What's Behind Apple's MobileMe Meltdown

Whenever I see Apple stumble, which admittedly isn't often, I stop and wonder why. And then I remember it's because the same invariant laws of computing, which apply to us all, also pertain to the Mac maker. With MobileMe, and the severe outage which accompanied its attempted launch last Friday, we see that no company can, er, mess with the cloud. Here's what I think really happened.MobileMe is of course Apple's service to allow users to sync up their Macs and iPhones, and share contact and calendar info with Mac-enabled family members (6pm: feed the kids; 11pm: other stuff) via the cloud. (This was previously known as Internet connectivity with a lot of server capacity -- sometimes leased -- accessible at the vendor's end. Or it was just called the Internet. Or just computing. But I digress.) Unfortunately, the cloud did not take kindly to what seems to have been Apple's lack of preparation. Customers are still steamed about outages which lasted over the weekend, which made it difficult for many users to access their e-mail at all, or convert their old .mac addresses to the service's new .me domain. (As penance, Apple is giving customers a 30-day extension on the $99-per-year service.)

Anyway, so while the blogosphere has pretty much regurgitated what happened, and repeated Apple's nonexplanation explanations -- the widely quoted company e-mail saying the transition was "rockier than we had hoped" -- no one has taken a stab at figuring out what really went wrong.

My assessment -- maybe I'm wrong, but, hey, I'm a blogger; that's my job -- is that what we have here is a failure to properly understand the difference between peak versus average capacity.

Probably MobileMe is amply designed to handle its expected daily load. It's surely configured to support the average day as well as an additional 20% capacity. This is so it won't crap out on those really big days, when people are synching their e-mails and appointments up the wazoo. (Strangely, I don't know what day that would be in the world of computers. It's not Christmas; that's the one day everyone's not online because they're charging up the batteries in their newly gifted laptops.)

However, what MobileMe clearly wasn't designed to handle was a day when the system load was average plus 1,000-percent (i.e., 11 days of load packed into a single 24-hour period).

Now, I know what you're thinking: Even if Apple's designers had considered this peak-versus-average conundrum (and they must've; they're not idiots, right?), there's no way they could have anticipated such a peak load. Right, and wrong.

As in, it's correct to assume that Apple's engineers did indeed do peak-versus-average planning. Of course they did. (Remember, they're not idiots.) However, it's wrong to give them a pass on my second point and say that they couldn't have foreseen the heavy launch-day load.

Hey, we're not talking about customers who've just bought an LG cell phone or a Gateway notebook, and are content to take it home and have dinner first. Nope, these are Apple customers we're talking about! These folks wave their new iPhones and MacBooks as they exit the iStore! Of course they're gonna plug the things in and access MobileMe. Probably every Mac owner in the world who wasn't getting ready to head to the Apple Store to buy a new iPhone 3G when it opened at 8 a.m. on Friday, July 11, jacked in to MobileMe simultaneously when it was set to launch at 3 a.m. that same morning.

So what I'm saying is, Apple's engineers anticipated the need. What they knew they couldn't do, because it wasn't financially sustainable (maybe not technically, either, but mostly moneywise) was build out anything which could support that very unusual one-time spike.

So what should they have done? Well, Apple should have taken a page from the New York Giants, who are currently in the process of screwing their customers, but are doing it in a way which minimizes the load on what they anticipate -- because of said screwing -- will be a highly stressed customer-service operation.

It's called a staged roll-out.

In the case of the Giants, they're hitting their loyal season-ticket holders with something called a personal-seat license (PSL). This means that, before you even get to buy your football tickets, you have to part with from $1,000 to $20,000 (depending on the location of your seat) simply for the right to buy the tickets to your seat. (QED! That's why it's called a PSL!)

The Giants could have notified all 55,000 ticket-holders at once. Under that scenario, though, those people actually flush enough to pay -- "Mr. Mara, it's Andy Rooney on line one" -- wouldn't even be able to get through amid all the complaints clogging up The Meadowlands' VoIP lines.

Instead, the team is sending out notifications in batches of 5,000, the better not only to field complaints, but to offer financing and down-selling options (there's a sales tool for a recessionary economy!) so as to get balky ticket holders off the fence and into their soon-to-be-personal seats.

Apple should likewise have staged the conversion from .mac to .me. It not only makes technical sense, but it fits in with the Apple ethos. Think of being able to tell your friends: Not only did I get an iPhone 3G last Friday, but I got my MobileMe invitation in the very first batch! And the e-mail was signed by Steve!

Because if you're an Apple acolyte, what could be better than that?

Think you've detected an anti-Apple bias? (It's not there; I own an iPhone!) Then please leave your comments below or e-mail them to me directly at [email protected].

Like this blog? Subscribe to its RSS feed, here.

For a mobile experience, follow my daily observations on Twitter.

Check out my tech videos on this YouTube channel.