Dear WTF Chuck - Twitter has $5M, why can’t they hire a Sys Admin?
By Steve Poland • February 1, 2008
Dear WTF Chuck,
Why is Twitter always down, despite them being flush with cash ($5mm), and all of their users whining for the past year about how much downtime they experience? Yet, we the users, go with the flow and don’t punish them (by leaving their service), because some of us are addicted to their service!
Go buy some servers and scale your operation! Hopefully moving hosts yesterday will better things.
-FRUSTRATED TWITTER ADDICT IN BUFFALO, NY
This is a new type of post on my Site called ‘Dear WTF Chuck‘. This is a spoof on ‘Dear Abby’ and is for you to rant to the world (or at least readers of my blog) on topics relating to web tech, affiliate marketing, etc. Feel free to submit your own rant using my contact form (select ‘Dear WTF Chuck Post Submission’ as the subject of your email) and use the above post as an example of the proper format (title, question/rant, possible solution, signature that includes or does not include a link to your website).
Comments
2 Responses to “Dear WTF Chuck - Twitter has $5M, why can’t they hire a Sys Admin?”
Got something to say?
I’m pretty sure that hardware is not the issue here… its cheap and easy to manage.
I’m willing to bet it has to do with Ruby on Rails. Great way to launch something real quick. But not that great to scale…
Not easy to re-architect to scale while still growing and addressing issues that goes with it. I’m sure they are working their asses of at Twitter HQ to get it in to shape.
Hiring bodies to trow at the problem is just not that easy… specially with talent being very sparse right now to say the least.
Something charming about Twitter, and in ways the downtime adds to the charm in an odd way.
I am pretty sure it doesn’t have much to do with Ruby on Rails. It scales really well. With a share-nothing approach, it is horizontally scalable.
Most of their problems have been with scaling their push architecture and with MySQL. They really need a different data architecture and more memory on their hardware to support caching. Basic RDBMS has become a limiting constraint on many varieties of distributed architecture, and they didn’t seem to have a good grasp of which scheme they were going to use early on.