Faster storage, faster server: measuring, comparing and optimising disk performance on a VPS
Christopher | 14 Apr 2025
Google Gemini
The performance of Virtual Private Servers (VPS) is largely influenced by the performance of the underlying disk systems. Especially for data-intensive workloads, the speed of the storage determines whether applications run smoothly or stall. To assess the performance of a VPS, measuring disk performance is therefore essential. Both write and read speed per second and the number of Input/Output Operations per Second (IOPS) are analysed in detail – two central metrics that provide deep insight into the behaviour of storage systems. In addition to established measurement tools, various approaches are available to specifically improve disk performance.
Disk performance as a critical factor in VPS performance
Disk performance is one of the central factors affecting the overall performance of a Virtual Private Server. In virtualised environments, where multiple users share the same physical hardware, the storage device can quickly become the bottleneck. Applications that process many small files or continuously access data – such as databases, web applications or CI/CD pipelines – are particularly dependent on fast write and read operations. Besides the raw transfer rate, the number of possible I/O operations per second (IOPS) is especially important, as it provides insight into the system’s ability to handle many concurrent accesses efficiently. Insufficient disk performance can manifest as longer load times, delays in script execution or sluggish response times across entire applications. Anyone who wants to accurately assess and optimise the performance of their Virtual Private Server should therefore pay particular attention to the storage component.
Tip: Also see CPU performance for Virtual Private Servers
Difference between I/O operations per second (IOPS) and write/read throughput per second
IOPS and write/read throughput are two distinct but closely related metrics for evaluating disk performance.
IOPS (Input/Output Operations Per Second) indicates how many input and output operations a storage system can process within one second. It is about the number of individual operations, not the volume of data – regardless of whether 4 KB or 1 MB is being read or written. This metric is particularly relevant for applications with many small accesses, such as databases or log files.
The write and read throughput per second, usually expressed in megabytes per second (MB/s), measures the actual amount of data that can be written or read per second. This metric is especially important for large, sequential files – for example when copying large media files or loading substantial software packages.
Summary:
- IOPS measures the number of operations → important for many small accesses.
- MB/s measures the data volume per second → important for large, sequential accesses.
An efficient system should ideally deliver good results in both areas – high IOPS for responsive applications and high MB/s for fast data throughput.
Tools and methods for analysing disk performance on VPS hosting
To assess the disk performance of a Virtual Private Server (VPS) precisely, various tools and test methods are available. They provide important metrics such as the maximum transfer rate, the number of I/O operations per second (IOPS) and the average access time. Depending on the deployment scenario, different measurement methods are appropriate.
iotop
For a quick overview of current read and write rates, the command-line tool iostat, which is part of the sysstat package, is suitable. It provides real-time information on block device utilisation and shows how heavily the disk is currently being used. Complementing this, iotop offers a live view of which processes on the server are accessing the disk most intensively.
Example evaluation of disk measurements with iostat
A practical use case for analysing disk performance on a VPS hosting instance is the evaluation of the measurements provided by iostat. The Device line gives a quick overview of the activity of the available storage media. Particularly relevant are the columns:
- tps (transactions per second)
- kB_read/s (read throughput)
- kB_wrtn/s (write throughput)
- kB_dscd/s (discard/trim data on SSDs).
In the example measurement on an Ubuntu VPS hosting instance, the result shows a typical load situation without anomalies:

| Device | tps | Read | Write | Note |
| sda | 1.36 | 8.88 kB/s | 21.97 kB/s | Normal, low utilisation on the main disk |
| loop0 | 0 | 0 | 0 | Temporary system mounts, e.g. snapshots |
| sr0 | 0 | 0 | 0 | Virtual CD-ROM drive |
Notably, the main drive (sda), with a transaction rate of 1.36 I/O operations per second and very low read and write rates, is only minimally utilised. This indicates a low current system load and shows that the disk does not represent a bottleneck in this situation.
Other devices such as loop0 and sr0 are present for system reasons but do not cause any relevant I/O load.
Especially for VPS hosting with shared storage, iostat can provide important indications under load as to whether a high number of transactions or slow throughput rates point to an overload of the storage system. For administrators and users of virtual servers, regular analysis of these values is a simple and effective method to realistically assess the performance of the storage system.
ioping
ioping is used to measure access times. It works similarly to a network ping but targets the storage system. Especially for database workloads that cause many small write and read operations, ioping provides valuable indications of potential latency.
Example evaluation of the disk measurement with ioping
For assessing the response speed of the storage system, the tool ioping is suitable as it performs targeted measurements of disk access times. In a short test on the VPS hosting instance under test, the measured values ranged between 0.96 and 1.88 milliseconds per access to 4-kilobyte blocks. The average latency was 1.19 milliseconds, which indicates an SSD-based storage solution and is a fairly common value for typical entry-level VPS hosting offerings.

| Metric | Value | Assessment |
| Minimum latency | 0.96 ms | Good (normal for SSD-based VPS hosting) |
| Average latency | 1.19 ms | Solid performance |
| Maximum latency | 1.88 ms | Not critical |
| IOPS value (test) | 838 | Acceptable for random reads |
The results of the measurement with ioping show an average access time of around 1.2 milliseconds. For VPS hosting with SSD storage this value is typical and falls within a practical range. While dedicated NVMe SSDs in high-performance environments can achieve even faster sub-millisecond values, the measured result provides a solid foundation for web applications, smaller databases and everyday server workloads.
fio
For synthetic load tests and targeted benchmarks, the tool fio is the practical standard. It enables the simulation of a wide range of access patterns – from sequentially writing large files to random read access with small block sizes. This allows typical workloads to be reproduced realistically and I/O performance to be assessed under load conditions.
Analysis of the fio measurement for disk performance
In the present example, the read performance of the VPS was tested with the benchmark tool fio. A random-read scenario (randread) with a block size of 4 kilobytes was used – a typical scenario to evaluate storage performance under many small accesses. This test procedure simulates real-world conditions such as those found in databases or web applications handling many small files.

| Metric | Measured value | Meaning |
| IOPS | 501 | Number of read operations per second |
| Throughput | 2007 KiB/s (≈ 2 MB/s) | Amount of data that could be read per second |
| Average latency | approx. 7.9 milliseconds | Time taken to serve a read operation |
| 99.9th percentile latency | approx. 12 milliseconds | 99.9% of all accesses were below this delay |
| Maximum latency | 214 milliseconds | An outlier in access duration |
The measurements show comparatively low storage performance – typical for many VPSs using shared storage.
- An IOPS value of around 500 is at the lower end for SSD-based hosting and may be barely acceptable for production applications with many accesses.
- The data throughput of around 2 MB/s is also low and indicates that the disk or storage system is heavily virtualised, with multiple customers sharing the same resources.
- The average latency of just under 8 milliseconds is still acceptable in practice, but noticeably higher than with dedicated SSD systems.
- The high util value of 99.72% in the disk stats also shows that the disk was almost fully utilised during the test.
The fio benchmark results illustrate how strongly the performance of VPSs can be limited by the storage infrastructure used. Especially in virtualised environments with shared storage, IOPS values in the range of 500 to 1000 and data throughputs of just a few megabytes per second clearly point to an entry-level solution with limited headroom. For data-intensive applications, using servers with dedicated SSD resources or an optimised storage backend is therefore recommended.
For a sound assessment of disk performance, a combination of several tools and tests is recommended. This captures different aspects of storage performance and provides a realistic evaluation of a VPS hosting instance's performance.
Disk performance of VPS hosting instances: How we test with fio
To evaluate disk performance of VPS hosting instances realistically and comparably, we rely on the benchmark tool fio (Flexible I/O Tester) in our VPS hosting tests. This programme is part of the test script YABS (Yet Another Bench Script) and simulates different access patterns on the storage system. Among other things, it performs random read and write operations with small block sizes — a practical scenario that places particular strain on data-intensive applications such as databases. The metrics obtained, including IOPS (Input/Output Operations per Second), throughput and latencies, are then incorporated into our performance evaluation of the VPS hosting instance.
What disk benchmarks on VPS hosting tell you — and what they don't
Disk benchmarks such as fio or ioping provide important indications of the storage system's capabilities on a VPS hosting instance. High IOPS figures and short access times point to a fast SSD- or NVMe-based setup and are especially beneficial for data-intensive applications.
However, the measurement results are always snapshots taken under specific test conditions. In shared hosting environments, performance can vary depending on the time of day or the load from other customers. Benchmarks therefore show the technically possible performance range — but do not guarantee consistently stable performance in real-world operation.
Which values for disk performance are good?
The following guideline values for read/write speed and Input/Output Operations per Second (IOPS) are considered good, medium or poor:
The measurement for disk read/write speed must be in the range of over 1500MB/s to be rated as good. Values between 1500MB/s and 1000MB/s are rated as medium. All values under 1000MB/s are considered poor.
The measurement for disk Input/Output Operations must be in the range of over 25.000k IOPS to be rated as good. Values between 25.000 IOPS and 15.000k IOPS are rated as medium. All values under 15.000k IOPS are considered poor.
Find out which VPS in our performance test scores best for disk performance:
The fastest providers from our vServer test
| Provider ↕ | Single Core ↕ | Multi Core ↕ | Read Write ↕ | IOPS ↕ | Overall Score ↕ |
|---|---|---|---|---|---|
| Fasthosts | 1.385 | 2.300 | 997 MB/s | 15,58k IOPS | EXCELLENT |
| Infomaniak | 1.180 | 2.024 | 63 MB/s | 0,99k IOPS | VERY GOOD |
| IP-Projects GmbH & Co. KG | 743 | 1.450 | 2.085 MB/s | 32,59k IOPS | EXCELLENT |
| netcup | 1.134 | 2.069 | 3.545 MB/s | 55,40k IOPS | EXCELLENT |
| Hetzner Online GmbH | 745 | 1.283 | 2.077 MB/s | 32,46k IOPS | EXCELLENT |
| Contabo GmbH | 2.096 | 5.248 | 3.328 MB/s | 52,01k IOPS | EXCELLENT |
Optimisation: Get more performance out of the VPS
Optimising disk performance on a VPS is often only possible to a limited extent, especially with shared storage. Nevertheless, targeted measures can help noticeably improve performance or avoid bottlenecks.
Some approaches to optimisation are:
- Adjust file system options: The noatime option can prevent unnecessary write operations to the file system, since access times are no longer recorded.
- Use caching strategically: Database and web server software such as MySQL or NGINX offer extensive caching features that reduce disk accesses and improve response times.
- Minimise swap usage: Sufficient RAM prevents the system from falling back to slow swap storage.
- Optimise processes and cron jobs: Regularly scheduled cron jobs should be timed so they don't run during periods of high load.
- Choose higher-tier plans: VPS hosting plans with guaranteed SSD or NVMe resources often provide significantly better I/O performance than entry-level models.
- Review provider choice: Specialist hosts sometimes offer plans with IOPS guarantees or storage tuning – ideal for data-intensive applications.
For particularly high demands, we ultimately recommend switching to a Dedicated Server with guaranteed resources. Only then can disk performance be reliably and permanently maintained at a consistently high level.
Conclusion: Assessing VPS disk performance correctly
Disk performance is a crucial component for the overall performance of a VPS – especially for data-intensive applications. At the same time, storage is among the components whose actual speed can vary most in virtualised environments. Factors such as the storage technology used, the virtualisation layer and the load caused by other customers can significantly influence the measurable values.
Benchmark tools such as fio, ioping or iostat provide important comparative metrics and reveal typical strengths and weaknesses of the storage system. However, they always represent only snapshots and cannot guarantee that the measured values will be consistently achievable under real-world load.
For a realistic assessment of disk performance, results should therefore always be considered in conjunction with your own usage scenario. If you regularly rely on fast I/O access, you should look for hosting plans that offer dedicated resources, SSD or NVMe storage, and fair storage allocation. Additionally, workload-optimised settings at the system and application level help to get the most out of your VPS.
Only by deliberately measuring storage performance, interpreting it correctly and optimising it when necessary can bottlenecks be detected and avoided early.
Tip: Find out about the fastest VPS hosting in Germany and how they perform in terms of storage performance.
The fastest providers from our vServer test
| Provider ↕ | Single Core ↕ | Multi Core ↕ | Read Write ↕ | IOPS ↕ | Overall Score ↕ |
|---|---|---|---|---|---|
| Fasthosts | 1.385 | 2.300 | 997 MB/s | 15,58k IOPS | EXCELLENT |
| Infomaniak | 1.180 | 2.024 | 63 MB/s | 0,99k IOPS | VERY GOOD |
| IP-Projects GmbH & Co. KG | 743 | 1.450 | 2.085 MB/s | 32,59k IOPS | EXCELLENT |
| netcup | 1.134 | 2.069 | 3.545 MB/s | 55,40k IOPS | EXCELLENT |
| Hetzner Online GmbH | 745 | 1.283 | 2.077 MB/s | 32,46k IOPS | EXCELLENT |
| Contabo GmbH | 2.096 | 5.248 | 3.328 MB/s | 52,01k IOPS | EXCELLENT |
Our article is based on our own experience and research, as well as information from external sources.
Sources & further reading on the topic:
https://fio.readthedocs.io/en/latest/fio_doc.html (Benchmark software)
Write a comment
- Performance
Tags for this article
More web hosts
More interesting articles
Server speed - what matters and how can it be measured?
Why is a server's speed so important, how can server speed be measured and how can it be increased?
Fastest VPS hosting in Germany: Who delivers the best performance in 2026?
Which VPS hosting services deliver top performance? This article puts Germany's fastest VPS offerings to the test.
What is the cache and how does it affect server performance?
The cache helps with the speed of websites and stores them. We show you how the cache works exactly.
Pagespeed explained: How to measure and improve website load time
This article provides a comprehensive overview of the various methods for measuring and optimising website load time.
Virtual cores, real performance: measuring, comparing and optimising CPU performance on a VPS
This article explains how to precisely measure, compare and target improvements to the CPU performance of VPS instances.
Largest Contentful Paint explained: How to measure and improve LCP
This article provides a comprehensive overview of the importance of LCP and sets out practical ways to measure this metr...