StoreHub

StoreHub assessed whether the United States needs a dedicated facility for data storage research and what such a facility should provide. Two survey instruments, a community workshop, and a facility-by-facility analysis identified demand for systems that researchers can reconfigure, stress, and deliberately fault. The public report recommends a staged federation built on existing infrastructure.

Why a Storage Facility​

Storage systems research often requires administrator access, device reconfiguration, replacement file-system software, and controlled fault injection. Production supercomputers and commercial clouds cannot permit many of these experiments, while existing reconfigurable testbeds were not designed around storage. StoreHub measured the resulting infrastructure gap and the resources needed to close it.

What the Project Did​

Two survey instruments. The StoreHub Community Insight Survey (N=76) combined a longitudinal questionnaire (N=35), fielded online and at HUG'24, SC'24, and SSDBM'25, with a live poll (N=41) at the first StoreHub Community Workshop. According to the public report, respondents spanned fifteen universities and four national laboratories.

The first StoreHub Community Workshop. The hybrid workshop ran June 23 to 25, 2025 at The Ohio State University with SSDBM'25. It combined research talks, breakout discussions, an SPDK, DPDK, and CXL tutorial, an agent demonstration, and a student session with nineteen accepted posters. Five travel awards supported external student presenters.

Infrastructure and landscape analysis. The project compared storage capabilities across Chameleon, CloudLab, FABRIC, NSF and DOE production systems, the Open Storage Network, and Grid'5000. It also reviewed the research record across FAST, HotStorage, PDSW, MSST, SSDBM, and SOSP.

Findings​

The public report summarizes the combined survey (N=76):

• 62% of respondents require administrator privileges that production and cloud systems deny.
• 56% report research they cannot conduct for lack of specialized resources.
• 98% say early access to prototype storage hardware would be valuable.
• 81% use AI or machine learning in their research, and 47% count AI applications such as training data and checkpointing among their primary research focus areas.

The final report found no U.S. facility dedicated to storage-systems research. It identifies a gap between small reconfigurable testbeds and production systems: storage-focused infrastructure that researchers can modify and disrupt at hundreds-of-nodes and tens-of-petabytes scale, with modern media, fault injection, and administrator access.

Recommendation​

The principal output is the 37-page public assessment report, The State and Future Direction of the Data Storage Research Community: Evidence from 2024-2026 and the Case for National Research Infrastructure. It contains 153 citations and states what findings would weaken or reverse its recommendation.

Read the web edition of the StoreHub Landscape Report or download the PDF.

The report recommends a federated national storage research facility at roughly $20 million to $25 million over five years. It proposes building over existing infrastructure and proceeding in stages through NSF Mid-scale Research Infrastructure.

Educational Impact​

According to the final report, project findings informed CS 550, CS 570, and CS 546 at Illinois Tech. The award also developed a draft CS 590 syllabus. Its material is being delivered within CS 595 in Fall 2026.

Illinois Tech led StoreHub with The Ohio State University.