Will faster user bandwidth fix your website performance woes?

It's a commonly held misconception that "faster bandwidth will make everything better" (whether that's faster home connections over fibre, or the faster networks for mobile).

So what is the impact of different connection speeds on an average webpage?

Well, we can easily test this using the free WebPageTest tool which can simulate different connection speed profiles (different bandwidth/latency combinations that match common home user or mobile connections).

We tested a popular IT news website www.theregister.co.uk and you can see the results below.

We can draw some interesting observations from even this simple test:

  1. Even though download bandwidth increases 333% between DSL (1.5Mbps) and Cable (5Mbps)  the page load time performance gain is only 12% not 333% as some might expect (( 4.409sec - 3.87sec) / 3.87sec = 12% ).
     
  2. Similarly 1330% improvement in download bandwidth between DSL (1.5Mbps) and Fibre aka FIOS (20Mbps) the page load time performance gain is only 32%.
     
  3. The performance return on increasing bandwidth is diminishing… a 333% change in bandwidth gave us 12% but 1330% change only gave us 32% improvement… so even though relative bandwidth went up ~400% the page load performance only went up 260%.
     
  4. DSL and Mobile 3G have roughly the same download bandwidth (1.5Mbps versus 1.6Mbps) yet the mobile performance is 3 times slower… which shows that LATENCY (300ms versus 50ms) is a bigger determinant of performance than simple bandwidth!

Connection speed graph Figure 1 - load time for www.theregister.co.uk at different connection speeds 

 

Bandwidth and Latency

Load Time (Empty)

Load Time (Primed)

FIOS (20/5 Mbps 4ms RTT)

2.998

1.164

Cable (5/1 Mbps 28ms RTT)

3.87

1.346

DSL (1.5 Mbps/384 Kbps 50ms RTT)

4.409

1.664

Mobile 3G (1.6 Mbps/768 Kbps 300ms RTT)

13.111

5.637

Table 1 - load time for www.theregister.co.uk at different connection speeds

So what conclusions can we draw from this simple test?

Firstly, you need to test your websites at different connection speeds to understand how your site performs for your slowest connected users (NCC Group can monitor at different connection speeds or you can trial our performance analyser tool). Your site might be considerably slower for users of lower bandwidth / higher latency connections

Secondly, increasing bandwidth/lower latency WILL make your site pages load faster… but not as fast as you might think AND there is a diminishing return as latency becomes more important than bandwidth.

Thirdly, given that bandwidth increases aren't likely to solve all your performance woes for you, investing in web performance optimisations (WPO) like reducing the server round trips required to load your webpage, image compression and optimisation, and structuring your page for optimum delivery are even more important. If you want to learn more about "practical web performance" check out our web performance presentations and whitepapers!

Published date:  05 June 2013

Written by:  Website Performance Expert

comments powered by Disqus

Filter By Service

Filter By Date