I am finding that a number of us at DTS are regularly being asked by clients, and by potential clients during the sales process, what they can realistically expect in terms of speeds from UFB GPON services, and in particular, the Bitstream2 and 2a class of services.
Service schedules for these services may state that the circuits are capable of up to 100Mbps or 200Mbps symmetrical speeds, but given these services are largely best effort, what will actual speeds be?
DTS now has UFB handovers with Chorus in 10 regions, UFF in 3 regions, Northpower in Whangarei, and Enable in Christchurch. The first point to make is that across these 4 providers we do not see a wide ranging set of speeds experienced, that is to say, there is no clear difference in the speeds experienced by clients across regions and/or suppliers.
Almost without fail, the first thing almost every connected client does is run a speed test to a national destination, and in virtually every case the speed reported is within 10Mbps of the maximum attainable service speed in both directions.
There is certainly fluctuation from that point on, but we have not seen any circuit dip below half of its potential throughout in any tests run in any region.
That is the reality now, but as advertised circuit speeds continue to march towards 1Gbps and customer uptake increases resulting in splitter cabinets becoming more congested, do ISPs need to be re-enforcing the best effort nature of GPON services?
The first clear point we tell clients is that if they have a need for committed bandwidth over and above 2.5Mbps, the BS2 and BS2a services are not for them. In that case they need to cast their eyes over to BS3a and BS4/P2P/HSNS services (lots of different names for essentially the same thing).
But, if all they are looking for is faster browsing and fast access to non-real time applications, then BS2 and BS2a services are going to be better than any copper service they currently have, and are also likely to be much faster than existing 50Mbps committed services more often than not. What we can’t tell them is how much faster these services will be or for how long.
The other point we make is that a 100Mbps circuit does not mean that will be the speed reached to international speedtest servers or when downloading international content via a single thread. We often point people to this old post of ours for more information on that particular point.
All said and done, this is a best effort service, but speeds to date are pretty sharp, and certainly compare well against advertised headline speeds. The lack of clarity on the actual speeds one can expect over UFB services is offset by the cost savings it provides when compared against P2P class services such as HSNS.
Brendan Ritchie (@bcarmody)