[jzwinck]

You're assuming a constant arrival rate and zero overhead. Both of which are definitely not accurate.

As a wild guess, you're off by 1000x. You may say 400MBps is still not too crazy, but it's beyond the real capacity of many systems out there.

[gue5t]

Of course those aren't sane assumptions, but if the peak rate were the significant information, why would they title the article with the per-day rate? It smells like someone decided "250 million is a large enough number to scare people". Fuck that.

Even if peak "instantaneous" flow is 400MBps, considering that SMS already has seconds to minutes of latency, a bit of buffering shouldn't be a big deal.

They start the article by saying they make six (not necessarily serial) network roundtrips for each message which indicates that something is horribly wrong with the design of their system. Their code needs to bill each SMS to the sender, send the SMS to the telco, and log successes and failures in a way that's searchable later. The rest of the article describes their terribly overwrought approach to these tasks. With so much enterprise software and so little actual discussion of the problem domain, there's no way they're butting up against anything interesting enough to merit a HN submission.