Just looking for some tips/advice on using WebPagetest to compare different web servers. E.g. servers from Rackspace, Linode, DigitalOcean etc.
I have setup a server on each host with the same spec, running the same PHP based website.
What’s the best practice for using WPT to compare the speeds of these servers?
Thanks
James
To be honest, this is a fairly pointless task, as how you set your stack up has far more bearing on server performance than the hardware - obviously you do require adequate ‘hardware’!
Once set up and tuned, the final bottleneck on your server will be CPU speed ( probably true for any decently coded CMS ) but only once the RDBMS is fed and watered, web server is correctly configured, etc.
Use of a CDN will also improve performance on high traffic sites by lowering the stress on the network interface as well as disk traffic, leavingyour server just to do the ‘thinking’.
As an example, I’ve just ( literally in the last couple of hours - it’s Monday morning here! ) resurrected a Magento ecommerce site, which was taking 15 seconds to respond to each keystroke in ssh, and the website was completely unuseable to all intents and purposes. I did this by better feeding the PHP opcode cacher, clearing down a directory, and moving a couple of filesystems to tmpfs.
It’s not brilliant yet, but is is now functional… http://www.webpagetest.org/result/140727_K3_JNA/
So what I’m saying is to choose SP on reputation, and compare the hardware on paper. There are no hard and fast rules, as every site has a different load, so the only sure fire way of getting the bast out of your server is to try it out.
Because of this, factor in the available upgrade paths that your chosen SP provides, that way you’re also covered when your plans for world domination bear fruit.
It’s also worth noting that a lot of what WebPagetest tests isn’t necessarily useful when comparing hosting providers. You’re a lot better off with something like apache bench or a load testing platform (just make sure to use a representative transaction load).
Thanks both, that’s very useful. I won’t do too much cross hosting testing then!