Slash Your AWS Compute Costs through Causal ML automation with HNR Deliver and Causa

At HNR, we are constantly pushing the boundaries of what we can offer our clients through innovative technology, particularly when it comes to enhancing efficiency, generating cost savings, and reducing a business carbon footprint. Earlier this year, we launched HNR Discover, a tool that helps companies baseline their carbon footprint, power consumption, and financial costs across their digital operations. HNR Discover seamlessly integrates with a company’s existing Amazon Web Services (AWS) cost-and-usage report, providing immediate clarity on these critical metrics. This has laid the perfect foundation for the release of HNR Deliver, our latest offering.

As a business, we're deeply integrated into the Amazon Web Services (AWS) ecosystem, with a significant focus on serverless architectures and extensive use of Lambda functions. We have partnered with tech pioneers Causa to leverage their Causal Machine Learning (Causal ML) technology. This collaboration allows us to achieve notable cost savings in our Lambda operations for both ourselves and our clients.

Understanding AWS Lambda Functions

AWS Lambda functions are serverless compute services that run code in response to events. These functions are incredibly flexible, allowing you to execute code without provisioning or managing servers. However, the costs associated with Lambda functions can add up quickly, especially if not optimised properly.

One crucial factor affecting the performance and cost of AWS Lambda functions is memory allocation. This parameter determines how much memory your function has access to during execution and determines what CPU type will be used for execution. While more memory can speed up execution time, it can also increase costs. Striking the right balance is key to optimising both performance and expenses. And this is where HNR Deliver and leveraging Causal ML has proven crucial.  

Causal ML for Decision Optimisation  

If you've not heard of Causal Machine Learning (causal ML) before, the team at Causa provide an overview of causal AI, which includes causal ML, here.  

Standard AI/ML is designed purely for prediction, whereas causal ML is tailored for real-world decision making. This distinction stems from causal ML's ability to comprehend the underlying mechanisms influencing outcomes, a critical aspect for informed decision-making and optimisation. In the context of AWS Lambda, causal machine learning optimises decisions on Lambda memory allocation to impact both performance and cost. Unlike traditional machine learning that doesn't grasp cause and effect, causal ML delivers superior results through enhanced optimisations.  

How HNR Deliver Can Reduce your Lambda Costs

HNR Deliver excels in modelling complex issues using advanced causal machine learning techniques. At HNR, our team conducted rigorous tests in our lab, optimising memory allocation for our Lambda functions. This initiative resulted in significantly lower operational costs, a reduced carbon footprint, and consistent throughput. Our findings validate that impactful changes can stem from even minor configuration adjustments.

Our comprehensive tests included:

1. Deploying Lambda functions with various memory sizes.

2. Sending diverse workloads to these functions and analysing their performance.

3. Utilising causal ML models to recommend the optimal memory size for each function.

4. Implementing the recommended memory sizes and re-evaluating performance.

5. Comparing performance metrics before and after applying the recommendations.

These tests can be expanded virtually to millions of instances, boosting our confidence in the outcomes and enhancing our capability to optimise successfully.

Our findings have been remarkable, demonstrating a significant reduction in Lambda costs. In our controlled lab environment, we achieved an impressive 50% cost reduction, and we are confident that our clients can realise savings of 20-30% on their Lambda costs.

Benefits of Using HNR Deliver for Lambda Optimisation

Cost Reduction

By accurately modelling the relationship between memory allocation and costs, HNR Deliver can help you identify the most cost-effective configuration.  

Improved Performance

Optimising memory allocation isn't just about saving money; it's also about ensuring your functions run efficiently. HNR Deliver can help you find the ideal configuration where performance is maximised without unnecessary spending.

Lower Carbon Footprint

Reducing the resources your AWS Lambda functions consume also means lowering your carbon footprint. This is an increasingly important consideration for businesses looking to minimise their environmental impact.  

What Next?

The technology is deployment-ready with HNR Deliver. Clients can manually apply our recommendations to test on their HNR Lambdas or utilise our Node.js client within HNR to ZERO. Additionally, we can establish a retraining system to periodically update models and recommendations, ensuring optimal performance and sustainability. We call this process "Adaptive Tuning", and, for clients using this Lambda optimisation module, our goal is to conduct it monthly.

As this initial module becomes fully operational, the integration of additional modules becomes both simpler and quicker. We will soon start working on Causal ML optimisation modules for both EC2 and ECS.

At HNR, our mission is to enable data-intensive businesses to operate more sustainably and efficiently by reducing their technology footprint's carbon emissions and costs. Our latest solution, HNR Deliver, is designed to accomplish just that. Reach out to us to discover how HNR Deliver can transform your Lambda functions and deliver significant cost savings.