It was supposed to be an easy project, but the salesperson screwed up. What should have been a straightforward Flash site promoting Windows XP Media Center Edition nearly melted our servers and crushed us financially.
To be fair, it wasn’t entirely the salesperson’s fault. But allow me to go into a bit more detail, and you can decide for yourself. It started when one of our agency clients brought the project to us and asked what it would take to make it happen.
Our developers looked at the technical aspects and scoped everything out, from the interactive Flash game with high-score table, to serving up dozens of video files promoting Microsoft’s new operating system, and putting it all together in a slick, unified site. We gave our client a budget, which was met with approval and the project was underway. But something was left out.
Did anyone figure out how much bandwidth would be required? A closer look at the (now signed) contract showed that the fixed price we gave included hosting. Ouch. This wasn’t a small brand hoping to get a few thousand visitors to the site. This was Microsoft, launching its next big evolution of its flagship operating system.
“How much traffic do you think this could get?” we asked the account director and nervously waited for the reply. The answer came back all too quickly. “You’re familiar with Burger King’s subservient chicken site?” Faces went pale as we recalled how insanely viral that site went.
And did I mention that the entire site was essentially one big video file, where each thing you clicked on played a different video? We were screwed. Sure, we had load balanced servers with redundant hardware and a 100Mbps upstream to multiple tier 1 providers. You’d be crazy not to have at least that. But this could have easily saturated even that level of network bandwidth and caused the servers to buckle under the load.
Come to think of it, overwhelming the servers would make the site unavailable and thus reduce our bandwidth bill. But it would also effectively shut down the sites of all our other clients as well. Besides, that’s not the way we do business. We have an obligation to not only deliver a great site but to keep all our client sites running smoothly without interruption.
So we took the plunge and re-implemented the site on top of a distributed content delivery network which served up the video from a network spanning across multiple continents. Our developers worked feverishly to accommodate the necessary changes in time for the deadline and we were able to test and deliver the finished site on schedule. The site performed extremely well, and both our client and Microsoft were pleased with how snappy the video loaded. Then came the acid test.
The site went live and the visitors poured in. It continued to perform admirably, thanks to our investment in the network. And we sat there nervously, refreshing the bandwidth stats watching the gigabytes – and our cost – move steadily upwards.
Fortunately for us the site ended up being somewhat of a niche Microsoft site and didn’t turn viral, so the bandwidth bill didn’t entirely bury us. But the lessons learned here paved the way for our approach to high volume sites that we continue practicing today.
In addition to continually improving our own cluster of web and database servers, we still deploy static content to our content delivery network. Spanning multiple continents and multiple data centers, we can serve over 1Gbps of sustained throughput effortlessly. Besides, by offloading the static content it keeps the core servers responsive, ensuring a positive site experience for visitors. Finally, we monitor every site 24/7 to make certain they continue to operate at peak efficiency.
Our clients sleep better at night knowing we can handle anything they throw our way. And we don’t let the salespeople throw in unlimited hosting any more.
