Seen in a brochure at PayPal X:
LinkedIn uses Joyent infrastructure to run the largest Ruby on Rails app with over 2 billion monthly pageviews.
- I thought LinkedIn was Java, when did they switch to Rails?
- 2 billion monthly? When I left my last company, our (PHP) app was doing over 7 billion monthly, and wasn’t even in the top 10 PHP applications out there. Surely there are Rails apps bigger than that.
12 thoughts on “Largest Ruby on Rails app?”
I work on a social network in Spain, called Tuenti. It's PHP and we have 1 billion hits daily. I'm pretty sure this data of Linked In is incorrect, it can't be just 2 billion monthly.
2 billion monthly seems accurate for LinkedIn's total number of impressions if LinkedIn is only in a vertical and not a large market social network. This would probably put LinkedIn in the top 500 but out of the top 100. Recently, however LinkedIn has grown steadily into the top 25 by some metrics. To do this, you are correct and these sites are estimating way over 2 billion monthlies for the company.
In any case, if pages served by Rails is only a fraction of their total and that fraction is at 2 billion, there should be many sites that are bigger Rails sites than them.
Were they just talking about Signal, LinkedIn's mashup with Twitter? It is Rails running on top of JRuby. I think there are other parts of their site that also run Rails/JRuby. The backend to Signal is Scala, which like JRuby runs on the JVM. I think most of their backend is still Java, but I think they are generally friendly to anything that runs on the JVM.
Doubtful. 2 billion monthly sounds like LinkedIn's entire traffic. I've never seen LinkedIn high on any metrics sites until this year. (Then again, since they have broken into the top 100 this year, I will grant that I might be wrong and you right.)
They do make a good amount of money, however. 🙂
There's a tech talk / meetup tonight talking specifically about LinkedIn's transition from Java to JRuby… http://www.eventbrite.com/event/933822087
JRuby is not Ruby on Rails.
They're probably running Rails on top of jRuby so they can transition they're Java classes/libraries, etc.
Good point. My point is that if you run all your Java code on JRuby, you do not get to call yourself a "Rails site" let alone "the largest Rails site." If they've doing this for a few years now, I would not be surprised if the entire front-end is using a Rails framework. I'd be shocked if they made a complete transition in a couple months—architectural changes are costly and slow (ask Delicious and Friendster—both (IMO) failed transitions to PHP). At what point would you call it "Rails" and not "Java"? Seems like the heavy lifting is not in Rails, and if it is, you're doing something very wrong (see Twitter/Rails or Facebook/PHP or Yahoo Search/PHP or…)
In any case, nobody noticed that the bigger problem I had (in fact, the title of my post) was that I don't think 2 billion monthly counts as the "largest Rails app" by any measure. I've worked for at least two companies (not Rails) that have over 3x those monthlies and both employ less people than LinkedIn. Two billion monthlies is simply not an impressive number anymore—heck Facebook alone probably does many times more than 2 billion each day!
Even if we discount Twitter as a Rails site (I don't know if that's fair), we're still left with Hulu and YellowPages.com (off the top of my head) which I think would be excellent candidates for a Rails site breaking a 2 billion monthly count. There's got to be a couple Facebook apps/games written in Rails that do that right?
LI's traffic must be a lot higher than that. Alexa has them listed as #22 and eBay at #23. I know that we do more than 10 billion URL requests today per day. That doesn't count static files (images, css, js) but does count API calls and Ajax requests. So the number of "hits" is definitely lower, but it's still more than 2 billion per day, not to mention 2 billion per month. Or maybe Alexa is wack and LI should be a lot lower than #22.
My number of 7-billion/month for Tagged (when I left) comes from counting top banner ad refreshes. Statics are not counted. It takes multiple Ajax requests to generate a single page. If the user does a lot of activity on a page without leaving, the banner ad will be refreshed via ajax on a user action. Thus the number of dynamic requests is much, much larger than this page count, but only by a small factor: In other words, that comes out to well less than 1 billion/day according to your metric—as it should because that site barely scratched the top 100 at the time.
Alexa estimates that LinkedIn has the traffic of around 2 billion a day. I don't know where 2 billion/month that Joyent is reporting comes from. Perhaps it was a mistranslation? I find that a little weird since Joyent is in the business of monitoring traffic.
Most likely is your original guess: only a tiny fraction (say 1/30th) of LinkedIn's consumer facing infrastructure is on Joyent (and Rails) and that is doing 2 billion/month. In which case my questions are valid ones:1) Is LinkedIn on Rails? and 2) doesnt that mean there are a lot of larger Rails sites out there?
JRuby is just another Ruby interpreter. It can run Rails, and often does. Many Rails sites with no Java code in the app whatsoever run on JRuby for one reason or another.