The Response Time Analysis of Digital Broadcasting System The Response Time Analysis of Digital Broadcasting System

The Response Time Analysis of Digital Broadcasting System

  • 期刊名字:武汉大学学报(英文版)
  • 文件大小:179kb
  • 论文作者:WANG Shuo,ZHANG Jiang-ling,FEN
  • 作者单位:School of Computer Science and Technology
  • 更新时间:2020-12-06
  • 下载次数:
论文简介

Vol.10 No.3 2005 515-519WUJNSWuhan University Journal of Natural SciencesArticle ID: 1007- 1202(2005)03-0515-05The Response Time Analysis of DigitalBroadcasting System0 Introduction口WANG Shuo, ZHANG Jiang- ling,FENG Danroadcasting system is a core of television station’s digit-School of Computer Science and Technology,HuazhongB! al developments, having become a highlight of researchUniversity of Science and Technology, Wuhan 430074,Hubei, Chinaon computer application[.2]. It should meet the real-time keyperformance demand. The accidentally change of playbill existsAbstract: Digital broadcasting system has become a high-in the process of program broadcasting. The maximum r(light of research on computer application. To respond to thesponse time of the system when the playbill is changed mustchanges of the playbill in the broadcasting system in real .be gotten, so that users can modify playbills before the time.time, the response time of the system must be studied. ThereIn addition, having a certain response time, users can easilyis scarcely the research on this area currently. The influencefactors in the response time are analyzed; the model on thecheck whether the operations are proper. At the same time,response time of the system service is built; how the influencethe analyses on the response time benefits designer to optimizefactors affect the response time of the system service is valida-ted; and four improvement measures are proposed to minimizethe system functions and find out the strategy that can respondthe response time of system service.to changes of the playbill in real time. So it is very necessaryKey words: digital broadcasting system; response time; .to study the response time of the digital broadcasting systemservice.CLC number: TP 333However, there is scarcely the research on this area cur-rently [3-10]. The response time of the digital broadcasting sys-tem service will be studied from four aspects: its influence fac-tors, its model, how the influence factors affect the responsetime of the system service and the improvement measures.1 The Factors Influencing the ResponseTime of System ServiceKnown from the IO route in Fig. 1, the response time ofbroadcasting system service includes three parts. The first,Received date: 2004- 03-06the time to中国煤化Igue workstation to net-Foundation item: Supported by the National Natural Science Foun-dation of China( 60273074)work managYCNMHGS T. The second, theBiography: WANG Shuo(1971-), female, Ph. D candidate, researchtime to transter playbill trom network management worksta-direction: massive storage technology and digital broadcasting technol-ogies. E mail: ws_ beronws@ hotmail. comtion to the gigabit net card of storage and broadcasting server,515Wuhan University, Joumal of Natural Sciences Vol.10 No.3 2005Users|od on which disk is written.If adopting write back strategy, Vw in this system is .|Cataloguing workstation160 MB/s. If adopting writethrough strategy, Vw is215-315 Mb/s. Tw and T; will be various with the differ-↓playbillent writing strategy.Network manage workstationDue that CPU processes modifying and creating pro-plsyillgram chain. When there are n pieces of playbill, T。willStorage and broadcasting serverbe O(n).Reading the playbill and setting up the program| Broadcasting|chain, the server will send out read request according tothe playbill. Ta is related with disk transerring rate U,Fig. 1 The IO route in broadeasing systemsize of data block Bs and the position of disk head. It in-believe this time as T2. The third, the time that storagecludes T。 (the seek time), T,(the rotalodelay time), andand broadcasting server accepts and responds the play-the delivering data time T. Among of them, U can bebill,assume this time as Ts、gotten by disk parameters. Bg is the size of a frame mediaTl and T2 relate to Vq and B. Vq is the deliveringstream equaling with 0.924 42 MB. Assume the numberspeed of gigabit nets and B is the size of playbill. Ts con-of disk in RAIDO as N, the size of data block in RAIDOsistsof T,To,Tg,T。and Tp. T; is the receive timeas B. T, and Tr are random and have something to dothat the server receives the playbill from its net cardswith the position of disk head, so when reading a framememory to its disk, Te is the time modification and crea-of media stream, Ts isting program chain, T is the response time of RAID, Teis the decoder time for program data, and Tp is the pla-Ts =B.N(T.+ T, + B:/u)(1)ying time for a frame of media stream data.The media data is sent to decoder buffer to decode1.1 The Influence Factors in T and T2In a real environment, Vq of gigabit net cards is lessafter it is fetched the host memory from RAID, here thehan 1 000 Mb/s. The reason is that the network speed istransmission time that transferring data from memory torelative with the number of processes, TCP window size,memory could be neglected. T。 is related with the size offrame B。and the decoding speed Ve. The size of decodingand I0 bus and hard disk of the measured server [3-8] .This broadcasting system uses 10 pieces of 3C985B-frame is a frame of media data, and the decoding speedSX net cards, which are high performance server netscan be acquired from the decoding card parameter. This .card just for gigabit Ethernet environment. In this sys-system uses Targa 3000 as the decoding card; its decodingtem, the 3C985B-SX-performance is that the throughputspeed is 10-50 Mb/s.Program in this system is in the type of PAL( phasefrom client to server comes to 942. 204 Mb/s; the CPUutilization is 52. 917%. And the throughput from serverAlteration Line), so broadcasting velocity should be 25frame/s, then Tp is 40 ms.to client comes to 940. 655 Mb/s; CPU utilization is46. 333%.The playbill's size also influences the respond time2 Mathematics Model of Responseof system. The longer the playbill is, the longer the de-Time of Broadcasting System Servicelivers time of network is, and the longer the responsetime is too. In this system, a piece of playbill totally oc-Known from the IO path in the broadcasting systemcupies 36 B. So n pieces of playbills demand Fibre cardand influence factors in the system response time, it isnet delivering B= 36n B.gotten:1.2 The Influence Factors in T;Tr = T+T,+ T3Suppose the receive time that the net management中国煤化工the fitering time Tm.workstation receives the playbill from the net card memo-and:MYTHC N M H Ge playbill to the gagabitry to its disk as Tw. Tw and Tr are related with the writ-net card of storage and broadcasting server. T3 includesing speed of disk V.[9,10]. Vw is relevant with the meth-T,T, Ta, Te, and Tp. Ti=Tn=B/vq. Tite is the516WANG Shuo et al: The Response Time Analysis of Digital ...time that net management workstation filters the ilgalplaybill according to the program database. Tit belongs3 System Emulation and lmprovementto microsecond class and can be neglected because the fil-Measures to Minimize the Responsetering is handled by CPU.TimeT; is similar with Tw. Assume the disk writingspeed of storage and broadcasting server is equal to thatWe emulates the response time of system in theof net management workstation, thenthree circumstances with Matlab to investigate how theT= T+ T;+ B/Vwinflunce factors affect the response time. Shown as theSupposing the number of playbill as n,quantityFig. 2, with different B, and N, in the best and averageclass of T. is O(n). using this cofficient“a”to expresscircumstances, n influences the response time of the sys-0( n), thenem very lttle; In the worst circumstance, when nT.=an1 000, the response time of the system increases slowlyT。= B。/ve(5)with the augment of n.From Eqs. (1) and (3)-(5), Eq. (2) turns into:TrR = 2B/vq + 2B/vw +2(T.+ T,)+ an1800the average circumnstance1400+B.(T.+ T, +Br/u)+B./r.+T。 (6)留11000B.NIn Eq. (6), B= 36n, because T。belongs to microsecond600the worst circumstancethe best cireumstance Iclass, we suppose a=2. 0X10-6s, and in the system200Bs=B。=0.924 42 MB. The performance parameter of50010001 5002000 2500system is shown in Table 1.(a) B:=32kB, N=6Table 1 Performance parameter of systemthe average circumstanceu/V./v:/T./ms T/msMb. s-MB.s-Mb. s-1E 1000Max9423159.65 5. 982350.0the worst circurmstanceMin0.75 0. 0095.0the best circumstanceAverage 612_2655.20 2.992127. 51000 1500 2000 2500From Eq. (6) and Table 2 we can get: .(b) B_=I6kB, N=6In the worst circumstance (when Vq,Vw, U and ve1800 |are the min, and T。,Tr are the max):Tr/ms=6.72X10-3Xn; 1000| 924.42(B:/19+ 15.63) + 1 550.33(7)x(B,N)In the best circumstance (whenv,Vw,U and Ve arethe max,and Ts,T, are the min):Tr/ms=4.44X10-3 Xn(C) B,=32kB, N=9「924.42(Br/23 +0.75)+ 189.4 (8)Fig. 2 The response time curve with various nTherefore take n= 1 000 for calculation conven-In the average circumstance (when Vg,Uw,U,v,ience. To investigate various B, how to affect the re-T。and T: is the average):sponse time of svstem at the different N, we take respec-Tr/ms=5.11X 10~3 Xntively中国煤化工3. it can be gtten that「924.42]the respMYHCNMHGesquicklywhenB,en-| (B.N)(B3/21+8.19)+ 325.3 (9)larges from 0 to 32 kB, the response time of system tendsin saturation when B, increases from 32 kB to the bigger.517Wuhan University, Joumal of Natural Sciences Vol.10 No.3 2005And at the best circumstance, the variety of B, afects the2500response time very little.2000Take B,=32 kB, n=1 000, and Bg=32 kB, n=the worst circumstance旨15002000 to investigate the variety of N how to influnce the1000 F the best circumstanceresponse time of system. From Fig. 4, it can be seen that500N=6 is the saturation point of the system response time,the system response time reduces sharp when N enlarges17 20Nfrom 2 to 6,the system response time changes little(a) =1000when N increases again from 6. In sum,B, = 32 kB, .N= 6 can make the system response time minimum.1800自15001400卜1000上the best circumstance自10000600the best circumstancethe average circumstance7 20V00 t(b) n=200020406080100120140B,/kBFig. 4 The response time curve with various N when B, =32 kB(a) N=6time,the length of playbill is not over 3 000. if usersprepare to catalogue one week programs, the length ofplaybill is not over 7 000.1000 |From Fig. 5, we can acquire that when the length ofthe. average circumstance.playbill is respectly 1 000, 3 000 and 7 000, in the worst200circumstance, all response time are not over 2 s; in thegeneral circumstance, all response time are not over 500ms; in the best circumstance, the response time are not(b) N=9over 250 ms.1000_the best circumnstance自1000. the average cireumstanee .(c) N=160.20.40.0.81.0n/ 10*Fig.3 The response time curve with various B, when n= 1000Fig.5 The response time curve with various n when B, =32 kB,N=6Fig. 4 shows the changes of the response time whenn increases from 1 000 to 10 000, B:=32 kB, N=6.Known from Fig.2 to Fig. 5, and Eq. (6), to reduceExcept advertisement program and weather forecast pro-the system response time, the following measures can begram, most of programs would be over 5 min long. Sup-adopted:pose each program is 5 min long, there will be 288 pieces1) Set the decoding speed into maximum in the pro-of playbill one day if television station broadcasts 24 hcedure code.programs per day. If there is a advertisement in every中国煤化工nd N in the driver soft-two minutes program, the length of an one day playbillware olMYHCNMHGwill be 864. So we regard 1 000 pieces of programs as ut-3) Store the media data in sequence.most. If users prepare to catalogue 3 day programs every4) Adopt disk write-back strategy and FC gigabit518WANG Shuo et al: The Response Time Analsis of Digital ..net card. Thus, the system response time can reach the[3] Cruz R L. A Caleculus for Network Delay, II: Network Anal-ysis. Information Theory IEEE Transactions on, 1991,37minimum,(1):132- 141.4] Cruz R L. A Calculus for Network Delay I. Network EI4 Conclusionments in Isolation. Information Theory IEEE Transactionson, 1991, 37 (1):114-131.In this paper, the influence factors in the digital[5] Cidon I, Gopal I, Grover G, et al. Real- time Packet Swite-hing: a Performance Analysis. Selected Areas in Communi-broadcasting system service are analyzed, and the modelcations. IEEE Journal on,1988 ,6(9):1576- 1586.on the response time of the system service is established.[6] PhilpI R. LiuJ. End to End Scheduling in Real- time Pack-et switched Networks. Network Protocols, Proceedings,According to the theories analyzing, the optimum num-1996 International Conference on, 1996,1:23-30.ber of disks and the adequate number of blocks in RAID7] Song M, Chang N, Shin H. A New Queue Discipline forof the system are gotten. And four improvement meas-Various Delay and Jitter Requirements in Real-time Packet-ures are proposed to minimize the response time of systemswitched Networks. Real Time Com puting Systems and Ap-plications , 2000. Proceedings Seventh Internatiomal Confer-service. The model supplies the theory gist of designing aence on, 2000,2:191-198.digital broacasting system.[8] Liu Fong-hao, Yang C C, Lee Weir Tsong. The Building,Analyzing and Simulating of a Multimedia Network Based onReferencesCATV System. Broadcasting, IEEE Transactions on,2000, ,46(1):79-87.9] Feng Dan, Zhang Jiang-ling. I/ O Response Time of RAID[1] Usui K, Kai K,Ohya A, et al. Multimedia Services in thein Different Load. Journal of Computer Research & Derel-HDTV MUSE System. Broadcasing IEEE Transactionsopment, 2001, 38(9):1144-1148.1, 1996,42(3):194-199.[10] Chen Qiong, Zhang Jiang ling. Analysis of I/O Service Time[2] Kawai N, Namba S, Yamazaki S. Performance of Multime-in High Performance Disk Array. Mini Mirco System, 2000,dia Broadcasting Through ISDB Transmission System.21(3) :235-237.Bradcasting IEEE Transactions on, 1996, 42(3):151-158.中国煤化工MHCNMHG519Wuhan University Journal of Natural Sciences Vol.10 No.3 2005

论文截图
版权:如无特殊注明,文章转载自网络,侵权请联系cnmhg168#163.com删除!文件均为网友上传,仅供研究和学习使用,务必24小时内删除。