IONOS PostgreSQL Benchmark: OLTP Performance and Pricing
IONOS PostgreSQL is the first DBaaS from an EU-based cloud provider that enters our benchANT DBaaS OLTP Ranking. More European providers will follow in the next weeks. Under the SMALL ranking configuration, IONOS achieves throughput close to the available Google Cloud SQL configurations, yet at higher costs and therefore also lower price-performance.
For readers new to the ranking: we compare managed database services using a common workload and defined resource limits. The aim is to show how much performance different services deliver under documented conditions, with configurations and raw measurements available for inspection. The PostgreSQL comparison focuses on online transaction processing (OLTP), the short transactions used in applications such as order processing and payments. Our ranking introduction explains the broader methodology and scope.
IONOS’ Managed PostgreSQL offer
The managed PostgreSQL offering of IONOS CLOUD captures deployment, maintenance, backups and monitoring of database instances. In its current state, IONOS PostgreSQL neither includes nor offers performance optimisation services, be they schema-, index- or workload-specific. Such tasks remain with the customer, so (manual) application design and database tuning still matter. Clusters can be provisioned through the web-based Data Center Designer or the API.
Technical deployment options
The following options refer to the PostgreSQL v2 API, the latest API available when writing this article.
IONOS PostgreSQL allows CPU cores and RAM to be chosen and configured separately and hence does not provide pre-defined instance types as, e.g. AWS RDS does. This gives customers flexibility to match their CPU and memory requirements without choosing a fixed instance family. Storage capacity and instance count are also freely configurable.
Users cannot choose between different storage backends, but are limited to SSD Premium. The deployment interface does not expose a choice of CPU vendor, model or generation. When queried via SQL, the instance reports an x86_64 PostgreSQL build, which may be a hint, but of course does not identify the underlying processor.
Replication determines how the primary database forwards changes to standby nodes, which can take over if the primary fails. Asynchronous replication confirms transactions without waiting for a standby node, so recently committed transactions can be lost during failover. Strictly synchronous replication waits for a standby and stops accepting writes if no synchronous standby is available. This choice affects both commit latency and failure handling. The service also offers an optional connection pooling service to manage and reuse database connections.
Mapping IONOS to the SMALL specification
The methodology of the ranking defines a SMALL specification that allows each instance at maximum 8 vCores and 32 GiB RAM, targets 500 GB of SSD storage, and enables high availability. These limits provide a common starting point; they do not imply identical processors, storage performance or total cluster costs.
The tested IONOS deployment uses PostgreSQL 17.11 in the Frankfurt region, with three instances, each configured with 8 cores, 32 GB RAM and 500 GB storage. One primary handles the benchmark traffic; two standbys provide redundancy. Standbys cannot be used for read replicas and hence, the set-up is not suited to scale readers.
We choose three nodes because this is the minimum number required to enable STRICTLY_SYNCHRONOUS replication for a IONOS deployment. We choose this mode to align the comparison with AWS RDS PostgreSQL Multi-AZ DB instance deployments, which synchronously replicate the primary to a standby in another Availability Zone. Both setups wait for replication, although their implementations and fail-over behaviour differ. The hyperscalers listed in the ranking are configured to use two nodes in HA mode, which matters when comparing cluster costs.
Connection pooling is disabled, and PostgreSQL parameters otherwise retain the provider configuration. We therefore measure the service without workload-specific database tuning.
We apply the established benchmarking suite BenchBase with its TPC-C workload. BenchBase TPC-C models order processing through a mixture of read and write transactions. The dataset contains 150 warehouses and 75 terminals act as concurrent clients. Once started, the clients submit transactions for 30 minutes with a SERIALIZABLE level isolation. It is important to note that BenchBase TPC-C does not use simulated user pauses specified in TPC-C.
The load generator runs on an IONOS VM with 16 vCores in the same region.
Disclaimer: BenchBase TPC-C is not an official TPC-C implementation and the results are no official TPC results. In particular, BenchBase reports throughput requests per second across the transaction mix. These measurements are not official, audited TPC-C results and also are not comparable to the tpmC metric proposed by the official TPC-C.
Performance and price-performance
The comparison below covers the nine public entries in the ranking snapshot from 15 October 2026. Throughput shows the processing rate achieved during the benchmark. P95 latency reports the response time within which 95% of the measured requests finish; lower is better.
With IONOS delivering 1,792 ops/s, it reaches a throughput comparable to Google Cloud SQL which delivers between 1,634 and 1,779 ops/s. IONOS trails Azure by 28.9% and the listed AWS configurations by 32.5–56.3%. Its 113 ms P95 latency is close to the older Cloud SQL configuration and higher than Cloud SQL N4, Azure and AWS.
The IONOS calculator gives a compute-and-storage cost of approximately $2,079 per 30 days, using 720 operating hours and excluding backup storage. This covers all three nodes, each with 8 cores, 32 GB RAM and 500 GB SSD Premium storage (shown as Premium in Figure 2). Traffic, backup and support costs are not considered in the scope of the ranking.
Price-performance divides throughput by monthly cluster cost: a higher value means more measured throughput relative to cost. IONOS achieves 0.86 ops/s per monthly dollar, the lowest value among all entries of the public part of the ranking. Its throughput is close to Google Cloud SQL, but the IONOS’ cluster is more expensive. The third node adds redundancy and cost without handling any shares of the workload.
Additional technical observations
The recorded PostgreSQL settings show synchronous_commit=remote_apply and synchronous_standby_names=any 1 (*). In this setup, a write transaction waits until one standby has applied its changes before the client receives confirmation. As PostgreSQL documents, waiting for this extra step adds commit latency. Testing other replication settings would help establish its contribution to the measured performance. Provisioning also stood out during our tests: deploying a single PostgreSQL instance took IONOS approximately 40 minutes, compared with roughly 10 minutes for AWS, Azure and GCP.
Possible Tuning Options
With instance size, storage capacity and topology fixed, replication mode is the main exposed deployment setting to investigate. An asynchronous run would show the performance impact of no longer waiting for a standby.
Further experiments can examine permitted PostgreSQL settings, such as work_mem and planner parameters, or the optional connection pooler. Pooling may offer little benefit here because the workload already reuses persistent connections. Each experiment would keep the workload unchanged and document the configuration changes, so any improvement can be assessed against this baseline.
Summary and Outlook
With IONOS we add a European cloud provider to the public OLTP comparison. The tested configuration reaches Google Cloud SQL’s throughput range, while its higher cluster cost results in lower price-performance. For readers evaluating IONOS, the published result offers a starting point for tests with their own application workload and availability requirements.
We will extend the ranking with additional EU providers, workloads and deployment-time measurements. Explore the live ranking and the IONOS configuration details and raw measurements to examine the comparison in more detail.