Estimating your cloud-computing costs? An AWS client? You're doing it wrong

Home

News: Estimating your cloud-computing costs? An AWS client? You're doing it wrong

  1. Slashdot pushed out an interesting and timely discussion last Friday based on Nancy Gohrin's IT World article entitled Why Cloud Infrastructure Pricing is Absurd.

    Estimating, provisioning and pricing is always insane, and no matter which route one takes, be it an in-house data center or a cloud-based provider, cost estimates rarely equal the dollar value of the check that gets written at the end of the month. Are cloud computing costs high? That question can only be answered if you have something with which to accurately compare it, which logically would be the cost of running an in-house data-center. But of course, how many organizations can honestly and accurately peg the cost of hosting applications in house? Sure, it's easy to amortize the cost of a blade server over five years, and perhaps the additional salary of a system administrator when ball-parking the in-house data costs. And it's equally as easy to leave out all of the hidden costs, be it the annual cost to power the machines, or the monthly cost to pay for the square footage consumed by the server room. Are the organizations that are using cloud computing services getting a good deal, or are they getting taken for a ride? It's tough to say when it's an apples to orangutangs comparison.

    Inspired by the Amazon AWS conference, TheServerSide ran a couple of stories on how various vendors are helping organizations reduce their cloud computing costs. Vittaly Tavor, founder and vice president of products at Cloudyn, shared his thoughts on the self-destructive habit IT professionals have in terms of needlessly costing themselves money by overprovisioning in the cloud. "No IT professional has ever been fired for overprovisioning," said Tayor rather sarcastically. It's true, and it's a practice that just puts more money into the wallet of Amazon's AWS product line.

    Overprovisioning in AWS? Cost-control tools and strategies can help

    Following a similar theme, TheServerSide also spoke with  Pete Adams, the chief operating officer at Cloudability, who not only waxed poetic about the proclivity of AWS users to over-provision, but he also explained why it is that there is such a mammoth divide between how much AWS clients are currently paying, and how much they could be saving if they were able to peg their systems to the correct reserved server or pricing tier. “There are over 2000 different types of reservations you can buy, depending on the instance type, the availability zone you buy it in and the operating system, and each one of those reservations has a different break-even point," said Adams.

    Overprovisioning, underutilization, cloud bursting and discovering the correct AWS pricing tier

    The pain point of grappling with the monthly cloud computing bill was highlighted at the Amazon re:Invent conference that was held last month in Las Vegas by the fact that every third vendor on the exhibition floor was pitching a Rube Goldberg machine that could put your computing needs through a variety of hoops and eventually spit out the spot servers and pricing tiers that would optimize service and minimize costs. The fact that so many vendors are competing in this space underscores the fact that indeed, cloud computing pricing is absurd, and if you're organization isn't in sync with the insanity, your optimized cloud-computing model might not nearly be optimized enough.

    Follow Cameron McKenzie on Twitter: @potemcam

  2. Cameron what all these vendors will fail to tell you that there are two aspects to cost in the cloud.

    The first and least interesting (though easier to marketing) is that you spent more in buying capacity than needed.

    The second, and probably much more relevent to the typical TSS reader, is that you consumed (incurred) more capacity (cost) because users or code drove the processing. You don't control costs you control the causes of such cost...which you could say consumption. Now the reason this aspect does not get the airplay that it deserves (far more than the first option) is that this requires insight into which users, tasks, activities, interactions and code paths drive costs...all largely in the system and application domain.

    I wrote an activity based costing & metering engine that does this in 2008...5 years ago. Unfortunately though control is critical and largely absent in enterprise operations it requires someone to think beyond stopping and starting instances and worse look to balance cost with business value of such activities. Already most developers and operations are sleeping at this stage then you mention activity based costing of shared services that costs all the way back through the service supply chain and it is game over...the walking dead makes to the door. IT services should be run like a business onto itself but how many engineers get to have power to create code as well as control the code. Very few. Probably because the majority don't get two degrees and when they do they get promoted up as far as possible away from the code that represents the business (of IT) to controlling people.

    If you are sill reading these articles might be of interest in costing cloud servcies:

    http://www.jinspired.com/wp-content/uploads/2012/02/MeteringInTheCloudVisualizationsPart1.pdf

    http://www.jinspired.com/site/using-system-dynamics-for-effective-concurrency-consumption-control-of-code

    http://www.jinspired.com/site/controlling-runaway-threads-in-the-jvm-using-resource-metering-quotas

     

  3. Thank you[ Go to top ]

    I would like to thank cameron and the server side for sharing and keeping us , like java folks on path.

     

    great job guys.

     

    -Pradeep Reddy