video streaming systems that can serve hundreds to thousands of concurrent users are already widely available. however, to deploy metropolitan-scale video streaming services for potentially tens of millions of users, current video streaming systems are still limited in capacity, and expensive in cost. to tackle this challenge, this study proposes a new Dynamic Consonant Broadcasting (DCB) architecture that exhibits three desirable properties. first, while existing broadcasting and multicasting streaming architectures are optimized either for light load (close-loop architectures) or heavy load (open-looparchitectures) conditions, DCB can be applied to systems of any load conditions and yet outperform both existing close-loop and open-loop architectures. second, while existing architectures commonly have specific client access bandwidth and client buffer size requirements, DCB can be configured for any client access bandwidth and client buffer size. last but not least, DCB cansupport the streaming of both constant-bit-rate (CBR) as well as variable-bit-rate (VBR) videos. more surprisingly, our results obtained from 300 full-length VBR video traces show that streaming VBR video using DCB in fact requires less network bandwidth than streaming an equivalent CBR video of the same average video bit-rate. this last observation suggests that we not only can provide more consistent visual quality with the use of VBR encoding, but also reduces the network bandwidth requirement when streamed using the proposed DCB architecture.
- reference: W. C. Liu and Jack Y. B. Lee, “Dynamic Consonant Broadcasting - A Unified Architecture for Variable Bit-rate Video Streaming”, Technical Report CUHK-MCL-2003-04.
