Amazon RDS Multi-AZ DB Clusters offer near-Single-AZ performance by offloading write workloads to local NVMe disks, while maintaining high availability with semi-synchronous replication across Availability Zones, allowing for tunable replica lag to balance performance and consistency requirements.

<div>
<p>This article benchmarks the performance of three Amazon RDS for PostgreSQL deployment options: Single-AZ DB instance, Multi-AZ DB instance, and Multi-AZ DB Cluster. It uses HammerDB's implementation of the TPC-C benchmark to measure New Orders per Minute (NOPM) under an OLTP workload.</p>
<p>Specifically, the article covers:</p>
<ul>
<li>Overview of the three deployment options and their replication methodologies</li>
<li>Benchmark settings, including EC2 instance type, RDS instance type, database size, and workload parameters</li>
<li>Benchmark results showing Multi-AZ DB Cluster and Single-AZ DB instance outperforming Multi-AZ DB instance in terms of NOPM</li>
<li>Analysis of the performance differences based on replication methods and write commit paths</li>
<li>Conclusion highlighting the balance between performance and availability across the deployment options</li>
</ul>
</div>


Related articles