Scaling a video on demand server

Early performance estimation is key to successful implementation

Illia Cremer, CoFluent Design
EETimes (4/27/2011 9:51 AM EDT)

Abstract

In a growing and more competitive video on demand (VoD) market, system designers face new challenges in VoD server infrastructures definition and sizing. Early performance estimation thanks to abstract modeling is a key enabler for providing best quality of service and compelling user experience.

This article illustrates how to model and simulate an example model of a RTP/RTSP video on demand server using the method, notations and tools provided by CoFluent Design.

The objective is to determine the client’s frame rate deviation and the average power consumption for different server configurations. The frame rate deviation is the difference between the expected theoretical frame rate and the actual frame rate of the video stream. It directly impacts the user’s watching experience and should be kept as much as possible close to zero.

The impact of different hardware elements of the server such as HDD type and server buffering is studied. The example also illustrates how to model multiple instances of the same function, and how to define an abstract network of computers.

Click here to read more ...