Whn I interviewed at Facebook I was kind of thrown back by the lack of engineering discipline. If I remember the quote accurately (this was about 1.5 years ago), they averaged 1 hour of downtime per subdomain (i.e. <college>.facebook.com) per week, which may or may not effect all users associated with that domain. To make scaling issues easier, race conditions are also all over. There is no data integrity of any sort. Some/all of these design decisions I could see being justified, because it is afterall just social profiles... but I was a little surprised. The weird thing is they release some really cool open source stuff and have some pretty great engineers there, so I'm not sure where this stems from.
Comments
Whn I interviewed at Facebook I was kind of thrown back by the lack of engineering discipline. If I remember the quote accurately (this was about 1.5 years ago), they averaged 1 hour of downtime per subdomain (i.e. <college>.facebook.com) per week, which may or may not effect all users associated with that domain. To make scaling issues easier, race conditions are also all over. There is no data integrity of any sort. Some/all of these design decisions I could see being justified, because it is afterall just social profiles... but I was a little surprised. The weird thing is they release some really cool open source stuff and have some pretty great engineers there, so I'm not sure where this stems from.