I've made a few recent changes to save on our Azure expenses and wanted to share them with you all. Here's a brief rundown of what we've done:
1. **Single Front Door:** We replaced multiple App Gateways with one Front Door, which involved setting up a private link to our internal Kubernetes load balancer and switching from Azure Application Gateway Ingress Controller to NGINX.
2. **Custom API Gateway:** We removed Azure API Management since we developed our own solution for managing API traffic, which we found unnecessary.
3. **Front Door Consolidation:** We consolidated our numerous Front Doors into just one, simplifying our setup while ensuring we stay within the limits for endpoints.
4. **Log Level Adjustment:** We tweaked our log levels to reduce the volume of unnecessary logs being ingested, focusing only on errors.
5. **Burstable VMs:** We switched to using burstable VM series in our Kubernetes cluster to cut down costs a bit.
Next, we're looking to replace multiple SQL Servers with a single SQL server and elastic pool. I'm curious if anyone else has more tips for cost-saving measures in Azure!
Also, I'd love to know what VM series others are using for their Kubernetes node pools. Right now, we are on the Standard_B4ms for user node pools, and I wonder if there's a better option for cost.
3 Answers
If you haven't already, consider using reserved instances or savings plans for your long-term resources. They can help lower your costs significantly. Just keep in mind that you’ll be committing to a certain capacity for usually one or three years. It’s worth discussing with your team if you expect to run these services for a while!
I totally agree with you! It's a solid option if you’re planning to keep your setup for the long haul.
Also, if you’re running multiple databases, consider moving those that aren’t accessed often to a cold database tier. You won't be paying if they aren't in active use, and you can further reduce costs!
That’s a smart move! I didn’t realize how much I could save by adjusting to usage patterns.
Yes, it’s all about optimizing usage for sure!
Another idea would be to switch your Log Analytics tables to 'Basic' if possible. This helped cut costs by about $2k per month for one project of ours. Also, don’t forget to resize your VMs to match your actual needs and implement auto-shutdown on non-production environments to save more!
That's a great tip! Logging can get super pricey if you're not careful.
Right-sizing workloads can really make a difference, especially for VMs.
Yeah, but don’t forget to balance that with the potential for spot instances for any non-critical workloads. It could really help with costs!