Technology / WORLD

AWS study finds PostgreSQL prewarming cuts restore delays

AWS compared Amazon RDS for PostgreSQL instances restored from the same snapshot. One 1,200-GiB database on gp3 was prewarmed with 15 parallel pg_prewarm sessions; another hydrated passively. The first instance completed storage initialization in about five hours.

Reading sizeSources ↓Share story ↗
Conceptual illustration of technology, processors and connected information.Illustration
Symbolic illustration

AWS compared Amazon RDS for PostgreSQL instances restored from the same snapshot. One 1,200-GiB database on gp3 was prewarmed with 15 parallel pg_prewarm sessions; another hydrated passively. The first instance completed storage initialization in about five hours. Both instances used db.m5.4xlarge with gp3 storage configured at 20,000 IOPS and 600 MiB/s throughput.

The passive instance had reached only 23% after five hours, with completion estimated beyond 24 hours. In later tests, a large table scan took 44 seconds on the prewarmed instance and 571 on the other; a complex join took 123 and 1,531 seconds. AWS says results are specific to its test configuration.

Data

Timeline

AWS study finds PostgreSQL prewarming cuts restore delays

AWS compared Amazon RDS for PostgreSQL instances restored from the same snapshot. One 1,200-GiB database on gp3 was prewarmed with 15 parallel pg_prewarm sessions; another hydrated passively. The first instance completed storage initialization in about five hours. Both instances used db.m5.4xlarge with gp3 storage configured at 20,000 IOPS and 600 MiB/s throughput. The passive instance had reached only 23% after five hours, with completion estimated beyond 24 hours. In later tests, a large table scan took 44 seconds on the prewarmed instance and 571 on the other; a complex join took 123 and 1,531 seconds. AWS says results are specific to its test configuration.

  1. · Occurred

    AWS study finds PostgreSQL prewarming cuts restore delays

    AWS compared Amazon RDS for PostgreSQL instances restored from the same snapshot. One 1,200-GiB database on gp3 was prewarmed with 15 parallel pg_prewarm sessions; another hydrated passively. The first instance completed storage initialization in about five hours. Both instances used db.m5.4xlarge with gp3 storage configured at 20,000 IOPS and 600 MiB/s throughput. The passive instance had reached only 23% after five hours, with completion estimated beyond 24 hours. In later tests, a large table scan took 44 seconds on the prewarmed instance and 571 on the other; a complex join took 123 and 1,531 seconds. AWS says results are specific to its test configuration.

Methodology

Reported values and benefits are attributed to the source; no independent measurement was made.

Updated:

Structured coverage

Subjects and places

Subject proposals

Taxonomy version: IPTC Media Topics 2026-Q2

Continue reading

Technology

AWS publishes AgentCore example for source-grounded presentations

The sample creates PowerPoint presentations from financial and compliance documents. In AWS’s internal test, it produced six slides from the 90-page Amazon 2024 Annual Report in 70–90 seconds.

AWS Machine Learning Blog1 min read Data included
Technology

AWS publishes deployment and evaluation workflow for JumpStart

An October 14 AWS guide shows how SageMaker Python SDK v3 can find and deploy a model from JumpStart, then measure it with sample prompts. The example records latency, approximate output length and single-request generation rate.

AWS1 min read Data included