By David Barclay, National Practice Manager, Data#3
One of the key premises for moving workloads to a Public Cloud service is based on the idea that you only pay for what you use. The assumption in this statement is that this is cheaper than outlaying capital to buy, configure, manage and maintaining your own infrastructure.
This idea continued to gain ground as vendors battled for market share and drove the per hour price down for raw compute services. Every few months it felt like there was an announcement from large Public Cloud providers on the latest round of price drops.
However, these headline costs are only a small part of the total cost of deploying and running applications in the cloud. This can be seen when looking at comparisons between the major providers, as published by various research organisations. For example, 451 Research’s Cloud Pricing Index measures the cost of a typical application over time based on a range of provider’s virtual machines, storage, database, bandwidth and support quotations.
The Cloud Pricing Index found the average monthly price for a typical application was $2.56 per hour in October 2014 and only dropped to $2.53 in December 2014 – about 1%. These price drops don’t have a linear effect on the total price of deploying applications – you need to do a lot more work to get a true comparison.
However, even after doing a detailed comparison, there are other hidden costs that if not managed properly, can quickly blow out your cloud budgets. If your Test and Dev teams lack operational disciplines around shutting down workloads on existing virtualised infrastructure, or suspending/shutting down the virtual infrastructure at the end of the business day or weekend, you may get a shock when you get your first Public Cloud bill. A good analogy of this is a tap left running overnight. The bills will continue to mount – effectively negating any cost savings realised by moving to cloud in the first place.
Amongst the many benefits of cloud computing is the ease with which new virtual machines can be configured and deployed – it only takes minutes and can be done with a company credit card. However, sizing those machines correctly is not a trivial exercise and it’s very easy to over-configure them, resulting in you purchasing too much power. Even if you’re not utilising all that power, you’re still paying for it all, hour after hour.
In addition, the security requirements such as the secure connection to the Public Cloud infrastructure and the policies that follow the information are also critical to consider. Application design also plays a part with developers now building cost awareness into applications to ensure that valuable computing resources aren’t wasted over time.
The bottom line is that you can’t just assume that a pay-as-you-go cloud model will save you money over the alternatives. And even if it can save you money, can your organisation account for variable monthly costs? You need to do your homework and enlist the help of a trusted advisor like Data#3 who can help you determine a true cost comparison and assist you with operations processes before making a commitment.