Data-intensive workflows break under pressure when traditional storage can’t keep up. Whether you are running AI training, HPC simulations, or high-resolution media projects, performance bottlenecks, file system fragmentation, and metadata overhead can bring progress to a halt.
Pixstor is a software-defined, high-performance scale-out filesystem with NAS capabilities. It delivers consistent throughput, low latency, and seamless multi-protocol access, supporting mixed workloads, massive file counts, and demanding use cases without compromise. With intelligent data tiering and a unified namespace, Pixstor simplifies management while ensuring your data is always fast, accessible, and secure.
- Supports any storage technology: NVMe, SSD, SAS, NL-SAS, etc.
- Offers multi-protocol support including SMB, NFS, SFTP, POSIX, S3, NVMe-oF
- Scales from terabytes to multi-petabyte environments with billions of files
- Delivers up to 180 GB/s performance in a 4U form factor
- Engineered for full-speed performance even at 99% capacity utilization
- Enables fast data access, file search, and lifecycle control
- Deploys on-premises or in the cloud, with support for hybrid workflows
Use Cases
HPC workloads requiring fast, parallel processing and access to large datasets at extreme I/O rates
AI/ML training environments requiring fast, low-latency access to large datasets from GPU clusters
Life sciences and genomics workloads involving parallel sequencing, scanning, and analysis
High-throughput video editing and VFX rendering in media and entertainment pipelines
Secure, multi-tenant file system for creative teams or research groups with controlled access and quotas
Organizations replacing legacy NAS or scale-out file systems that can’t maintain performance at scale
Features
Pixstor delivers a high-performance, POSIX-compliant, parallel file system designed to handle billions of files with precision and control. Its unified global namespace and distributed metadata architecture eliminate bottlenecks while supporting large-scale, mixed-protocol environments.
- Unified namespace across all disk types, tiers, and locations
- File-level policies for quotas, ACLs, snapshots, and metadata
- Simultaneous multi-protocol access: NFS, SMB, SFTP, POSIX, S3
- Supports NVMe-oF and GPU Direct for ultra-fast, low-latency data paths from storage to compute and GPU nodes
- Namespace isolation through logical project-level segmentation (e.g., Filesets)
- Local caching for accelerated access to hot or reused data
- Native support for billions of files and directories
- Optimized for large block sizes up to 8 MB without compromising small file performance
Pixstor automates intelligent data placement with fine-grained tiering policies that move data between performance and capacity layers without disrupting workflows or access permissions.
- Policy-based tiering driven by file age, size, access time, location, owner, or tags
- Transparent tiering across NVMe, SSD, HDD, which can be extended to object storage, public cloud, tape and other storage systems with DataCore Ngenea
- Seamless movement of data with preserved access paths and permissions
- Files are transparently recalled from lower tiers when accessed
Snapshots and clones in Pixstor are lightweight and built for scale, enabling rapid rollback, test environments, and full recovery scenarios across sites.
- Copy-on-write snapshots with minimal performance overhead
- Writable snapshot clones for staging, testing, or temp workflows
- Windows Previous Versions integration for user-accessible restore points
- Versioned snapshot retention on backup targets
- Mountable backup sets for immediate DR access
- Site-to-site replication and DR sync using snapshot-diff technology
Pixstor’s search engine goes beyond filenames to enable deep discovery, tagging, and automation using metadata and content analysis across all tiers – even offline or archived datasets.
- Built-in metadata and content indexing with Pixstor Search
- Indexing of file attributes (name, size, owner), content (EXIF, text), and AI/ML tags
- Custom metadata tagging via extended attributes (xattrs)
- Search and preview across live, archived, and remote data
- Exportable results for editorial or pre-staging workflows
- Actionable search cart: search → select → act (copy, move, archive)
- Accessible via web UI or programmable API
Pixstor optimizes storage usage with inline compression and intelligent deduplication logic that identifies redundant content and reduces footprint without impacting performance.
- Inline compression at the time of write
- Transparent decompression on read
- Policy-driven identification of duplicate files based on search results
Security is built into Pixstor at every layer, from role-based access to encryption and tenant-level segmentation ensuring your data is always protected and compliant.
- Role-based access control (RBAC) for granular user management
- Secure NAS containers with full data, metadata, and access isolation
- Authentication via Active Directory, LDAP, OKTA, and OAuth2
- Immutability and data retention policies using snapshots and ACL enforcement
- Snapshot-based rollback for recovery and audit scenarios
- TLS, SSH encryption in transit and optional encryption at rest
- Per-tenant isolation for access, network paths, and policy scopes
Pixstor includes built-in observability and automation, giving you full visibility and control from dashboard to CLI to programmatic workflows.
- Real-time metrics for bandwidth, latency, CPU, and queue depth
- Per-client and protocol-level I/O analytics
- Integrated Logstash/Elasticsearch for events and alerts
- Centralized logging and audit trails for events and file actions
- Email notifications for system events and thresholds
- File-based automation hooks (e.g., file drop triggers job)
- SaltStack for system configuration and version control
- Lifecycle automation via web UI, REST API, and Python SDK
- Compatible with workload schedulers like Slurm via POSIX/NFS access
Pixstor supports cloud-based workflows when combined with Ngenea, enabling transparent tiering, search, and policy-based data management across on-premises and cloud storage.
- S3-compatible object access, including native integration with DataCore Swarm
- Transparent tiering to and from AWS, Azure, GCP, and other object stores
- Hybrid workflows with intelligent caching and push/pull logic
- Search and preview across cloud and on-prem tiers
- Policy hooks to enforce cost-aware tier usage and retention
File Usage Analytics
Observability & Monitoring
Policy-based Data Management with Ngenea
Benefits
Accelerate Time-to-Results
- Slash job start delays with instant access to datasets
- Pre-stage data intelligently for compute workflows
- Eliminate storage-related bottlenecks in critical pipelines
Work Smarter with Automation
- Set and forget lifecycle policies based on real usage
- Automatically move data between fast and deep storage
- Reduce manual file management and guesswork
Find What You Need, Fast
- No more hunting across systems: unified view of all files
- Search by project, user, or tags even if data is archived
- Actionable search: copy, move, archive in one step
Lower Costs Without Compromise
- Keep expensive storage focused on active work
- Maximize ROI by using commodity hardware
- Predictable costs: pay only for what you deploy, scale on your terms, no hidden fees
Protect Data Without Slowing Down
- Snapshots and DR built-in, not bolted on
- Keep users productive while ensuring compliance
- Restore files quickly without calling IT
Plug Into Any Workflow
- Works with your existing apps, workflows, and infrastructure
- Integrates with job schedulers and pipelines
- No vendor lock-in: data stays in your control
Architecture
DataCore Pixstor and Ngenea Reference Architecture for HPC/AI Environments
Deployment
Pixstor and Ngenea can be deployed on-premises, in the cloud, or as a hybrid solution, giving you the flexibility to support any workflow or infrastructure strategy.
Licensing
Pixstor uses a per-drive licensing model, based on the number of NVMe or SAS drives deployed. There are no per-terabyte charges and no tier-based surcharges. This approach supports flexible multi-tier performance and capacity configurations while keeping costs predictable as the system scales.