A new study into the impact on broadband ISP download speeds of peak (afternoon) vs off-peak (morning/daytime) internet usage at the largest six broadband providers in the United Kingdom has revealed that Plusnet suffers the biggest proportional drop in performance, while oddly TalkTalk got faster.
Home broadband is somewhat of a “Best Efforts” service, which may sacrifice guaranteed bandwidth and a low contention ratio in order to be more affordable for residential consumers (i.e. network capacity is shared between many users and thus the performance will vary). Naturally this often becomes more noticeable during periods of heavy usage, such as when more people are online and actively using the connection.
Residential broadband connections typically come under most strain in the late afternoon and evening periods (i.e. after work), while business connections are more likely to be impacted during the daytime period (normal working hours). Exceptions can of course occur, such as during school holidays or when a major event (e.g. sport) causes a surge above normal traffic levels (can occur at almost any time of day).
Broadly speaking the impact tends to be fairly small and it’s arguably going to be less noticeable this year because more ISPs will be upgrading their capacity in order to cope with future / faster technologies, such as G.fast and FTTP/H. Nevertheless we have done a bit of new analysis using Thinkbroadband‘s latest Q1 2018 (March) consumer speedtest data for BT, Plusnet, EE, Sky Broadband, TalkTalk and Virgin Media to check.
We should point out that the following are median averages (Megabits per second) that have been extracted from the multi-download (x6) speedtest results. Take note that this may have been impacted by the Easter holidays in late March and it won’t be able to separate business from residential traffic. Off-peak here is defined as Midnight to 5.59am, while Peak reflects 6pm to 11:59pm.
Caution is also advised when looking at upload performance because those are of a lower speed and at this scale it doesn’t take much to produce a big percentage shift one way or the other, even though the actual impact upon end-users will be fairly small.
Average Median Speed Peak vs Off-Peak by ISP (Mbps)
ISP | Off-Peak Download | Peak Download | DL Difference |
BT | 35.2 | 33.2 | -5.68% |
EE | 13.1 | 13.7 | 4.58% |
Plusnet | 27.90 | 25.70 | -7.89% |
Sky Broadband | 19.1 | 19.1 | 0.00% |
TalkTalk | 17 | 19.4 | 14.12% |
Virgin Media | 82 | 76.7 | -6.46% |
ISP | Off-Peak Upload | Peak Upload | UL Difference |
BT | 7 | 6.8 | -2.86% |
EE | 1 | 2.5 | 150.00% |
Plusnet | 5 | 4.8 | -4.00% |
Sky Broadband | 3.7 | 4.1 | 10.81% |
TalkTalk | 2.7 | 3.3 | 22.22% |
Virgin Media | 8.8 | 7.3 | -17.05% |
A quick glance across the variable results shows that there’s no particularly obvious / shared pattern, which perhaps isn’t too surprising given all of the aforementioned factors that could affect the outcome. Admittedly Plusnet did suffer the sharpest decline, although from an end-users perspective it’s negligible.
On the flip side TalkTalk actually gained +2.4Mbps (14.12%) of download speed and it’s unclear how that occurred, although such a jump could in theory be caused by a large migration of users to a faster package or an upgrade in overall network capacity (assuming the network was previously suffering from a shortage of capacity).
The results can also be impacted by the proportion of people choosing to test their connections at different times of date, which may help to explain EE’s rather odd result for upload performance (perhaps more of their FTTC base than ADSL2+ were testing during peak periods etc.).
Last year’s fixed broadband speeds report from Ofcom also helped to demonstrate the kind of impact this can have (here), as visualised below across several different times of day and split between ISP packages. This is arguably one of the most accurate representations available, despite suffering from a relatively small sample size (c.2,000 homes).
UPDATE:
Shifted the Ofcom chart to the bottom to reduce confusion.
Plusnet never change, they’ve always had capacity issues at peak times, it just depends how much. Plusnet are just an ISP experimenting on how little capacity they can get away with, against how many special introductory offers they can advertise to obtain new customers to replace those that leave. You get what you pay for.
I’m on Plusnet fibre unlimited and I’m a heavy downloader/uploader and I’ve never had any drops in performance.
Hmme
Is there not another interpretation of the data?
VDSL and GFast have pretty sophisticated mechanisms for reducing cross interference between pairs. Personally I have seen big improvements in throughput on existing VDSL2+ connections as these have been implemented.
There is no such tech on ADSL and very little on ADSL2+.
So could we be seeing a build up of cross talk between the pairs as activity levels increase impacting performance? This is a documented effect.
So there needs to be a second correlation (matrix) between the observed peak/offpeak and solely ADSL/ADSL2+. Or of the stats with ADSL and ADSL2+ stripped out.
Just a thought.
Well @A_Builder, ignoring that there is some “by technology” data available, it wouldn’t explain it for Virgin Media using your randomly picked way of excusing it.
@A_Builder
ADSL and FTTC crosstalk isn’t affected by how much data is transferred. ADSL and FTTC modems just constantly exchange packets, if we aren’t passing any data those packets are empty. Crosstalk is a function of “having sync” between the two modems, and not by the data itself.
Something doesnt add up here.
On the graphs the visual drop off is visually huge on cable compared to FTTC.
The % are as follows for the 3 tiers 84% 79% 75% yet on average median download speeds performance loss it only equates to 6.46%?
That’s partly because of the differences between Cable and FTTX and where contention kicks in.
With Cable there’s a finite number of channels on a given node, which can cover hundreds of houses. Each channel can do around 50Mbps. This means that the last mile is a shared, contended, resource. If several people on the same node, with a channel set that overlaps yours, are hammering their connection the speed of everyone using those channels will suffer.
FTTC however multiplexes data at the street side DSLAM into a fibre link, which can easily do multiple gigabits. Each cabinet runs up to 255 connections. The last mile is not contended, and the link back to the BT POPs will be large.
The issue THEN becomes peering – how many gigabits of traffic your ISP can send and receive because of arrangements with other ISP’s and transit providers.
If you have 1000 customers on a 100Mbit service, but only have a 1Gigabit transit agreement to get out to the wider internet, then if each customer is using 1Mbps of traffic, well, things are going to slow down, probably packet loss etc.
So, technology plays a part, but also upstream capacity.
Nice simple way to explain things by TBB coloured figures per tech used, the Median list shows average percent across products per ISP.
If i wanted to move and wanted ADSL or FTTC id choose Sky from that list. If i wanted something that can out pace FTTC id choose Virgin
Those Virgin TBB stats must be exclusively from VM’s 200mbs + services which isn’t representative of the whole VM service. I’m on 100/6 mbs service and many are on the cheapest 50/3mbs service which now looks to be the same price at £29pm as the 100/6 + phone + basic tv service, and I don’t get that median upload of 8.8 or 7.3.
Peak upload on the faster packages is 12 or 20 mbs. The variation on adsl is much more variable and impossible to correlate with the graph above.
http://www.virginmedia.com/ourspeeds
Makes perfect sense the main upload tiers are 50/3Mb, 100/6Mb, 200/12Mb and 300+/20Mb. If there were even amounts of users on all 4 of those tiers the average would still be 10.25Mb upload.
So the 8.8Mb upload figure for upload peak average is probably about right, cerrtainly not a million miles out, does demonstrate most are on 100Mb or 200Mb Mid tier packages.
As the news item also points out…
“Caution is also advised when looking at upload performance because those are of a lower speed and at this scale it doesn’t take much to produce a big percentage shift one way or the other, even though the actual impact upon end-users will be fairly small.”
All seems about right to me.
You can always leave for FTTC if you want a faster average upload but the average download compared to what you can get on VM will suck balls. Pays your money and take your choice as they say… Moaning either way or trying to read more into things though will not alter the figures.
I dislike public ownership of anything,but there is a strong economic case for nationalising Openteach and extending Fibre Optic technology wherever it has to go,charging accordingly,and by available speed.
Anything else is just idle chattering whilst Rome burns.
R
Why not nationalise Virgin Media for this?
I am in West Wales and suffered from intermittent Broadband. Plusnet suggested Fibre – it would then be like lightening speed – first class for FaceTime chats with family. There would be no need to use the Extender. Wrong ! …don’t see much of a change at all….extender shows when service is being cut back and there are times when there is no service at all.
Are you sure this is not a hardware issue, rather than a line issue?
I’ve never had any peak time contention issues with Plusnet, in fact performance (latency/jitter) has been substantially better than it was with BT.
Very nice article, keep sharing these type of interesting and informative articles. Thank You.