Warning! Disturbing Images Of website gore. If you read this post, you will be tormented by terrible nightmares of your precious websites been eaten by social multifaced beasts.
I just had my first site crashed by the sheer power of the facebook like button.
Most webmasters are familiar with the “digg effect”, or at least the effect of any major social site that suddenly sends thousands of hits per second to your site. It doesn’t have to be digg.
In order to explain this post better, I whipped up some crude shots of how I imagine the two effects.
This is how I imagine the Digg effect:
Yeap, this is what happens to a site when it gets frontpaged on Digg, or any other big social cesspool like Reddit or Stumble. Of course there are the pictures of burned servers etc, but that is the aftermath. This is the action shot.
Now, here is what happens to a site when it reaches the level of 4000 likes per hour, the Like effect:
It doesn’t seem that bad, does it? Well, lets go deeper and you will understand. When you implement the open graph protocol, you basically add some metadata to your pages so that facebook knows what to show on a user’s stream.
On your server logs, you will see this dreaded line of text:
Browser Type: facebookexternalhit/1.1 (+http://www.facebook.com/externalhit_uatext.php)
That is not a visitor, that is the facebook bot pulling the metadata, every time someone loads a facebook page with an external like line.
So, if for example 100 people liked your page, and for simplicity’s sake lets say 2 of each of those people’s friends visited their profile page, you would have instant 300 metadata pulls off your site. I am sure that facebook has some sort of a short-lived caching in place, but from the practical data that I see, the metadata gets pulled quite often.
On my case, my moderate VPS crashed 10 times in a single day when people started liking and visiting and re-liking and spreading the word. I think, that facebook’s multiple servers were pulling data from my website from different locations, like the hands on the picture. Simply put, a DOS attack.
And if that doesn’t seem that bad, remember that you actually do get visitors from facebook likes. And the like plugin loads data from facebook again, to show the like count. And images get reposted on facebook streams (when the user likes and leaves a comment, RTFM), with an average of 200 friends per user. So the images get hotlinked, the site loads for the users, the metadata gets pulled every few seconds. I believe the scientific term for this is “clusterfuck”.
The worst of all is that the caching plugin didn’t help that much. Usually, with a cache you get to survive a digg effect, but in the like effect, it pulls on the server CPU and the cache does not save the day.
We still don’t know if the facebook likes have any SEO value, and the traffic is the standard low quality social one. From what I have seen, it is not viable to maintain a powerful server for viral content with low conversion rates.
Has anyone experienced the Like effect? Have you found a way to ease the impact?
Popularity: 1% [?]
