Single-thread performance and predictable latency

Best OCI Instances for Databases

Databases have different bottlenecks than stateless workloads: single-threaded query planning, buffer pool management, and write-ahead logging all depend heavily on single-core speed and memory bandwidth. While value-per-vCPU still matters, you should also weigh single-thread PassMark scores and networking bandwidth for replication and client throughput. Bare Metal shapes eliminate hypervisor overhead entirely, giving you consistent sub-millisecond latency — critical for OLTP databases like PostgreSQL and MySQL running under heavy write loads.

What to look for

Database query planning and execution are often single-threaded. High single-thread PassMark and strong networking bandwidth reduce query latency and replication lag. For PostgreSQL and MySQL, Bare Metal shapes are worth the premium for consistent, jitter-free latency.

  • Single-thread PassMark
  • Value score (PassMark/vCPU ÷ $/vCPU/hr)
  • Networking bandwidth (Gbps)
  • RAM (GB)

Ranked instances — 51 shapes

#Shape$/vCPU/hr
1$0.0100
2$0.0100
3$0.0125
4$0.0125
5$0.0125
6$0.0125
7$0.0125
8$0.0125
9$0.0150
10$0.0150
11$0.0150
12$0.0150
13$0.0150
14$0.0150
15$0.0150
16$0.0150
17$0.0150
18$0.0150
19$0.0270
20$0.0375
21$0.0069
22$0.0069
23$0.0070
24$0.0319
25$0.0319
26$0.0319
27$0.0319
28$0.0319
29$0.0319
30$0.0069
31N/A
32N/A
33N/A
34N/A
35N/A
36N/A
37N/A
38N/A
39N/A
40N/A
41N/A
42N/A
43N/A
44N/A
45N/A
46N/A
47N/A
48N/A
49N/A
50N/A
51$0.0069

More use cases