• Is this your RBI speed 50%

  • Get this with Wireless 100%

Long have I argued that the government’s RBI (Rural Broadband Initiative) contract was deficient in its specification of the RBI Broadband Service.

Not only are the specifications for the service inadequate, but given the way that mobile broadband demand (as distinguished from rural fixed broadband demand) is growing, the performance of the service can only degenerate over time.  The performance backstops written in to the contract are at dial-up speeds and too-often, are those speeds attained.

From the contract:

The RBI Broadband Service will initially be based on HSPA+ technology operating in the 900 MHz band that is designed to provide the following peak speeds:
Downstream 21Mbps; and
Upstream 5Mbps.
Contention ratio
In satellite and mobile networks, where high fixed costs require it for the business case to work, the contention ratio is high. For satellite, high can mean as many as 400 users on one circuit. For 3G mobile networks, contention ratios are not published and are confused by the mixing of voice and data on the same network. In wired networks, copper and to a lesser extent fibre, contention ratios vary by service provider. With wholesale prices now largely fixed and limited opportunity for service providers to differentiate themselves by type of services offered, any price competition is paid for by a higher contention ratio.
The lowest-cost ISPs are likely to have the highest contention ratios and therefore, the slowest service. In days past, this mattered less than it will in future days where the high bandwidth demands of video will reduce the quality of service offered unless contention ratios are reduced.
An example of this is the after-school broadband slowdown.

The mobile telcos no longer quote the 21Mbps speed because they know that users can never see those speeds.  So what speeds should we expect?

If only one person is using the cell site, then peak throughput speeds may be around 16Mbps.  The difference between this and the 21Mbps peak speed is used for the overhead information required in a digital service carrying user data.

A similar argument applies to the new 4G LTE services where the telcos have stopped talking up LTE speeds of  100Mbps, and instead quote speeds “10 times faster than its third-generation counterpart” (Vodafone) and “…people are experiencing speeds of over 5x faster than 3G” (Telecom).

As an approximation, if two people use 3G networks at the same time for the same purpose (for example downloading video), then the bandwidth available to each will be half of the peak throughput rate – or about 8Mbps each.  Double that to four users (each downloading video at the same time, and the speeds drop by a half again, and so on.  The real-world situation is not so simple because rarely do people do the same things at the same time.  So there is a statistical relationship between peak available throughput (the 16Mbps) and the throughput that individual users will experience.  All telcos use that statistical relationship to determine the ratio of users to the number of circuits that  people can use the service at the one time.  This is the contention ratio.

From the RBI contract:

The service is designed to have a peak throughput of at least 5Mbps downlink and 500Kbps uplink, and dimensioned to deliver a 99.9% probability of providing to any End User a minimum average throughput of 45Kbps downlink during any 15 minute period

Whew!  What it means is that the throughput you experience on an RBI service, can drop to below dial up speeds as long as it does so for no longer than 15 minutes.  Even then, Vodafone have a way out of providing any performance guarantee through the 99.9% probability aspect.

Again from the RBI contract:

…has a mean one way packet delay of <1 second; and has no specific metric for packet delay variation or packet loss ratio.”

This is latency (one way packet delay) and jitter (packet delay variation).  Latency matters for voice over Internet (VOIP), on-line gaming and cloud services.  When the latency on my broadband service exceeds 100mS, my cloud connection is prone to dropping out.  The RBI contract provides for a latency of 10 times this value.


Vodafone Speedtest: aftershool

So that’s the background, what is the practical experience?

The Vodafone Speedtest: afterschool graphic shows a download speed of 2.55Mbps, around half of what the government promised and just over half the speed I normally get on weekday mornings.  To argue that this speed is an improvement on what most rural people get is true enough but deliberately misses the point that in tomorrow’s broadband society, this speed will not be sufficient.  It will not enable use of advanced web-based education, health and government services.  Worse, the government have no plans, no concept, of what will be done in the future to improve rural people’s access to true high speed broadband.


Vodafone Traceroute result

This speed is what is coming to define the new ‘digital divide’.  Already, our average broadband speeds (Truenet surveys put this at 11.5Mbps nationally as at February 2013) are around four times this.  And much of this is to do with the contention ratio built in to the network design.

More importantly, at least to me, is that latency at 84mS and jitter at 22mS are both high.  Latency is the delay between you pressing the ‘Go’ button, and something actually happening. The high latency on Vodafone 3G services is highlighted here by the timeouts (“*”) on this traceroute result.  The record was made immediately after the speed test was completed.

How might this be improved?

By moving to a different technology.

Like WiFi that operates over unlicensed spectrum.  Or better still, like ‘Super WiFi’ systems that would operate over the rural-friendly 700MHz digital dividend band.


InfoNet Speedtest result

The next chart records the speed tests, measured within a few minutes of the previous Vodafone speedtest.

The figures speak for themselves.

Need assistance with these issues?  call John on 021 46 36 86


InfoNet Traceroute result