It wasn't so long ago we discovered that, under certain circumstances, Twitter can add a punch to television shows' ratings. But what wasn't seen until just recently was the corollary to that discovery: that under certain circumstances, television shows can add a punch into Twitter's (News - Alert) ratings as well. That particular corollary made itself very well known recently, however, to the point where Twitter actually reached a one-second peak of Tweets per second.

The discovery was made recently, though it ties back to an event that happened on August 3, a Saturday, in Japan. A Japanese network was airing “Castle in the Sky,” and one moment in the film prompted such feeling that Twitter saw a record-breaking spike in its overall use levels. The spike in question constituted 143,199 tweets per second.

That number sounds large in isolation, but it's when the number is compared against standard levels of Twitter use that its true scope can fully be appreciated. Normally, Twitter traffic represents a little over 500 million tweets every day. That means, typically, about 5,700 per second. This means that the “Castle in the Sky” spike represented right around 25 times what normally is seen on Twitter.

Normally when a website is hit with 25 times its normal traffic in so short a span, service outages can take place that leave the site slow, or otherwise unreachable, until traffic tapers off to normal levels. Indeed, Twitter itself fell prey to similar effects at the 2010 World Cup. The sheer number of tweets coming in for penalty plays, shots on goal, and everything else made Twitter unavailable for many users in that time frame. But, perhaps learning from this incident, Twitter ramped up its internal processes, and built out impressively.

For instance, Twitter began to realize that Ruby on Rails just wasn't up to the kind of task that Twitter was after. When the 2010 World Cup events took place, there were fully 200 engineers contributing to Twitter's Ruby on Rails installation, and Twitter was running storage for its tweets on a temporarily-sharded MySQL storage system. It became clear to Twitter, though, that most of its current infrastructure was terrific, but for a system of limited size and scope. Twitter discovered that it needed more, and that sent Twitter's engineers essentially back to the drawing board in a bid to find what to do.

Twitter thus embarked on a variety of changes, bringing in a system it refers to as “Decider” that can allow for things like making certain features available only to certain parts of the Twitter community, as well as making systems more independent and distributed to better handle large loads. Improvements were made to the storage system and several other key systems, and the results have been astonishing. Not only is the system now better equipped to take on large spikes in usage—just like the “Castle in the Sky” spike—but the whole system in general is much more available, at last report approaching four nines uptime, or 99.99 percent.

Twitter, clearly, is a very popular service. Its user base is getting a lot out of the service, and clearly taking it to levels that even Twitter hadn't planned for, at least in the early going. But Twitter is clearly taking customer service seriously, and is working hard to keep the user base engaged and satisfied, making the tough calls and the big changes in support of a social networking service that's rapidly gaining in popularity and respect from its users.