View video

Look – this is beta, it might not work, it might not scale, etc, etc, but Amazon has shown a great track record with their web services (AWS) thus far. First was S3 – storage on demand. Next was EC2 – computing on demand. Lately they’ve added FPS – a flexible payment service, and they’ve had SQS – simple queue service – for a while now. Major sites are running one or more of these services in production right now, sites like New York Times and SmugMug. I even read somewhere that Microsoft is using S3 for storage on one of their sites.

SimpleDB sounds very similar to the CouchDB project I’ve covered in the past. It’s a schema-less data repository of name-value pairs with automatic indexing. It scales instantly, it’s pay per use, and it looks like it uses an API based on REST. Initial commentary indicates it’s using Erlang to accomplish this. More details regarding the API, performance, etc, will come out over the next few weeks as people start to exercise this.

I spend more and more of my time procuring equipement, retro-fitting equipment, dealing with power, cooling, and space issues, budgeting for neeeded equipment, estimating expenditures on future equipment – less and less time building cool stuff. Being able to get up and running in a day on these new services, trying out ideas at low cost, then tossing away bad products or expanding successful products – that’s the future – and it’s all coming into focus right now.

From my point of view, all Amazon is missing now is a proper load-balancing service. I’m guessing that is on tap for them in 2008. If they go the extra mile and make that load-balancer allow geo-graphical distribution of traffic to unique data centers, that would push this stuff into a whole other stratosphere.