Today's 802.11b Networking News is sponsored by IIR's Public Access Wireless LANs
conference, 1-3 October 2002, Lisbon

The above is a paid, sponsored link. Contact us for more information.

[ Other News, below ]

The back-end bandwidth for a hot spot may become as important as its very existence. One of the nails in the coffin of MobileStar's deployment of hot spots in Starbucks stores was their insistence (prompted by Starbucks) on using dedicated T1 lines to bring bandwidth to the locations. Their competitors and others in the industry sniggered: why spend several hundred to several thousand dollars a month per location to bring in a T1 when comparable DSL service might only be a few hundred at most?

The answer is, of course, quality of service. No matter who provides the DSL -- a regional telco, Covad, or a local phone company -- it's not their top priority. Circuits go up and down. The service at times is pumped into overloaded ATM (asynchronous transfer mode) circuits and then shot thousands of miles before emerging at a network operation center or NOC. (Smarter ISPs manage this better, placing points of presence or POPs closer to major DSL customer bases.)

And who cares if the DSL goes down? It's usually an 8 to 5 proposition, weekdays, holidays excluded. You lose your circuit on Friday at noon, and if you're lucky enough to reach someone who can troubleshoot it and get a trouble ticket issued, you might get it solved by the end of the day -- but as likely as not, no.

I speak from experience on both sides: I've had almost 100 percent reliable (knock photo-wood-grain finish) DSL in several locations, but I've also had the Friday-afternoon experience.

Contrast this with T1: a T1 line is point-to-point. That is, service originates at one location coming out of a CSU/DSU which converts a router's output into T1 frames. Four wires snake from one location to a central office where the circuit is connected to another circuit and back into another office, CSU/DSU, and router. Because the quality of service (QoS) is maintained throughout, and because T1 is priced as a business service that may actually cover a telco's cost of operations, the choke points can be serviced right away.

When I had a T1 line back in 1994-1996 running an early Web development business, even in those heady early days my service calls were escalated. I paid the phone company $250/month for what I now pay them, over DSL, about $20 to $30 per month.

The context of T1s and QoS is equally important to hot spots. Starbucks demanded a robust business network over which to run their store-to-HQ ecommerce network, training systems, and incidentally to allow MobileStar to provide wireless-to-Internet service. MobileStar's mistake, apparently picked up by VoiceStream-cum-T-Mobile's deep pockets, is to subsidize Starbucks IT (information technology) expense as part of their cost of placement in a premium location.

For T-Mobile, it may make sense: spending $100 million on a national network of several thousand points all of which proudly display their wireless data/cellular brand could easily result in more cellular customers, and it's a bully launching point for any seamless, integrated cell/Wi-Fi tests and rollouts.

For most other locales, DSL could suffice, but the ups and downs of managing multiple DSL circuits, or relying on your hot spot customers (if you're deploying the service or reselling it), could swamp the uninitiated who base DSL performance on their own experience.

Other options are available, such as frame relay, an interesting technology in which you bring varying speed lines to facilities scattered around a number of central offices back to a central frame cloud. A circuit can be stuck in the middle of that cloud, essentially, pumping the frame traffic off to the Internet. The frame network can be cheaper to deploy because you don't need to bring Internet service to every location, just to the frame cloud itself. You also can offer a higher level of performance, because frame is a business service, and a higher pool of bandwidth.

Overseas, some providers already seem to be using bandwidth as a selling point, like NTT in Japan promoting the notion that the backhaul from their Wi-Fi hot spots will be multimegabits per second, rather than a more typical T1 speed (1.5 Mbps) or much slower (DSL speeds of 128 to 768 Kbps).

As conference centers start to roll wireless networking into their array of services to trade shows, they'll face more and more resistance from their $2,000 charges for a few days of Ethernet access at T1 speeds. A colleague recently asked me privately about how trade shows might respond to these enormous facilities charges while attendees at the show were spending $10 per day for full Internet access.

My response was that QoS and packet shaping (ratcheting bandwidth based on type of service, and even specific users) could allow conference centers to charge reasonable amounts for service (nothing like the ridiculous margins of today, I would expect).

The conference center wireless network can be segregated so that trade show booths and speaker media have guarantees for available bandwidth to aid in streaming, demonstrations, and simply using corporate VPNs. Average attendees might be subject to bandwidth throttling, lower priority of their packets, or other limits to further distinguish service.

In the majority of hot spots, 512 Kbps service will continue to be the average rule of the day, and most hot spot networks don't have the Starbucks burden weighing on their backs. But as bandwidth use increases across many kinds of applications, hot spots will increasingly need to plan for bandwidth front and center, rather than as a last detail in deployment.

Other News

iPass proposes standard gateway interface for roaming: the company hopes its proposal becomes a de facto method for a client to access any arbitrary network for authentication.

Motorola to extend Canopy networking to include unlicensed 5 GHz band: some technical details, but no pricing or availability. Motorola's thrust is odd: anyone can become an ISP. Yeah, if they have quite a lot of technical and business expertise. There are plenty of existing ISPs that will be delighted to have yet another tool in the final mile, point-to-point arsenal. [via Alan Reiter]