VoDKA: Developing a Video-on-Demand Server using Distributed Functional Programming

Abstract

In this paper, we present some experience of using the concurrent functional language Erlang to implement a distributed video-on-demand server. For performance reasons, the server is deployed in a cheap cluster made from off-the-shelf components. The demanding system requirements, in addition to the complex and ever-changing domain, suggested a highly flexible and scalable architecture as well as a quite sophisticated control software. Functional programming played a key role in the development, allowing us to identify functional abstractions throughout the system. Using these building blocks, large configurations can be defined using functional and process composition, reducing the effort spent on adapting the system to the frequent changes in requirements. The server evolved from a prototype that was the result of a project supported by a regional cable company, and it is currently being used to provide services for real-world users. Despite our initial concerns, efficiency has not been a major issue.

A versatile multiplexing algorithm exclusively based on the MPEG-2 systems layer

“…Between two PCR values in the generated TS, the bitrate is 1 Note that not every TS packet carries 184 bytes of payload. A TS packet with last data chunk of a PES packet may be filled with stuffing, which does not enter in the decoding buffer.…”

“…In a video server [1], where different types of multimedia data are stored in PESs (Packetized Elementary Streams), or even grouped in single programs in PSs (Program Streams), it is necessary to combine a set of them into one SPTS (Single Program Transport Stream) to serve it when a user sends a request to the server. This SPTS must be compliant with the STD (System Target Decoder) defined in the MPEG-2 systems standard [2] in order to avoid buffer overflows or underflows in later stream operations (such as demultiplexing and decoding it).…”