If this was aiming for the scale of twitter replacement, the realistic answer is: read as much as possible from what was published about the current twitter architecture and try to replicate that while avoiding things they used in the past. If you're thinking in terms of a specific database rather than queues that write into partitions distributed around the world, you've probably already lost. Maybe think really hard about splitting out mostly-readers and pop-stars and whether they need input/output that's architected differently. Find people who design event stream architectures.
Because let's be serious. Apart from youtube comments, I don't think there's anything the scale of twitter for short messages at the moment. (and even that's much simpler problem)
Because let's be serious. Apart from youtube comments, I don't think there's anything the scale of twitter for short messages at the moment. (and even that's much simpler problem)