Introduction
Hammerspace is on a mission to accelerate AI Anywhere. Our data plane unifies your data estate into one global namespace and automates data movement to GPUs & AI models. With metadata-indexed discovery and a standards-based approach, Hammerspace eliminates vendor lock-in and accelerates AI workloads that need access to data across sites, clouds, and existing storage silos from any vendor.
Hammerspace has partnered with Supermicro to deliver Hammerspace Ready Nodes – Hammerspace Data Platform solutions based on Supermicro server hardware. Hammerspace Ready Nodes deliver the power of Hammerspace on pre-qualified hardware that is optimized for price-performance, with the convenience of a simplified purchasing model.
Ready Node Types and Options
Hammerspace can be configured to address a broad range of performance and capacity needs, including the extreme levels of performance required in AI and HPC workloads. Metadata and data servers are scaled independently using the node types shown below.
Customers may also include simple Linux Storage Servers (LSSs) in any Hammerspace configuration, using their choice of hardware. Any physical or virtual Linux server with attached storage exported via NFSv3 may be used as an LSS.
Hammerspace can also directly utilize existing storage systems, wherever they reside. File and object metadata are brought into the Hammerspace Global File System using a unique data-in-place metadata Assimilation process – no data migration required.
| Node Types | Metadata Nodes | Data Nodes | |
|---|---|---|---|
| Description | Anvil Metadata Nodes | Data Services (DSX) Nodes | EC-Groups Nodes |
| Function | Runs Hammerspace Anvil metadata services | Provide client-side access to data on any storage via standard NFS, SMB, & S3 protocols, with scale-out flexibility Provide automated cross-platform data services for data placement, data protection, file copy management, workflow automation and more, with linear scalability | Used to scale performance and capacity using erasure code–protected NVMe Flash and/or HDD-based storage |
Metadata Node Specifications
Hammerspace Anvil metadata nodes can be deployed using these pre-qualified Supermicro configurations, as software on customer-provided bare-metal servers, as VMs, or in the cloud.
| Metadata Node Specification | Model | |
|---|---|---|
| A405 | A610 | |
| Use Case | Single and Multi-Site Deployments | Massive/Global Multi-Site Deployments |
| Form Factor | 1U | |
| Qty. Required | 2x: Standard HA, 3x: 3-Zone HA | |
| Max File Count | Metadata Capacity-Based | |
| Metadata Drive Count | 2, 4, 6, 8, 10, 12 (Mirrored) | |
| Metadata Drive Size | 3.2TB, 6.4TB, 12.8TB | |
| Metadata Drive Type | Mixed-Use NVMe Flash | |
| Boot / OS Drive | 2x M.2 960GB (Mirrored) | 2x M.2 1.92TB (Mirrored) |
| CPU | 2x 24-Core Intel 6527P 3.0GHz | 2x 32-Core Intel 6737P 2.9GHz |
| RAM | 512GB | 1024GB |
| NIC (Management) | Dual-Port 10GbE RJ45 | |
| NIC (Metadata) | 1x Mellanox 100GbE 2x QSFP56 or 1x Mellanox 400GbE 1x OSFP VPI | |
| Warranty | 3 or 5 Year NBD | |
Data Services (DSX) Node Specifications
Hammerspace DSX nodes can be deployed using these pre-qualified Supermicro configurations, as software on customer-provided bare-metal servers, as VMs, or in the cloud. When DSX nodes are used solely for data orchestration, no internal data storage is required.
| DSX Node Specification | Model | |
|---|---|---|
| D100 | D200 | |
| Use Case | Single and Multi-Site Deployments | Massive/Global Multi-Site Deployments |
| Form Factor | 1U | 2U |
| Qty. Required | Minimum 3x D100 or D200 Recommended | |
| Data Capacity Raw | 0 – 737 TB | 0 – 1475 TB |
| Data Capacity Mirrored | 0 – 368 TB | 0 – 737 TB |
| Data Drive Count | 0-12 | 0-24 |
| Data Drive Size | 3.84 TB, 7.68 TB, 15.36 TB, 30.72 TB, 61.44 TB | |
| Data Drive Type | Read-Intensive NVMe Flash | |
| Boot / OS Drive | 2x M.2 1920GB (Mirrored) | |
| CPU | 2x 24-Core Intel 6527P 3.0GHz | |
| RAM | 1024GB | |
| NIC (Management) | Dual-Port 10GbE RJ45 | |
| NIC (Data) | 2x Mellanox 200GbE 2x QSFP112 or 2x Mellanox 400GbE 1x OSFP VPI | |
| Warranty | 3 or 5 Year NBD | |
EC-Group Node Specifications
Hammerspace EC-Group configurations may be flash or HDD-based and consist of controller nodes with optional internal NVMe storage and optional HDD-based JBODs.
EC-Group Controller Nodes
| Controller Nodes Specification | Model | |
|---|---|---|
| ECG1 | ECG2 | |
| Function | EC-Group Controller w/Optional NVMe Storage | |
| Form Factor | 1U | 2U |
| Qty. Required | 4 or More ECG1 or ECG2 | |
| EC Options | 4+2, 1 or 2 Failure Resilient, 8+2, 1 or 2 Failure Resilient | |
| Data Capacity Raw ea. | 16 TB – 676 TB | 16 TB – 1413 TB |
| Data Capacity EC ea. | 10 TB – 514 TB | 10 TB – 1074 TB |
| Data Drive Count | 2 – 11 | 2 – 23 |
| Data Drive Size | 7.68 TB, 15.36 TB, 30.72 TB, 61.44 TB | |
| Data Drive Type | Mixed-Use NVMe Flash | |
| Boot / OS Drive | 2x M.2 1920GB (Mirrored) | |
| CPU | 2x 32-Core Intel 6737P 2.9GHz | |
| RAM | 1024GB | |
| NIC (Management) | Dual-Port 10GbE RJ45 | |
| NIC (Data) | 2x Mellanox 200GbE 2x QSFP112 or 2x Mellanox 400GbE 1x OSFP VPI | |
| JBOD Connectivity | 1x SAS 12Gb/s PCIe 4.0 8-Port HBA | |
| Warranty | 3 or 5 Year NBD | |
EC-Group JBODs
| EC-Group JBOD Specification | Model | |
|---|---|---|
| ECG-JBOD60 | ECG-JBOD90* | |
| Function | Optional HDD-Based EC-Group Capacity | |
| Form Factor | 4U | |
| Qty. Required | 1:1 Relation to Qty. of ECG Controller Nodes | |
| EC Options | 4+2, 1 or 2 Failure Resilient, 8+2, 1 or 2 Failure Resilient | |
| Data Capacity Raw ea. | 240 TB – 1440 TB | 240 TB – 2160 TB |
| Data Capacity EC ea. | 160 TB – 1094 TB | 160 TB – 1642 TB |
| Data Drive Count | 20 – 60 | 20 – 90 |
| Data Drive Size | 12 TB, 16 TB, 20 TB, 24 TB | |
| Data Drive Type | SAS HDD | |
| Connectivity | 12Gb/s SAS | |
| Warranty | 3 or 5 year NBD | |
Example Ready Node Configurations
The configurations in this section are examples showing several possible ways to deploy Hammerspace Ready Nodes. Many other configurations and combinations are possible that incorporate existing legacy storage and/or net-new storage. This enables Hammerspace to meet performance and capacity requirements at any scale. Contact Hammerspace or a Hammerspace channel partner for a custom Ready Node configuration tailored to your requirements.
Example: Mirrored Configurations
These examples use DSX nodes with embedded NVMe flash to provide high-performance storage and add Hammerspace data management and data orchestration features to existing or new storage environments
Example: EC-Group Configuration
This example combines DSX nodes containing embedded NVMe flash with EC-Group HDD arrays. This enables adding both high-performance and high-efficiency storage to new or existing environments, along with Hammerspace data management and data orchestration capabilities.
Example: Customer-Supplied Storage Configurations
Hammerspace is software, and it works with any new or pre-existing storage from any vendor. These examples show how to bring Hammerspace data management and data orchestration features to existing storage environments, including deploying Hammerspace with Linux storage servers to create a high-performance parallel file system for AI and other performance-intensive use cases.
Using Hammerspace with existing storage has many benefits, including:
- Extending the useful life of the existing assets, keeping them out of the landfill longer
- Reducing the carrying cost of older storage by moving to hardware-only support
- Increasing the performance of existing storage using parallelism and intelligent routing
- Eliminating data migration disruption – migrate at your own pace, transparently
