I’m very skeptical about the effectiveness and long-term benefits of the healthcare system put in place by the Affordable Care Act (Obamacare). However, I’m not going to join in with the people who are proclaiming that Obamacare doesn’t work because they’ve visited HealthCare.gov, poked around for a bit, and found busted web sites.
It’s the first day of production operation for a highly-anticipated new website. Suddenly systems that have only ever been subject to synthetic test loads are being hit with a huge number of real-world visitors. That’s a hard thing to get right, and almost everybody screws it up a bit. (For example, Grand Theft Auto V is having a bit of a wobbly.)
Stuff just happens. Maybe somebody loads last week’s test data into the pricing tables instead of this week’s final data. Maybe the server farm that produces the drop-down list of plans for your location turns out not to scale — running the system on 10 times as many servers supports 10 times as many users, but running on 100 times as many servers only supports 35 times as many users, so the system hangs at at that step, and everybody presses F5 to reload the page, doubling and tripling the load in minutes.
Maybe a database administrator accidentally places two highly-active database tables on the same hard drive. Maybe a jQuery graphics plugin used on only three pages turns out to load assets from a third-party server run by people who had no idea they were about to support a nationwide healthcare rollout. Maybe the batch job that should have loaded the new graphics assets stopped after only updating half the servers and nobody’s noticed because they’re all working on some other problem. Maybe the training video includes a cute kitten and Ellen DeGeneres tweets the link to 22 million followers, overloading the video servers. Those who can’t see the video press F5 to try again.
And so on. This kind of stuff just happens.
Sure, there are a few companies who get this sort of thing right, but they tend to fall into a few special categories such as (1) server farms with a business plan, such as Google or Amazon, for which a sudden surge of 30 million visitors per hour may not even be a reportable event for their automated provisioning system, (2) companies such as Microsoft and WordPress, which operate mature, well-understood systems where even the peak loads have predictable effects, and (3) companies where day one performance is critical, perhaps because the release is timed to a marketing plan and if the site isn’t there when customers come looking, they’ll give up and never ever come back.
The Obamacare sites don’t fall into any of those categories, so it shouldn’t be surprising that they’re a little glitchy on the first day. Everybody has at least three months to sign up, so there’s still plenty of time to get things right. If you’re having problems enrolling now, just try it again in about six to eight weeks. By then they should have eliminated any teething problems.
Note that I’m not saying that they will get things right. For all I know, the marketplace sites launched too early and are doomed to months or years of failure. But day one is just too soon to tell.
Leave a Reply