Hammerspace data orchestration.
Hammerspace is the data orchestration layer of the Hammerspace Data Platform — a single global namespace and policy-driven data services layer that spans heterogeneous storage and presents it to applications as one system. In the LStore joint solution set, Hammerspace runs in production above LStore at ACCRE: it orchestrates placement across every tier and, on the fast NVMe and SSD tiers, moves the data itself, while LStore holds the data on its depots and manages it on tape — so that active data stays close to compute while durable capacity scales horizontally beneath it. Hammerspace is a data orchestration partner in the LStore ecosystem; this page describes the joint solution set, what each side contributes, and how the integration works in production.
In partnership with Hammerspace, Inc., Redwood City, California · hammerspace.com
Hammerspace publicly reported that ACCRE expects to reduce its average cost of storage by 48 percent through this architecture, while providing faster, more flexible data access to the Vanderbilt research community.
LStore is the capacity core. Hammerspace is how it becomes a platform.
LStore is a fault-tolerant distributed file system — the durable core that holds an institution's data on its depots and stages it to tape for long retention. What an operating institution needs around that core is a layer that unifies the namespace across all of its storage tiers, presents the result to applications as one coherent system, and moves data through the fast front tiers as access patterns change. That is the role Hammerspace fills in the joint solution set, and it is the reason this page exists separately from the LStore section.
LStore by itself is powerful. It is also open: the platform is developed in the open, available from public repositories, and an institution with the appetite to do so can download and run it on its own. What that institution then has is a distributed file system — a capacity-and-durability layer. What it does not yet have is a unified data plane across that layer and the other storage tiers an operating environment requires: scratch capacity for in-flight computation, fast persistent storage for working datasets, archive media for long retention. Each of those tiers is a separate system without something to coordinate them.
Hammerspace is that coordinating layer. It presents a unified namespace that spans all of the storage tiers — the LStore depots and tape among them — and orchestrates placement across all of them as access patterns change; on the fast NVMe and SSD tiers it moves the data itself, through its Anvil and DSX servers. Applications see one coherent system rather than several separate stores. The architecture stops being “a fast tier, a capacity tier, an archive tier” and starts being one system that happens to have multiple tiers underneath it, with policy — not manual data movement — determining which tier holds each file at any moment.
This is the gap that closes when LStore is deployed with Hammerspace rather than alone. The same files that LStore stores durably are reachable through a global namespace and standard protocols, and move between tiers continuously without application disruption — Hammerspace orchestrates the placement across every tier and performs the movement on the fast NVMe and SSD tiers, while LStore performs the movement on the tiers it owns, holding the data on its depots and managing it on tape. The institution's storage decisions become policy decisions rather than directory-structure decisions; the platform underneath is what makes that change in posture possible.
Hammerspace can be configured in many ways; this page describes the LStore joint solution set specifically, and other implementations may differ.
Four capabilities, one platform.
The Hammerspace Data Platform combines four capability pillars that together produce the “one coherent system” experience above the storage tiers underneath. Each is described here in Hammerspace's own framing; each appears in the production deployment at ACCRE.
Single Global Namespace
A global file system that spans sites, clouds, and storage — eliminating the barriers between data silos and giving users and applications a single, secure way to see and access data across incompatible storage types and multiple locations. For the LStore deployment, this is the layer that presents the entire tiered architecture to applications as one filesystem.
Standards-Based Parallel File System
A standards-based parallel file system — Hyperscale NAS — that combines HPC performance with enterprise NAS simplicity, built on Linux-standard pNFSv4.2 with Flex Files. No proprietary client software is required beyond standard Linux. Applications and compute clusters reach the data through protocols they already use.
Data Orchestration Services
Policy-driven services that unify siloed data and automate the flow of data between local and remote compute resources. In the ACCRE deployment, this is the layer that runs the named objectives that place files across the fast front tiers and the LStore capacity core based on age, size, and access patterns — described in Section 4.
Multi-Protocol Access
POSIX-compliant access through pNFS, NFS, SMB, and S3 — without proprietary clients or agents — with CSI for Kubernetes integration and API access for direct application integration. The data is reachable from the protocol the application expects, not the protocol the storage vendor prefers.
The architectural posture across all four capabilities is the same: software-defined, deployable on any commodity Linux server, integrating with storage from any vendor, with no proprietary clients required on application or compute nodes. This is the property that makes the integration with LStore tractable in the first place — Hammerspace consumes LStore as a Linux-server-attached NFS-exported path, the same way it consumes other storage in the architecture, with no platform-specific adapter required.
An open file system, and a complete data plane.
LStore is an open, production-proven file system: the source is available, the documentation is comprehensive, and it runs at scale at ACCRE today. How complete a data plane to compose around it is the institution's choice — and the ecosystem is built to make that easy, with Hammerspace and the storage specialists each contributing layers an institution can elect to add.
LStore standalone
An open, production-grade file system.
A capacity-and-durability layer with per-file placement policy and erasure-coded fault tolerance — open, mountable, scriptable, and operable, and proven in production at ACCRE. Institutions with strong in-house engineering run it directly and compose their own environment around it; the platform's openness is designed to make exactly that possible.
It is equally a foundation to build on — its openness deliberately leaves room for institutions to integrate it on their own terms, and for the ecosystem's specialists to add capabilities around it.
LStore + Hammerspace
A complete data plane on day one.
The same LStore capacity core, with Hammerspace's global namespace, policy-driven orchestration, multi-protocol access, and operational data services delivered as integrated layers above it. The architecture is complete from the first deployment; tiering, namespace, and access-protocol decisions are configuration rather than engineering. Operational in production at ACCRE.
An institution choosing LStore + Hammerspace is choosing a complete, integrated solution that has been proven at scale — with the open file system at its core preserved as the openness commitment the platform's authors made.
Both paths are first-class, and the choice is the institution's. LStore stands on its own as an open, capable platform; Hammerspace is an option an institution can elect to add an integrated data plane on top of it. And the ecosystem's storage specialists remain available either way — Spectra Logic tape libraries and storage appliances can be part of an LStore deployment whether or not Hammerspace is adopted. Institutions compose the combination that fits their workload.
Hammerspace orchestrates. LStore stores. The tiers compose.
The production architecture at ACCRE composes the Hammerspace Data Platform on top of an LStore-anchored tier stack. The diagram is a reference shape; the durable LStore tier is the architectural requirement, and additional tiers are deployed based on workload. The named data orchestration objectives below show how Hammerspace orchestrates placement across whichever tiers a deployment includes.
Policy-driven placement, by objective
Hammerspace expresses tier placement as named objectives at the policy layer. In the ACCRE production deployment, three of them carry the working pattern:
accre-create-on-scratch-vg — directs all newly created files to the Tier 0 scratch tier. The default write path. Files arrive on the fastest available medium first, regardless of their eventual destination.
accre-push-to-persistent-tiers — migrates inactive files off the scratch tier after ten minutes of idle time, routing by size. Files at or below 1 MB land on the Fast tier; files above 1 MB move to the LStore durable tier. The threshold reflects the storage economics: small files held on SSD cost a small premium, large files held on SSD cost considerably more.
accre-create-on-fast-vg — an alternative initial-placement objective for workloads where the scratch tier does not improve the write path. New data goes directly to the Fast tier, bypassing scratch.
The objectives are configuration at the orchestration layer rather than fixed behavior. A deployment with different workload characteristics — more small files, longer working-set residency, different scratch economics — lands at different thresholds, different idle windows, and possibly different tier compositions. The mechanism is the same; the parameters are tuned to the institution's actual data. The technical detail of how this orchestration interacts with LStore's own placement model is on the Operations & Lifecycle page.
Workloads at the intersection of scale, performance, and retention.
The LStore + Hammerspace joint solution fits institutions whose data is large enough to need a durable capacity core, demanding enough to need parallel-file-system performance on top of it, and long-lived enough to benefit from the policy-driven tiering that makes the economics work. The use cases below are the alignment points between Hammerspace's published industry segments and the institutions LStore is engineered for.
An operational integration, delivered as one solution.
The LStore + Hammerspace integration is operational, in production, and being introduced to evaluation partners now — with a broader production launch in coordination with the August 2026 release of the expanded tape-integration tier. Institutions evaluating the joint solution set engage through Unique Checksum as the lead integrator for the LStore commercialization, with Hammerspace as the data orchestration partner in the engagement.
For a prospective customer institution, the engagement path follows the broader LStore engagement model. An initial conversation covers workload profile, durability and retention requirements, performance characteristics, and any residency or compliance constraints. A reference architecture is sized to the institution's actual data; a pilot deployment is scoped to a bounded production-representative installation; the production deployment follows once the pilot has validated the architecture against the institution's workload. The Hammerspace layer is integrated alongside LStore as part of that delivered solution.
For institutions whose deployment includes the tape preservation tier, the integrated solution — LStore as the durable capacity core, Hammerspace as the orchestration layer, Spectra Logic as the preservation tier — is the production architecture at ACCRE today, and is the solution Unique Checksum delivers to institutions whose retention horizons make tape a working tier of the architecture. The dedicated Spectra Logic page covers the preservation tier and the tape library family.
For a conversation about an LStore + Hammerspace deployment, write to sales@uniquechecksum.com. Hammerspace's own product pages, datasheets, and technical materials are available at hammerspace.com.