A BROADCASTING PROTOCOL FOR COMPRESSED VIDEO

Similar documents
Cost Control of the Transmission Congestion Management in Electricity Systems Based on Ant Colony Algorithm

Handout #5. Introduction to the Design of Experiments (DOX) (Reading: FCDAE, Chapter 1~3)

Combining Pay-Per-View and Video-on-Demand Services

An Interactive Broadcasting Protocol for Video-on-Demand

A Dynamic Heuristic Broadcasting Protocol for Video-on-Demand

Comparative Study of Word Alignment Heuristics and Phrase-Based SMT

Efficient Broadcasting Protocols for Video on Demand

Use the template below as a guide for organizing the text of your story.

A Realistic E-Learning System based on Mixed Reality

A Genetic Programming Framework for Error Recovery in Robotic Assembly Systems

Improving Bandwidth Efficiency on Video-on-Demand Servers y

Music Performer Recognition Using an Ensemble of Simple Classifiers

Real-time Scheduling of Flexible Manufacturing Systems using Support Vector Machines and Neural Networks

Exploiting the Marginal Profits of Constraints with Evolutionary Multi-objective Optimization Techniques

Scheme For Finding The Next Term Of A Sequence Based On Evolution {File Closing Version 4}. ISSN

11 Hybrid Cables. n f Hz. kva i P. Hybrid Cables Description INFORMATION Description

Improving Server Broadcast Efficiency through Better Utilization of Client Receiving Bandwidth

Heterogeneous Talent and Optimal Emigration 1

An Efficient Implementation of Interactive Video-on-Demand

Simple VBR Harmonic Broadcasting (SVHB)

A variable bandwidth broadcasting protocol for video-on-demand

Trace Adaptive Fragmentation for Periodic Broadcast of VBR Video

Improving Video-on-Demand Server Efficiency Through Stream Tapping

Logistics We are here. If you cannot login to MarkUs, me your UTORID and name.

Lossless VBR Video Broadcasting with User Bandwidth Limit using Uniform Channels

Tabbycat: an Inexpensive Scalable Server for Video-on-Demand

Rank Inclusion in Criteria Hierarchies

EE260: Digital Design, Spring /3/18. n Combinational Logic: n Output depends only on current input. n Require cascading of many structures

Video-on-demand broadcasting protocols. Jukka Leveelahti Tik Multimedia Communications

1. Introduction. SPIE/ACM MMCN2003, Santa Clara, CA, Jan An Efficient VOD Broadcasting Scheme with User Bandwidth Limit

Instructions for Contributors to the International Journal of Microwave and Wireless Technologies

Image Intensifier Reference Manual

Video-on-Demand Broadcasting Protocols: A Comprehensive Study

16.5 Media-on-Demand (MOD)

A Lossless VOD Broadcasting Scheme for VBR Videos Using Available Channel Bandwidths

A Proactive Implementation of Interactive Video-on-Demand

Recognizing Names in Biomedical Texts using Hidden Markov Model and SVM plus Sigmoid

Providing VCR Functionality in Staggered Video Broadcasting

Simon Sheu Computer Science National Tsing Hua Universtity Taiwan, ROC

Seamless Workload Adaptive Broadcast

Mullard INDUCTOR POT CORE EQUIVALENTS LIST. Mullard Limited, Mullard House, Torrington Place, London Wel 7HD. Telephone:

Positive-living skills for children aged 3 to 6

Supply Quantitative Model à la Leontief *

Polychrome Devices Reference Manual

T-25e, T-39 & T-66. G657 fibres and how to splice them. TA036DO th June 2011

AREA (SQ. FT.) BREAKDOWN: 1. SALES AREA: 2. ENTRY VESTIBULE (EXT.): 3. SERVICE: 4. TOILET ROOM: 5. OFFICE: 6. STAIRWAY/REAR EXIT: 7.

Video Cassette Recorder

US B2. ( *) Notice: Subject to any disclaimer, the term of this patent is extended or adjusted under 35 U.S.c. 154(b) by 0 days.

S Micro--Strip Tool in. S Combination Strip Tool ( ) S Cable Holder Assembly (Used only

Romeo And Juliet II. Paris. Lord Capulet. Tybalt Benvolio. Juliet. Romeo. Mercutio. Juliet s Nurse. Shakespeare On Stage

Read Only Memory (ROM)

Pattern Smoothing for Compressed Video Transmission

T541 Flat Panel Monitor User Guide ENGLISH

A. Flue Pipes. 2. Open Pipes. = n. Musical Instruments. Instruments. A. Flue Pipes B. Flutes C. Reeds D. References

tj tj D... '4,... ::=~--lj c;;j _ ASPA: Automatic speech-pause analyzer* t> ,. "",. : : :::: :1'NTmAC' I

THE IMPORTANCE OF ARM-SWING DURING FORWARD DIVE AND REVERSE DIVE ON SPRINGBOARD

PROBABILITY AND STATISTICS Vol. I - Ergodic Properties of Stationary, Markov, and Regenerative Processes - Karl Grill

For children aged 5 7

RIAM Local Centre Woodwind, Brass & Percussion Syllabus

Color Monitor. L200p. English. User s Guide

Technical Information

DIGITAL DISPLAY SOLUTION REAL ESTATE POINTS OF SALE (POS)

A Quantization-Friendly Separable Convolution for MobileNets

Our competitive advantages : Solutions for X ray Tubes. X ray emitters. Long lifetime dispensers cathodes n. Electron gun manufacturing capability n

BesTrans AOC (Active Optical Cable) Spec and Manual

current activity shows on the top right corner in green. The steps appear in yellow

Academic Standards and Calendar Committee Report # : Proposed Academic Calendars , and

Craig Webre, Sheriff Personnel Division/Law Enforcement Complex 1300 Lynn Street Thibodaux, Louisiana 70301

Movies are great! Within a passage, words or phrases can give clues to the meaning of other words. This

Chapter 7 Registers and Register Transfers

Conettix D6600/D6100IPv6 Communications Receiver/Gateway Quick Start

Content storage architectures

THE HIGH-BANDWIDTH requirements and long-lived

An optimal broadcasting protocol for mobile video-on-demand

DIGITAL SYSTEM DESIGN

Relative frequency. I Frames P Frames B Frames No. of cells

Correcting Image Placement Errors Using Registration Control (RegC ) Technology In The Photomask Periphery

Sealed Circular LC Connector System Plug

Production of Natural Penicillins by Strains of Penicillium chrysogenutn

QUICK START GUIDE v0.98

Cost-Aware Fronthaul Rate Allocation to Maximize Benefit of Multi-User Reception in C-RAN

Working with PlasmaWipe Effects

Organic Macromolecules and the Genetic Code A cell is mostly water.

PROJECTOR SFX SUFA-X. Properties. Specifications. Application. Tel

The UCD community has made this article openly available. Please share how this access benefits you. Your story matters!

Why Take Notes? Use the Whiteboard Capture System

Hybrid Transcoding for QoS Adaptive Video-on-Demand Services

Motivation. Analysis-and-manipulation approach to pitch and duration of musical instrument sounds without distorting timbral characteristics

THE Internet of Things (IoT) is likely to be incorporated

Energy and Exergy Analysis for Single and Parallel Flow Double Effect Water-Lithium Bromide Vapor Absorption Systems

Voice Security Selection Guide

3. Sequential Logic 1

NIIT Logotype YOU MUST NEVER CREATE A NIIT LOGOTYPE THROUGH ANY SOFTWARE OR COMPUTER. THIS LOGO HAS BEEN DRAWN SPECIALLY.

To Bean or not to bean! by Uwe Rosenberg, with illustrations by Björn Pertoft Players: 2 7 Ages: 10 and up Duration: approx.

Feasibility Study of Stochastic Streaming with 4K UHD Video Traces

Energy-Efficient FPGA-Based Parallel Quasi-Stochastic Computing

A STUDY OF TRUMPET ENVELOPES

References and quotations

Appendix A. Quarter-Tone Note Names

PowerStrip Automatic Cut & Strip Machine

Transcription:

Proceegs of the EUROMEDIA 99 Coferece (Much, Aprl 6-8, 1999), pp78-84 A BROADCASTING PROTOCOL FOR COMPRESSED VIDEO Jeha-Fraços Pârs 1 Departmet of Computer Scece Uversty of Housto Housto, TX 7704-475 pars@csuheu Keywors: veo o ema, veo roacastg, MPEG, compresse veo ABSTRACT Veo roacastg protocols ca mprove the effcecy of veo o ema servces y reucg the awth requre to strute the veos that are smultaeously watche y may vewers Whle may roacastg protocols have ee propose, oe of them takes to accout the fact that the server wll roacast compresse veos whose awth requremets epe o the rate of chage of the splaye mages We preset here a ew roacastg protocol specfcally talore for the task of roacastg compresse veo Our protocol cotuously roacasts the frst segmet of each veo real tme It the roacasts the remag segmets of the veo at frequeces esurg that each segmet ca e completely owloae y the tme the customer has fshe watchg the prevous segmet Hece these segmets o ot requre excess awth 1 INTRODUCTION Despte all the attractveess of the cocept, veo o ema (Wog 1988) has yet to succee o the marketplace Oe of the explaatos for ths lack of success s the fact that veo o ema has to compete wth cheaper, well-estalshe rvals such as veo retal a pay-pervew Veo roacastg s oe of the several approaches that have ee propose for reucg the cost of veo o ema It s ot a paacea sce t oly apples to veos that are lkely to e watche at the same tme y may vewers Rather tha watg for vual requests, veo roacastg cotuously trasmts these veos over several ata streams such a way that o customer wll have to wat more tha a few mutes efore eg ale to start vewg the veo The potetal savgs are qute mpressve as t has ee estmate that at least 40 percet of the vewers wll e orerg the same 10 to 0 popular veos (Da et al 1994; Da et al 1996) All the veo roacastg protocols that have ee evelope so far assume that the veos wll have a fxe awth correspog to a fxe veo cosumpto rate Ths assumpto s ot correct ecause the server wll roacast compresse veos whose awth requremets epe o the rate at whch the mages eg splaye chage (Garrett a Wllger 1994; Bera et al 95) For stace, aytme acto scees a cartoos wll requre more awth tha slower movg scees a ght scees To esure jtter-free elvery of veo a system allocatg a fxe awth to each veo, we woul thus have to set the veo roacastg awth to the maxmum t rate requre y the fastest momets of the fastest pace scees of the veo As a result, a sgfcat fracto of the awth coul rema uuse most of the tme We propose a ew veo roacastg protocol that avos ths rawack Our Varale Bawth Harmoc Broacastg protocol (VBHB) ves each veo to e roacast to fxe sze segmets Oly the frst segmet of each veo s roacast real tme; the other segmets are reroacast at a suffcet frequecy to esure that they ca e completely owloae y the set-top ox y the tme the customer has fshe watchg the prevous segmet of the veo Hece these segmets o ot requre 1 Most of ths work was performe whle the author was vstg the Laoratoro Nacoal e Iformátca Avazaa Xalapa, Veracruz, Méxco 78

excess awth As a result, excess awth oly occurs the frst segmet of each veo The remaer of the paper s orgaze as follows Secto revews relevat roacastg protocols Secto scusses varale awth roacastg protocols a trouces our VBHB protocol whle Secto 4 compares the awth requremets of the VBHB protocol wth those of other roacastg protocols Fally Secto 5 cotas our coclusos Stream 1: Stream : Stream : S S,1 S, S,1 S, S VIDEO BROADCASTING PROTOCOLS The smplest roacastg protocol s staggere roacastg (Da et al 1994, Almeroth a Ammar 96) Staggere roacastg cotuously retrasmts each veo over k stct streams at equal tme tervals The oly rawack of the approach s the farly large umer of ata streams t requres to guaratee a reasoale watg tme Coser, for stace, a veo lastg two hours, whch happes to e farly close to the average urato of a feature move Guarateeg a maxmum watg tme of te mutes woul requre startg a ew roacast of the veo every te mutes a a total of twelve streams Much more effcet protocols have ee propose, amog whch are Vswaatha a Imelsk s pyram roacastg protocol (Vswaatha a Imelsk 96), Aggarwal, Wolf a Yu s permutato-ase pyram roacastg protocol (Aggarwal et al 1996), Hua a Sheu s skyscraper roacastg protocol (Hua a Sheu 1997), Juh a Tseg s harmoc roacastg protocol (Juh a Tseg 1997) a ts varats (Pârs et al 1998a; Pârs et al 1998) All these protocols ve each veo to segmets that are smultaeously roacast o fferet ata streams Oe of these streams trasmts othg ut the frst segmet of the veo The other streams trasmt the remag segmets at lower awths Whe customers wat to watch a veo, they frst wat for the egg of the frst segmet o the frst stream Whle they start watchg that segmet, ther set-top ox (STB) also starts owloag ata from the other streams Each vual roacastg protocol orgazes these ata a fferet fasho ut all guaratee that the STB wll e ale to play each segmet of the veo tur The oly rawack of the approach s the fact that the set-top ox must have eough local storage to store up to 40 percet of the veo I the curret state of storage techology, ths mples that the STB must have a local sk S,1 S, S, S,1 Fgure 1: The frst three streams of a veo uer harmoc roacastg The most promsg roacastg protocols are the protocols ase o the harmoc roacastg protocol (Juh a Tseg 1997; Pârs et al 1998a) ecause o other protocol ca elver the same maxmum watg tme wthout requrg much more awth Harmoc roacastg (HB) ves each veo to segmets of equal urato = D/ where D s the total urato of the veo It repeately roacasts each segmet S, for 1, o a separate ata stream wth a awth /, where s the cosumpto rate of the veo expresse, say, ts/s (see Fgure 1) Whe a customer requests a veo, t wats for the start of a stace of the frst segmet a the starts owloag ata from every stream of the veo That meas that the clet a the server must e ale to support a awth of B HB ( ) = = H( ) = 1 where H( ) s the th harmoc umer Sce the frst segmet s roacast at a awth equal to the veo cosumpto rate, the maxmum amout of tme customers wll have to wat efore vewg a veo s gve y the urato = D/ of that frst segmet Ufortuately HB caot always elver all ata o tme To uersta that, let us efe frst a susegmet as the fracto of a segmet the clet receves urg tme uts The frst segmet oly has oe susegmet, the segmet tself a every other segmet S has equal susegmets, S, K, 1 S Coser the the frst two streams 79

Fgure 1 If the clet makes ts request tme to receve the seco stace of a starts recevg ata at tme t 0, t wll ee all of the ata for S 1 y tme t0 + / However, t wll ot receve all of that ata utl tme t0 + As t turs out, HB wll ot work uless the clet always wats extra tme uts efore cosumg the ata Several varats o HB o ot mpose ths extra watg tme Cautous Harmoc Broacastg (CHB) uses 1 streams a roacasts the frst segmet of the veo o ts frst stream a smlar fasho as HB Its seco stream alterates etwee roacastg S a S at awth The the remag streams roacast segmets S 4 to S a maer such that the -th stream trasmts segmet S +1 at a awth / ( 1 ) As efore, the clet wll start owloag ata from all streams whe t starts segmet Hece the total awth requre y the CHB protocol wll e gve y 1 BCHB( ) = + H ( ) 1 = + 1 = that s, roughly / more tha the orgal HB protocol Aother harmoc protocol, Polyharmoc Broacastg (PHB) (Pârs et al 1998), mposes the same fxe elay to all customers watg to watch a gve veo Sce ths elay s the same for all customers, the protocol ca take avatage of t to reuce the trasmssos of all segmets, clug the frst oe As a result, PHB requres more ata streams ut less total awth tha CHB to acheve the same maxmum watg tme VARIABLE BANDWIDTH PROTOCOLS We wll coser veos of urato D a characterze each veo y ts average cosumpto rate, say, ts per seco a y the mmum trasmsso rate m that ca guaratee -tme elvery of all frames, always expresse the same uts as Note that m wll ot e ecessary equal to the effectve awth of the fastest pace momets of the veo max ecause the STB coul have receve ahea some ata from the fastest pace scees Ths wll typcally happe f the prevous scees of the veo requre a effectve awth feror to the actual trasmsso rate Hece, we wll always have m max The rato β =m / wll e always greater tha or equal to oe It represets the awth overhea resultg from the varatos the effectve cosumpto rate We wll call t the awth overhea coeffcet (BOC) of the veo Broacastg protocols that requre the STB to owloa large parts of the veo whle these parts are vewe y the customer are poorly sute to compresse veo ecause these parts wll have to e trasmtte at a hgher awth to esure jtter-free elvery of the veo Coversely, protocols that owloa most parts of the veo ahea of tme wll perform etter ecause they o ot have to guaratee o-tme elvery of each vual frame The CHB protocol satsfes these requremets ecause the frst two segmets of the veo are the oly oes that ca s e watche whle they are trasmtte We wll frst see how t ca e talore to accommoate compresse veos a later erve from t a eve etter roacastg protocol To evaluate CHB awth requremets for compresse veo, we ee to characterze each segmet of the veo y ts average cosumpto rate a the mmum trasmsso rate m that ca guaratee flcker-free elvery of all frames We wll always have m > We ca ow specfy the awths of all ata streams the followg fasho: a) stream 1 wll trasmt segmet at awth 1m that s the lowest awth guarateeg o tme elvery of all frames of that segmet; ) stream wll alterate etwee trasmttg segmets S a S at awth equal to max( m, ) that s the maxmum of the lowest awth guarateeg o tme elvery of all frames of segmet S a the average cosumpto rate of segmet S (we o ot have to worry aout elverg o-tme the frames of S sce the clet mache ca always owloa t ahea of tme); c) all other segmets of the veo wll e trasmtte y streams to -1 a stream wth 1 wll trasmt segmet S +1 at awth The total awth requre y the CHB protocol wll the e gve y 80

B ( ) = + max(, ) + CHB 1m m 1 = The Varale Bawth Harmoc Broacastg protocol (VBHB) we propose here mproves two ways upo the CHB protocol Frst, the VBHB protocol roacasts segmets S a S o separate ata streams at suffcet awths to guaratee that these two segmets wll e always owloae ahea of tme Seco, t uses all the awth avalale the frst stream to crease the sze of a reuce that of S Stream 1: Stream : Stream : S S S, S S full awth 1m awth =, eff awth = / S,1 S, S,1 S, Oserve that the frst segmet of the veo ees to e roacast every tme uts to mata a maxmum watg tme equal to tme uts It ees also to e roacast at a mmum awth 1m to guaratee o tme elvery of all frames The excess awth 1m 1 ca the e use to trasmt ( 1 m extra ata, whch woul crease the urato of segmet y ( 1m tme uts where s the average cosumpto rate of the veo urg that terval As a result, stream wll have to trasmt a smaller segmet S whose urato wll e ow equal to ( m ) 1 1 To esure that the whole segmet s always owloae ahea of tme, t ees to e reroacast every ( 1m + tme uts The ecessary awth s the gve y =, eff ( 1 m ( 1 m + ( 1m 1) = ( + ) 1m 1 Fgure : The frst three streams of a veo uer VBHB Let us llustrate ths computato a example Assume a awth overhea coeffcet β=15, equal average awths 1 a for the frst two segmets a a maxmum watg tme = 10 mutes We woul the have 1m = 15 1, whch meas that we woul e ale to trasmt 15 mutes of ata o the frst stream each 10 mutes Sce we wat segmet S to start wth a 0 mute elay, segmet S wll oly cota fve mutes of ata These ata wll ee to e trasmtte whle s eg watche, that s, over a terval of 15 mutes The awth requre to trasmt these ata woul e oe thr of the average awth of segmet S Gve a veo of urato D parttoe to segmets of average urato =D/, our protocol wll allocate ts chaels as follows: a) stream 1 wll trasmt a segmet of total urato ( 1m + at awth 1m ; ) stream wll trasmt segmet S of urato ( 1 m 1 ) at awth,eff ; c) all other segmets of the veo wll e of equal urato a e trasmtte y streams to so that stream wth wll trasmt segmet S at awth 1 The total awth requre y our VBHB protocol wll the e gve y 81

Bawth 1 10 8 6 Staggere Pyram Skyscraper VBHB CHB 4 0 005 01 015 0 Watg tme as fracto of total veo urato Fgure : Bawth requremets of the VBHB protocol for β=1 compare to these of other roacastg protocols BVBHB( ) = 1m +, eff + = 1 4 COMPARISON WITH OTHER PROTOCOLS To compare the awth requremets of CHB a VBHB, we oserve that the expressos for B ( CHB )a BVBHB ( ) ffer the awth they requre to trasmt segmets S a S oly We ca thus compare the two protocols wthout havg to take to accout ether the umer of segmets of the veo or the awth requremets of segmets S 4 to S BCHB ( ) BVBHB( ) = max( m, ) (, eff + ) To smplfy ths expresso, we wll assume that segmets S a S have the same average cosumpto rate a a equal awth overhea coeffcet β The the effectve awth of chael smplfes to a we have ( β ) β, eff = = + ( β ) β β 1 β + β 4 BCHB ( ) BVBHB( ) = β ( + ) =, β β whch meas that our ew protocol wll requre less awth tha the CHB protocol to prove the same maxmum watg tme wheever β 119, whch s very lkely to e the case practce The worst case for our protocol s clearly β = 1, whch correspos to a fxe awth veo sgal We woul the have BVBHB ( ) BCHB ( ) = a our protocol woul requre / uts of awth more tha the CHB protocol 8

Fgure shows the awth versus clet watg tmes for staggere roacastg, Skyscraper Broacastg wth a maxmum wth of 5, Pyram Broacastg, Cautous Harmoc Broacastg a our VBHB protocol wth β=1 To elmate the factor D represetg the urato of the veo, the maxmum watg tmes o the x-axs are expresse as percetages of the veo legths Eve uer these worst case cotos, the VBHB protocol outperforms Skyscraper Broacastg a Pyram Broacastg protocols whe the watg tme oes ot excee 0 percet of the total urato of the veo, that s a watg tme of less tha 4 mutes for a two-hour veo As we have see t earler, VBHB outperforms CHB for all values of β 119 Aother aspect of the performace of a roacastg protocol s the amout of sk space t requres the clet STB Sce our ew protocol oly ffers from the CHB protocol the way they roacast segmets S a S, the two protocols wll have etcal requremets for all values of greater tha, say, 5 or 6 To evaluate these requremets, we ca follow the same approach as Juh a Tseg ther aalyss of the HB protocol (Juh a Tseg 1997) Let R e the amout of ata the clet STB receves whle t s playg segmet S It s gve y R = k k=+ 1 for < The average amout of ata cosume whle the STB s playg segmet S s gve y C = a we ca efe B as the amout of ata the clet has ts uffer after t has fshe playg segmet S, a calculate t as B = B + R C 1 for < The maxmum value of B wll gve a goo approxmato of the storage requremets of the two protocols Ths maxmum value ecreases slowly wth a ever excees 45 percet of the veo for all values of >10 Ths s ot sgfcatly fferet of the storage requremets of other harmoc protocols (Juh a Tseg 97, Pârs et al 1998) 5 CONCLUSIONS Veo roacastg protocols am at reucg the cost of veo o ema servces y strutg more effcetly the veos that are hgh ema a are thus lkely to e vewe at the same tme y may vewers Ufortuately oe of the exstg veo roacastg protocols takes to accout the fact the server wll roacast compresse veos whose awth requremets epe o the rate at whch the mages eg splaye chage As a result, these protocols wll overallocate the awth requre to roacast each veo We have propose a veo roacastg protocol that avos ths rawack Our protocol cotuously roacasts the frst segmet of each veo real tme It the roacasts the 1 remag segmets of the veo at frequeces esurg that segmet S ca e completely owloae y the tme the customer has fshe watchg segmet S 1 Hece these segmets o ot requre excess awth More work s stll eee to evaluate the performace of the protocol usg real compresse veos a to vestgate other ways to map the veo segmets to the ata streams ACKNOWLEDGMENTS We wsh to thak here Dr Mara Crsta Loyo a all her colleagues at the LANIA for ther very gracous hosptalty We are also very grateful to Messrs Steve W Carter a Cyrus D Vesua for ther commets for < wth a, more geerally, B B = k = 4 k=+ 1 k 1 = k 1 REFERENCES Aggarwal, C C; J L Wolf; a P S Yu 1996 A permutato-ase pyram roacastg scheme for veo-oema systems I Proceegs of the Iteratoal Coferece o Multmea Computg a Systems (Hroshma, Japa, Jue), 118 16 Almeroth, K C a M H Ammar 1996 The Use of Multcast Delvery to Prove a Scalale a Iteractve 8

Veo-o-Dema Servce, IEEE Joural o Selecte Areas Commucatos, 14, o 5 (Aug): 1110-11 Bera, J; R Sherma; M Taqqu; a W Wllger 1995 Log-rage epeece varale t-rate veo traffc IEEE Trasactos o Commucatos, 4: 1566 1579 Da, A; P Shau; D Staram; a D Towsley 1994 Chael allocato uer atchg a VCR cotrol veo-o-ema systems, Joural of Parallel a Dstrute Computg, 0():168 179 Da, A; D Staram; a P Shahau 1994 Scheulg polces for a o-ema veo server wth atchg I Proceegs of the 1994 ACM Multmea Coferece (Sa Fracsco, CA, Oct), 15 Da, A; D Staram; a P Shahau 1996 Dyamc atchg polces for a o-ema veo server Multmea Systems, 4, o (Jue): 11 11 Jeha-Fraços Pârs otae hs Igéeur Cvl egree from the Uversté Lre e Bruxelles, hs Dplôme Etues Approfoes from the Uversté e Pars VI a hs Ph D from the Uversty of Calfora, Berkeley Dr Pârs has ee wth Purue Uversty a the Uversty of Calfora, Sa Dego He s ow Assocate Professor of Computer Scece at the Uversty of Housto From Septemer 1997 to Jue 1998, he was o saatcal leave at the Uversty of Calfora, Sata Cruz Hs research terests clue strute systems, the maagemet of replcate ata a struto protocols for veo o ema Dr Pârs s memer of the ACM a the IEEE Computer Socety From 1989 to 199, he was char of the IEEE Computer Socety Techcal Commttee o Operatg Systems a Applcato Evromets Garrett, M a W Wllger Aalyss, moelg a geerato of self-smlar VBR veo traffc I Proceegs of the ACM SIGCOMM 94 Coferece (Aug), 69 80 Hua, K A a S Sheu 1997 Skyscraper roacastg: a ew roacastg scheme for metropolta veo-oema systems I Proceegs of the ACM SIGCOMM 97 Coferece (Caes, Frace, Sep), 89 100 Juh, L a L Tseg 1997 Harmoc roacastg for veo-o-ema servce IEEE Trasactos o Broacastg, 4, o (Sep): 68 71 Pârs, J-F; S W Carter; a D D E Log 1998a Effcet roacastg protocols for veo o ema I Proceegs of the 6 th Iteratoal Symposum o Moelg, Aalyss a Smulato of Computer a Telecommucato Systems (Motréal, Caaa, July), 17 1 Pârs, J-F; S W Carter; a D D E Log, 1998 A Low awth roacastg protocol for veo o ema I Proceegs of the 7 th Iteratoal Coferece o Computer Commucatos a Networks (Lafayette, LA, Oct), 690 697 Vswaatha, S a T Imelsk 1996 Metropolta area veo-o-ema servce usg pyram roacastg Multmea Systems, 4, o 4 (Aug): 197 08 BIOGRAPHY 84