MrOwl Case Study
Background
Company
MrOwl is a cloud storage solution that enables its customers to store files and search on their content. MrOwl’s product also enabled the ability to share files and make those files publicly accessible and searchable. MrOwl utilizes AWS as their public cloud of choice for hosting their infrastructure. They view their AWS implementation as a cost center, by reducing costs we’re able to have an immediate impact on their bottom line.AWS is the public cloud of choice for MrOwl, accounting for their whole infrastructure. AWS is viewed fully as a cost center, and a cost reduction is a benefit to the bottom line.
“It takes an outside point of view to really take a look and assess what's happening in your
accounts” - Becky Raichur, Co-Founder & CTO
The Review
Initial Overview
During our initial review, we identified their key areas of overspend: Compute and Storage. In most environments, these
are the most common areas to have overspend, as misconfigurations have a significant impact on costs. In their
environment, the largest impact on their monthly spending was underutilized instances and misconfigured data retention
policies.
On initial review we found overspend in a few key areas: Compute and Storage. This is not a surprise, as these tend
to be the areas hardest hit by misconfigurations. Core to these areas are low utilization and misconfigured data
retention practices.
EC2:
We performed an in-depth analysis of their instance sizing to ensure that machines were appropriately utilized. Across the board, cluster sizes were decreased, and storage configurations were modified. Additionally, unnecessary resources were terminated.RDS:
Their environment was utilizing an out-of-date instance class, along with severe underutilization allowing us to take advantage of AWS’s lower-cost gravitron instance families.
Elasticache:
Their clusters were heavily underutilized, due to our ability to understand infrastructure from a holistic perspective we were able to recommend consolidation of the clusters along with a decrease in cluster size significantly reducing their AWS Spend.
S3:
We identified misconfigurations in S3 Buckets resulting in data usage that did not appear directly to the clients, as objects were being deleted the data was being retained resulting in ballooning S3 Storage costs.
DMS:
We identified an unutilized DMS Cluster that was previously performing synchronization between their backend data stores which was no longer active, we recommended terminating the unused resources.Support:
AWS Support is a percentage of their monthly AWS spend with their existing plan, by lowering our overall AWS costs we
were able to reduce the support costs significantly.
Areas of Improvement
While MrOwl did not ask us to perform a deep dive architecture review, we still provide an overview of recommendations
with regards to architecture. Such as moving from a standalone SSL VPN Solution to AWS managed SSL VPN. Along with that
moving from Bastion hosts to the newly released CloudShell. Furthermore, implementing an AWS Savings Plan to lower their
costs even further.Conclusion
Total Savings
| Service |
First Pass Savings | Total Savings |
|---|---|---|
| EC2 | 39.6% | 59.8% |
| RDS | 77.8% | 83.9% |
| ElastiCache | 72.8% | 77.1% |
| S3 | 88.5% | 88.5% |
| DMS | 100% | 100% |
| Support | 37.7% | 46.7% |
| Total | 54.4% | 66.7% |
“AWS Costs are hidden, and they make it hard to understand” - Becky Raichur, Co-Founder and CTO
We saw the most savings in RDS and S3. We have found that these tend to be the more misconfigured services across our customers as they are the services that tend to scale costs exponentially. RDS instance types should be tuned to real-world database events and scaled accordingly, and S3 based backups should align with your organization's high availability goals, not exceed them.