lena.finn@scb.se Camilla Bergeling +46 8 506 942 06, camilla.bergeling@scb.se

More information

INTRODUCTION TO FORECASTING

INTRODUCTION TO FORECASTING INTRODUCTION TO FORECASTING INTRODUCTION: Wha is a forecas? Why do managers need o forecas? A forecas is an esimae of uncerain fuure evens (lierally, o "cas forward" by exrapolaing from pas and curren

More information

Relationships between Stock Prices and Accounting Information: A Review of the Residual Income and Ohlson Models. Scott Pirie* and Malcolm Smith**

Relationships between Stock Prices and Accounting Information: A Review of the Residual Income and Ohlson Models. Scott Pirie* and Malcolm Smith** Relaionships beween Sock Prices and Accouning Informaion: A Review of he Residual Income and Ohlson Models Sco Pirie* and Malcolm Smih** * Inernaional Graduae School of Managemen, Universiy of Souh Ausralia

More information

Full-wave rectification, bulk capacitor calculations Chris Basso January 2009

Full-wave rectification, bulk capacitor calculations Chris Basso January 2009 ull-wave recificaion, bulk capacior calculaions Chris Basso January 9 This shor paper shows how o calculae he bulk capacior value based on ripple specificaions and evaluae he rms curren ha crosses i. oal

More information

A Universal Pricing Framework for Guaranteed Minimum Benefits in Variable Annuities *

A Universal Pricing Framework for Guaranteed Minimum Benefits in Variable Annuities * A Universal Pricing Framework for Guaraneed Minimum Benefis in Variable Annuiies * Daniel Bauer Deparmen of Risk Managemen and Insurance, Georgia Sae Universiy 35 Broad Sree, Alana, GA 333, USA Phone:

More information

A Resource Management Strategy to Support VoIP across Ad hoc IEEE 802.11 Networks

A Resource Management Strategy to Support VoIP across Ad hoc IEEE 802.11 Networks A Resource Managemen Sraegy o Suppor VoIP across Ad hoc IEEE 8.11 Neworks Janusz Romanik Radiocommunicaions Deparmen Miliary Communicaions Insiue Zegrze, Poland j.romanik@wil.waw.pl Pior Gajewski, Jacek

More information

Hotel Room Demand Forecasting via Observed Reservation Information

Hotel Room Demand Forecasting via Observed Reservation Information Proceedings of he Asia Pacific Indusrial Engineering & Managemen Sysems Conference 0 V. Kachivichyanuul, H.T. Luong, and R. Piaaso Eds. Hoel Room Demand Forecasing via Observed Reservaion Informaion aragain

More information

WATER MIST FIRE PROTECTION RELIABILITY ANALYSIS

WATER MIST FIRE PROTECTION RELIABILITY ANALYSIS WATER MIST FIRE PROTECTION RELIABILITY ANALYSIS Shuzhen Xu Research Risk and Reliabiliy Area FM Global Norwood, Massachuses 262, USA David Fuller Engineering Sandards FM Global Norwood, Massachuses 262,

More information

MTH6121 Introduction to Mathematical Finance Lesson 5

MTH6121 Introduction to Mathematical Finance Lesson 5 26 MTH6121 Inroducion o Mahemaical Finance Lesson 5 Conens 2.3 Brownian moion wih drif........................... 27 2.4 Geomeric Brownian moion........................... 28 2.5 Convergence of random

More information

Small and Large Trades Around Earnings Announcements: Does Trading Behavior Explain Post-Earnings-Announcement Drift?

Small and Large Trades Around Earnings Announcements: Does Trading Behavior Explain Post-Earnings-Announcement Drift? Small and Large Trades Around Earnings Announcemens: Does Trading Behavior Explain Pos-Earnings-Announcemen Drif? Devin Shanhikumar * Firs Draf: Ocober, 2002 This Version: Augus 19, 2004 Absrac This paper

More information

Stochastic Recruitment: A Limited-Feedback Control Policy for Large Ensemble Systems

Stochastic Recruitment: A Limited-Feedback Control Policy for Large Ensemble Systems Sochasic Recruimen: A Limied-Feedback Conrol Policy for Large Ensemble Sysems Lael Odhner and Harry Asada Absrac This paper is abou sochasic recruimen, a conrol archiecure for cenrally conrolling he ensemble

More information

DDoS Attacks Detection Model and its Application

DDoS Attacks Detection Model and its Application DDoS Aacks Deecion Model and is Applicaion 1, MUHAI LI, 1 MING LI, XIUYING JIANG 1 School of Informaion Science & Technology Eas China Normal Universiy No. 500, Dong-Chuan Road, Shanghai 0041, PR. China

More information

Hedging with Forwards and Futures

Hedging with Forwards and Futures Hedging wih orwards and uures Hedging in mos cases is sraighforward. You plan o buy 10,000 barrels of oil in six monhs and you wish o eliminae he price risk. If you ake he buy-side of a forward/fuures

More information

17 Laplace transform. Solving linear ODE with piecewise continuous right hand sides

17 Laplace transform. Solving linear ODE with piecewise continuous right hand sides 7 Laplace ransform. Solving linear ODE wih piecewise coninuous righ hand sides In his lecure I will show how o apply he Laplace ransform o he ODE Ly = f wih piecewise coninuous f. Definiion. A funcion

More information

Acceleration Lab Teacher s Guide

Acceleration Lab Teacher s Guide Acceleraion Lab Teacher s Guide Objecives:. Use graphs of disance vs. ime and velociy vs. ime o find acceleraion of a oy car.. Observe he relaionship beween he angle of an inclined plane and he acceleraion

More information

What do packet dispersion techniques measure?

What do packet dispersion techniques measure? Wha do packe dispersion echniques measure? Consaninos Dovrolis Parameswaran Ramanahan David Moore Universiy of Wisconsin Universiy of Wisconsin CAIDA dovrolis@ece.wisc.edu parmesh@ece.wisc.edu dmoore@caida.org

More information

Impact of Human Mobility on Opportunistic Forwarding Algorithms

Impact of Human Mobility on Opportunistic Forwarding Algorithms Impac of Human Mobiliy on Opporunisic Forwarding Algorihms Augusin Chainreau, Pan Hui, Jon Crowcrof, Chrisophe Dio, Richard Gass, and James Sco *, Universiy of Cambridge Microsof Research Thomson Research

More information

Why Did the Demand for Cash Decrease Recently in Korea?

Why Did the Demand for Cash Decrease Recently in Korea? Why Did he Demand for Cash Decrease Recenly in Korea? Byoung Hark Yoo Bank of Korea 26. 5 Absrac We explores why cash demand have decreased recenly in Korea. The raio of cash o consumpion fell o 4.7% in

More information

Spectrum-Aware Data Replication in Intermittently Connected Cognitive Radio Networks

Spectrum-Aware Data Replication in Intermittently Connected Cognitive Radio Networks Specrum-Aware Daa Replicaion in Inermienly Conneced Cogniive Radio Neworks Absrac The opening of under-uilized specrum creaes an opporuniy for unlicensed users o achieve subsanial performance improvemen

More information

Markit Excess Return Credit Indices Guide for price based indices

Markit Excess Return Credit Indices Guide for price based indices Marki Excess Reurn Credi Indices Guide for price based indices Sepember 2011 Marki Excess Reurn Credi Indices Guide for price based indices Conens Inroducion...3 Index Calculaion Mehodology...4 Semi-annual

More information

The Impact of Surplus Distribution on the Risk Exposure of With Profit Life Insurance Policies Including Interest Rate Guarantees.

The Impact of Surplus Distribution on the Risk Exposure of With Profit Life Insurance Policies Including Interest Rate Guarantees. The Impac of Surplus Disribuion on he Risk Exposure of Wih Profi Life Insurance Policies Including Ineres Rae Guaranees Alexander Kling 1 Insiu für Finanz- und Akuarwissenschafen, Helmholzsraße 22, 89081

More information

A Note on Using the Svensson procedure to estimate the risk free rate in corporate valuation

A Note on Using the Svensson procedure to estimate the risk free rate in corporate valuation A Noe on Using he Svensson procedure o esimae he risk free rae in corporae valuaion By Sven Arnold, Alexander Lahmann and Bernhard Schwezler Ocober 2011 1. The risk free ineres rae in corporae valuaion

More information

As widely accepted performance measures in supply chain management practice, frequency-based service

As widely accepted performance measures in supply chain management practice, frequency-based service MANUFACTURING & SERVICE OPERATIONS MANAGEMENT Vol. 6, No., Winer 2004, pp. 53 72 issn 523-464 eissn 526-5498 04 060 0053 informs doi 0.287/msom.030.0029 2004 INFORMS On Measuring Supplier Performance Under

More information

4. International Parity Conditions

4. International Parity Conditions 4. Inernaional ariy ondiions 4.1 urchasing ower ariy he urchasing ower ariy ( heory is one of he early heories of exchange rae deerminaion. his heory is based on he concep ha he demand for a counry's currency

More information

Economics Honors Exam 2008 Solutions Question 5

Economics Honors Exam 2008 Solutions Question 5 Economics Honors Exam 2008 Soluions Quesion 5 (a) (2 poins) Oupu can be decomposed as Y = C + I + G. And we can solve for i by subsiuing in equaions given in he quesion, Y = C + I + G = c 0 + c Y D + I

More information

Chapter 4: Exponential and Logarithmic Functions

Chapter 4: Exponential and Logarithmic Functions Chaper 4: Eponenial and Logarihmic Funcions Secion 4.1 Eponenial Funcions... 15 Secion 4. Graphs of Eponenial Funcions... 3 Secion 4.3 Logarihmic Funcions... 4 Secion 4.4 Logarihmic Properies... 53 Secion

More information

How To Predict A Person'S Behavior

How To Predict A Person'S Behavior Informaion Theoreic Approaches for Predicive Models: Resuls and Analysis Monica Dinculescu Supervised by Doina Precup Absrac Learning he inernal represenaion of parially observable environmens has proven

More information

Chapter 8 Student Lecture Notes 8-1

Chapter 8 Student Lecture Notes 8-1 Chaper Suden Lecure Noes - Chaper Goals QM: Business Saisics Chaper Analyzing and Forecasing -Series Daa Afer compleing his chaper, you should be able o: Idenify he componens presen in a ime series Develop

More information

The Impact of Surplus Distribution on the Risk Exposure of With Profit Life Insurance Policies Including Interest Rate Guarantees

The Impact of Surplus Distribution on the Risk Exposure of With Profit Life Insurance Policies Including Interest Rate Guarantees 1 The Impac of Surplus Disribuion on he Risk Exposure of Wih Profi Life Insurance Policies Including Ineres Rae Guaranees Alexander Kling Insiu für Finanz- und Akuarwissenschafen, Helmholzsraße 22, 89081

More information

Nikkei Stock Average Volatility Index Real-time Version Index Guidebook

Nikkei Stock Average Volatility Index Real-time Version Index Guidebook Nikkei Sock Average Volailiy Index Real-ime Version Index Guidebook Nikkei Inc. Wih he modificaion of he mehodology of he Nikkei Sock Average Volailiy Index as Nikkei Inc. (Nikkei) sars calculaing and

More information

Investigating Effects of Drug Therapy for HIV Infection by Double Compartments Cellular Automata Simulations

Investigating Effects of Drug Therapy for HIV Infection by Double Compartments Cellular Automata Simulations Recen Researches in Modern Medicine Invesigaing Effecs of Drug Therapy for HIV Infecion by Double Comparmens Cellular Auomaa Simulaions SOMPOP MOONCHAI,3 YONGWIMON ENURY,3 * Deparmen of Mahemaics, Faculy

More information

A Re-examination of the Joint Mortality Functions

A Re-examination of the Joint Mortality Functions Norh merican cuarial Journal Volume 6, Number 1, p.166-170 (2002) Re-eaminaion of he Join Morali Funcions bsrac. Heekung Youn, rkad Shemakin, Edwin Herman Universi of S. Thomas, Sain Paul, MN, US Morali

More information

Idealistic characteristics of Islamic Azad University masters - Islamshahr Branch from Students Perspective

Idealistic characteristics of Islamic Azad University masters - Islamshahr Branch from Students Perspective Available online a www.pelagiaresearchlibrary.com European Journal Experimenal Biology, 202, 2 (5):88789 ISSN: 2248 925 CODEN (USA): EJEBAU Idealisic characerisics Islamic Azad Universiy masers Islamshahr

More information

Heuristics for dimensioning large-scale MPLS networks

Heuristics for dimensioning large-scale MPLS networks Heurisics for dimensioning large-scale MPLS newors Carlos Borges 1, Amaro de Sousa 1, Rui Valadas 1 Depar. of Elecronics and Telecommunicaions Universiy of Aveiro, Insiue of Telecommunicaions pole of Aveiro

More information

Mobile and Ubiquitous Compu3ng. Mul3plexing for wireless. George Roussos. g.roussos@dcs.bbk.ac.uk

Mobile and Ubiquitous Compu3ng. Mul3plexing for wireless. George Roussos. g.roussos@dcs.bbk.ac.uk Mobile and Ubiquious Compu3ng Mul3plexing for wireless George Roussos g.roussos@dcs.bbk.ac.uk Overview Sharing he wireless (mul3plexing) in space by frequency in 3me by code PuEng i all ogeher: cellular

More information

Chapter 6: Business Valuation (Income Approach)

Chapter 6: Business Valuation (Income Approach) Chaper 6: Business Valuaion (Income Approach) Cash flow deerminaion is one of he mos criical elemens o a business valuaion. Everyhing may be secondary. If cash flow is high, hen he value is high; if he

More information

Factors Affecting Initial Enrollment Intensity: Part-Time versus Full-Time Enrollment

Factors Affecting Initial Enrollment Intensity: Part-Time versus Full-Time Enrollment acors Affecing Iniial Enrollmen Inensiy: ar-time versus ull-time Enrollmen By Leslie S. Sraon Associae rofessor Dennis M. O Toole Associae rofessor James N. Wezel rofessor Deparmen of Economics Virginia

More information

The Greek financial crisis: growing imbalances and sovereign spreads. Heather D. Gibson, Stephan G. Hall and George S. Tavlas

The Greek financial crisis: growing imbalances and sovereign spreads. Heather D. Gibson, Stephan G. Hall and George S. Tavlas The Greek financial crisis: growing imbalances and sovereign spreads Heaher D. Gibson, Sephan G. Hall and George S. Tavlas The enry The enry of Greece ino he Eurozone in 2001 produced a dividend in he

More information