One of our customers has a spiky workload with an EC2 On-Demand equivalent workload averaging $4.36M. They want to reduce their total spend on computing and explore the possibilities of modernizing their applications, with the aim of reducing their EC2 footprint for the future.
With the actual schema of Saving Instruments, they don’t have the flexibility to reduce the EC2 workload because the Saving Plans commitments are rigid.
Although there is good coverage (87.30%) using Saving Plans, the savings are not maximized.
The covered workload uses a mix of 1-year and 3-year Saving Plans (full upfront, partial upfront, and no upfront), the first expiring in 100 days and the last one in 560 days.
But, the workload has a lot of peaks, and increasing the coverage will also increase the waste.
The implemented solution had the following issues:
- Inability to reduce commitment
- Large remaining contract commitment (liability)
- Low discount on 1-year contracts
Cloudwiry proposed replacing the expiring Saving Plans with a mix of Convertible RIs which will increase the savings by an additional 35% and reduce the liability.
Improving the utilization of Convertible RIs will be easy using Cloudwiry’s RI Maximizer. With Convertible RIs it’s possible to use 3-year terms with higher savings and improve flexibility as described below.
Parallelly, Cloudwiry recommends covering the peaks with Spot instances, averaging 50% cost savings (mixed Microsoft Windows and Linux workload), adding additional 16% savings in Compute costs, bringing the total to 51%.
It’s possible to reduce the exposition to risks by 15% during peaks.
Another improvement of this solution is that risks are also reduced with refinancing by replacing the SPs with Convertible RIs because it’s possible to extend the duration of the commitments hence reducing the monthly amount.
Reservation Elasticity will protect the company in case the demand shrinks in the upcoming months.
Finally, all these Convertible RIs are non-upfront, so no need to lock investment and perform complex amortizations.