VMblog recently had the opportunity to learn more about AuriStor during the 64th edition of the IT Press Tour in New York. During an engaging technical deep-dive, Jeffrey Altman, Founder and CEO, and Gerry Seidman, President & Member of the Board, demonstrated how their technology is transforming distributed storage for modern enterprises.
“AuriStor is a technology company first and a product company second,” explained Jeffrey Altman. “Our focus is on developing the tools our existing end users require as they evolve their computing infrastructures. Whether that be conversions to a containers-first world or deployment within new supercomputers.”
Evolution and Background
The company’s journey is particularly interesting. While many startups chase trendy markets, AuriStor took a different approach. Founded in October 2007 as Your File System, Inc., the company emerged from the high-energy physics computing community. What started as a project to enhance the OpenAFS file system evolved into a commercial venture developing next-generation distributed file system technology.
As Gerry Seidman noted during the presentation: “When hybrid and cloud were the words everyone scrambled to say they had, all we had to do was put it on our marketing material. The product is fundamentally designed to be distributed across different servers, different campuses – it just naturally works in a hybrid multi-cloud environment. There’s no special sauce to make it hybrid multi-cloud, it’s just what it is.”
What Makes AuriStor Different?
AuriStor File System (AuriStorFS) stands out by providing:
- Advanced Security: Implements combined identity authentication, multi-factor authorization, and AES-256 encryption
- High Performance: Supports up to 8.2 Gbit/second per listener thread with up to 16 threads per service
- Massive Scalability: Can manage up to 2^64 volume IDs per cell and objects up to 16 exabytes in size
- Cross-Platform Support: Works across Windows, Linux, macOS, and other major operating systems
- Seamless Integration: Maintains backward compatibility with IBM AFS and OpenAFS installations
Revolutionary Performance Improvements
One of the most fascinating technical revelations from the presentation was Simon Wilkinson’s work on the RX protocol. The team achieved a stunning 10x performance improvement in data transfer speeds. As Seidman explained: “We observed the following performance while copying a 1g file from local disk into AFS:
- AuriStor Client -> OpenAFS server: 3m.11s
- AuriStor Client -> AuriStor Server (2021): 1m
- AuriStor Client (2025) -> AuriStor Server (2025): 30s”
Real-World Impact
The practical implications of these improvements are significant. And the company has found success across several key verticals:
Financial Services
One global financial institution implemented AuriStorFS across approximately 80 cells and 300 servers, supporting 175,000+ clients and over 1.5 million volumes. The system handles continuous software deployment with updates every few minutes.
Research and Academia
SLAC Linear Accelerator Center utilizes AuriStorFS alongside IBM Spectrum Storage for research data management, providing worldwide access to compute resources and experimental data.
Government Agencies
The U.S. Geological Survey deploys AuriStorFS for regional distribution of critical natural hazard information, requiring high availability and real-time content replication.
Altman shared a compelling example: “Production AuriStorFS fileservers regularly service more than half a million client connections simultaneously. We have servers that are servicing anywhere from 15,000 to 30,000 clients per machine.”
Cloud Migration Success Stories
The U.S. Geological Survey case study demonstrates AuriStor’s sophisticated approach to cloud migration. “They no longer run their own data centers – everything’s now in a variety of federally contracted clouds,” Altman explained. “For them, we simply added servers in the cloud, for which we granted them a free license during the transition period. We didn’t charge more money or require them to enter new contracts. The clients simply discovered that the data was now in the cloud, as opposed to in the data center in Virginia or Colorado or California, and continued doing their daily job.”
What makes this particularly noteworthy is the system’s ability to handle real-time content distribution across multiple regions. The USGS deployment takes snapshots every 10 minutes, automatically replicating critical natural hazard information across multiple AWS regions. This infrastructure is intentionally distributed to ensure data availability even if entire regions become unavailable during natural disasters. Clients automatically connect to the nearest available region, with seamless failover if their primary region becomes inaccessible.
AuriStor’s cloud implementation also addresses a common challenge in cloud environments – packet loss and reordering. The system has been specifically enhanced to handle consistent packet loss rates of up to 0.4% in AWS and 0.9% in GCP, particularly when traversing long paths. These improvements, combined with path MTU discovery and proportional rate reduction, enable consistent performance even in challenging cloud network conditions.
Container Integration and Modern Workloads
The team shared fascinating insights about their work with containers, particularly relevant for VMblog readers. As Seidman pointed out: “When people have to transfer data from one organization to another, they typically upload to Globus where it can be downloaded. With AuriStor, if you want to move from cloud provider A to cloud provider B, you do have to copy the data, of course, but the clients do nothing – it just works.”
The container integration story goes much deeper. AuriStor has developed a Container Storage Interface (CSI) driver that integrates directly with Red Hat OpenShift, addressing a critical challenge in container deployments. Modern machine learning container images often exceed 40 gigabytes, yet typically only 5% of that data is actually accessed during execution. Rather than pulling entire container images to each node, AuriStor’s implementation allows containers to mount AFS volumes directly, fetching only the required data on demand.
This approach is particularly powerful for organizations transitioning to containers. One financial customer is using this capability to migrate 180,000 systems to containers while maintaining their existing software distribution workflow. Their software distribution trees often exceed 400 gigabytes, with updates occurring every few minutes. Traditional container image distribution would be impossible at this scale, but AuriStor’s incremental replication and on-demand data access make it feasible.
The integration extends to the Linux kernel itself – the built-in AFS client (kafs) now supports AuriStorFS protocols and is distributed in major Linux distributions including Fedora, Ubuntu, Debian, and RHEL 9.2+. Recent developments even enable booting Linux systems directly from AFS volumes using standard kernel tools, opening new possibilities for containerized workload management.
Technical Architecture Deep Dive
During the presentation, Altman revealed an interesting technical detail about their approach to file system architecture: “We are not really a cross-file system – we’re more an object system. Every volume has a unique ID and every object, every directory, every file has a unique ID within it. So if you think about volume as an object bucket, the protocol that goes under the wire is really an object protocol, which has a lot of benefits.”
The system excels in several key areas:
Enhanced Security Framework
- Implements YFS-RxGK security class using IETF standard components
- Provides federated mutual authentication
- Supports perfect forward secrecy
- Enables combined identity authentication (user@machine)
Performance Improvements
- Increased maximum window size to 8192 packets (11.2MB) compared to traditional 32 packets
- Implements advanced congestion control mechanisms
- Supports multiple listener threads for improved throughput
Scalability Features
- Handles up to 400 access control entries per ACL
- Supports directory inheritance
- Manages volumes up to 16 zettabytes
Future Outlook
Looking toward 2026, AuriStor is tackling ambitious challenges. “We’re enhancing RX RPC with a CUBIC replacement for New Reno congestion control,” Altman shared. “More than fifty percent of our research effort is spent on network optimization. If we can’t make the network fast, it doesn’t matter what anything else does.”
Other key areas of focus include:
- Implementing Unicode directory entries and extended attributes
- Deeper integration with OpenShift and container orchestration systems
- Enabling diskless system booting from AFS volumes
Pricing Innovation
In an industry obsessed with storage-based pricing, AuriStor takes a refreshingly different approach. As Seidman humorously noted during the presentation: “We give you the first 100 exabytes for free!” The company’s pricing model is based on servers and user identities rather than storage capacity, making costs more predictable for enterprises.
Conclusion
For data center professionals dealing with distributed storage challenges, AuriStor represents more than just another storage solution – it’s a fundamentally different approach to global data access. As Altman emphasized, “We are a technology licensing company. We do not host or manage storage. Every end user deployment is unique.”
The system’s ability to maintain backward compatibility while introducing advanced features makes it particularly valuable for organizations with existing AFS deployments looking to modernize their infrastructure. With its focus on performance, security, and scalability, AuriStor is well-positioned to address the evolving needs of modern data centers and cloud deployments.
##
