According to a Bain & Company report, 84% of on-premise workloads are overprovisioned. This implies that when companies migrate a workload to the cloud, they’re sending excess computing and storage capacity right along with it. Instead of becoming more efficient, they’re merely transferring their existing inefficiencies to a new location.
Source: Bain & Company
According to AWS, companies that take rightsizing seriously save 36% on their cloud IaaS costs. To regularly identify underutilized resources and make thoughtful decisions, right-sizing must become an ongoing process within an organization.
However, enablement is tough. To regularly identify underutilized resources, take decisions that may impact production environments, hand-pick among 350+ instance types, and carry out the execution, can be challenging. At Cloudwiry, we believe it is not the best use of anyone’s time and resources – “Time spent optimizing, is time missed on innovation.”
Customer Use Cases:
Here’s a peek into the monthly percentage savings leveraged by our customers monthly giving Cloudwiry the necessary permissions to handle the rightsizing:
| Customer | EC2 cost | EC2 Rightsized Cost | Savings | Savings % |
| 1 | $350,000 | $249,000 | $101,000 | 29% |
| 2 | $1,124,527 | $635,154 | $489,373 | 43.52% |
| 3 | $319,458 | $206,630 | $112,828 | 35.32% |
| 4 | $4,500 | $2,280 | $2,220 | 49.33% |
| 5 | $35,467 | $23,654 | $11,813 | 33.31% |
Data-Driven Approach to Rightsizing
TSO Logic conducted a wide-scale rightsizing identification for Is Cholesterol a Steroid? steroids uk.com innocan pharma to participate in the lift & co. expo’s business growth panel on november 19th, 2021 around 100,000 instances in the North America region and identified a saving opportunity of around $55 Million dollars annually – roughly 36% of the cloud bill.
At this moment, a question comes to mind. What are the numbers behind it? Can my organization incur similar savings?
Here are 4 simplified real-world examples from our analysis:
Example 1:
| Name | Instance Type | Avg CPU | Max CPU | Avg Memory | Max Memory | Recommended | Savings |
| Prod-1 | c4.8xlarge | 5% | 17% | 40% | 80% | r5.2xlarge | 68% |
| Linux | 36 vCPU, 60 GB | 8 vCPU, 64GB | Monthly – $782 | ||||
| us-east-1 | $1.591/hr | Total Hours | 720 hr/mo | $0.504/hr | Annual – $9,522 |
The client was facing performance issues and decided to upgrade in the same family to fix the same. However, the analysis showed that the CPU utilization was very low compared to memory. In this scenario, the client achieved 68% of savings by migrating to memory-optimized series.
Example 2:
| Name | Instance Type | Avg CPU | Max CPU | Avg Memory | Max Memory | Recommended | Savings |
| Prod-2 | r5.xlarge | 35% | 80% | 5% | 17% | c5.large | 66.27% |
| Linux | 4 vCPU, 32 GB | 4 vCPU, 8GB | Monthly – $120 | ||||
| us-east-1 | $0.252/hr | Total Hours | 720 hr/mo | $0.085/hr | Annually -$1,463 |
This client was similarly facing performance issues and was upgrading in the same memory-optimized family. However, data-driven analysis showed that the issue was with the CPU and not with the memory. Moving to a Compute Optimized series, they were able to achieve savings of around 33% without any performance impact.
Example 3:
| Name | Instance Type | Avg CPU | Max CPU | Avg Memory | Max Memory | Recommended | Savings |
| Prod-3 | r5.xlarge | 15% | 90% | 20% | 45% | t3.xlarge | 33.97% |
| Linux | 4 vCPU, 32 GB | 4 vCPU, 16GB | Monthly – $62 | ||||
| us-east-1 | $0.252/hr | Total Hours | 720 hr/mo | $0.166/hr | Annual – $750 |
This instance was rightsized on CPU and a bit oversized on Memory. But, according to our ML and AI algorithms, the average CPU was very low and max CPU peaked only at certain times of the day, which made this a perfect candidate for burstable instances. The client was able to see around 34% of the savings after moving to the burstable series.
Example 4:
| Name | Instance Type | Avg CPU | Max CPU | Avg Memory | Max Memory | Recommended | Savings |
| Prod-4 | r5.24xlarge | 4% | 20% | 7% | 8% | c5.9xlarge | 69.55% |
| Windows | 96 vCPU, 768 GB | 36 vCPU, 72GB | Monthly – 502$ | ||||
| us-east-1 | $10.465/hr | Total Hours | 69 hr/mo | $3.186/hr | Annual – 6,026$ |
This is a non-production instance that is running an average of 69 hours a month. In this case, rightsizing resulted in savings of $6,026.18/month.
These examples are, of course, simplified and barely scratch the surface of the entire process. For a more nuanced perspective on your business, feel free to reach out to me on my email. Hope this was useful!
For more strategies on how to optimize your cloud, schedule a call with our cloud experts or drop us an email at hello@cloudwiry.com
LISTEN »
