---
last_verified: 2026-03-14
description: Memory sizing guidelines for OCI compute and database workloads.
---

compute_memory:
  flexible_shapes:
    min_gb_per_ocpu: 1
    max_gb_per_ocpu: 64
    default_gb_per_ocpu: 16
    recommendation: |
      Start with 16 GB per OCPU for general workloads.
      Java applications: 8-16 GB per OCPU.
      In-memory caches (Redis, Memcached): size for data set + 20%.
      Database client/app servers: 8 GB per OCPU usually sufficient.

  common_ratios:
    general_purpose: 16  # GB per OCPU
    memory_intensive: 32
    compute_intensive: 8
    java_application: 12
    web_server: 8

database_memory:
  adb_s:
    notes: "Memory managed automatically by ADB. Not directly configurable."
    sga_estimate: "~7 GB per OCPU"
    pga_estimate: "~2.5 GB per OCPU"

  exacs:
    notes: "Memory allocated per VM cluster from Exadata infrastructure pool"
    recommendation: "Size VM cluster memory based on SGA+PGA requirements"
    typical_sga_per_ocpu: 8
    typical_pga_per_ocpu: 3

  dbcs:
    notes: "Memory = compute shape memory. Choose shape accordingly."
    recommendation: |
      Total DB memory = SGA + PGA aggregate
      SGA = 60-80% of available memory
      PGA = 10-20% of available memory
      OS overhead = 10-15%

sizing_from_source:
  on_prem_oracle_to_oci:
    steps:
      - "Get current SGA + PGA sizes from AWR/ASH"
      - "Add 15% for OS overhead"
      - "Map to compute shape with sufficient memory"
      - "For ADB: translate to OCPU count (memory scales automatically)"
    notes: "ADB auto-tunes memory, so OCPU sizing is the key input"

  aws_rds_to_oci:
    steps:
      - "Note current instance class memory"
      - "Map to equivalent OCI shape memory"
      - "For ADB: divide memory by 10 to estimate OCPU count"