Twitter’s lead architect Blaine Cook, whose main task had been to enable Twitter to scale better, has left the company.
Michael Arrington, taking a book from Valleywag and making this into a personal story, thinks this means that the “Amateur Hour” at Twitter is over.
Blaine says he is moving to England so that his partner who doesn’t have a US work permit can pursue her career. Fair enough. I’m not very interested in the personal side of this (Adam Ostrow has some interesting thoughts about it over on Mashable, though).
What is far more important here is that Twitter just hired two new engineers with experience in scaling big applications, including Steve Jenson who worked on Google’s blogging platforms.
While this weekend’s outage was most definitely too long and Twitter’s communication about it was flawed, Twitter actually had very few problems in the last few weeks.
To keep its current user base happy and to become more mainstream, they have to keep up that record. With our without Blaine.
There are enough competitors just waiting for them to stumble.
They are all learning the same lesson over and over again. RDBMS systems that cater to high write rate events with transient small joins are the most brittle critters. I know this first hand.
Now, Fitzpatrick solved this years ago at Livejournal and Typepad. He is now a Google man. he get’s it.
There is such an investment in the off-the shelf mysql databases that has literally created a whole class of scaling problems. As a matter of fact, the clustering options for these write throughput problems are worse that what they are trying to cure!
They didn’t think this through - it’s too easy to set up a few database servers and wait for the madness. “Gee harry, things seem to be fine, for now.”
Don’t get me started on what conventional wisdom regarding wen applications and RDBMS has wrought on my vertical systems and mobile messaging systems work. Thank G-d I found http://www.db4o.com/default.aspx