I85PREP
⌘ K
Oracle Cloud AR study materials
Oracle Cloud AR study resources

Oracle Cloud AR: Designing Database, Observability and Cost-Optimization Solutions on OCI

Oracle Cloud AR · Work through your exam, one topic at a time.

All guide sections

Designing Database, Observability and Cost-Optimization Solutions on OCI

This domain tests database selection and the observability stack that proves a design works, then asks you to cut spend without breaking SLOs.

Database choices

  • Autonomous Database (serverless or dedicated) handles patching, tuning and backups; supports auto scaling and Autonomous Data Guard. Choose it for minimal-ops OLTP and mixed workloads.
  • Base Database on VM or bare metal DB systems gives full control (SIDs, custom parameters, RAC); Exadata for extreme performance and consolidation.
  • MySQL HeatWave for MySQL-compatible analytics; Data Guard and sharding for scale and DR on Base DB.

Observability

  • Monitoring: metrics, dimensions, alarms, and notification topics. Memory and disk metrics need the Management Agent - they are not free by default.
  • Logging: service logs and custom logs in log groups, searchable and retained per policy; Service Connector Hub streams logs and events to Object Storage or Streaming.
  • Events for resource state changes; APM for traces; Operations Insights and Database Management for capacity and SQL-level tuning.

Cost optimization

  • Cost Analysis sliced by compartment and cost-tracking tag; budgets with alert rules; quotas to stop runaway provisioning.
  • Right-size flexible shapes, autoscale, and stop idle instances - you still pay for their boot volumes.
  • Object Storage lifecycle and auto-tiering; lower block volume performance tiers where IOPS allow; BYOL where licensing permits.

Worked example

A dev tenancy spends heavily on always-on 8-OCPU VMs and Standard Object Storage. Move to 2-OCPU flexible shapes with schedule-based autoscaling for business hours, stop non-production instances at night, add a lifecycle rule moving objects older than 30 days to Infrequent Access, and set a budget alert at 80% of forecast. Reasoning: each change attacks a different cost driver - idle compute, idle capacity, and storage class - without touching production SLAs.

Traps

Alarms do nothing without a notification topic and subscription. Monitoring Agent must be installed for guest OS metrics. Stopping an instance does not eliminate boot volume charges. Deleting a compartment requires removing all resources inside it first.