UFB speed tests

A number of schools have reported to us that they don't get their access speed when running speed tests on their browsers after they connect to fibre.  

1. The stated speed os the sum of the low and high(CIR) rates... and the native speed test will only measure the low rate.
Your plan will consist of an access speed that consists of low priority (untagged) and high priority (tagged) traffic with a CIR (guaranteed rate) usually 10Mbps.  
For example a 100Mbps connection will consist of 90Mbps low priority and 10Mbps high priority.  Thus the theoretical maximum on a connection with 100Mbps access speed for untagged traffic which is usually your web etc is 90Mbps.  
Even if you do not have any high priority traffic using the 10Mbps CIR your low priority traffic cannot use this.  
If you were able to test the two types of traffic at once with two different speed tests then the sum of these would be a theoretical maximum of 100Mbps.  
This means that if your school has the lower speed 30Mbps or 50Mbps plans then you would expect 20Mbps or 40Mbps theoretical maximum speed on your speed tests.

2. Most online speed tests use layer 3 protocols which are slower.
The plan speeds are calculated at Layer 2 - the ethernet layer.  Because most speed tests work at Layer 3, TCP, then you need to allow for an overhead for the encapsulation and that can vary from between 2% and 13% depending on packet size.  So this reduces your speed test results.

3. Latency is a factor especially with multimedia.
Latency (the time taken for the traffic to travel through the fibre and the ISP infrastructure to the test server and back) will also affect the maximum speed attainable.  This is usually noticeable when it exceeds 10ms.  This can be tested by pinging the test server.  By multi-threading the TCP tests this can be minimized but most speedtest's just use two threads so are susceptible to this.
This was not noticeable on DSL speedtest's as the speeds were much slower.

4.  Test from your firewall.
The higher speed of fibre has moved the bottleneck from the broadband connection to the school server or local network in many cases.  To avoid this problem the only real test is directly into the fibre connection device/router with the rest of the school network removed.

5. Test every part of your end to end network to identify bottlenecks.
Any congestion between your RSP's network and the speed test server can slow things down.  That's why you should only test to a server within your RSP's network (such as to speedtest.callplus.co.nz, for example).

6. RBI speeds will appear higher than UBI speeds.
As standard RBI plans have a 2.5Mbps high priority CIR then standard low priority traffic speeds will be slightly higher than for 100Mbps Urban UFB.

So, to summarise, if your fibre connection is working OK then schools should expect the following range of maximum speeds on a speed test (with normal low priority traffic):

100Mbps connection: 65 - 85Mbps
50Mbps connection: 30 - 40Mbps
30Mbps connection: 15 - 20Mbps

I would challenge anyone to tell the difference between a 70Mbps and 100Mbps connection by looking at browsing speeds.

The best test of your connection is how well it delivers the required experience for your users.  This is where caching, DNS performance, and other factors also get tested that speedtest's won't pick up.