I’m a recent graduate working on my personal project, and I chose NEON as my production database. Everything was great until I received a $100 bill, which is a significant amount for me in Mongolia. So, I started analyzing my costs and decided to reduce my vCPU from 1 to 0.25. While I expect this will lower my bill, I’m concerned about potential impacts on query response times.
From my research, it seems that lowering the vCPU count doesn’t significantly affect response times.
Can anyone confirm this?
I’d appreciate any insights or experiences regarding the impact of vCPU allocation on query performance in a serverless environment like NEON.
Hi! If you have efficient queries/indexes and load that doesn’t experience extreme spikes, and the amount of memory offered by the 0.25 compute unit size is sufficient, then as you noticed your Postgres instance will continue to provide acceptable response times.
If your application is not always actively receiving traffic, you can use Neon’s auto-suspend feature to reduce your bill.
Our billing documentation has an overview of how pricing is calculated and sample estimates: Billing metrics - Neon Docs
The basic formula is:
Compute time cost = Compute size * Active time * Compute-time price
Compute size and Active time are the variables you can control to reduce your compute bill. Your selected region also has an impact. This table lists pricing per region: Billing rates - Neon Docs
I hope this helps!