It’s the latency, you see 20 Aug 07

I saw this question about Brightbox on a forum today:

what would be the advantages of hosting with a UK company over a US based company

One simple repercussion of hosting a site for a European audience on an ISP in the USA concerns the Atlantic ocean, which happens to be quite big and that means packets take a long time to traverse it. This results in higher latency. If your ISP is on the West Coast it’s even further away.

Some casual ping testing from a machine in the UK to a number of Rails hosts in the US showed between 95 and 161 millisecond round trip times. With HTTP ping testing I was seeing up to 327ms, which is likely due to the TCP handshake consisting of at least 2 packets before the request is even issued. From the same machine to a Brightbox (different ISPs of course) I saw around 15 milisecond ping RTT and 33 millisecond HTTP.

Average Ping RTT graph (smaller is better)

Average HTTP RTT graph (smaller is better)

With HTTP features such as persistent connections you can reduce the impact of the TCP handshake, but with these hosts your European visitors will still see up to 161ms delay before your Rails application even starts processing the request. Fragment caching won’t help you much there. With the new breed of AJAX web applications this is a huge impact on how responsive things will feel to your visitors.

This wasn’t a scientific test with double blind trials, lab coats and conical flasks but it gives you an idea of what a difference 5,500 kilometers makes.