Understanding Hetzner SSD VPS Performance and Best Practices

#cloud #devops #performance

Understanding Hetzner’s SSD-Backed VPS Performance

Hetzner Cloud VPS run on hypervisor servers equipped with NVMe SSDs, often configured in RAID10 for optimal performance. This hardware stack delivers enterprise-grade speed, with independent benchmarks showing sequential read/write speeds peaking at multiple gigabytes per second.

For instance, one midrange CPX server test recorded throughput of approximately 1.4 GB/s^[1], while newer cost-optimized instances have demonstrated up to 4.46 GB/s on large block transfers^[2]. In 2023, Hetzner tied for the top score (9.5/10) in disk I/O in a comparison of popular VPS hosts^[3], underscoring its strong storage capabilities.

Shared vs Dedicated Storage in Hetzner Cloud

Like most clouds, Hetzner’s VPS are multi-tenant, meaning storage is shared across virtual machines on the same physical host. This can introduce “noisy neighbor” effects, where intensive I/O from one VM temporarily affects others.

However, Hetzner's Dedicated General Purpose (GP) plans provide unshared hardware resources to eliminate such contention. These plans are designed for low latencies and consistent high CPU, network, and disk usage, making them ideal for production workloads.

In summary: For consistently high disk I/O performance, choose a dedicated/GP instance or even bare-metal servers rather than smaller shared-core instances.

Local NVMe SSD vs Network-Attached Block Volumes

Hetzner offers two core storage types:

Local SSD (True NVMe attached directly to the host)
Volumes, which are network-attached SSD block storage

Benchmarks reveal a significant performance gap. On a typical Hetzner Cloud VM, the local SSD can sustain around 39.6k IOPS and ~155 MiB/s for 4KB reads, whereas a 4 GB cloud volume achieves only about 7,500 IOPS (~29 MiB/s)^[4].

This roughly 5× better IOPS on local SSDs is expected since local NVMe storage directly utilizes the host’s controller and bus, whereas volumes traverse Hetzner’s internal network with additional overhead and redundancy.

Best Practice: Use local NVMe SSD storage for I/O-intensive workloads such as databases and file processing, and reserve Volumes for backups or data sharing.

Performance Tuning Recommendations

Maximize your disk performance by applying standard Linux tuning techniques:

Mount SSD partitions with noatime,nodiratime options to avoid unnecessary metadata writes, significantly improving IOPS^[5].
Use modern I/O schedulers optimized for SSDs such as mq-deadline or none.
For databases or write-heavy applications:
- Batch commits
- Increase buffer sizes and cache writes
- Enable asynchronous I/O if supported
Keep swap usage minimal to avoid degrading disk performance.
Consider tmpfs or in-memory caches (e.g., Redis, Memcached) for high-speed temporary data storage.
---

Scaling Strategies for Disk I/O

Avoid bottlenecks by distributing workloads horizontally:

Add more VPS instances to spread I/O load.
Use load balancers, caches (CDNs or memory caches), and database sharding.
Offload static assets to object storage or CDN.
Employ read replicas for databases.
Use batch jobs spread over multiple nodes.

Hetzner’s generous network allowance (20 TB traffic included on EU sites) makes multi-server architectures practical.

If vertical scaling is necessary, upgrade to larger VPS plans with more CPU, RAM, and NVMe SSD storage, or transition to Hetzner’s dedicated server lineup.