It seems to be a favorite pastime of the web to moan about the uptime of Twitter. In March of 2008, their uptime according to pingdom was a mere 99.57%. Wah, wah, wah. Sure it could be better. But I’ve felt more connection problems with AIM that I’ve had with Twitter. Even my Comcast network connection seem less reliable than Twitter at times.
That doesn’t exonerate them from building a more stable service. Especially not considering that they have five million dollars of other people’s money to do it with and a few years of practice.
But I for one am pretty damn impressed with what they’ve managed to accomplish so far. They’ve built a simple service that tons of people have become so dependent on that the slightest flicker of the light sends them into terminal withdrawal symptoms. Talk about building something people want!
Now the problem, of course, is that they have a hit on their hands, but there’s no obvious paths to turn that hit into dollar bills (short of waiting until someone offers that 10x+ that their investors probably are savaging for).
If the growth in Twitter usage was mirrored by an equal growth in Twitter profits, the necessary investments needed for infrastructure would be self-evident. But when the money pot is an ever-shrinking gift-with-strings-attached, you can’t just blow your way out of the issue with cash.
Regardless, this is in my opinion the best of what we get from the startup culture of the valley today. An awesome new way to communicate, built very quickly, and addictive enough that people get a little loopy in their expectations. I want my free beer cold, gawd dammit!
So cheers to the Twitter guys. You guys deserve a world of credit for making that Something Awesome. Now please figure out a way to make scaling problems a cause for celebration (more users = more money) rather than condolences (more users = less money).
Jared
on 23 Apr 08If only Twitter had been built using Erlang. 5 nines of reliability beats no nines (or would you call it a point-five-seven of nines?) any day :)
All kidding aisde, that’s a pretty amazing uptime for a Rails-based app. I’d be very interested to see the future of Twitter on mod_rails, because I think that’s going to set the baseline for the community.
Marc
on 23 Apr 08But Merb doesn’t have the same problems as Rails, no.
Matt
on 23 Apr 08How does Twitter make money? No paying for accounts and no advertising. Am I missing something?
AllanG
on 23 Apr 08Pingdom’s measurement of twitter’s uptime can’t be accurate. Pingdom is probably not able to measure failures in the backend resulting is service interruption while the systems were still reachable.
JD
on 23 Apr 08I would also, as a fairly recent adopter of Twitter, like to see a logical progression of feature adds (more customization of your broadcasts, ways to stretch the character limit, etc…). Some more seamless integration with other, broader-scope social networking platforms seems only natural. As well as a proprietary package of mobile clients that take full advantage of said feature adds would be great. Am I expecting too much “uptime” or is web2.0 sort of married to the iterative release model and they’re too focused on scaling to improve the product?
JD
ceejayoz
on 23 Apr 08AllanG makes a very important point – much of Twitter’s downtime and other trouble hasn’t involved the sort of failure that Pingdom would detect (i.e. completely down HTTP).
Instead, most of the problems have been things like API access being down due to “temporary overload”, messages showing up repeatedly, massively late, or not at all, etc. If Pingdom could measure that, you’d have a much more accurate – and lower – metric.
Tim Connor
on 23 Apr 08Uh, Jared, by my count that is 2 nines (on the way to 3 nines). You think 5 nines is 99.99999% ? Hahahhahha, run the math.
Matt Radel
on 23 Apr 08I always wondered how Twitter made any money. They did just part ways with their Chief Architect, btw.
Bill Ferrante
on 23 Apr 08I echo both AllanG and ceejayoz. During this last round of Twitter issues, the website was up, I could load my personal twitter page, but I received no updates.
That outage would not have showed up in any ping type test.
Nick
on 23 Apr 08@matt: good question. Does anyone know if they actually make any money?
The way I look at it they have a gold mine of marketing information. I buy a product or service and like/dislike it I will twit it. I see a show, I twit. This combined with millions of users is a collective intelligence that any company would want to harness and understand.
Steven Fisher
on 23 Apr 08Twitter’s problems are usually internal database problems and “Something is wrong!” web page responses, which I very much doubt would trigger pingdom. Although I welcome correction on this…
Paul Lloyd
on 23 Apr 08Loving the theme of the last few posts – you’re saying all that I’ve been thinking about the Valley for years. It’s madness is something that should be questioned (as you are doing), not foolishly celebrated as it is in magazines like Fast Company.
Andy
on 23 Apr 08He he… is this a response to Techcrunch almost implying Rails apps didn’t scale?!
Rahsun McAfee
on 23 Apr 08I’m with ‘Nick’, I think there is some value in there for sure.
I honestly thought they were taking advantage of those aspects already somehow. Isn’t that how you get money in the first place?
Peter Cooper
on 23 Apr 08I’d be very interested to see the future of Twitter on mod_rails, because I think that’s going to set the baseline for the community.
I’m not sure if you’re just trying to drop buzzwords in here, but for an extremely high traffic, scalable, custom app, why would mod_rails have anything to do with it? You’d be running daemons at the lowest level; not faffing around with Apache and mod_rails.
Frank
on 23 Apr 08@Doug
My understanding, no – Merb does not have scaling problems like Rails.
Joe Grossberg
on 23 Apr 08AllanG +1
Responding to pings is not the same as responding to meaningful HTTP requests that actually use the service.
Plus, won’t pingdom have false-positives, when Twitter.com is responding by displaying that “Twitter is being upgraded” page and the like?
Ontario Emperor
on 23 Apr 08Twitter’s issues aren’t technical issues, but business issues. Twitter has communicated repeatedly that they’re not even going to work on a monetization plan until they can offer a stable, growing system. So right now they’re taking zero dollars in (although they may be getting something from their Japanese partner).
More critically, Twitter has issues in communicating status to its users. Compare the near-immediate responses (with apology) from Disqus and GroupTweet today to Twitter’s paltry two communications (without apology) over an entire weekend.
memespring
on 23 Apr 08charging people for > 250 SMS’s a month would be an obvious start on the way to becomeing a bit more sustainable. i find it crazy that they currently just cut people off.
DHH
on 23 Apr 08Frank, I’m all ears for your understanding considering that just about all Ruby web frameworks use the same infrastructure and approach to scaling (while some may be more efficient and others have less features, this has nothing to do with scaling).
Anonymous Coward
on 24 Apr 08What’s the problem with this Merb vs. Rails thing?
I’ve heard a lot of Merb users attacking Rails…people attacking the “enemy” is not the best way, comparing and getting good things to float is.
My last marketing class was about not bitching at other people’s work, is less productive.
Javier[EmaStudios]
on 24 Apr 08Actually, I believe Twitter will REPLACE MSN/AIM/Gtalk, etc. Today facebook launched a new chat feature that is just the same than the Gtalk on Gmail. I believe Twitter is the easiest way to communicate with everyone, eliminates the “stop talking to me!” problem that chat has and works making everyone aware of where the other person is or it’s doing.
Twitter replaces chat, is that simple. I’m already buying stocks!
Javier EmaStudios
Jason T Johnson
on 24 Apr 08I can forgive downtime a lot more easily than I can forgive horrible security issues where other users are allowed to log into my Twitter account, access my settings (including my mobile phone number), and post embarassing tweets posing as me.
When I submitted a support request, the only response was a form letter to the effect of “Yeah, we know. We fixed it. Clear your browser cache and let us know if it happens again.”
Sorry for the negative tone. I’m just trying to point out that downtime is not Twitter’s biggest problem. It’s customer service and unapologetic carelessness with security.
Sunny
on 24 Apr 08Well its up alright. Except sometimes my timeline shows me tweets that were posted 48 hours ago. Uptime in twitter’s case is not everything. Timely and instant delivery of tweets would be a better metric.
Carter
on 24 Apr 08David, as much as I like Twitter, I think it is a centralized solution to a big gap in the write-a-blog/reads-many-blogs ecosystem. Speaking generically, people who post to twitter are “blogging”. When they follow people, they are “subscribing”. This is all very familiar. The functionality gap comes arises because my blogging tool can’t tell me who is following/subscribing me.
I think an interesting and distributed solution would be to clone part of Twitter’s API not for centralized services (Pownce, FriendFeed, etc) but for ALL blogging software. If blogger A subscribes to blogger B, let the subscription tool for A notify the blogging tool for B.
More: http://cubanlinks.org/articles/2008/4/22/twitter-is-proprietary-rss
Vicky
on 24 Apr 08DHH,
You need to shamelessly promote your twitter id. Where are you on twitter?
The masses want to hear what you have to say!
Vicky
on 24 Apr 08Not to leave out JF, the masses want to hear what you have to say too!
I actually follow DHH on twitter & no updates for 6 mos… Or actually 1 update 6 mos ago. Shame on DHH.
I see you guys made the Webware 100 this year again. Congratulations!!!!!
How is 37 Signals missing this one?
PS Do follow 37Signals on twitter, but want more than product updates. Feed me! :) :)
Chinmay
on 27 Apr 08The best thing about the current Twitter uptime issues is the fact that they have less money!
Does someone remember when the only way to build a reliable website was to pay Oracle/BEA a ton of money? That changed not because BigCo Corp wanted to reduce it’s IT cost; but because a bunch of startups wanted something that was reliable; and cheap.
IMHO most breakthroughs in technology come by because of constraints.
If twitter manages to create scale with less cash, we’d all benefit!
kenobi
on 30 Apr 08As a fan of twitter, the best outcome for me would be an acquisition by the likes of Yahoo to fund it as a loss leader. I don’t mind the odd contextual ad either. As long as it keeps going and is left relatively pure (as Yahoo did with flickr) who cares who owns it?
Of course, that could all change with a Microsoft acquisition.
This discussion is closed.