Re:Reddit now tracks all outbound link clicks by default, ex...
by jedberg on Friday July 08, 2016 @04:42AM
(#12054263)
As one of those talent, I have to respectfully disagree. First of all, R/W to the database was solved pretty early on. We had queues many years ago for the writes and tons of read slaves for the reads. Also, the servers that serve comments and listings were not the same ones that took in votes and comments. It was most definitely a "throw servers at it" problem, with a few bottlenecks on caching that were solved as they came up.
The biggest issue was that we didn't have the budget (both money and manpower) to deal with scaling a global transaction system, which is just plain hard. Reddit gets almost as much traffic as eBay, but eBay has thousands of engineers and Reddit doesn't even have 100.
Back when I was there, the site worked well enough, there were only a few of us, and we spent about 50% of our time on spam and about 25% on community management. That didn't leave much time for scaling and performance. For context, we were serving a billion pages a month on a budget of $50K/mo.
Today they are making great progress in scalability because they finally have the budget to hire the people to make it great.
The biggest issue was that we didn't have the budget (both money and manpower) to deal with scaling a global transaction system, which is just plain hard. Reddit gets almost as much traffic as eBay, but eBay has thousands of engineers and Reddit doesn't even have 100.
Back when I was there, the site worked well enough, there were only a few of us, and we spent about 50% of our time on spam and about 25% on community management. That didn't leave much time for scaling and performance. For context, we were serving a billion pages a month on a budget of $50K/mo.
Today they are making great progress in scalability because they finally have the budget to hire the people to make it great.