Cost Savings idea and metrics to find idle SQL instances and overprovisioned Instances separately.

Hello PIYUSH SOLANKI,

Thank you for reaching out to the Microsoft Q&A forum. 

When investigated In Azure the “RDS‐equivalent” PaaS services are:

• Azure SQL Database – fully managed single or pooled databases
• Azure SQL Managed Instance – near 100% SQL Server compatibility in a PaaS model

Below is a two‐step approach showing you how to find

(1) idle databases/instances and (2) over-provisioned ones, and then how to right-size them using the Azure CLI.

Identify idle SQL Databases / Managed Instances

• Key metrics to watch: – cpu_percent (average CPU usage) – dtu_consumption_percent (for DTU‐based DBs) – vcore_consumption_percent (for vCore‐based DBs) – data_io_percent, log_write_percent (optional)

• CLI example to pull the last 7 days of avg CPU for a single database:

az monitor metrics list \
  --resource /subscriptions/<sub>/resourceGroups/<rg>/providers/Microsoft.Sql/servers/<server>/databases/<db> \
  --metric cpu_percent \
  --interval P7D \
  --aggregation Average \
  --output table

•If you see Average CPU < 5 % (or DTU % < 10 % consistently) it’s a good candidate for decommissioning or pausing (serverless).

• For serverless tier you can also query the “paused” state:

az sql db show \
  --resource-group <rg> \
  --server <server> \
  --name <db> \
  --query "sku.tier, status"

Find over-provisioned instances

1. • Compare actual usage vs provisioned capacity. Rule of thumb: if avg CPU < 30 % and peak CPU < 75 % over a 30-day window, plan to downsize.

• CLI example to pull peak CPU and DTU for all DBs in a server:

for db in $(az sql db list --resource-group <rg> --server <server> --query "[].name" -o tsv); do
  echo "DB: $db"
  az monitor metrics list \
    --resource /subscriptions/<sub>/resourceGroups/<rg>/providers/Microsoft.Sql/servers/<server>/databases/$db \
    --metric cpu_percent dtu_consumption_percent \
    --interval P30D \
    --aggregation Average,Maximum \
    --output table

Right-size with CLI

• Change service objective (DTU model) or vCore count (vCore model) – DTU example:

az sql db update \
  --resource-group <rg> \
   --server <server> \
    --name <db> \
    --service-objective S0

– vCore example:

bash az sql db update \ --resource-group <rg> \ --server <server> \ --name <db> \ --edition GeneralPurpose \ --family Gen5 \ --capacity 2 

• For Managed Instance:

az sql mi update \
  --name <mi-name> \
  --resource-group <rg> \
  --vcores 8 \
  --storage <new-size-in-gb>

(Optional) Automate idle shutdown of Managed Instances • Use start/stop schedules to eliminate vCore charges during known idle windows:

https://docs.microsoft.com/azure/azure-sql/managed-instance/instance-stop-start-how-to

Reference list

1. Plan and manage costs for Azure SQL Database https://docs.microsoft.com/azure/azure-sql/database/cost-management
2. Azure SQL Database architecture best practices (cost optimization) https://docs.microsoft.com/azure/well-architected/service-guides/azure-sql-database#cost-optimization
3. Architecture best practices for Azure SQL Managed Instance https://docs.microsoft.com/azure/well-architected/service-guides/azure-sql-managed-instance#cost-optimization
4. FinOps best practices for Databases (ARG queries) https://docs.microsoft.com/cloud-computing/finops/best-practices/databases#sql-databases
5. Azure CLI – az monitor metrics list https://docs.microsoft.com/cli/azure/monitor/metrics#az_monitor_metrics_list
6. Choose between the vCore and DTU purchasing models https://docs.microsoft.com/azure/azure-sql/database/purchasing-models
7. Optimize Reserved Instance usage for existing resources https://docs.microsoft.com/azure/cost-management-billing/reservations/reserved-instance-purchase-recommendations

Let me know if any further queries - feel free to reach out!