More Time Computing. Less Time Waiting for Data.
Simulations, research computing, and engineering analysis deserve storage matched to their most demanding I/O stages. SiliconDisk provides DRAM-based shared block storage for active, latency-sensitive HPC working sets.
Published appliance-level specifications; actual results depend on workload and configuration.
Not every HPC bottleneck is a compute problem.
Many HPC environments are designed around fast processors and parallel execution, but some workflows repeatedly access small files, temporary data, or shared working sets. Storage delay in these stages can affect overall efficiency.
Add a specialized performance tier.
SiliconDisk uses DRAM-based block storage and high-speed Ethernet connectivity to accelerate selected working operations. It can complement parallel filesystems and capacity storage when the software stack supports the required integration.
Where Faster Storage Can Make a Difference
Serve the active data that benefits from the shortest storage wait.
Accelerate active working sets
Provide a responsive storage tier for simulation and analysis data that is repeatedly accessed.
Support intensive temporary I/O
Target repetitive small-block reads and writes, scratch operations, and intermediate processing.
Enable flexible host access
Use NVMe-oF over compatible Ethernet fabrics to build specialized multi-host storage designs.
Analyze the storage path
Review device and network activity with Insight Analytics™ during performance tuning.
Use Cases Worth Evaluating
- Scientific simulations
- Engineering modeling
- High-speed scratch areas
- Repetitive numerical analysis
- Computational research
- Interactive data visualization
Performance Is More Than a Specification.
SiliconDisk combines memory-based storage with ATTO hardware acceleration, high-speed connectivity, and the engineering insight to evaluate real workloads.
xCORE™ acceleration
Hardware-based I/O processing designed to move data efficiently.
NVMe-oF / RoCE v2
High-speed shared block-storage access over compatible Ethernet networks.
Insight Analytics™
Detailed visibility into storage and network performance.
What Could Your Compute Cluster Do With Less I/O Waiting?
Review the I/O characteristics of your HPC workflow with ATTO engineers.
