October 2013 Broadband Report
Telecom Tops VDSL Speed Measure
Telecom VDSL is now the best performing VDSL provider.
Ultra Fast Broadband using fibre technology show speeds far superior to any other technology.
Vodafone Cable had the shortest Webpage Download times from the TrueNet test pages. On live pages, Orcon fibre is the best.
International Latency has increased significantly from last month, with results across all technologies showing increases reaching over 200 milliseconds.
Summary of Performance Measures - All ISPs
TrueNet uses Medians to demonstrate results for each probe due to the skewed nature of the distribution of test results.
Table1 test results show Fibre 50Mb/100Mb to be significantly faster than any other product, but still under reaching the advertised speed.
Vodafone Cable 10Mb/100Mb is experiencing problems in reaching advertised speeds on half of the probes. However, Vodafone Cable test page download times are the shortest.
30Mb/s fibre service is reliable and predictable. For both Orcon and Snap, TrueNet measure peak evening speeds at better than 99% of the off-peak speed. Although peak speed of 27 or 28Mb/s is less than the advertised 30Mb/s.Table 1: Summary of all Performance Measures
*8pm Mb/s speed for ADSL and VDSL is the average for all connections of each technology.
Speed (File Download Only)
Capped & Uncapped ADSL Speed
Capped & Uncapped test results are combined again because performance of both is very similar. ADSL tests are based on a Wellington 300k file downloaded every hour by every ADSL probe.
Chart 1: The top performers are Snap, Xnet, Telecom and Orcon, with all four ISPs achieving consistently high File Download speeds even during busy periods (evening). Xnet's ADSL service, during the evening busy period, has improved noticeably from last month.
Vodafone's ADSL service, during the slowest hour, has improved from last month to exceed 91% of their maximum speed by time of day, placing them ahead of Slingshot which dropped from 93% to 90%.
All October ISP results are above 90% again, which was once used as the target for acceptable performance. Continuing improvements like this demonstrate the value of TrueNet reporting.Chart 1: ADSL Speed by Hour
VDSL tests are based on 1MB file downloads from Auckland and Wellington every 5 hours by every VDSL probe.
This month the best download performance for each test from our servers in Auckland and Wellington is selected. The results show the median of best download times.
Chart 2: VDSL results for Telecom are a major improvement on last month as predicted, with a performance superior to that of Snap which in previous reports was the best.
Unfortunately we still have one too few Voyager VDSL volunteers to report results. We would like more to volunteer here.Chart 2: VDSL Speed by Hour
Fibre & Cable Speed
This month Fibre and Cable have been combined because we show actual speeds for these technologies, and until more speed options become available for fibre the chart is not cluttered.
Neither fibre nor cable are dependent on the location of the home as is the case with DSL allowing direct comparisons of speed.
We compare Fibre & Cable with average VDSL to show a direct comparison, but note that VDSL is dependent on location and average speed does vary probe to probe.
Chart 3 uses a single colour for each technology, the dotted lines show the slower speeds of each technology
VDSL average speed from an Exchange is slower than VDSL from a cabinet (FTTN). This is because copper line lengths are usually shorter from a cabinet than an exchange.Chart 3, Fibre, Cable and VDSL Actual Speeds
Cable test results for the 100Mb/s service cannot show the true picture. TrueNet uses the median of all tests, but in this case there are two nearly equal sized groups of results, also known as bi-nodal. This is shown in Chart 3a.
One group has speeds of about 100Mb/s, the other achieves just 20Mb/s. Specific checks with these volunteers confirm that they pay for 100Mb/s, they are connected correctly, and they are located in both Wellington and Christchurch. Further research is required to identify the cause.Chart 3a, bi-nodal tests
In chart 3a showing the percentage of tests in 10Mb/s divisions, there are two "nodes" for tests from the Wellington Server clustered around both 20Mb/s and 100Mb/s, this is a bi-nodal result. A median or average would be around 60Mb/s which provides little information about what users are experiencing.
Tests that reach speeds of 100Mb/s are confined to the Wellington server, suggesting that any website hosted in Auckland, and only peered at APE, is likely to be no better than 20-30Mb/s for all cable services.
Vodafone is working on identifying the issue(s), but it looks likely to be related to the article we published in September. Half of our probes appear to have been corrected since September.
Webpage Download Time
TrueNet Webpage download tests are separated into two groups, test pages and live pages. The charts show time to download all pages, measured in seconds and summed together in a total download time. Speed has a minor impact on webpage download times, latency and DNS response times are also factors.
Chart 4: Testpages are slightly faster on Cable and Fibre, and between the two, show similar total download times. Vodafone Cable is the quickest overall.
VDSL & ADSL testpage results are similar, but Snap VDSL stands out as quickest overall, with ADSL being similar for all apart from the poor performing Telecom ADSL and VDSL.
Note Telecom are slowest due to the time taken to download international testpages, otherwise they achieve similar download performance to other ISPs for NZ pages.
The sample size for Maxnet is too small to be included in this analysis - we need more Maxnet volunteers.
Chart 4: Testpage Download Time in Seconds - National & International
Chart 5: The live webpages downloaded in October include a few pages from very popular websites (Trademe, Stuff & AirNZ and a selection of banks). The sample size for this analysis is limited to those volunteers allowing larger samples to be collected, this reduces the number of participating ISPs further, and eliminates Xnet.
Fibre probes download the pages a lot sooner than other technologies, with Orcon fibre slightly faster than Snap.
Vodafone Cable takes less time than their ADSL, with a performance similar to VDSL.
Telecom & Snap VDSL services download webpages quicker than their respective ADSL services, suggesting a consistent advantage to VDSL vs ADSL for browsing.
The range of results for ADSL is large, with Snap significantly quicker than all other ISPs.
Chart 5: Live Webpage Download Times in Seconds - National & International
For some ISPs International Latency to the USA (Dallas) has increased significantly from last month. eg Snap has increased from 180ms to 240ms. Only Orcon and Xnet have Latency results less than 200ms.
TrueNet will undertake a longitudinal study later.
Chart 6: Latency Performance
Domain Name Server (DNS) Response Time
The conversion from a Website URL (www.truenet.co.nz) to its IP address (126.96.36.199) can slow webpage download times. Investment by some ISPs in more local DNS servers improves performance.
TrueNet measures DNS response times to provide additional information about the variables contributing to webpage download delays.
Chart 7 and 8: For this report TrueNet measures the time taken to receive a response from all of the ISP's DNS servers, and records the best result for each hourly test.
Chart 7 splits results by ISP, and Chart 8 splits them by Technology.
Missing bars indicate insufficient probes in that region.
Orcon are best in Waikato, Snap are best in all other regions.
It should be noted that Snap, despite being located in Christchurch, have the fastest DNS reponse time in Auckland.
Chart 7: Domain Name Server Response Time by ISP
In summary, faster technologies are quicker at DNS lookup. ADSL is always worst in all regions, and fibre fastest. DNS lookup times are likely to be dependent on upload speeds, and this is evident here, with the order of performance matching upload speeds.
Chart 8: Domain Name Server Response Time by Technology
Volunteer to test in Australia here - (any technology)
Details on how we measure are available on our Technical page.
ADSL, VDSL, DSL - the standard broadband service provided over a telephone line from an exchange or a cabinet (FTTN), VDSL is a faster version than ADSL. They use similar technology and backhaul, so sometimes DSL is used when referring to both.
Capped Plans - the most common ADSL service, where you have a monthly plan having a GigaByte (GB) limit of usage each month before your speed is slowed or you must pay more.
Unlimited Plans - ADSL service where there is no monthly limit on the amount of data used. Specifications for this service include that it may be "Managed" and have "performance reductions applied during peak demand periods."
Cable - Cable is offered by Telstra & Optus, and is available in a limited number of cities.
DNS - Domain Name Server. As the Internet is based on IP addresses, a DNS service translates domain names into the corresponding IP addresses.
DSLAM - the exchange or cabinet based equipment that your modem is connected to, over the pair of copper wires that are exclusively allocated to your premises.
Ethernet - The wiring used to connect computers to a network, typically an Ethernet cable is coloured (often blue), with small square connectors at each end.
ISPs - TrueNet has probes measuring almost 20 ISPs but only reports on those where there are 5 or more probes working during any particular month.
Latency - The time for a packet of data to be returned by a remote server to the probe when a "Ping" command is issued. TrueNet sets targets for maximum median latency that are known to be achievable.
Median - The Median is found for each probe and this is input to any analysis to calulate the average performance. This means that any result represents the “middle” performance measure applicable for that probe. Using median ensures that the result is more representative due to the often skewed nature of measurements by probe.
Speed - Throughput or the median peak connection speed achieved during our standard test downloading an image from our test servers. TrueNet normally reports speed as a comparison at low vs high demand times to show any capacity constraints evident in speed performance, often called the Time of Day analysis.
UFB Fibre (NZ) - Ultra Fast Broadband connections are the service offered by some ISPs over the Fibre to the Home (FTTH) network built by LFCs over government subsidised fibres. Services now being offered include 100Mbps and 30Mbps.
NBN Fibre (AUS) - "NBN co is a single entity rolling out fibre nationwide and then wholesaling it to ISP’s" from a good comparison here
FTTN: is based on fiber-optic cables run to a cabinet serving a neighborhood. It uses existing coaxial or twisted-pair infrastructure to provide connections from the cabinet to the home.
FTTH: Premises are connected using a gigabit passive optical network (GPON). A fibre cable, known as the "drop fibre", goes from the premises to the street. The "drop fibre" cable joins a "local network" which links a number of premises to a splitter in the fibre distribution hub.
LFC - Local Fibre Company. These companies are rolling out FTTH connections subsidised by the government, but must sell services through ISPs.
Webpage Download - TrueNet maintains a Standard Test page which is used for measuring the time to download the entire page. This page is visible here, we use a copy located on our test servers for test downloads. The time to download excludes the time for a browser to generate the page on a screen, some are faster than others.