Over the past 8 years, ProsperOps has autonomously generated $5 billion in savings for our customers across AWS, Azure, and Google Cloud. Historically, customers have optimized to save.
Today, with AI on page one of every company’s agenda, customers optimize to innovate.
Our mission is to unlock optimization dollars to fund this innovation. We’ve historically accomplished this through automated rate optimization, but our vision has always been bigger.
Coming together with Flexera gave us access to other key FinOps capabilities that enhance our ability to impact optimization outcomes.
Ocean and Elastigroup are Flexera products that perfectly complement ProsperOps through expanded rate optimization (automated spot instance management) and workload optimization (automated Kubernetes cluster optimization and intelligent virtual machine autoscaling).
While all three products reduce different aspects of cloud waste, they are disparate solutions. Without shared context and coordination, their independent optimization actions can battle each other and deliver overall suboptimal outcomes.
What if a “single brain” could holistically automate the broad optimization problem set to ensure savings outcomes were optimal?
Introducing Unified Autonomous Optimization.
Single Brain for Rate and Workload Optimization
With Unified Autonomous Optimization, customers who use ProsperOps and Ocean/Elastigroup achieve higher savings and Effective Savings Rates through holistic automated outcomes.
Let’s see how this works.
Example 1: Spot and commitment management
Before, Ocean/Elastigroup’s spot management algorithms had no visibility into ProsperOps-managed commitment expiration dates, their ability to be exchanged, and what future commitment patterns would look like (e.g. Adaptive Laddering™). Without that context, moving compute workloads to spot could inadvertently make commitments unutilized, which would generate less savings. When working independently, the algorithms were conservative, leaving savings on the table.
Now, with shared context and coordination, the systems cooperate to continually automate the optimal commitment and spot portfolio mix to achieve maximized financial outcomes.
Example 2: Kubernetes optimization
Prior to this new capability, Ocean’s automated pod rightsizing, bin packing, and node autoscaling algorithms would work to reduce compute usage. However, if commitments existed for the compute nodes being removed or scaled down, the workload optimization actions might yield no savings.
Unified Autonomous Optimization enables ProsperOps and Ocean to analyze and adapt the commitment portfolio and cluster nodes together.
The following autonomous rate and workload optimization capabilities have been unified and are now holistically optimized:
- Pod rightsizing: Adjust pod resources to match container needs in run-time.
- Bin packing: Pack right-sized pods more efficiently across nodes, freeing capacity.
- Node/VM autoscaling: Scale nodes/VMs in response to pod and workload demand.
- Node/VM scheduling: Turn on or off nodes/VMs based on preset schedules.
- Commitment management: Manage the optimal mix of commitment types (Reserved Instances, Savings Plans, Committed Use Discounts) and terms to maximize savings and flexibility.
- Spot management: Intelligently and safely layer in spot instances (which offer the highest discounts) where possible.
Real World Results
Our customers have enabled Unified Autonomous Optimization in production, and we monitored their Effective Savings Rates to quantify impact.
Duolingo is the world’s most popular language-learning platform. They enabled this capability for their largest production Kubernetes cluster. Prior to the change, their cluster-specific All Discounts* Effective Savings Rate (ESR) was 55.8% (which was already high compared to industry peers).
After the change, their ESR increased to 60.8% (net of ProsperOps and Ocean charges).
To illustrate the impact, imagine you are running $1 million in monthly cluster usage. An increase of 5 percentage points in ESR would generate an incremental $50k/month or $600k/year in savings for reinvestment.
That’s one cluster. What kind of results can be achieved at scale?
Tealium, the premier customer data platform, enabled this capability across their entire cloud estate. Their overall compute All Discounts* ESR for the month was 62.4% (again, net of ProsperOps and Ocean charges), meaning that for every 1 dollar of on-demand compute spend, they only paid their cloud provider ~37 cents.
These results are truly world-class and can only be achieved through a unified system sharing context and coordinating autonomously.
* Note: The All Discounts Effective Savings Rate scope measures the efficacy of all rate optimization efforts including commitments, spot, and custom pricing, net of management fees.
Start Optimizing
Unified Autonomous Optimization is generally available starting today. If you’d like to quantifiably understand the incremental savings opportunity for your environment, we invite you to run our free savings analysis.
New customers can sign up here. For current ProsperOps customers, please reach out to your customer experience advisor. If you are a Flexera customer, please contact your account manager.
Less waste and improved optimization outcomes are always a good thing. Continually finding ways to drive higher Effective Savings Rates is how we do our part in helping customers prosper even more.
Prosper on! 🖖
Erik
