The weakest link

Not long ago, I confessed that I had become a subscriber to BT’s Infinity (FTTC) service.   The upgrade in capacity was substantial – in my case, increasing download speeds from about 8 Mbps to more than 70 Mbps.  I’m obviously pleased to have so much more power under the bonnet, and I’m still amazed BT managed to squeeze that extra power out of an old fashioned village network. And my turbo-charged system certainly does do some things a great deal faster.  And yet…  And yet, if I’m honest, the majority of my web browsing seems little affected by the extra power available.  I still see frequent messages like ‘Waiting for XYZ website..’ and ‘Loading…’  This is all just a bit unexpected – and a trifle disappointing. 

Fortunately, there is some analysis of this phenomenon in an article in this week’s TechPolicy Daily, itself based on data from Akamai’s respected ‘State of the Internet’ reports.  The thrust of the article is that ‘browsing speed’ is the product of both network capacity and non-network factors – the responsiveness of web servers, web browsers and end user devices.  According to the author, this means that “increasing network capacity has a minimal effect on the user experience of surfing the web.  In fact, upgrading from a 10 Mbps broadband connection to a faster one is unlikely to produce a perceptible effect most of the time”.

 While the fact that network capacity is not the only determinant of a faster web seems pretty obvious, the surprise is that those non-network factors are far more influential in the browsing experience – up to 15-20 times more important. Akamai has logged these ratios internationally for the first quarter of 2014 and some of their results are shown below:

 

Nation	Ratio
Japan	7.4
Canada	8.9
Germany	8.3
US	11.2
Italy	6.4
France	8.3
UK	16.0
Ratios of non-network to network factors in web page load time. (Source: Akamai)

 Two results from this analysis are particularly striking: first, the dominant contribution of non-network factors applies equally in markets with high speed networks – such as Japan – and those, like Italy, where broadband speeds are generally lower.  Secondly, the UK appears particularly susceptible to non-network delays in browsing: the ratio of 16.0 reflects the fact that while UK network delays average only about 0.3 seconds, non-network factors add a whopping 4.8 seconds.   

Sadly, the reasons for the UK’s extreme result are not explored in this US publication but the overwhelming message from the analysis is clear: a faster internet is going to rely not only on network capacity but, even more crucially, on ”faster servers, faster laptops and smartphones, better organised web pages and faster browsers”.  Our current infrastructure planning, dominated as it is by network and broadband issues, perhaps needs to acknowledge that more clearly.