Okay, class. Eyes up front. Today in Computer Science III, we will be reviewing the word "dogfooding" and illustrating why it is such a bad idea. (Some of my readers already know what's coming....)
So, "dogfooding" is a term that originated at Micro$oft. It happens when a company, either when it acquires a new business or partners with another, tries to chase out competitors products from the business environment. M$ staffers referred to M$' desire to do so as wanting to "eat their own dog food", and the term "dogfooding" was born. M$ has long been known for dogfooding. When they bought Hotmain back in 1995, they kicked out the Sun servers running things and replaced them with NT servers. We all know how THAT worked out. WebTV, likewise, was powered by Sun. M$ decided to transition it to Windows Server and WinCE operations. We all know how THAT worked out. When Apple bought NeXT in 1997, M$ partnered with Dell to revamp its WebObjects-based web store. We all know how THAT worked out. There was even talk of creating an auction site to compete with eBay, but the servers would be Windows machines instead of the Sun Sparcs eBay has. Aaaaaaand we all know how THAT worked out.
When M$ created the Zune music player in an attempt to dethrone the Apple iPod as the king of music players, they bungled it spectacularly. Why deal with M$ and it's stupidity when you could just get an iPod? M$ has been desperate to keep the Zune afloat somehow, and they are way behind in the mobile phone market. The result was Project Pink. It was to create a Zune-based phone. Problem: M$ doesn't know jack shit about developing. Answer: buy an allready existing developer and use them. M$ dug around, and found themselves an acquisition target -- Danger.
Danger is best known to people as the folks behind the T-Mobile Sidekick (a.k.a. the Hiptop in Australia and Europe). Danger runs the servers that the Sidekicks interact with. Yes, it's that dreaded boogeyman, cloud computing. The Sidekicks, when they power on, download information from the cloud servers, such as photos and so on. Lots of phones store their information in a cloud server. The iPhone and MobileMe do it, too. The key difference -- the iPhone uses iTunes to back up your data. MobileMe uses Windows' built-in Sync software. When you power down your phone, it usually retains the configuration. But, if something hiccups, it starts with a blank table and has to rebuild everything. iPhones and MobileMe have either the cloud or your computer to rebuild the table. The Sidekick, however, was specifically designed with the assumption that it would always have access to the cloud. There's an old Polish proverb that says, "Nature always sides with the hidden flaw." I mean, really. This is such a basic mistaken assumption I can't believe it made it out of the brainstorming committee, let along onto the market.
So, M$ buys Danger in 2008, and realizes the road isn't as smooth as they thought. Danger had contractual obligations to T-Mobile to keep developing the Sidekick line. M$ couldn't just wait, they were too far behind as it was. So they went and set up their own staff to move Project Pink forward, complete with unrealistic deadlines, blue skying, and overpromised technology that would eventually be cut to make deadlines (you know, just like EVERY OTHER M$ PROJECT EVER!) By the time Danger was freed up enough to work on Project Pink, it was what is known in the business as "a total cluster fuck." Realizing they were doomed, Danger employees started jumping ship. M$ still had to maintain those servers for T-Mobile, and used their own data centers to help host the information. M$ still had a rock in their shoe about the Danger servers, though -- they worked on a stack of Sun Solaris, Linux, and Oracle Real Application Cluster. I have it on reputable authority that this set-up was an absolute nightmare to maintain, but it worked.
And it's here that the tale gets weird.
In late September, service outages started occuring. On October 2, 2009, the MicroSoft Data Loss happened. Danger's cloud servers horked and the network didn't go down so much as plummeted. Sidekick users were advised not to turn off or allow their units to lose power until the service interruption could be fixed. Services started being restored on October 6, but on October 10, everyone threw in the towel. The servers had what they would call on MythBusters a catastrophic failure. ALL data was lost and would never be replaced.
T-Mobile was already made at M$ for basically taking Danger away from the Sidekick project. This has only made things worse. T-Mobile has a SLA (Service Level Agreement) that says they are entitled to monetary restitution for each minute the network goes down. This was approximately two weeks with all customer data lost and damage to the brand name. M$ is looking at the All Time Grand Champion Hall Of Famer lawsuit for this. Plus, the fact that no one is really saying what went wrong with the servers (which would at least reassure people that the problem wouldn't happen again) combined with Danger being a small company with lots of employees having high level access to servers implies this might not have been dogfooding (M$ was trying to transition, but they wouldn't make that bad of a mistake) but sabotage (Joel and the 'bots: gesundhiet!). The idea of backing up, plus the nightmarish but effective architechture, points to someone gaining root access and putting in a time bomb to chew up all data and backups. M$ has been touting Azure Services, their own cloud computing solution to compete with Google and Amazon's offerings. Guess who isn't getting their calls returned. T-Mobile is trying to tie Sidekick to Google's Android in an effort at damage control. On October 14, 2009, a class action lawsuit was launched against M$ and T-Mobile over all of this.
Folks, how many times do I have to bitch about cloud computing?!? This year, Blackberry went down, Google went down, and now this is the highest profile crash, period. It seems people are finally realizing that "the cloud" is not unicorns and rainbows as promised, but very much a liability. There's a reason computers evolved away from dumb terminals talking to a central server.
Those who don't learn from history...go on to run multi-billion dollar corporations, apparently.