I’ve noticed a trend, and apparently AWS has too.
Multi-cloud. There, I said it. The M word was strictly forbidden while I was at AWS, now there’s documentation around it.
Like it or not, the world is moving to multi-cloud and lots of change.
Change meaning a lot of uncertainty in terms of economy and the tech landscape with the rapid advancement of AI.
As organisations move towards cutting costs, while still rapidly experimenting with AI, how would you as a builder show your value to your manager and their bosses?
The answer is easy.
Make money appear in front of them.
But, I don’t work in Sales or Marketing you might say.
You don’t have to.
The answer lies in cost optimisation across your cloud workloads to "make money appear from no-where".
At AWS I helped customers find where some of the levers were to help save costs e.g. right-sizing, using Savings Plans and Reserved Instances etc.
There are multiple dashboards that provide customers visibility into what their spend is on AWS.
Great! But many customers were stuck with what they needed to actually action with this data.
With most Enterprise customers being multi-cloud, do they need to do the same thing across multiple clouds?
Yes, and that meant more time spent on analysing individual dashboards and compiling a list of actions for each Cloud Service Provider.
After achieving the Linux Foundation’s FinOps Certified Practitioner and FOCUS Analyst certifications I had a lightbulb moment.
What if we can convert all the CUR (Cost and Usage Report) data from each Cloud Provider, convert it to a FOCUS format data and analyse it?
As it happened, I was curious around playing around with Snowflake and jumped on a “Zero to Snowflake in 90 mins” hands-on labs.
Well done on a really interactive session Snowflake! I've delivered and participated in many hands-on lab sessions in my time both virtual and in-person, but that was one of the most engaging sessions I've attended.
I looked around for FOCUS sample datasets that contain multiple Cloud Service Providers which I found on GitHub here.
Then I was looking around for any Snowflake tutorials that I can run cost optimisation queries for, but I couldn’t find any.
It re-ignited being able to build something quickly and intuitively on Snowflake using AWS behind the scenes on a free Snowflake account.
So initially I was having a look at what’s possible with the visualisations via SQL statements.
I was happy with it initially but then I realised, Snowflake dashboards are quite simple and are not for full fledged visualisation like Tableau or even Amazon QuickSight.
So I went one step further by creating a Streamlit app via the in-built Snowflake feature and had a FinOps chat interface.
This makes it more intuitive as you can select quick actions, or ask it specific questions using the same data in the Data Warehouse using Cortex.
Here are some of the highlights of the Streamlit app.
One of the first questions is around modernisation to help with cost optimisation. A lot of organisations that I worked with still had a lot of tech debt, with monolithic apps running on Amazon EC2s. Modernising them into microservices gives customers the benefit of optimising costs as you move to containers and serverless services.
Some people spend the weekends going outside, but since it was rainy here I figured I’d spend my time building something useful.
It’s all about showcasing your value to your organisation by thinking big and inventing and simplifying, based on customer needs.
Whether that customer is an internal customer or external, it doesn't matter.
You should be continuously improving by keeping up to date with your technical skills, just like AI is by adding data to the vector database in a RAG model as your "context".
This blog is about FinOps with AI tooling that helps with visibility.
If you want to learn more about FinOps for AI workloads across multiple environments, have a look at this awesome overview https://www.finops.org/topic/finops-for-ai/
Top comments (0)