1. 14
  1.  

  2. 5

    The graph type chosen to present the relationship between cost and instance type is misleading. The x-axis has categorical, not temporal data, so a line graph is inappropriate. A bar graph would have been a better choice. In addition, there should be another scale on the right for the $/GiB-hour measure.

    1. 4

      There are many programmers who aren’t at all concerned with the costs of one particular cloud services company. I am one of them.

      1. 2

        I also object to the idea that the specific prices Amazon charges for their cloud service are fundamental pieces of knowledge every programmer ought to know, in the same way that facts about the latency of physical pieces of hardware are. There’s maybe a steelman of this position that having a rough guess of the raw cost of a given amount of computation is a useful thing to know (maybe it helps you know what a fair price for compute resources is when you buy them from one of Amazon’s competitors or provide those resources with your own hardware), but I don’t think that raw cost is well reflected by the publicly-available numbers Amazon charges.

      2. 4

        I already found the more legitimate versions of these articles to be silly, that concerned themselves with, say, timing information for very specific types of hardware.

        I find this article and its naming scheme that imitates those other articles to be drivel. I can’t write for anyone, sans myself, but I’ve never used this AWS nonsense and never plan to. I find so many willing to recentralize the Internet to be disconcerting. I suppose these numbers would be fine to have if you actually used this AWS nonsense, but why would you be purchasing something before you knew the cost thereof? Is it normal for people to not only rent from a behemoth instead of a VPS or the more respectable self-hosted server and then, in addition, not know what they’re doing until they get a bill?

        1. 11

          I can’t write for anyone, sans myself, but I’ve never used this AWS nonsense and never plan to

          That’s fine, but I don’t think you need to be overly antagonistic at the article author.

          I suppose these numbers would be fine to have if you actually used this AWS nonsense, but why would you be purchasing something before you knew the cost thereof? Is it normal for people to not only rent from a behemoth instead of a VPS or the more respectable self-hosted server and then, in addition, not know what they’re doing until they get a bill?

          You are assuming a lot here. You are making the assumption that companies which requisition AWS’s services are not aware of their TCO (total cost of ownership) by the mere presence of this blog post? I have to disappoint you, but when I was pitching to one of my previous CTO’s about using Redshift, I had to create a spreadsheet of pricing calculations and projected cost depending on varying usage scenarios. We were very aware of what we were getting ourselves into, and the cost/benefit ratio. I imagine any company that likes to stay in business is also aware of such factors.

          You also miss out on one crucial use case that is almost always a net win with cloud infrastructure: extremely bursty workloads.

          Netflix–I can only imagine–must save hundreds of millions annually by leveraging AWS’s elastic nature to spin up insane amounts of compute during Friday evenings, only to spin them down come Monday. It would be extremely wasteful to be paying for tens (hundreds?) of thousands of idling machines during Monday -> Thursday, under the scenario that they have to own enough compute power that commensurate with their peak load.

          1. 1

            Netflix could make a cool buck renting out those servers when they’d otherwise be idle.

            1. 2

              They’d also have to pay a buck to cool them ;)

              1. 1

                They have an internal spot market afaik

            2. 4

              One of the best arguments against (or for) centralizing is cost. This article helps attempt to give better information about that for people that might want to host things themselves.

              Your dismissal of it and, more importantly, the people who might find it useful could drive off more folks than you would have convinced if it had been friendlier.

            3. 1

              Although I understand what this article is trying to do I find it slightly misleading saying things such as “1 GB RAM 10 $/month”. I understand that is a median cost, but a median cost can be very skewed as you pass the “bang for your buck” sweet spot.

              Using the AWS calculator I got that an ec2 instance with 2 vCPU, 1 GB of RAM and a 30GB SSD would cost around 9.86 USD. Which is less than the median for just the ram itself. Of course the same would apply on costlier VMs making 1Gb of ram for 10 USD look like a bargain but I just personally don’t like the approach.