F-FDN: Federation of Fog Computing Systems for Low Latency Video Streaming

Abstract

Video streaming is growing in popularity and has become the most bandwidth-consuming Internet service. As such, robust streaming in terms of low latency and uninterrupted streaming experience, particularly for viewers in distant areas, has become a challenge. The common practice to reduce latency is to pre-process multiple versions of each video and use Content Delivery Networks (CDN) to cache videos that are popular in a geographical area. However, with the fast-growing video repository sizes, caching video contents in multiple versions on each CDN is becoming inefficient. Accordingly, in this paper, we propose the architecture for Fog Delivery Networks (FDN) and provide methods to federate them (called F-FDN) to reduce video streaming latency. In addition to caching, FDNs have the ability to process videos in an on-demand manner. F-FDN leverages cached contents on the neighboring FDNs to further reduce latency. In particular, F-FDN is equipped with methods that aim at reducing latency through probabilistically evaluating the cost benefit of fetching video segments either from neighboring FDNs or by processing them. Experimental results against alternative streaming methods show that both on-demand processing and leveraging cached video segments on neighboring FDNs can remarkably reduce streaming latency (on average 52%).

SMSE: A Serverless Platform for Multimedia Cloud Systems

“…The conventional approach of streaming through preprocessing all the multimedia segments ahead of time is not feasible for interactive media streaming for two reasons: (A) In live streaming, media are not available for pre-processing; (B) Even for the on-demand streaming, given the long-tail access pattern to the contents and the ability to stream the contents with a combination of interactive services, it is costprohibitive to pre-process all the contents for all the services. For instance, just to cover most common display standards, for each media stream, some 90 to 270 versions must be pre-processed and stored [28], [37]. In fact, pre-processing of streams is performed only for common and frequentlyused interactive services and the rest should be processed in an on-demand (i.e., lazy) manner.…”

“…The caching module determines the hotness (i.e., popularity) at the segment granularity, as discussed in our prior study [36]. That is, the segments (tasks) that are anticipated to be streamed again in the near future are cached for reusing in a local caching server or on the CDN [37].…”

LwTE: Light-Weight Transcoding at the Edge

“…Wang et al [24] proposed an algorithm to determine edge servers for transcoding operations to improve perceived quality by clients. A distributed platform at the edge named Federated-Fog Delivery Network (F-FDN) has been proposed in [25]. F-FDN stores only one bitrate of non-popular videos and leverages the edge server computing capability to provide requested bitrates at the edge to minimize video streaming latency.…”

A Comprehensive Study of Caching Effects on Fog Computing Performance

“…V.Veillon et al [90] Dedicated to providing a consistent streaming experience with low latency and no interruptions. The paper proposes a Fog Delivery Network (FDN) architecture and federation methods (named F-FDN) to reduce video streaming latency.…”

“…According to test results, the cache location improved by the suggested methodology enhances the STP by roughly 18% over the best existing caching distribution. V.Veillon et al [90] 2019 Files-Papadopoulos et al [97] presented an improved simulation framework that (a) supports graph-based network topologies, (b) requests have been reconstituted for differentiation of requirements, and (c) statistics were computed per site and network metrics per link, improving the granularity and parallel performance. They also proposed a two-phase optimization scheme that used simulation outputs to guide the search for optimal cache placements.…”