This article examines the cost of Input/Output Operations Per Second (IOPS) in cloud vs. on-premise storage environments. Using examples from Microsoft Azure, AWS, and 45Drives’ Storinator and Stornado systems, it compares the practical costs of achieving reliable performance across different types of storage.
The goal is to understand which option makes more sense for workloads that depend on consistent, high-speed access to data.
When you’re dealing with serious data-intensive workloads like analytics, AI, or any task that depends on rapid access to data, performance often comes down to one key metric: IOPS. Every read, write, and transaction comes with a cost.
As data and storage processing power continue to increase, the need for high IOPS has become more apparent. The ability to handle large amounts of data quickly is now critical to performance.
Today, there are two main ways to store and access data: in the cloud or on-premise, where IOPS ultimately occur. As such, this blog compares the cost per IOP between two major cloud providers (Microsoft Azure and AWS) vs. on-prem setups that use both spinning disks (Storinator XL60) and high-performance NVMe (Stornado F16) from 45Drives.
When comparing the cost per IOP for cloud and on-premise systems, challenges arise because cloud services are billed monthly, whereas on-premise infrastructure involves a one-time purchase. To make a fair comparison, we must evaluate the monthly provisioned cost of IOPs for the cloud and convert the capital cost of on-prem equipment into a monthly equivalent, assuming a five-year lifespan. Although many durable systems operate well beyond five years, further lowering their effective cost, this assumption serves as a baseline comparison.
In this instance, the cost per IOP will be determined by dividing the total storage cost by the number of IOPS the system delivers, then converting that value into a monthly rate for an even comparison across platforms. The following calculations were retrieved directly from AWS EBS i02, Microsoft Azure Ultra Disk, and 45Drives.
Here are the cloud prices *can slightly vary depending on region:
AWS EBS io2 Pricing: $0.065 per IOP (per month, up to 32,000 IOPs)
(source: https://aws.amazon.com/ebs/pricing/?utm_source)
Azure Ultra Disk: $0.0052 per IOP (per month)
(Source: https://azure.microsoft.com/en-us/pricing/details/managed-disks/)
Here are the on-prem prices (Monthly over a 5-year span):
Storinator XL60 total cost per IOP: $4.33/60 months = ($0.072 per month, using 60 x 30TB HAMR drives, 1.8PB)
Stornado F16 total cost per IOP: < $0.01/60 months = (0.00017 per IOP per month, using 16 x 30.72TB Micron Pro drives, 491.52TB)
As you can see, the cost per IOP for the NVMe Stornado F16 system is substantially lower than that of cloud-based and on-prem spinning disk options. These savings can increase even further with caching, as storing frequently accessed data in RAM or solid-state media reduces the number of IOPS required. This means you can achieve the same performance with fewer operations, further lowering overall operational costs.
While the Storinator XL60 has a higher cost per IOP for high-performance workloads, its total cost of ownership remains substantially lower than cloud in many cases. To read more, refer to our Total Cost of Ownership blog.
Newsletter Signup
Sign up to be the first to know about new blog posts and other technical resources
