Research And Design Of DASH Scheduling Algorithms Based On SVC Coding

In recent years,with the continuous development of the Internet video industry and the increasing number of mobile terminals such as mobile phones and tablets,new applications and functions such as video upload,on demand,live broadcasting using mobile devices have emerged,which has increased the proportion of video content traffic in the whole Internet.Currently,streaming is prevalently used to provide video service for Internet users,but high network dynamics caused the problem of unstable media transmission.Heterogeneous terminal screen size and network access also pose challenges for terminal adaption.DASH(Dynamic Adaptive Streaming over HTTP)can to some extent solve the above problems,and soon became one of the mainstream Internet video transmission technology.However,traditional AVC encoding based DASH has some inevitable problems,such as low ratio video chunk reuse and high rate of quality oscillation.Nowadays,a lot of terminals have multiple network interfaces.Based on these,this thesis proposes a DASH seheduling algorithm that employs multiple network interfaces and SVC encoding,which adapts better to terminal heterogeneity and network dynamics,improves user,s quality of experience,and alleviates the server's storage as well as network bandwidth pressure.Experimental results show that compared with AVC coding,our scheme can reduce the server storage by approximately 50%,and compared with the traditional DASH based on bandwidth and buffer with a single interface,our scheduling algorithm can improve the video playback quality by about 45%,and reduce the number of video quality switches by about 37%.In addition,in DASH,an MPD file describes the availabld video qualities and their average bitrates.However,the coarse-grained average bitrate is insufficient for an accurate client-side scheduler.For VBR,video bitrates usually exhibits drastic temporal vibration,which may cause either overestimate or underestimate of bandwidth requirements for future-to-fetch chunks.This could result in suboptimal scheduling decisions and possibly drain the playback buffer.Unfortunately,for AVC encoded videos,there is little choice that rate-based schedulers can rely on to make scheduling decisions.In this thesis,I first observed that with SVC,video segment size across different layers exhibits high correlation.Then I showed that enhancement layer segment size can be predicted by the corresponding base layer segment with reasonable accuracy.Based on this observation,I proposed SIZE-Predict,a simple greedy scheduling algorithm that utilizes this correlation to partially tackle the bitrate underestimate or overestimate problem.Preliminary experimental results show that compared with the rate-based scheduling that is agnostic to bitrate variability,SIZE-Predict can improve the average playback bitrate by up to 20%and reduce playback stalls.The performance is only slightly worse than prophetic scheduling algorithms with future knowledge.The results demonstrated that a size predictor should be added to SVC scheduling algorithms to augment performance gain,analogous to the bandwidth predictor in DASH.