How do you streamline rightsizing when migrating to the cloud?

We rely on Azure Migrate for our analysis prior to migration. Sometimes I even use LiveOptics to get a clearer picture of what we need. Streamlining the process has been really beneficial.

We focused on using the cloud provider's tools to analyze our applications before making the actual migration. As we transitioned different customer services, we made sure each one operated correctly and adjusted configurations as needed. Over the first year, we closely monitored recommendations from the cloud advisor and found that a lot of our initial sizes were way too large, allowing us to significantly downsize. After the migration, many customers started refactoring their setups to move from VMs to services as they realized they didn't need to run multiple virtual machines for simple tasks.

The right approach really boils down to acquiring and acting on the right data. Cloud engineering requires a level of precision that many might not be used to. So there’s no secret rule; it's all about getting it right based on solid information.

This process demands some thoughtful analysis. You should evaluate each VM to see if it can run on a smaller footprint in terms of RAM, disk space, or CPU. I recommend shifting focus from servers to services. For instance, if you're mainly using databases, consider options like RDS or Aurora to let AWS manage the infrastructure. Moving to an app-centric approach will help alleviate some of the cost surges associated with a direct, one-to-one migration.

When I joined my current job, the team was already deep into migrating to AWS without much planning. I took inventory of what was to be moved and categorized them based on their functions. After assessing current resources, I found suitable instance types using comparison tools. I created default sizes for each category and verified them with cost optimization tools. You do have to be careful though, as while cloud services can be cheaper, it requires effort to manage effectively.

From my background in cloud services, it's clear that the decision-makers often aren't the ones implementing the changes. You can approach this perfectly or just aim for 'good enough' and tweak it later. Using historical data is useful, but don't expect a perfect 1:1 transition because things change in the cloud environment. It's crucial to anticipate flaws when replicating existing infrastructure and make rapid adjustments. Most organizations I worked with found out that they had more fundamental issues with their applications rather than cloud-specific problems. Continuous modification is key to finding that sweet spot between cost and performance.