Hi all, I would like to use ns-2 to investigate the impact of various QoS mechanisms on different networks. Before I get too deeply into it, I wanted to canvas peoples thoughts, to see if I can do what I would like to (and whether it's worth it).
The networks simulated would be Linux/BSD based. Is anyone aware of how "close" the ns-2 implementation of the network stack is to the one in the kernels? I would like my simulation to offer comparable results on the real systems if at all possible. I know Linux has some built in support for packet tagging (via netfiler) and queuing disciplines, so does anyone know how these things match up in ns2? How much effort will it be to get the two to be comparable enough for meaningful results. Essentially my research will try to answer the question of should we bother implementing some form of QoS over our network, or will the system work better if we just let each application fend for itself. If QoS would help, how much will it cost (time, and money) to implement it, etc. Does anyone have any thoughts on this subject? Thanks, lee --
