I was checking out Paul Vallee’s comments about MySpace’s definition of uptime. It seems others are seeing spotty uptime with this poster child of the Web 2.0 phenomenon.
I’m watching MySpace for other reasons though. They have deployed the Isilon IQ Clustered Storage solution for serving up the video content. Isilon is a competitor of my company, PolyServe. Isilon is good at what they do—read intensive workloads (e.g., streaming media). I don’t like the fact that it is a hardware/software solution. I’m a much bigger fan of being free of vendor lock-in. In the end, I’m an Oracle guy and Isilon can’t do Oracle so that’s that.
Anyway, another thing that is interesting about Web 1.0 and now Web 2.0 shops is the odd amount of “IT Street Cred” they seem to get. Folks like Amazon, eBAY and now MySpace are not IT shops, really. They have gargantuan technology staff, and their IT budget is not representative of normal companies. Basically, they can take the oddest of technology combinations and throw tremendous headcount of very gifted people at the problem to make it work. Not your typical COTS shop.
Now, having said that, are these shops solving interesting problem? Sure. Would any normal Oracle shop be able to do things they way, say, Amazon does it? Likely not. Back in 2004, Amazon admitted to an IT budget of USD $64 Million before some $16 Million savings realized in one way or another by deploying Linux.
Kevin Closson's Blog: Platforms, Databases and Storage 
a superb point, the one about Amazon et all not being normal IT shops!
Paul Vallee recently blogged an extremelly interesting presentation from an ebay tech strategist on what they do for their systems, how they use Oracle, and so on.
I often rant about j2ee and its inadequacy for large volumes despite its used and abused “scalability” claims. Only to get slapped in the face with the monotonous, fundamentally wrong and moronic “but ebay and other big sites use it”.
Fact is, and confirmed by that presentation: ebay only uses servlets from the entire j2ee edifice. Even the supposedly very efficient cache mechanism inbuilt into j2ee containers had to be thrown out and rewritten inhouse! The rest of j2ee was found to be a complete elephant that fit nowhere and couldn’t be used for anything in any usable timeframe.
Of course: they can afford an enormous technology misfit like that because for them to rewrite even an OS is peanuts.
The absolutely horrible problem of course is that normal, budget-limited sites out there are sold on the j2ee mantra by unscrupulous and dishonest vendors as if it was a panacea for all development. And they absolutely cannot afford this sort of approach to development.
The result? Failed projects, missed deadlines, complete IT development wrecks. They have been widespread over here in the last 6 years.
But of course the whole litany of incompetent excuses takes over: it’s never the problem of the development technology of course. “ebay uses it!”, it can’t possibly be wrong!
And the horrible, inefficient and wasteful show goes on: too many people with vested interests…
Hey Kevin! Very good point about “normal” vs. “not normal” IT shops. Thanks!