3 Nov 2009
I have recently been conducting a little research into hosting companies/ISPs/data centres to understand more about their speed.
One hosting provider in the UK, UKFast, has recently been marketing the advantages of speed as a prime factor. Consequently, they have allegedly invested 25% of their profits year on year into improving their internet connectivity while at the same time ensuring that they never exceed [by that I infer “sell more than”] 40% of total bandwidth available*. Fair play – we all like stuff to be faster. I was also pointed to a 3rd party web site who provide speed measuring of UK-based web hosting providers – WebCop.
* I was told this by a UKFast sales representative.
I was interested by WebCop’s claims, namely that by distributing their testing servers across UK-based hosting centres, they eliminate bias of one or another datacentre and concentrate instead of the actual, average throughput delivered by them. It’s a fair claim, but there could be issues. Today, I sent them this message:
I’m interested by your web hosting speed statistics, for two main reasons.
Firstly, there isn’t much info on your site about how you conduct tests – e.g. which web sites are used to measure the hosting companies relative speed. This concerns me, as hosting companies can easily make the most prominent web sites the fastest, putting them on the front line of the data centre, while allocating less bandwidth to smaller web sites.
Secondly, you don’t mention from where you test each hosting company’s sites/servers. So, for example, you could be testing a London-based server using servers in Manchester and Leeds, but the contention in one direction may be significantly higher than in the other direction. Therefore, you could have skewed results. In addition to this, if one hosting provider/ISP has a faster network, how can you prove this by testing on their competitors’ slower networks?
I’m looking forward to hearing back from them. Currently UKFast appears to have leapt ahead in terms of the speed ratings, according to WebCop.
Good question. I ran a #whois on webcop.co.uk and found that the domain is registered by a company in the Netherlands who has a POBox address in Gibraltar! Because whois output is subject to Nominet copyright, I cannot redistribute it here. But if you want to see it, try www.123-reg.co.uk.
I have tried to dig a little deeper; the web is very unrevealing of a company that seemingly wants to stay hidden. I did find out that UKFast’s sister brand, GraphiteRack.com, registered their domain name through ENom, the same registrar that WebCop used, but nothing more.
The public-facing WebCop server seems to be hosted by Tagadab.com, a Clara.net Group company. Interesting that a company (WebCop) with testing servers distributed across the UK, use a London-based ISP with only 6 IP addresses allocated from IANA and some very “comptetitive” prices. Perhaps they want to keep their web traffic well away from testing servers…
5 Nov 2009
Not heard anything from WebCop yet…
9 Nov 2009
I got a reply from WebCop:
Our testing servers are located on many different networks and effort has been taken to ensure that they are also geographically evenly located throughout the country. This means that if we test a server located in London it will be tested from all over the country and the average result given. This allows us to display results that are averaged not only across different provider’s networks but also across different geographical locations.
As for your first point, we are currently addressing this and looking to find the best way to ensure that providers don’t cheat in the same way we know they do for Webperf testing. Currently for the larger providers we test a server located in the standard customer space and not their main website, and for smaller providers we test their company website server. We are looking for a way to make this fairer and are working with some of the larger providers to do this.
On the surface this is a fair strategy. However, it’s very, very easy for a data centre to prioritise traffic to/from a particular machine. My feeling is that this could be happening already although, of course, I can prove nothing.
My gut instinct tells me that if the majority of datacentres in the UK felt they could increase sales by claiming the fastest network connectivity, they would.
However, every UK datacentre (apart from one) seems to hover around the same speed of connectivity, which suggests that either the system of tests is not respected amongst the datacentre community (in other words, it isn’t perceived as being particularly accurate), or the service provided by one is much faster than the bigger ISPs with which it peers… which seems rather unlikely.
I respect the WebCop team for this endeavour, but strongly feel that until the testing methodology is properly published for the networking and datacentre community, there can be little value in its findings.