[sclug] Wireless Broadband in Slough and Windsor areas

[Roland Turner]

James wrote:

> I think you've got a point, we use a small wireless provider at work and
> to be frank the latency is pretty crap. They only serve a small area, so
> can't really afford to buy a decent leased line. Result is that lots of
> people end up fighting for little bandwidth. 
>
> User, bandwidth ratio was also a problem when I worked for a slightly
> larger wireless ISP a couple of years ago.

after Simon wrote:

>> The prices seem similar (well ok, a little less) and the area they are
>> targetting gets ADSL already.  I'd have thought you might have lower
>> latency over ADSL than wireless.

Simon didn't explain his reasoning, but it occurs to me that he (or others) may be confusing the high-latency experienced in satellite (particularly geostationary satellite) service compared to wired/fibre service with higher latency in wireless services in general. (Apologies in advance if I'm mistaken.)

It is certainly the case that a bidirectional service with a geostationary satellite (think 2m dish) introduces about 0.5s of round-trip latency in speed-of-light delays alone. Given that first tier backbone providers typically guarantee maximum latencies of 55ms intra-continent and 80ms trans-atlantic, wire/fibre connections are often close to an order of magnitude less latent than satellite connections.

The same problem does not apply to wireless local loops in which the distances are relatively tiny. (In fact, for such services, wireless has a marginal advantage; an EM signal (light, radio, X-ray, ...) passes through the Earth's atmosphere at essentially the same speed that it passes through hard vacuum: C or 3*10^9m/s. An electrical signal passing through copper is constrained to copper's impulse tranfer speed, which is about 0.7C or 2.1*10^9m/s. This difference is truly marginal though; to see a single ms of difference, you'd need parallel single-hop wired and wireless paths of 9*10^8/10^3m ~= 900km.)

A bigger issue is the technology used to control shared access to the medium: TDMA, CDMA and other similar schemes typically used by mobile telephony introduce no latency. CSMA/CD (Ethernet) and the like (802.11*) do introduce local latency which increases rapidly with sustained increases in local traffic beyond about 80% of available local capacity. ADSL providers avoid this particular problem entirely, at the not-insubstantial cost of setting a relatively low upper bandwidth limit (2Mb/s?). Wireless providers using 802.11* and wired providers using CSMA/CD (e.g. broadband IP over cable) are both susceptible to this (but N.B. when I lived in Montreal I was able to sustain 8Mb/s transfers over cable... With cable, the limiting factor tends to be the choices made by the carrier, not the technology itself).

Netvigator claims not be using WiFi bandwidth - it claims to have seperately licensed spectrum - but it's website lacks clues about the technology in use. The nett result is that we can't infer much about latency simply from the fact that they are wireless.

The problem that James refers to applies to any access provider; if the upstream bandwidth/user ratio is too low, then users will perceive both a drop in bandwidth and an increase in latency. A wired provider with too low a ratio will perform far worse than a wireless provider with an adequate ratio. What we probably can assume is that many WiFi providers are under-resourced and/or under-clued (assuming that knowledge about IP, routers and APs is enough to build an access provider is a pretty easy mistake to make, and the low cost of entry means that such providers are likely to appear from time to time) and therefore have poor ratios, but this is not an unavoidable consequence of wireless, and is not indicated in the case of someone who has gone to the expense and difficulty of securing a spectrum license, as Netvigator claims to have done. I'll be watching them with interest.

- Raz