They’re doing a very good job these days. I send live cricket score updates of the ongoing ICC World Cup via http://www.twitter.com/crictimes and had some problems at the beginning.

Now, I just love the service (and all other cricket fans too)

all’s well that ends well

Rails definitely can scale, I put together this presentation for BarcampLondon2:

http://www.slideshare.net/Geor.....r-web-app/

Hey thanks George. Great resource and information.

Sure Rails scales but it requires more powerful (and expensive) hardware comparing to other platforms. Mantain a huge Rails site can be a really nightmare. In my opinion Twitter developers will soon rewrite the entire app with another language/framework.

I have tried rails, and in the end i went back to the old and good apache/php framework!

Simple, cheaper and robust!

WAAAAAAAAAAAAAAAAAA!!

Sure you can process more requests per second in PHP (compared to rails), but having programmed both languages I believe rails is definitely quicker to develop in. In Western Countries people are much more expensive than a few more servers.

I’d be very interested if anyone has any figures to back up the claims about rails/PHP time to market.

my business partner’s Twitter is also missing. He can receive twitters via SMS also.

http://www.fannation.com (owned by Time) is probably a bigger RoR impementation that Twitter….

So George, your choosing quicker development times even given the slower response time to users? Hmmm… inmates immediately come to mind.

@ George. I shifted from PHP to rails a couple of months ago. Rails is definitely faster to develop on. Of course the servers are more expensive. I’m not very sure about the slower response thing though. And scaling on rails is really not a huge issue.

-
At no point should any serious developer who is developing for anyone other than himself - pursue the quickest developing time instead of the best user experience -

If you are pursueing the other - you are more a money whore than a developer

uhh, i had some issues around the same time: http://www.flickr.com/photos/b.....466672534/

this looked more like a scheduled downtime than a deletion or whatever. i am 80% positive that the message soon after did change to “we can’t find what you’re looking for” but it did come back up earlier this week.

I could not log into Twitter. Tech support said it was a bug, they fixed it and I should clear my cache.

Check out our site to see how fast Rails goes. http://www.metadot.net

I’m still struggling to figure out what the value behind Twitter is. Twitter seems to me to be a temporary business opportunity that delivers minimal value to users. No one has thought of a compelling use for Twitter that would drive thousands of viewers to “follow” that person’s Twitter. Microblogging is a step backwards and soon, everyone will have smart phones and have great mobile internet access. Why make so much of an intermediate solution that doesn’t really solve anything?

With the exception of a couple of unemployed people who have nothing better to do than read what kind of drink you’re having at the pool or what color your stool was this morning, no one else really cares. No one cares that the founder of Twitter is going to Whole Foods for the fifth time in my life. The level of self-importance of someone who uses Twitter is astounding and it’s all a big joke. Most Twitter users are tech bloggers or vloggers who have high belief in their importance and all their “friends” and “followers” are in the same sphere. Echo chamber anyone?

@George, that seems like an odd place to locate a SAN (i see it’s just for static files but still) and also that i would have to be convinced that memcatch (i’ll have to go study up on that) is reliable enough to be such a portal where you have it in the architecture.

But definitely interesting, going to read more on this. Thanks for posting your slides on RoR.

So we should be watching Twitter as the largest RoR web app and MySpace the largest .NET one, [although MySpace hasn't been fully migrated.]

@mark - In our environment we seeing RoR hitting around 500-900req/s. I met an architect from yahoo who said they run PHP servers at around 2000req/s (there’s no fair comparison here as the servers won’t be the same, or running the same code etc but taking the values as typical), then in the worse case the RoR code takes 1/500s to process and the PHP 1/2000s. The network IO is way more than this and so we believe the performance is comparable. Of course if you overload the server then you will start to see a problem.
@pallet jack - I’d much rather take half the time to do an equally good job. Would you rather take 15mins or 30mins commuting to work from the same house to the same office? In our experience RoR allows us to write equivalent code much quicker. Not matter how nice the theory everyone in the real world is pressured by either time, quality or money constraints (the classic you choose two, I’ll choose the third balancing act)
@Amy - MemCached is used by LiveJournal, Facebook and is widely accepted as very fast and reliable. In our architecture if its down then its now issue as it just falls back to the DB. The SAN was just an example for rails partials that have been cached (so can be shared between workers). For plain static files such as images they would be on your load balancer - apache/nginx/squid

The slides are slightly out of date right now, we’ve made a few minor adjustments lately so I’ll be doing an update over the summer

I attended the San Jose Ruby conf lat week where Blaine Cook from Twitter gave a presentation on ‘Scaling Twitter’. Interesting stuff..

Ezra Zygmuntowicz also gave an great talk on Merb, which is basically using Mongrel and Ruby without the Rails to vastly improve performance at the cost of simplicity

both presentations are here: http://www.slideshare.net/tag/sdforum

@george. thanks for the link

I keep reading all these posts about how Ruby doesn’t scale and not once has anybody mentioned using LiteSpeed instead of Mongrel. LiteSpeed is MUCH faster than Apache and also faster than Mongrel in my testing. This would have been the first thing I would have done because it wouldn’t require any changes to code. Also, I would have looked at Postgresql instead of MySQL. MySQL has been shown to have scalability problems when using multiple cores. I know they are working on it, but I don’t think they’ve gotten it completely fixed yet.

I enjoy Twitter, but I’ve been frustrated, because I can’t seem to *receive* updates via IM.

I can add new updates to Twitter using IM, no problem… but it never sends me updates when my friends post. I’m using OSX and iChat.

Has anyone else successfully received updates from their Twitter friends using iChat?

I too admit that I don’t really get Twitter, but I bet my parents would love having tabs like that on my siblings and I. I can already hear my dad saying “I see that you haven’t had time to get your passport….”

Apparently, all that you need to do to fix the IM Bot is to post a comment on TechCrunch.

For some reason, it works now - let’s see how long it lasts…

It seems like that had hard time scaling Twitter. But their experiences are very valuable to any Rails coder.

I wrote about 2 all time best rails scaling case studies here:-
http://www.webforth.com/2007/0.....he-maximum

I think 37signals gets some pretty heavy traffic across their apps so I guess it scales. the problem with twitter is most likely the way its been developed and the sheer volume of traffic that is hitting its servers.

a great article on scaling is the one about myspace http://www.cioinsight.com/arti.....842,00.asp

The Twitter hype has arrived in Germany: texteln.de, frazr.com, faybl.com and probably some more try to get the twitter-market in Germany.

Hope they solve their scalability problem with rails - our service is also programmed with rails

Agree Arun, I’m completely stoked about Twitter, not because I use it because I don’t. I love the fact that they are really blazing a trail in rails scalability.