BeeGFS and Parallel File Systems in High‑Performance Computing: Evolution, Market Trends, and Technical Overview

Traditional file systems have rarely been the spotlight of the IT industry, but in the high‑performance computing (HPC) sector the landscape is shifting as new workloads and storage demands emerge.

HPC vendors have long relied on GPFS and Lustre for parallel file solutions. Recent changes—such as the rise of large‑scale analytics, machine learning, the expansion of HPC into mainstream enterprise applications, and the growth of cloud storage—have introduced new challenges for file servers, making them more complex and harder to manage.

Intel’s decision in April 2017 to discontinue the commercial version of Lustre sparked concerns about Lustre’s future and opened opportunities for alternatives. Around the same time, the open‑source BeeGFS (originally FhGFS) began gaining traction. Initiated in 2005 by the Fraunhofer High‑Performance Computing Center in Germany, its first test release appeared in 2007, with a commercial version launched in 2009. In 2014, Fraunhofer spun off ThinkParQ to promote BeeGFS worldwide.

ThinkParQ offers BeeGFS as free, open‑source software together with support, consulting, and system‑integrator partnerships. Development of the parallel file system remains largely within Fraunhofer.

Since 2017, ThinkParQ and BeeGFS developers have expanded collaborations with cluster‑management vendors (Bright Computing), HPC solution providers (Penguin Computing), hardware manufacturers (Ace Computers, QuantaCloud), and have extended their reach to regions such as Russia and Japan. At SC17, BeeGFS v7.0 was announced, featuring a new storage‑pool design, SSD/HDD hybrid support, and enhanced data placement policies.

In Japan, Fujitsu announced the AI‑bridged cloud infrastructure (ABCI) supercomputer, which will use BeeGFS onDemand (BeeOND) to achieve accelerated performance comparable to HPE’s Tsubame 3.0, delivering up to 1 TB/s with a 1 PB NVMe buffer.

According to ThinkParQ’s CEO Sven Bruener, interest in BeeGFS has grown because it addresses market concerns about Lustre’s future and offers a competitive, performance‑focused solution. Users often switch from Lustre or GPFS due to maintenance complexity, and they appreciate BeeGFS’s ease of deployment and superior performance.

Scalability is a key differentiator: BeeGFS can start on two servers and scale instantly by adding components, comparable to PanFS and even recognized by IBM. Its architecture allows building small systems with few components and expanding to exabyte scale without technical limits.

While BeeGFS users are primarily in Europe, ThinkParQ is rapidly expanding in Russia, the United States, and Japan, with notable deployments at national labs such as Oak Ridge and institutions handling multi‑petabyte workloads in bioinformatics and academia.

Historically, GPFS has existed for over 25 years, focusing on data management, whereas Lustre originated as a research project 17 years ago. BeeGFS benefits from lessons learned in those systems, offering a user‑space solution that runs on top of the local Linux file system, simplifying deployment and management.

BeeGFS also supports a Burst Buffer technology called BeeGFS on Demand (BeeOND), which mitigates I/O spikes by using flash storage to maintain consistent performance.

Bright Computing has collaborated with BeeGFS for several years, simplifying deployment on Bright clusters and demonstrating that BeeGFS is lighter and easier to install than GPFS or Lustre.