When simple isn’t simple
Amazon recently blogged about their new simple AWS cost calculator. The image above is the after version, not the before. But, I don’t think it’s simple at all. You want to see a simple cloud services pricing model? Try Rackspace ::
I’m not interested in comparing Rackspace and Amazon feature for feature, dime by dime, cloud outage for cloud outage. What I am interested in, is why Amazon continues to make their pricing model more and more complex.
Amazon is walking a fine line between making the cloud economical and accurately metered for usage versus making it easy to understand and compare. The calculator does give you one space to accurately total up all the mythical servers you might have. And yes, it lets you choose every possible combination of cloud services so that you can have the most accurate estimate possible.
The issue is that it becomes extremely difficult to compare with actual hardware or other virtual server solutions. Say for example, I want to setup one of the new RDS instances (SQL server in the cloud). I then have to decide what size machine I need (easy), how much disk space to allocate (pretty easy), estimate the bandwidth (not too hard), backups space (guessable), I/O requests (whaaa?). And, then I need to do the same exercise for my web servers, my “static” IPs, virtual load balancer, do I want static files served off the CDN-lite system that can sit on top of S3? Oh, the madness!!!
When I’m using a real machine I pay for the machine, the bandwidth, and maybe the power in fairly easy to estimate ways. I don’t have to worry about paying for an IP that I’m not using, or how many I/O events my applications make.
I suspect some of the more esoteric costs are to dissuade people from being jerks, but I have to think that there is a better way to deal with it. Putting a tax on holding a static IP unused, probably makes it cost prohibitive to DoS the cloud’s address-space. Can’t you just cancel the account and blacklist the credit card? Amazon, please punish the jerks, not the rest of us!
“But, you only pay for what you use!” That may be, but metering I/O requests is about as obscure to me as metering me by machine instructions, or jmp instructions that my application makes. How about malformed packets? We need to make sure people are using proper network protocols after all. Number of LED blinks made by the server my AWS instance is sitting on?
To be a little fair, I do think Amazon is trying to make is easy for us to all use their platform in a fair and economical way. But, they’re doing themselves, and us a disservice by having such a precise and complex pricing scheme. I do think that there’s a definite cost advantage being on AWS today, but when I’m asked I continuously have to say that “It’d cost X and Y to move to the cloud, but we don’t know how much A and B and C will cost because they’re hard to measure”. I really, really hope that changes soon.
