Storage Benchmark Kit

Storage Benchmark Kit (SBK) is a Java framework for engineers who need repeatable throughput and latency measurements across many storage backends. It is strongest for teams building their own benchmark workflows, retaining metrics, or coordinating runs across hosts; it is less suitable as a quick, desktop-only disk speed checker.

Its breadth and configurable workload engine make it a capable open-source foundation, provided your environment can run JDK 25 and you protect the unsecured local console.

Overview

SBK applies one workload engine to object stores, message systems, databases, file systems, caches, and local queues. That common approach helps teams compare unlike systems with consistent controls instead of maintaining a separate utility for each backend. The coverage is not universal in a ready-to-build form: ChromaDB, HaloDB, and Ignite are in the source tree but disabled in the aggregate build.

The project is Apache License 2.0 software, with bug reports and feature requests handled through GitHub Issues and usage or design questions through GitHub Discussions. It is therefore a sensible choice for organizations comfortable operating and adapting open-source tooling rather than expecting a commercial support desk.

Key features

The harness exposes controls for concurrency, duration, rate, payloads, timestamps, and reporting. Those knobs let an engineer model sustained, rate-limited, or payload-sensitive workloads and make runs reproducible. They also add setup responsibility: meaningful results depend on choosing parameters that represent the system being evaluated.

Output can go to the console or CSV, or through a local web console, Prometheus, and gRPC. PrometheusLogger makes benchmark measurements consumable by Prometheus, while SBK Dashboard supports Grafana dashboards and persistent time-series history. The combination suits teams that need to compare runs over time, not just read a one-off terminal result.

For distributed work, SBK-GEM copies and starts SBK on remote hosts over SSH and aggregates measurements with an embedded SBM instance. SBK-YAL loads single-node arguments from YML; SBK-GEM-YAL does the same for distributed arguments. These components provide a path from a local experiment to coordinated multi-host measurement without abandoning the same workload model.

SBK requires JDK 25. Its wrapper can fetch a pinned OpenJDK 25 release and verify its SHA-256 checksum, reducing installation friction and version drift. Automatic installation covers Linux x64/AArch64, macOS x64/AArch64, and Windows x64. The Local Web Console uses plain HTTP with neither authentication nor TLS, so it belongs only on a trusted benchmark network or behind an SSH tunnel.

Pricing

Apache License 2.0 — free. SBK is open-source software with no paid plans listed. That makes it appropriate for broad internal use without seat fees or usage quotas, but your team takes on deployment, configuration, and operational support itself. There is no commercial tier to provide a supported alternative to the JDK and network-security requirements.

Platforms

SBK provides an API and command-line interface for self-hosted use on Linux, macOS, and Windows. Automatic JDK setup supports Linux x64 and AArch64, macOS x64 and AArch64, and Windows x64. This is a practical cross-platform base for engineering teams, while the self-hosted model means you must manage hosts, SSH access, and benchmark-network isolation.

Who it's for

Choose SBK if you are comparing storage technologies, developing storage drivers, preserving benchmark history, or running coordinated measurements across remote machines. Its workload and reporting controls fit performance engineers and infrastructure teams that need repeatability and integration with existing Prometheus/Grafana monitoring.

Look elsewhere if you want a minimal graphical disk test, a turnkey hosted service, or a console secured by default. SBK's JDK 25 runtime, source-level breadth, and plain-HTTP web interface require more engineering discipline than a small desktop utility.

Pros and cons

Pros

• One harness for varied backends: object stores, messaging, databases, file systems, caches, and queues can be assessed with shared workload concepts.
• Detailed workload control: concurrency, rates, durations, payloads, timestamps, and latency/throughput reporting support tailored scenarios.
• Strong reporting path: CSV, Prometheus, gRPC, Grafana dashboards, and persistent time-series history support both immediate analysis and trend tracking.
• Distributed option: SBK-GEM uses SSH to launch remote workers and aggregate measurements.
• No license charge: Apache 2.0 licensing removes paid-plan and seat constraints.

Cons

• JDK 25 dependency: environments standardized on older Java versions must accommodate a newer runtime.
• Console security gap: plain HTTP without authentication or TLS requires a trusted network or SSH tunnel.
• Uneven backend availability: ChromaDB, HaloDB, and Ignite are not enabled in the aggregate build.
• Self-service operation: support flows through GitHub rather than a paid support plan.

Alternatives

For a focused disk benchmark, fio is a free Linux, macOS, and Windows option under GNU GPLv2. Choose it when you want a dedicated disk-testing tool rather than SBK's multi-backend framework and distributed harness.

AJA System Test is a free macOS and Windows utility; its Lite edition is suited to straightforward standalone testing, while the full version adds PCIe bandwidth testing.

simple-disk-benchmark is a free, MIT-licensed command-line tool for Linux, macOS, and Windows, a better fit when a small CLI is all you need.

Parkdale offers a free download for Android, macOS, and Windows, making it preferable when mobile coverage or a simpler desktop workflow matters.

AmorphousDiskMark is a free Mac-only choice; AS SSD Benchmark is free Windows freeware. DaVinci Resolve and Buddi serve different software categories rather than storage benchmarking.

For a broader starting point, browse the Disk Benchmark Software list.

Verdict

Storage Benchmark Kit is a strong choice for engineers who need configurable, scriptable storage measurements with Prometheus/Grafana retention and an upgrade path to distributed runs. Choose it for that breadth and control. Look elsewhere for a quick desktop disk check, or if requiring JDK 25 and securing a plain-HTTP console would outweigh the value of its open, extensible framework.