On the face of it, reserving cloud capacity in advance seems like a no-brainer. After all, the promised cost savings can be compelling. Customers purchasing Amazon EC2 Reserved Instances, for example, can save up to 75% of the costs of buying on-demand instances when needed.
As is so often the case with such tantalizing benefits, however, the cost savings of reserved instances come with a catch: purchasing the wrong types or the wrong number of instances can, over time, prove confining and even wasteful. You don’t want to be locked an instance’s older technology as new innovative instances materialize. Nor do you want to pay up front for capacity that ends up exceeding your actual needs.
There are ways to lessen your exposure to some of these risks. Amazon Web Services (AWS), for example, offers three types of reserved instances (RIs) – Standard RIs, Convertible RIs, and Scheduled RIs. With Convertible RIs, customers can change the attributes of an RI, and with Scheduled RIs, they can reserve capacity in advance for specific time windows. Of course, the steepest price discounts are tied to the inflexible Standard RI option.
Clearly, the best of both worlds is to get the maximum cost advantages of reserved capacity without forfeiting the flexibility and agility that cloud computing can provide. To achieve the best balance between these two, sometimes-at-odds, objectives requires careful analysis and realistic forecasting.
As should be obvious, a company can’t hope to make intelligent decisions about reserved instances if it doesn’t have good visibility into its current infrastructure capacity usage and application needs. This requirement is hardly restricted to reserved instance decisions. As we’ve discussed in prior posts, detailed application and infrastructure analytics are also a prerequisite for right-sizing cloud instances and for implementing automated scaling.
Along the same lines, it’s also critical to not assume that the types of cloud instances or in-house data center infrastructure you’re currently using is something you should simply replicate in your reserved instances. If your current utilization rates are subpar or your application performance is borderline, it makes no sense to lock in those some inefficiencies and limits for the next two or three years.
With its ability to apply predictive analytics to determine workload patterns and infrastructure utilization, Densify can help organizations make informed decisions about how to find the best balance in their reserved instance investments. For companies already locked in to reserved instance contracts, Densify can provide month-by-month plans for what new instances to add to the reserved mix as existing instances sequentially expire.