My thought on FinOps focused AWS Announcements - part 2

As promised, here's the second part of curated FinOps announcements from AWS between Thanksgiving and AWS re:Invent. While many updates were announced, nothing groundbreaking has emerged yet. Part one for reference.

Quick Context

Before diving in, it's worth noting that Amazon Bedrock (AWS's LLM service), Q (their integrated AI chatbot), and Connect (AI-powered customer service product) were major focal points. However, we'll focus on the more FinOps-relevant updates.

Disclaimer: This summary is based on pre-re:Invent announcements and may not capture all updates.

Key FinOps-Focused Announcements

1. Declarative Policies

A powerful new feature that allows organizational leaders to create robust guardrails around service configurations and usage. This will allow for better control over member accounts and potentially improve security and cost management. If done the right way- predefined configuration can be established across all existing and new member accounts. The fine line here is to not make it so restrictive that engineers can’t develop quickly. For example you can now specify predefined lists of EC2 instances that can be launched which would not allow for accidental large instances to run unchecked. On the contrary, this can become too restrictive and slow down development. There is a fine line and it is up to the engineers responsible for setting up these declarative policies to do so.  More here.

‍

2. EKS Auto Mode

This is probably the star announcement for EKS. One step further in the automation of infrastructure, this will autoselect EC2 instances, configurations, and scale based on demand. I would categorize this feature as the more mature versions of EKS for overall usage. A lot of engineers might want to have more control in the same aspect “advanced filters” exist in some dashboards, but there is a lot of space for making EKS deployment easier with this. More here.

3. Enhanced Cost Anomaly Insights

FinOps heavy update- Enhanced Cost Anomaly insights- instead of showing service level information. This new feature will be able to detail the exact grouping and filter causing the anomaly. 

Let me explain this feature in more detail, as it seems to be one of the few real FinOps features in this long list. Traditionally when you get an anomaly you usually look at cost by the service. So you don’t know what region, account, or machine type is causing the cost increase. Usually you’ll do multiple actions to find that out. Service A cost increase > let’s find out what member account > let’s find out what region in that member account > let’s find out what the exact machine by arn or name > let’s find out what exact usage type caused that issue. Aha! It’s the data transfer costs, as usual. This new feature should remove the multi-step process and deliver a clearer analysis of the root cause.

If you read this far, you deserve to know what is brewing behind the scenes for us. Our internal anomaly detection will get some enhancements too, where we will run our own internal detailed insights powered by some improved natural language explanations based on your previous trends, historical usage, and suggested solutions. Stay tuned! 

4. AWS Billing and Cost Management Data Exports

Word for word here- AWS Billing and Cost Management Data Exports for FOCUS 1.0 is now generally available. Great effort from all cloud providers to come into an agreement on columnar names, this will benefit cross-cloud teams to align on naming conventions. 

5. Improved AWS Pricing Calculator

A more intelligent tool for cost forecasting:

• Considers existing discounts
• Helps manage workload migrations
• Enables more accurate cost planning

We’re curious to try this out. However, not to keep plugging our services here- but the benefit of our monthly flexible commitments is that there is no need to spend all this time planning based on existing commitments and compromise, you can add more savings commitments as needed and remove it just as easily. 

6. SageMaker Improvements

• Scale inference to zero during low usage
• Potential for significant cost savings

7. Scheduled Reserved EC2 Capacity

As explained in my previous post this has nothing to do with reserved instances vs on-demand discounting. This is provisioning of EC2 capacity for critical workloads in a specific AZ, which can be scheduled in the future.

‍

New Instance Types (honorable mentions):

• I8G: Enhanced real-time storage performance
• M8G and R8G RDS instances in US and European regions

If anything was missed, please let me know and would love to include it in the next one, as I am sure there are probably more announcements to come. 

Try app.north.cloud for free. The savings you need, when you want them.