| Magnus on Wed, 11 Jun 2003 13:25:19 -0400 |
|
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Keep in mind the economics of this. The very cheapest I've ever heard of a T1 line running is $500 per month. It's more typically $800 per month. None of us are going to pay that out of pocket. Setting up two wireless access points with directional antennas should cost under $1000. One time fee. And if you need more than a T1 capacity, (cheap) wireless can't hack it. Period.
802.11b on average provides 2Mbps over long haul links. That already trumps a T1. 802.11a and 802.11g can potentially offer much higher throughput than this. For a short link like Prospect Park to Swarthmore (if the trees weren't there) 10Mbps should be a pretty reasonable expectation. The killer to all this, as in all networks, is going to be topology.
It is the same problem which the Cable companies face. As long as nobody uses the service, then there is more than enough bandwidth for all. But, assume for argument's sake that you actually have 10 meg bandwidth, if you have 10 users, that means each one can only ever get 1 meg each. Once you start aggregating say 20 or 30 or 50 users performance is going to take a hit to the point where people are going to be constantly complaining. Yes, bandwidth hogs will at some point be a problem. But the math is more complex than what you're presenting. If you have 10Mbps of bandwidth, the most any one user can actually get is 10Mbps. If ten users are hammering the network, they may have an effective throughput of 1Mbps each. That's still 66% of a T1 for one person. How is that such a bad thing? In a mesh network topology, the very act of adding nodes increased available paths from any one point to any other one point in the network. The larger repeater sites connecting one town to another are natural bottlenecks. But within the town itself, the network actually becomes more robust as nodes are added to it, because each of them acts not only as a client but also as a router. You can play the game, "Oh, they don't all use it at the same time." That's called the AOL Actually the ISP term for it is "overcommitment", and it is a universally accepted business practice. It is a true statement, just as my initial statement is -- as long as nobody uses it, there is plenty of bandwidth. It's also misinformed. If an ISP provided 56Kbps of bandwidth to the internet for every dialup user connecting at 56Kbps, and had a 56Kbps modem dedicated to every subscriber, they would soon go out of business. The simple fact of life is that most users are idle most of the time, and about 10% of your users will be responsible for 90% of the utilization. Most people's habits, when they are online, involve browsing web pages which is actually very bursty traffic with long periods of silence in between. Email works in much the same way. Put another way -- a wireless back-haul, in an Urban environment, is a waste of time and energy. In a Suburban environment, it might make sense, maybe. In a rural environment, it will likely make sense. OK. I disagree. When we get this thing up, if our vision is proven out, will you turn down the opportunity to hook up to it? Are you just here to say this won't work? Or do you have anything constructive to say to get us around any architectural bottlenecks? - --Magnus -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.3.2 (Darwin) iEYEARECAAYFAj7nZbgACgkQYPuF4Zq9lvbgbQCghOpNHpadvVkr37wu4kmqCxPA g7sAn3IbasmAmPjUnNfps1FI7Hzt8n7Y =ZKuO -----END PGP SIGNATURE----- _________________________________________________________________________ Philadelphia Linux Users Group -- http://www.phillylinux.org Announcements - http://lists.netisland.net/mailman/listinfo/plug-announce General Discussion -- http://lists.netisland.net/mailman/listinfo/plug
|
|