It is true that AWS and other elastic compute providers are constantly lowering their pricing for compute resources on-demand. It is also true that AWS clients have the ability to reduce their rates even more by committing to minimum compute utilization thresholds. I’ve run the numbers with CFO’s and CXO’s a few dozen times now however over the past 2 years and there is still no doubt about it, elastic computing is far better suited for dynamic work loads and standard in-house or hosting solutions are far better suited for static and predictable work loads.
Maybe more people will start believing this now that NetworkWorld is saying the same thing in the article they pushed out recently linked below? I recommend checking it out as it links to a bunch of relevant and tangible case studies and statistics singing the same song.
While cost is the primary driver for most CFOs, there are other reasons why AWS is a bad play. The quality of the hardware that they employ is not competitive with the current state of the art in the industry. Their redundancy at the hardware level is pretty much non-existent. Anyone who has every had an AWS image choke up on them and had to wait for it to restart on a different piece of hardware is painfully aware of the impact of this.
On the cost side, it only goes UP the more you utilize it. In most cases, once you’ve gotten to the point of spending $20-$30K per month with them, your ROI had you bought the equipment up front and housed it in a decent colo or on premise is usually in the range of < 6 months.