Anywhere services need redundancy, not single points of failure

I've got to chime in with Jon Gruber of Daring Fireball on this: No matter how much Research in Motion promotes their NOC approach to running its Blackberry service, it's still a single point of failure for all Blackberry subscribers. And given that this weakness has been demonstrated to Blackberry subscribers with two multi-hour outages in the last 11 months, at some point, businesses are going to scream "Fix it!" I'm surprised someone from the high-availability computing world hasn't pilloried RIM already.

For those who don't think RIM's outages are any big deal, here's a fun fact. If RIM were trying to meet a 99.999% availability for its Blackberry service, the three-hour outage on February 12, 2008 would have used up its allowed downtime for the next 34 years. Oops.

It's easy to forget that until the Internet came along and demonstrated that distributed and decentralized networks really were more reliable, most of the major computer companies built networks with centralized command and control systems, yet those networks never achieved anything like the resilience of the Internet. It's a shame Anywhere business customers using Blackberrys are going to have to learn that lesson again the hard way. It's not a question of if that will happen; it's a question of when.

Meanwhile, everyone who wants to avoid the Anywhere school of hard knocks should repeat after me:

I will not accept single points of failure in my Anywhere service.

I will not accept single points of failure in my Anywhere service.

I will not accept single points of failure in my Anywhere service.

....