Distributed Active Element for High-Performance Data Distribution

Requirement on multi-point data distribution in IP networks assumes some distribution service, be it implemented as a part of network native services (IP multicast) or user-empowered solution (data reflectors). In our previous work, we have introduced user-empowered distribution networks based on Active Elements (AE) [1], which scale well in terms of number of clients connected. It is however not sufficient in terms of scalability with respect to bandwidth of each single stream distributed, i. e. it is not suitable for distributing streams whose bandwidth exceeds capacity of each single AE [2]. In order to improves the scalability with respect to the bandwidth of a stream, we propose a concept of distributed AE, suitable for implementing on computer clusters with low-latency internal interconnection. Its architecture is based on parallelizing whole AE architecture including listener and sender modules. which however brings problems with packet reordering in sending part. While packet reordering is largely unwanted for general network element (e.g. router) behavior, as it severely tampers performance especially of TCP-based applications, it is more acceptable for multimedia application that rely on UDP protocol and thus need to handle potential packet reordering anyway. Simple solution is a design with all distributed modules except for sender module—while this would help for computationally intensive operations on the streams, it is of little use for high-bandwidth streams. In this paper we show, that even when having multiple sending modules with no explicit synchronization, the reordering introduced by the distributed AE has an upper bound for real-time synchronous applications under certain assumptions. It can be further reduced by implementing proposed Fast Circulating Token protocol.