Slashdot

For some time now I've been taking part in WIMAX trials here in Hamilton Ontario. [community-media.com] This too was trumpeted as a glorious thing that would change the face of our city, bring us into the high tech 21st century etc.

In practice although WIMAX seems to work OK (aside from a real lag much of the time, which may just be bad server configuration by Primus Communications), My sense is that the company isn't really committed to it. I doubt that there will be a serious public roll out.

The idea seems great - a wireless Internet connection that works wherever you are. The reality seems a bit less rosy, and my guess is that a city wide wireless network will need a good level of customer support - not Primus' strong point by a long shot.

>Not all WiMAX operators are unhappy.
>
>Internode says an Airspan-supplied network is providing consistent average speeds of 6Mbps at >distances up to 30km, with CEO Simon Hackett describing the platform as "proven."

So where exactly lies the problem? Implementation?

I remember when a bunch of wireless guys got invited down to Intel for a private overview three or four years ago and we spent most of the couple hours trashing most of their basic assumptions about the technology. Their major response was, "well by the time it is deployed we will have figured it out"

WIMAX isn't going to be the success that it should be because I think it was driven more by marketing than technology.

I'm going to fiddle my fingers until they have a few more disasters till they get it working. In the meantime mesh will definitely deflate the momentum WIMAX needs right now.

Maybe someone can clear this up- does Clearwire use WiMax or not? Wikipedia didn't make it clear. My experience with them was that they didn't either have the infrastructure or the bandwidth to support their meager customer base. The thing worked just fine during the day when nobody really used it, but during busier hours you had significant lag and flow problems- however, the download rate was still good, but you can't play games with a ping of over a second.

To me, WiMax is the future version of 56k.

NLOS performance depends on a number of things, including how well the underlying technology can handle multipath and otherwise distorted signals. But the main thing is probably frequency; the higher the frequency, the worse the NLOS performance. WiMax is designed to run at many different frequencies, and the article fails to mention which one was in use.

The issues with latency and jitter, though, probably aren't as dependent on frequency.

I chose as my ISP in Mexico City E-go [ego.net.mx], co-owned by Alestra [alestra.com.mx], the Mexican AT&T subsidiary. It started offering WiMax connection in 2003 in limited areas of Mexico City (I understand nowadays it covers most of the Central, Western and Southern parts), before even WiMax was standardized. Clients get a NextNet [nextnetwireless.com] RSU unit [nextnetwireless.com], which is basically a network bridge.
The latency complaints you state are simply not true - I get consistent ping response times of 100ms in average (with minimum response times of around 50ms) to hosts in Mexico City, 200ms to hosts in the USA. Yes, this is about 80ms higher than wired equivalents - but it's not so much of a killer. What I do get, of course, is way higher packet loss - About 5% when things are optimal, and it sometimes gets up to 50%. But yes, I'm located at a relatively poor reception area, at one of the lower-income (this means, no incentive to place many antennas nearby) neighbourhoods in the South of the city, where the mostly flat valley where most of the city is located begins to become quite hilly. The RSU unit does not provide any means (for the client) for monitoring connection, to help choose the best possible location. It only has five LEDs (and no, they are not blue, just an unfashionable old green. Bummer.) indicating signal strength, and I always get one or two of them. I have seen signal quality significantly better when at a five-leds connection.
Prices and speed are more or less in-par with Mexico's near-monopoly TelMex; I'm paying about US$40 for a nominal 1Mbps/128Kbps connection (512K guaranteed, whatever that means). The upstream data flow _is_ shaped to 128k, but the downstream speed is not - when the network smiles on me, I get up to 2Mbps. It is not common, though.
I understand E-go (back then called I-go, don't ask me why) was praised as the world-first massive WiMax deployment - Even before the standard was finalized. There are several aspects of the installed network that show clearly the gear is pre-standard (i.e. extreme sensibility to position changes - If I move my RSU over two centimeters, it has to resynchronize with the antenna. This process takes around two seconds, so no big deal).
To me, clearly, the reason it hasn't got more popular is because it is owned by a relatively small company, and has not had the muscle to stand in front of Telmex's publicity machine.
Of course, we benefit more than DSL users from having a low client density :) E-go owns 20MHz of spectrum, which allows it to give a theoretical maximum of 70Mbps to a given area. If many too people were to subscribe, each client would have much less effectibe bandwidth alloted.

There's nothing magical about WiMax. Other frequency ranges, other protocols, that's about it.

The only interesting thing about it is that it's not operated by traditional telcos.

But remember, what traditional telcos sell is not telecom, they're SELLING UBIQUITOUS telecom.

An untraditional telco would have to sell at a nonzero price ubiquitous. If they sell at zero price (or truly flat rate), a smartass will monopolize all access and resell it at real market price (what people are truly willing to pay). If the service is only sporadically available, no one will want to pay for it, or they would be better off setting up a fix line connection at the only place it works. If they comply to the two conditions, they are definitely traditional telcos.

In the long run, WiMax is bad for the consumer. As I explained above, the business model behind WiMax can only be the "traditional telco model". But now we have two technologies with incompatible end user hardware, incompatible operator hardware. Nokia and Alcatel Lucent will sell less copies of their products to operators, thus the price will rise. Nokia and Alcatel Lucent will ask for higher fees from the opco, guess who will pay the bill. Nokia and Motorola will sell less copies of their products to end users, guess who will pay for the relatively higher cost.

Furthermore, with WiMax vs 3G, there are now not one, but two markets for mobile data and voice. Barrier to jump from one to the other market is nonzero for the consumers. Each of the individual markets is also smaller, hence less competitive.

Fuck WiMax

He claimed ... latency rates reached as high as 1000 milliseconds. Poor latency and jitter made it unacceptable for many Internet applications and specifically VoIP, which Buzz has employed as the main selling point to induce people to shed their use of incumbent

Sounds like they didn't configure it right, on one or both of two issues.

First: WiMAX has a frame rate that is an exact multiple of the 8000 frames/second rate of the telephone networks' digital carriers (and A/D converters). While this was obviously intended to allow it to carry telephone TDM signals and their associated timing (which normally isn't an issue for IP transport), WiMAX has its own, unrelated, timing issues that mandate the base stations be synchronized - to each other and preferably to a telephony network clock or a GPS-derived clock.

The base stations assign timeslots to each remote. They measure the propagation characteristics and (depending on the sort of base station) may adjust signal strengths, modulation rates, and/or antenna aim for the associated timeslot to obtain good communication, and may pick a timeslot that is currently "quiet" on the antenna / antenna-aim appropriate for the remote in question.

The problem is that multiple subscriber stations between two base stations (perhaps not adjacent ones) that are reusing a channel may both be "audible" to both base stations - perhaps due to using non-directinal antennas, perhaps due to reflections. If the base stations assign overlapping timeslots to their peered subscriber stations they will interfere. So the base stations try to assign their subscriber stations "quiet" slots - i.e. slots that don't already have interference from another nearby base station's remotes.

Now that's just fine if the base stations' clocks are synchronized. The timeslots hold a constant relationship to each other and a quiet slot stays quiet. But if the base stations are not synchronized their relative framing drifts. So one base station's subscriber's slot may drift into that of another base station's subscriber, resulting in a drop of the link quality. Then the base stations readjust the configuration - perhaps moving the subscriber stations to new slots. But these do the same thing. Over and over. Result: Links keep flaking out and control traffic is massive.

With the base stations synchronized and the subscriber stations carrying VoIP or other fixed-rate stream traffic, the stations will tend to hold on to quiet slots that march along with the stratum-III timing regularity of telephone carriers.

The second Quality of Service issue is packet priority. The routers at both the subscriber and base stations should be identifying the VoIP (or other fixed-bandwidth streaming) flow and giving its packets priority over other traffic on the link. That way the (limited and constant bandwidth) voice packets can take the preallocated slots every time while any additional variable traffic waits for the necessary additional slot allocation. If this is not done, other traffic (such as file transfers and web browsing) will keep "stealing" the time slots out from under the time-critical VoIP / streaming packets, resulting in long and variable latencies - horrendous jitter. If it IS done (and the link is stable due to the base-station timing synchronization), the VoIP flows will have jitter characteristics virtually identical to those of telephony TDM networks.

(This, by the way, is why "network neutrality" can't be reduced to "treat all packets the same" if you want to share the same IP network between streaming services such as video and VoIP and best-effort services such as file transfers and browsing.)