In the on-premises world, you have to provide your own capacity, which requires a delicate balance. Because you pay for all the hardware and software regardless of whether you use it, you don’t want to tie up budget if you don’t need it. On the other hand, purchasing and standing up new servers takes effort and time; smart planning helps minimize that impact on the team, because they’ve already got a lot on their plate, and on the organization, because application performance can’t suffer. When you move applications to the public cloud, you no longer have to make those capital and human investments in your infrastructure. But that doesn’t mean you can just let it rip.
The dark and expensive side of “pay for what you use”
In the public cloud, you pay for what you use. Too often, we like to insert the word “only” in there—as in you only pay for what you use—but this creates a false sense of budgetary security. Because while that statement is true, so is this one: You have to pay for what you use. Just like on-premises. Except there’s one big difference. Because in the data center you’ve essentially pre-paid, there’s a cap on your costs. Of course, you may pay a price in the form of performance and availability. In the cloud, however, you don’t necessarily have that ceiling. Which means if you’re not paying attention, or if you’re not making the optimal choices, you could be on the hook for skyrocketing cloud bills. And, while you can get some pretty hefty discounts by making up-front commitments via reserved instances (Azure or AWS) or committed use discounts (GCP), you could ironically end up in a situation where you’ve paid for capacity you’re not using.
Smart planning—AKA optimization—is as important in the cloud as on-premises
Cloud optimization means that your instances and full application stacks are rightsized to the ideal configuration to meet your organization’s needs. But there are different needs in play here. Obviously, you must meet your current and future performance and availability SLAs based on workload characteristics (CPU, disk, I/O, memory), but not at any cost—you also want to meet your budgetary requirements.
On top of that, you need to meet your enterprise risk tolerances. If you know you will need extra capacity, but not when you’ll need it, you have some decisions to make. Will you compromise performance or user experience in favor of cost and stick with what you need 95% of the time? Or should you increase instance size to handle the expected, but mostly unused, capacity and eat the additional cost? This last question is critical when evaluating the alternatives, especially since there are so many choices available at different price points. There may be options that look very similar at first glance but could have different ramifications over the long term. For example, one selection might be slightly cheaper but has a higher risk of performance bottlenecks above a certain threshold which could cost the enterprise thousands of dollars, while another costs a few hundred dollars more but doesn’t have that threshold. There’s no definitive right answer here, only the right choice for your organization.
If you can plan in advance, reservations will save you money. But done wrong, you could end up needlessly spending money for capacity you rarely—or worse, never—use. But making comparisons is difficult given the plethora of choices, and further compounded by the fact that provider offerings change constantly and costs can vary by location. That’s why you need to be able to winnow the options based on your usage, budgets, and risk profile and then evaluate the short list by understanding the impact of anticipated or even unexpected changes on performance and cost—all before making any financial commitments so you’ll be rightsized right out of the gate.
Select and configure for efficiency—with Virtana Workload Placement
Virtana Workload Placement delivers the instance and full application rightsizing capabilities you need to select the best public cloud infrastructure provider and instance type to support your specific and unique workloads to ensure the most efficient configuration for your risk tolerance, performance requirements, and resource consumption.
Nate Attella
