Finances Friday
I'll be a grown-up one day, probably
2025-12-27, by DrFriendless AWScosts
Being unemployed, particularly at this time of year, can cause a bloke to lose track of what day it is. DrMrsDrFriendless is away overseas so she’s not around to tell me either. It’s up to me to be a grown-up and figure it out by myself.
One discipline I’m trying to instill in myself is the concept of “Finances Friday”, where I pay some attention to how much money I’ve got. This is a problem because I hate thinking about such things. They do not engage me at all. But today is Finances Friday, and as you can see it’s a Saturday so it’s not really going according to plan yet.
The good news is that when I started this last Friday I discovered an AWS cost that could be avoided, and coded around it. Then today I discovered a graph in the AWS Billing section that shows me that that worked. Here is the graph of AWS costs per day for the last 4 weeks or so.
I don’t know for sure what that spike on December 1 was, I think it might have been a domain name registration. But as you can see, since the change on about December 19, costs are running about a dollar per day lower. Woohoo, it’s going to take me longer than before to become poor!
Just to confirm that it was the removal of the VPC endpoint which caused that change, I graphed the costs per hour per service.
The colours are hard to discern, but the tallest lines are:
- kinda blue - Relational Database Service
- kinda red - Virtual Private Cloud
- kinda green - EC2 (servers)
The most recent set of lines is smaller because it’s for today, which isn’t finished yet. Otherwise, the graph shows that the VPC costs decreases dramatically, which is what I planned.
The database remains by far the biggest cost. I’d love to decrease its size and hence cost, but it’s already down at the “you cannot seriously want a database this puny” tier, and AWS doesn’t offer much smaller. Nevertheless it is performing well.
One of the problems I had back in the old days was that database contention caused writes from Lambdas to fail, requiring them to be re-run, thus increasing write activity and hence database contention. The implementation of the inside queue plan using the EC2 and SQS (AWS queueing service) has fixed that. What we are afraid of is the green line coming down to the bottom, but as long as the orange line stays low we should be good.
So for this week I’m content with Finances Friday, albeit kinda Saturdayish. Let’s hope I am as diligent again next week.
