March 9, 2026 at 12:00 am
Comments posted to this topic are about the item Not Just an Upgrade
March 9, 2026 at 9:02 am
After some user complaints we got to upgrade the hardware where the reporting database server runs on, notorious for heavy cpu usage.
Going from 2019 hardware to 2025 gave us a 30% boost (cpu/ram) resulting in etl finishing 2 hours faster. Resulting in happy users: no more waiting on ETL to finish, less bottlenecking where one user would soup up the available cpu, faster response
March 9, 2026 at 3:57 pm
Amazing what hardware will do. Did you add CPUs/RAM?
I'm all for optimizing code, but also for considering resource changes.
March 10, 2026 at 9:26 am
All the work I do is in the cloud, and the Database As A Service model means that patching and upgrading are mostly outside of my jurisdiction. I can choose a size of DB instance, and largely that is that.
I also use Snowflake. Again, DBaaS, but we have a choice as to whether to use generation 1 or 2 warehouses.
Gen-2 cost 25% more, and for certain loads can be up to 40% faster. In such a case, Gen-2 will save you money. Gen-2 gives you a more recent processor and better memory, but if CPU and memory aren't constraining what you do, then Gen-1 might be the better bet.
We've looked at the cost of refactoring our data pipelines to reduce cloud costs and also reduce the need to upgrade to more expensive hardware. This is where Claude's code demonstrates value. We have a good idea as to what needs refactoring. We asked Claude to catalogue and quantify what needs to be done and to detect code smells. It took 1 hour to get set up and for Claude to produce its output, which has been accurate.
We have asked it to do more as a follow-up and are checking what it has produced. So far, it is looking promising. We couldn't afford the cost of doing this work the traditional way. It would take too long and distract us from revenue-generating activity. Even at this early stage, we can see real benefits. If what we do is successful, then it means that addressing years of tech debt becomes feasible and affordable.
In the case of self-managed database servers, I would suggest having an upgrade plan, whether you upgrade or not. That way, if the need to upgrade is thrust upon you, then you are not on the back foot.
Viewing 4 posts - 1 through 4 (of 4 total)
You must be logged in to reply to this topic. Login to reply