On 16/12/06 13:38 +0530, jtd wrote: <snip> > Opex is high with current business model. The wifi network is a DIY > non-business. > That doesn't make sense. You have two choices: 1) Put in a fat pipe (fairly expensive), even if you go wireless. Wireless has other issues as well, but will work fairly well in a rural environment. Then just keep growing and lighting up more of the fat pipes for a very small expense. Your per unit costs come down as you oversell.
2) Put in a narrow pipe, and then put in more pipes as you grow. Lowers your capex, increases your opex drastically. > > Can you show me how it would > > scale upto a few thousand nodes? > Eat the pudding test: > In my experience bw drops to 4 mbps and stays there with a 54Mbps AP > for 10 to 20 users (havent tried higher). Latency increases though, And dies beyond 25, from experience :). I suggest you try filesharing and something else on the same LAN with wireless. > but not by much. And we are talking of fat pipes for tens of nodes. > According to Fred pook, the city of Dharamstala is fully covered by > wifi - afaik 1000 nodes. Essentially, the capability to service 20000 people, who will be accessing basic non-voice/video services, and not be doing any heavy data transfers. > However Scalability to this size will definetly be a problem. > But scalability to lets say 3 to 4 nodes with one 10mbps pipe per > cluster, and 30 users per node. Should give u good performance. We And the 10 Mbps pipe costs you how much to lay? For a slightly higher cost, why not lay fiber (where the costs are in the termination, not in the pipe as opposed to being the other way rounf for copper), and get 100Mbps to the central node? The problem isn't in the fat pipe. The problem is in taking bandwidth from the termination of the central pipe to the edge. > are talking of rural and semiurban areas with poor phone penetration. > > > The problem in ISPs isn't one of bandwidth (that becomes cheaper > > per unit as you buy more), it's one of getting a reliable internal > > network which scales cheaply. > > Having said the above. Cellular and land networks are almost ubiquitos > in India. And a wireless infrastructre would be an non lucarative > business (given prevalent business models for content, voice and live Errr, you do realise the BSNL refuses to allow other telephone carriers to use their fiber backbone to even carry voice calls? And charges their customers more money for "interconnectivity" and that subsidy for rural connectivity? Honestly, our problem is stupid bureaucrats and their old, outdated business models. > media), but would help in providing competition the same way as libre > software has. Note libre software provides several viable business > models. One would have to think deeply about possible business models > for such disruptive wireless networks. > Me? I would make access a public good, and run out fibre to lots of places (cheap), copper to the last mile, and then lease it out to service providers. The closest I have seen to that model is in Scandanavia, South Korea, Japan and New Zealand. Devdas Bhagat -- http://mm.glug-bom.org/mailman/listinfo/linuxers
