Being simultaneously timely and just at the wrong time, arstechnica takes a well written birds-eye view of the cloud infrastructure space. Timely because I've just rolled out a piece of cloud infrastructure code I outsourced to Fruition Technology for "single click" publishing of mixes and the monitoring, billing and management thereof. This being 2010 the storage backend is Amazon S3 and deployment is onto a Linux server running on a VPS. This also being 2010 it's beginning to look in many ways like not the wisest approach - while S3's promise of basically bottomless storage and bandwidth is appealing, albeit actually quite expensive once you do the math, my favourite DNS provider now has not only entered the VPS hosting market but sells un-metered bandwidth to go along with it. $20/month per 256MB/8GB/5MBit slice. Development would have been cheaper, too.

Looking at the "smart" side MediaTemple will now run you a django container for (drumroll please) $20/month - which seems increasingly to be the magic number. This would be more per month than the VPS but Django can be not-entirely-simple to deploy (very much a solved problem at this end, now) and would have saved quite possibly heaps of programmer time.

So, I dunno. It's a complex and evolving market. Performance monitoring shows that slicehost can be lumpy, Amazon just plain bad. No single supplier can tick all the boxes and ... when they do ... I reckon they will win hands down. If I "had my time again" I think I'd be tempted to go for the "maximally hands-off" approach of grid containers and S3 storage purely to get development time down but then my django development arguably started more like two years ago and this stuff was just not around then. What is for certain is that I don't have time to sit around talking about it!