Scheduling for Variable-Bit. Rate Video Streaming By H. L. Lai 1
Contents Variable-Bit-Rate Videos Bit-Rate Smoothing Monotonic Decreasing Rate Scheduling Aggregated Monotonic Decreasing Rate Scheduling Conclusions Q&A 2
Variable-Bit-Rate Videos CBR vs. VBR Problems with VBR 3
CBR vs. VBR 2 types of video compression: CBR compression n n Constant bit-rate Variable visual quality VBR compression n n Variable bit-rate Constant visual quality 4
Problems with VBR Complex admission control and scheduling Hard to provide performance guarantee Solution: Smoothing 5
Bit-Rate-Smoothing Principle Design Considerations Review of smoothing algorithms 6
Principle 7
Design Considerations Lossless or lossy video? Stored video or live video? Zero or non zero playback delay? Deterministic or statistical performance guarantee? 8
Optimal Smoothing Algorithm Minimal variability Minimal peak rate J. D. Salehi, S. -L. Zhang, J. Kurose, and D. Towsley, “Supporting stored video: reducing rate variability and end-to-end resource requirements through optimal smoothing”, IEEE/ACM Transactions on Networking, pp. 397 -410, vol. 6, issue 4, Aug. 1998. 9
Piecewise Constant Rate Transmission and Transport Control the separation and no. of bit-rate changes J. Mc. Manus and K. Ross, “Video on demand over ATM: constant-rate transmission and transport”, Proceedings of IEEE INFOCOM, pp. 1357 -1362, Mar. 1996. 10
CBA & MCBA Minimal peak rate Minimal BW increases (CBA) Minimal BW changes (MCBA) Critical Bandwidth Allocation (CBA) W. Feng and S. Sechrest, “Critical bandwidth allocation for the delivery of compressed video”, Computer Communications, pp. 709 -717, vol. 18, no. 10, Oct. 1995. Minimum changes Bandwidth Allocation (MCBA) W. Feng, F. Jahanian and S. Sechrest, “Optimal buffering for the delivery of compressed prerecorded video”, ACM Multimedia Systems Journal, Sep. 1997 11
Rate Constrained Bandwidth Allocation Check all frame sizes Prefetch earlier if any frame exists BW constraint W. Feng, “Rate-constrained bandwidth smoothing for the delivery of stored video”, SPIE Multimedia Networking and Computing, pp. 58 -66, Feb. 1997. 12
Time Constrained Bandwidth Allocation Construct an upper bound curve with both buffer and time constraints Construct schedule with any other smoothing algorithms W. Feng, “Time constrained bandwidth smoothing for interactive video-on-demand systems”, International Conference on Computer Communications , pp. 291 -302, Nov. 1997. 13
ON-OFF Scheduling Single rate for whole system Send “as late as possible” R. -I Chang, M. C. Chen, J. -M. Ho and M. -T. Ko, “Designing the ON-OFF CBR transmission schedule for jitter -free VBR media playback in real-time networks”, Proceedings of the Fourth International Workshop on Real-Time Computing Systems and Applications , pp. 2 -9, Oct. 1997. 14
Other Studies Smoothing at multiple intermediate nodes J. Zhang, “Using multiple buffers for smooth VBR video transmissions over the network”, 1998 International Conference on Communication Technology , pp. 419 -423, vol. 1, Oct. 1998. Multiplexing optimally smoothed schedules W. Zhao and S. K. Tripathi, “Bandwidth-efficient continuous media streaming through optimal multiplexing”, Proceedings of International Conference on Measurement and Modeling of Computer Systems, pp. 13 -22, Apr. 1999. S. S. Lam, S. Chow and D. K. Y. Yau, “A lossless smoothing algorithm for compressed video”, IEEE/ACM Transactions on Networking, pp. 697 -708, vol. 4, issue 5, Oct. 1996. Scene based smoothing H. Liu, N. Ansari and Y. -Q. Shi, “Dynamic bandwidth allocation for VBR video traffic based on scene change identification”, Proceedings of International Conference on Information Technology: Coding and Computing, pp. 284 -288, March 2000. Re-arranging sending sequence of frames R. Sabat and C. Williamson, “Cluster-based smoothing for MPEG-based video-ondemand systems”, IEEE International Conference on Performance, Computing and Communications, pp. 339 -346, Apr. 2001. 15
Other Studies (cont. ) Lossless online smoothing J. Rexford, S. Sen, J. Dey, W. Feng, J. Kurose, J. Stankovic and D. Towsley, “Online Smoothing of Live, Variable-Bit-Rate Video”, International Workshop on Network and Operating Systems Support for Digital Audio and Video, pp. 249 -258, May, 1997. Controlling online encoding parameters with: n Buffer occupancy S. C. Liew and D. C. -Y. Tse, “A control-theoretic approach to adapting VBR compressed video for transport over a CBR communications channel”, IEEE/ACM Transactions on Networking, pp. 42 -55, vol. 6, issue 1, Feb. 1998. n Network status N. G. Duffield, K. K. Ramakrishnan, and A. R. Reibman, “SAVE: an algorithm for smoothed adaptive video over explicit rate network”, IEEE/ACM Transactions on Networking, pp. 717 -728, vol. 6, issue 6, Dec. 1998. 16
Monotonic Decreasing Rate Scheduler Motivation Constructing an MDR Schedule Performance Evaluation Admission Complexity Waiting Time vs. System Utilization Buffer requirement 17
Motivation Existing smoothing algorithms contains both upward and download bandwidth changes n n Complex admission to provide deterministic performance guarantee Upward changes may fail in mixed traffic environments Solution: transmission with downward bandwidth changes only – MDR Scheduler 18
Constructing an MDR Schedule 19
Performance Evaluation 274 VBR encoded DVD videos tested Avg. bit-rate: 6. 01 Mbps Avg. length: 5780. 7 s Round length: 1 s Requests generated according to Poisson process to select a random video Un-admitted requests put to FIFO queue 20
Admission Complexity 21
Waiting Time vs. System Utilization 22
Buffer Requirement 23
Aggregated Monotonic Decreasing Rate Scheduler Principle Bandwidth Over-allocation Admission Complexity Performance Evaluation Effect of Network Topology 24
Principle Specify a buffer requirement, B For streams with buffer requirement: n n <= B, deliver with MDR schedules > B, deliver with optimal smoothing and over-allocate bandwidth to maintain monotonicity of the aggregate system traffic 25
Bandwidth Over-allocation Rate New exceptional stream, smoothed using optimal smoothing. Time + Rate Current aggregate bandwidth utilization. Time = = Rate Aggregate bandwidth utilization and reservation after new stream is admitted. Bandwidth over-allocated here to maintain rate monotonicity. Time 26
Admission Complexity Unsuccessful admission comparisons = additions: O( +(1 )(g+1)) = O(1+(1 )g) Successful admission comparisons: O( +(1 )(g+1+w)) = O(1+(1 )(g+w)) additions: O(w) Where: is the proportion of videos served by MDRS g is the no. of bit-rate increases in optimal smoothing 27
Admission Complexity (cont. ) With 16 M of client buffer 28
Admission Complexity (cont. ) With 32 M of client buffer 29
Admission Complexity (cont. ) With 64 M of client buffer 30
Waiting Time vs. Client Buffer Size (cont. ) 31
Waiting Time vs. Client Buffer Size 32
Waiting Time vs. System Capacity To be completed… 33
Effect of Network Topology In practice, network topologies are likely to be more complex We simulate a network with two-level, tree-based topology The effect of maintaining monotonicity within each individual branch is studied Results: to be completed… 34
Conclusions Scheduling of VBR video streaming is a complex problem Smoothing can reduce the variability; but will not completely solve the problem The MDR Scheduler can provide deterministic guarantee with low admission complexity Performance is comparable optimal smoothing With a trade off in performance and complexity, the AMDR Scheduler adapt to any buffer size 35
Q&A Thank you 36
Скачать презентацию Scheduling for Variable-Bit Rate Video Streaming By H
90a6f98ac9931c17f58fbe93e0183544.ppt