Over the weekend, Don Park published an interesting analysis of whether he should use Amazon's EC2 and S3 services for a new web service given the new pricing scheme S3 has come up with. The killer is the one cent charge per 10,000 requests the service has instituted. Don has calculated that hosting significant portions of his app on S3 would lead to prohibitive costs of $8000/month while he could host his own under similar loads for $1000/month.
Shelley Powers voices similar concerns for a site she is planning.
The rub seems to be the request load these sites represent to Amazon. Both authors were essentially using Amazon as a secondary web host for serving images and other media files. Amazon is finding that it's too costly to do that for the original price scheme that did not include a per request charge. Based on my own personal usage for the sites hosted on this server, I've concluded that the Amazon fee would be higher than the hosting fee I'm currently paying. So, S3 is not for me.
Well, who is it for? Apparently people who serve files on average larger than 53.6 KB and who don't expose the service directly to the web. They apparently want you to impose your own site as an intermediary as a sort of buffer.
As Shelley Powers notes in an indirect way, the real issue in web scaling is not back end storage but dealing with the interaction. I suppose S3 could offer a nice storage solution, but to use it, you have to learn their API and then buffer them from too much user interaction. That sounds like a fairly lmited market.
Leave a comment