Since Google officially introduced page speed as a ranking factor in 2010 I thought I’d put together some of the things we do when optimizing a website for better performance. As a starting point (besides of actually accessing the site and get a feeling on how fast it really is) I’d usually go with the Google Webmaster Tools > Labs > Site performance tab:
GWT basically tells you what the average load time of that specific page is. In addition Google compares this to other sites being monitored. If you scroll down on that page you also see details on a per URL base to optimize. This might look like this:
So I suggest we try to get rid of those, shall we? All right, here we go:
1. Reduce number of HTTP requests
CSS: In general, CSS files should be put on top of a page – this helps pages appear loading faster because of progressively rendering. Also you really should have just one single CSS file not dozens of them. Remember: Its one request vs. X.
Images: Besides of just serving the images you really need there is one thing that does have a massive impact on performance – using CSS sprites. I’m not going to dive too deep into this one because it’s a quite complex topic. On a very high-level: Using sprites means combining an “unlimited” number of (small) images into just one and then only display parts of that “big-image” by using CSS (positions). This technique saves the overhead of having to fetch multiple images and the server does just have to do a single HTTP request. There is detailed explanation available over here.
2. Cache & compress your files
To make sure files that haven’t changed since a visitor downloaded them previously you need to work with appropriate header tags. Since web pages are getting more complex these days the number of components which have to be downloaded with each page view does increase. By implementing the correct header tags you are going to make sure those components are cacheable and won’t be downloaded on subsequent page views.
According to Yahoo!’s best practices for speed up your website: For static components: implement “Never expire” policy by setting far future Expires header. For dynamic components: use an appropriate “Cache-Control” header to help the browser with conditional requests.
Doing so you need to keep in mind, that you have to change the component’s filename whenever the component changes – otherwise the visitor will keep the old (and locally cached) file which might end in lack of functionality or other unwanted behavior.
Depending on the components size it’ll take a specific amount of time to download it to the visitor’s computer. Using GZip compression the load times can significantly be reduced which also results in less traffic per request. Since we already do have a great article on State of Search by fellow blogger Louis Venter I’m just referring to his post on how to set up compression in Apache or IIS.
Another way to speed up file delivery is using a CDN (content distribution network) – the idea behind a CDN is to distribute static files (like to above mentioned ones: CSS, images, flash, etc.) to a large amount of servers on different geo-locations. CDNs usually do the caching and compressing stuff anyways – so if you can’t do that stuff your-self it might be worth considering; even more if you have a lot of international visitors.
3. The server-side
Another area to tap in might be “the server-side” of your website. Obviously this very much depends on what setup you’re running (Unix vs. Win) but let’s assume we’re on a “classical” Unix box running Apache, PHP and mySQL.
Looking at this from a web-servers perspective: Of course Apache is pretty much the “to-go with”-solution if you want to keep it simple. However if your sites does serve a lot of request (or you’re on a smaller box but don’t want to change for whatever reason) you should really have a look at nginx which is a HTTP and reverse proxy server. A lot of high traffic sites including rambler.ru, wordpress.com and others do use it for a reason. Check out this article for a nice comparison.
Moving forward to PHP you’d probably consider some kind of PHP accelerator to speed things up even more. Most of these accelerators work by caching the compiled byte code of PHP scripts to avoid parsing and compiling the code each time a site will be requested. The code will then be stored in shared memory to reduce the amount of (slower) disk reads. Check out this list – if I’d choose one I’d go with APC.
And last but not least – the database: A lot of self-developed database-driven sites do suffer from poor performance. Mostly this happens because simple “SELECT“-queries don’t use appropriate indexes and stuff like that. As a reference, go check out this top-10 list on mysql.com.
In addition to that it might also be an idea to setup a static file cache before even serving dynamic PHP scripts accessing the database. If you’re on WordPress for example, go check out W3 Total Cache which does exactly this.