probably. But 155 m tweets/day is huge. I guess it's no more the birds carrying the blue whale. they have certainly improved their architecture, if not perfected it.
I did see the Fail Whale once or twice last week and my client (TweetDeck) occasionally hits API errors, so it isn't perfected quite yet. But handling 1800 tweets a second and pushing them to real-time streams is no small task.
handling 1800 tweets a second [...] is no small task
It's funny how people consistently overestimate that.
Actually, yes, it is a fairly trivial task and there's no justification for failwhaling over it when you have money to buy competent developers.
Many companies process orders of magnitudes more than that, under much harder constraints and much more complex requirements (think financial industry and telecoms).
Actually I have built similar systems, on a smaller scale.
Phone companies and financial institutions don't typically have messages which affect millions of accounts, for a start.
You're saying that as if it was a relevant metric. It's not. What matters is the number of concurrently online subscribers and the aggregate throughput. Twitter doesn't go and append to 1mio inboxes when akutcher sends a tweet.
Most of the companies you cite make on the order of dollars per transaction. It's easy to build out infrastructure to handle that volume when you're making that kind of money.
