Distributed Video Caching for Multi-Camera Applications

Abstract

EVS, specialised in solutions for video productions, provides multi-camera applications able to access some video content, acquired from a multi-camera system and stored on remote production servers. Sequences of video can be requested by operators using the multi-camera applications. Those operators could then access the same content many times, causing the production servers to be heavily occupied to handle similar video requests.

The goal of this thesis is to provide a software solution that allows to reduce this amount of requests, and to reduce the delay encountered by the operators requesting some video content.

Firstly we present a global view of the existing tools in content distribution, like Peer-To-Peer and Content Delivery Networks, but also in replacement policy and in indexing with Bloom Filters.

We continue by describing in details the context within which this solution is supposed to be used, as a local distributed caching service, and the already existing video storage system developed by EVS.

Then we explain our solution, called DVP(Distributed Video Provider), which is based on a local memory and disk caching solution build on top of the EVS video storage system, with a peer-to-peer video exchange between multi-camera applications within a same private network. We also discuss the limitations of such a solution.

Finally, we evaluate DVP through multiple tests and demonstrate that it reduces both video content access delay and production servers' load in terms of video requests.