Twitter was off line long enough yesterday for me to notice (I variously rec’vd 500 pages as well as ‘Connection reset; server too busy’ pages). No mention of it in the 20 or so blogs I follow.
The 500 error indicated a back-end problem; this wasn’t just a matter of too much front-door traffic. We’ve already seen the response to that in the form of the tweety-bird page. 500 errors are the server’s way of saying “Holy Shit. I give up.”
I think Ruby’s edges (or core maybe) are showing here. This has been discussed at length at Laughing Meme, but I haven’t seen much about this subject on Twitter’s blog. Lots of folks are using Ruby daily, with many of them adopting Rails, and all of them have a stake in its future.
Performance is a tough characteristic to engineer into a language after the fact. Is Ruby/Rails going to be able to pull this off, or will it remain “…dead slow” to the point of no longer being considered for scalable web development?
Certainly, this is not the first time Twitter has had an outage. The outages were so common for awhile that they were referred to as “…yet another period of downtime” in this posting at [Techcrunch] . But didn’t anyone else notice? More likely they don’t care. After reloading the page a few times & seeing the errors, I gave up & didn’t check in again until this morning.
If your service can be offline (or at least unusable) for that long without any impact, then monetizing it might be problematic.