One of the biggest obstacles making people hesitant in converting their companies to the cloud are initial costs. How to avoid costly mistakes?
It’s actually pretty easy as a developer to just jump in and try cloud computing. You can set up very inexpensive servers and storage, try out your code, and if something doesn’t work you just stop the service. You don’t have any long commitments or big up-front investments to worry about. On the other hand, if you’re a company making a strategic decision, you have to be more careful because you’re investing in training, tools and especially time that you’ll never get back if you make a mistake. More than one cloud customer has just chosen a provider based on little more than a whim, put applications into the cloud and then suddenly realized several months later that their cloud bill is way bigger than they expected it to be.
A few ways to avoid that are:
- Actually do an evaluation. When you buy a car do you just go grab the same one your neighbour has? Do you just buy the cheapest one on the lot? There are far more reasons to do a careful evaluation of cloud services than there are to evaluate different car models before you buy one.
- Know your applications. The specific characteristics of your application will determine your need for cloud infrastructure, service levels, elasticity, interoperability, portability, scalability, recoverability, functionality, response time and security, among other things. Not one of those things is the same from one cloud service to another. If you don’t know what your application needs, you won’t configure your service properly, and you may not choose the best provider. If you don’t know what workload your application generates, where the peaks and valleys are, and whether or not it can monitor itself and auto-scale using cloud functionality, you will end up paying too much, in some cases way too much. If, for example, you choose AWS reserved instances when you should be using on-demand or vice-versa, your charges could easily be two or three times higher than they need to be. It’s possible that your application isn’t even well-suited for the cloud, so migrating it is a recipe for disappointment. Keep in mind that application requirements can and do change over time, so you should revisit them regularly to see if your original decisions are still the best ones.
- Clean up your mess. Governance is just as important for cloud services as it was in the non-cloud world and in some cases more so. Your administrators need to get in the habit of looking for cloud resources that you’re paying for that have become underutilized but are still sitting there generating charges every month. Perhaps usage patterns have changed and instances that were once reserved would now be cheaper on-demand, or on-demand instances are now being utilized around the clock and would be better off being reserved. Maybe applications aren’t letting go of resources at night or on weekends, and those are costing you. Maybe there’s storage you’re paying for that isn’t associated with servers anymore, or maybe it’s time to migrate some data to cheaper storage just like you would do in the non-cloud world. Sometimes new resource types are released by the provider that are cheaper than the old ones, but you don’t get the cost benefit unless you reconfigure your applications to use them. Cloud Management software can help with most of these housekeeping tasks, so if you’re a large shop you should probably look at making the investment.
The author: Scott Feuless
Scott is one of the world’s foremost experts on measurement and evaluation of Cloud services. His work with Cloud Computing began in 1999 when he helped transform a small software Value Added Reseller into a public startup Application Service Provider (ASP) as the firm’s Chief Technology Officer. He also joined the Cloud Services Measurement Initiative Consortium (CSMIC), where he was the primary author of the metrics.
Read also Praim interview with Scott Feuless on how to evaluate and choose the best cloud solution for your company.