Optimized buffer allocation for video multicasting applications with virtual memory implementation

Abstract

Memory requirement is a key issue when servicing more number of nodes in a heterogeneous environment. We have to ensure an optimum buffer size in order to reduce the initial latency. In this paper, we propose a novel approach, which involves, augmenting a virtual memory (VM) with the existing physical memory of the video buffer. In addition to the benefits of improved queuing size, it also paves the way for better quantization, ordering of frames, and higher error resiliency. Instead of allocating the virtual memory as a single file, it is fragmented according to the number of nodes in a specific multicast group. This avoids the flushing of an entire buffer to begin a new streaming. Furthermore, this avoids the requirement of proxy servers for each multicast group. Hence, this reduces the burden of the server, when a new user joins the multicast session later, requesting same content. VM also facilitates a platform to support old and advanced graphical interface irrespective of device capability. A bit rate reduction technique is also applied to video streams which enables seamless streaming even in degraded network. Unlike existing buffer adaptation techniques, this method does not require a feedback loop, as most of the rate adaptation is processed by the VM. Extensive simulation results show that this research work helps in significant improvement in throughput, PSNR (peak signal-to-noise ratio) when compared to existing buffer management algorithms such as RBA (rate-based adaptation), BBA (buffer-based adaptation), Elastic, and FESTIVE (Fair, Efficient, and Stable adapTIVE algorithm).