Vous êtes sur la page 1sur 3

DistributedVideoTranscoding DistributedSystem IsmailSunni/138531 RischanMafrur/138173 Contents: Introduction Motivation DesignSystem Result Conclusion

Introduction Nowadays, video has become one of the most popular content in the world of internet. The quality of video is becoming higher and higher, as the video format become more diverse. People do not only watch the videoontheir computerortvbutalsoontheirmobile.Theproblem is most of their mobile devicesdonotaccommodatetoplayhighresolutionvideooronlybeable to play specific video format. From this issue, we need to do video transcoding. But another problem raises, video transcodingneedsenormouscomputingpowerthatwillconsumesmobile devices limited battery. So we need to move the transcoding processto another machine and our mobile devices can download and play it. Since video transcoding is high computation process,weproposetodoitdistributedlybyusingStormsothatitcanbedonefaster. Motivations 1. Videocontentisgettingbiggerandbiggerandavailableinvariousformat 2. Manydevicesthatneedmanyvideoformat 3. Videotranscodingneedsenormouscomputingpower SystemDesign Inthisprojectwecalculatetimeprocessfortranscodethevideofileindifferentmodethereare: 1. StandaloneMode 2. DistributedModeusingprogramthatweremade 3. DistributedModeusingtwitterstorm Beforeweexplainmoredetailsaboutthethreemodeabovewewanttodescriptaboutour environmentsystem. WehavethreePersonalComputer: 1. PCServer(168.131.152.160):Ubuntu,Intel(R)Core(TM)2QuadCPUQ9400@ 2.66GHz,RAM2GB 2. PCWorker1(168.131.152.160):Ubuntu,Intel(R)Core(TM)i73770CPU@3.40GHz, RAM8GiB 1

3. PCWorker2(168.131.40.44):Macbook,IntelCore2Duo2.4GHz,RAM2GiB ThePCServerandWorker1arePCinAdvancedNetworkLabandthePCWorker2isin Dormitory. VideoFileInformation file:stormvideo.mp4 duration:00:07:57.84 resolution:576x360 size:9.728MB Threemodethatwehaveforprojectsimplementation. 1. StandaloneMode Inthismodewejustdoingasimplework.Wetranscodetheonevideofileandthenwe calculatethetimeprocess. 2. DistributedModeusingprogramthatweremade Inthismodewedevelopsimpleprogramusingpythonprogramminglanguage.Thefirst programisforcookingthevideofilescontainthreefunctiontherearefunctionforsplitting thevideos,functionfortranscodethevideos,andfunctionforconcatthevideos,wecall thisprogrampyvid.py.Thesecondprogramisforsendingandreceivingvideofilefrom servertoclientandfromclienttoserveraftertheclienttranscodethevideofiles,wecall thisprogrampysendfile.py.Thesourceprogramweattachinthezipsource. 3. Distributedmodeusingtwitterstorm

Result Wehavethreeresultforthisprojectimplementation.Thefirstresultiswhenwetranscodethe videofileinstandalonemode,totaltimeprocessforthismodeis0m36.936s.Thesecondresult iswhenwetranscodethevideofileindistributedmodeusingprogramthatweremade,the resultintablebelow.


filesize before 9.728MB splitin server 0m12.473s sendto worker transcodein worker sendto server concatin server 0m0.456s filesize after 4.152MB

note PCServer PCworker1 (Lab) PCWorker 2(Dorm) Macbook

0:00:43

0m6.484s

0:00:00

0:00:44

1m4.308s

0:00:04

Thelastresultiswhenweusethestormlibraryfortranscodingthevideofile,inthis implementationwegotproblemwestuckwhenweimplementthecodewiththestormlibrary. Thereasonsare: 1. WhenweusetheJavathelibraryffmpegforjava(xuggler)istoooutofdateaswellasif weusethepythonlibraryforffmpeg. 2. Becausethereasoninnumberone,wemustcalltheffmpegdirectlyfromthecode(use subprocess)andwhenweusethesubprocesswestuckbecausethestormlibrary doesntsupportforsubprocess.

Conclusion Stormisnotsuitableforvideotranscoding Stormisforstreamdata,sowecantusestaticvideo Ifweusestreamvideo,itcantbedistributed ItsbettertouseHadoop,althoughbasedonapaper,itdoesntgiveagood performance Untilnow,wehaventfoundreferenceforthesameresearch(distributedvideo transcodingusingtwitterstorm)

Vous aimerez peut-être aussi