Data Centers

Data center infrastructure is no longer a separate conversation from the wired edge. The same operators running broadband access are running edge compute. The same regulators demanding CRA-aligned security at the gateway are asking the same questions about the storage fabric behind it. And AI-era workloads have collapsed the old distinction between “networking” and “storage” into a single problem of moving verifiably correct data with verifiably low latency.

The Foundation’s Data Centers track is organised around that reality.

What we’re working on

Storage acceleration at line rate

Modern storage architectures — NAS, SAN, hyperconverged, object — all face the same cost pressure: CPU cycles spent on compression, deduplication, encryption, and integrity checking are cycles not spent on workloads. The answer across the industry has converged on hardware-accelerated storage data paths, where transformation operations run off-host, in parallel, with real-time integrity verification and hardware-rooted cryptography.

The Foundation’s work here is not to build another accelerator. It is to integrate accelerator-class hardware cleanly into open storage stacks, publish reference integrations that operators can actually deploy, and ensure the resulting architectures meet the same CRA- and NIST-aligned security posture we ask of everything else.

Deterministic connectivity, core to edge

Data centers increasingly span a spectrum — a hyperscale core, a metro aggregation tier, and a proliferation of edge sites that each look like a small regional data center in their own right. Keeping connectivity predictable across that spectrum is the hard part.

Foundation reference architectures in this area cover the full span: deterministic Ethernet switching inside the rack, predictable aggregation between racks, and the wired-edge interconnect that ties branch, metro, and edge sites back to core storage and compute. The emphasis is on architectures an operator can actually commission, audit, and operate — not idealised whitepaper topologies.

Security as a first-order property

Storage and connectivity inside the data center inherit every obligation we already take seriously at the wired edge: end-to-end authentication, hardware-rooted secure boot, verifiable data integrity, and cryptographic protection of data at rest and in motion. What changes is the scale and the blast radius.

Our reference architectures carry an explicit data-center security annex, aligned with the same CRA-aware and NIST-aligned posture the Foundation applies to wired-edge products. That alignment is the point: a single security posture that travels from the gateway to the substation to the storage node, rather than a patchwork of per-segment baselines.

AI and HPC-adjacent workloads

AI/ML training and inference, HPC, and hyperconverged workloads stress storage and connectivity in patterns that did not dominate even five years ago. High concurrency, low-latency data transformation, peer-to-peer DMA between accelerators and storage without round-tripping through host memory, and sustained multi-terabit throughput are no longer exotic — they are the baseline for any 2026 design. Foundation work in this track keeps that baseline in view.

Operator-grade, not laboratory-grade

Two phrases we use carefully across this site: operator-grade and deployable.

Operator-grade means the work survives contact with a real operations team using real change-management processes. It can be rolled back. It can be audited. It can be handed to second-line support without a rebuild. It is documented in a form an auditor will accept.

Deployable means the reference architectures come with parts lists of known-good silicon, working conformance suites, a CRA-aligned security annex, and tooling that runs on a laptop. Not whitepaper topologies, not a booth demo: a design you can actually commission.

Every piece of data-center work the Foundation ships is measured against both.

Why a neutral foundation is the right home for this

Much of the strongest work in data-center acceleration and connectivity happens inside silicon vendors. That work is excellent, and the Foundation’s position is not to duplicate it. Our position is that operators, integrators, and regulators need a neutral venue where that work is integrated into open reference architectures, validated with public conformance evidence, and paired with a security posture that is consistent across the full wired infrastructure — from broadband gateway to data center core.

That role only works if the venue is neutral. The Foundation is the venue.

Members driving this track

Several member categories collaborate on the Data Centers track:

• Silicon vendors contributing storage and connectivity acceleration
• Data center operators — hyperscale, enterprise, edge, and colocation
• Network operators with edge-compute build-outs
• OEMs and system integrators shipping storage and server platforms
• Software partners across management, analytics, and AIOps

If your organisation ships into data centers and you want this track to move faster, join a working group or bring a concrete deployment problem to the contact page.