Hardware requirements
elDoc supports deployment on both physical (bare-metal) servers and virtualized infrastructure environments.
The platform is compatible with all major enterprise hypervisors, including Proxmox VE, VMware vSphere, and Microsoft Hyper-V, as well as leading cloud providers such as Amazon Web Services and Microsoft Azure.
elDoc can be deployed in:
- On-premises environments (including fully isolated air-gapped infrastructures)
- Private cloud environments
- Public cloud environments
System requirements depend on several factors, including:
- Enabled platform modules
- Number of concurrent users
- OCR and document processing workloads
- AI/LLM usage intensity
- High-availability and clustering requirements
AI-powered Intelligent Document Processing (AI IDP), OCR, and local LLM inference workloads require significantly higher CPU, RAM, and GPU resources. In contrast, deployments focused primarily on eSignature, File Management or BPM functionality can operate efficiently on more lightweight infrastructure configurations.
The elDoc platform architecture is organized into three primary infrastructure layers:
AI / LLM Servers
Responsible for AI inference, OCR, embeddings, vector processing, and AI agents. Required for deployments utilizing self-hosted AI / LLM infrastructure.Application Servers
Responsible for business logic, REST APIs, user interfaces, workflow execution, and integration services.Database & Search Servers
Responsible for document/file storage, metadata persistence, full-text indexing, and vector databases.
Infrastructure sizing should be calculated as the combined total of all selected layers.
For example, a fully self-hosted on-premises AI deployment with air-gapped operation typically requires all three infrastructure layers:
- AI / LLM Servers
- Application Servers
- Database & Search Servers
Smaller deployments using cloud-hosted AI services may require only the Application and Database layers.
Example deployment scenario cards
| Scenario | Components |
|---|---|
| File Management only | App + DB |
| AI File Management + AI Chat + Agentic RAG with Cloud LLM | App + DB |
| eSignature only | App + DB |
| BPM only | App + DB |
| Fully Air-Gapped deployment with AI Chat + Agentic RAG or AI IDP | AI + App + DB |
1. AI / LLM infrastructure layer
- CPU: Intel/AMD, 3+ GHz, 8-16 Cores
- RAM: 64+ GB DDR5
- Storage: 2TB NVMe
- GPU:
- 1-2 x NVIDIA RTX 5090 (32GB VRAM) or
- 1-2 x NVIDIA RTX A6000 (48GB VRAM)
- CPU: Intel/AMD, 3+ GHz, 8-16 Cores
- RAM: 128+ GB DDR5
- Storage: 4TB NVMe
- GPU: 96+ GB Total VRAM
- 3+ x NVIDIA RTX 5090 (32GB VRAM each) or
- 2+ x NVIDIA RTX A6000 (48GB VRAM each)
Note: Total GPU VRAM capacity is the primary factor for AI / LLM sizing and may be distributed across multiple GPUs or servers.
- CPU: Intel/AMD, 3+ GHz, 8-16 Cores
- RAM: 512+ GB DDR5
- Storage: 4TB NVMe
- GPU: 512+ GB Total VRAM
- NVIDIA RTX A6000 or
- NVIDIA H100
Note: Total GPU VRAM capacity is the primary factor for AI / LLM sizing and may be distributed across multiple GPUs or servers.
No dedicated AI / LLM servers are required when using cloud-based AI / LLM services.
2. Application infrastructure layer
- CPU: Intel/AMD, 2.5+ GHz, 6-12 Cores
- RAM: 16-32 GB DDR4/DDR5
- Storage:
- 128GB SSD/NVMe (system)
- 256GB SSD/NVMe (data)
- RAID recommended
- CPU: Intel/AMD, 2.5+ GHz, 8-16 Cores
- RAM: 16-48 GB DDR4/DDR5
- Storage:
- 128GB SSD/NVMe (system)
- 256GB SSD/NVMe (data)
- RAID recommended
Note: High-availability (HA) deployments require at least two application servers.
- CPU: Intel/AMD, 2.5+ GHz, 12-24 Cores
- RAM: 32-64 GB DDR4/DDR5
- Storage:
- 128GB SSD/NVMe (system)
- 256GB SSD/NVMe (data)
- RAID recommended
Note: High-availability (HA) deployments require at least two application servers.
3. Database & Search infrastructure layer
- CPU: Intel/AMD, 2.5+ GHz, 4-8 Cores
- RAM: 8-32 GB DDR4/DDR5
- Storage:
- 128GB SSD/NVMe (system)
- 1+ TB SSD/NVMe (data)
- RAID recommended
Note:
- High-availability database deployments require a minimum of two servers with equivalent capacity (MongoDB PSA deployment).
- A three-server deployment with equivalent capacity is recommended (MongoDB PSS deployment).
- CPU: Intel/AMD, 2.5+ GHz, 6-12 Cores
- RAM: 16-48 GB DDR4/DDR5
- Storage:
- 128GB SSD/NVMe (system)
- 2+ TB SSD/NVMe (data)
- RAID recommended
Note:
- High-availability database deployments require a minimum of two servers with equivalent capacity (MongoDB PSA deployment).
- A three-server deployment with equivalent capacity is recommended (MongoDB PSS deployment).
DB Deployment Option I (3 Servers)
Minimal production-ready high-availability deployment
3 DB Servers:
- CPU: Intel/AMD, 2.5+ GHz, 6-12 Cores
- RAM: 32-64 GB DDR4/DDR5
- Storage:
- 128GB SSD/NVMe (system)
- 5+ TB SSD/NVMe (data)
- RAID recommended
DB Deployment Option II (11-12 Servers)
Recommended production-ready high-availability deployment with horizontal scaling
3 DB Config Servers:
- CPU: Intel/AMD, 2.5+ GHz, 4-6 Cores
- RAM: 8-16 GB DDR4/DDR5
- Storage:
- 256GB SSD/NVMe (system + data)
- RAID recommended
6 DB Data Servers:
- CPU: Intel/AMD, 2.5+ GHz, 6-16 Cores
- RAM: 8-32 GB DDR4/DDR5
- Storage:
- 256GB SSD/NVMe (system)
- 3+ TB SSD/NVMe/HDD (data)
- RAID recommended
2+ DB Router Servers
- CPU: Intel/AMD, 2.5+ GHz, 4-6 Cores
- RAM: 8-24 GB DDR4/DDR5
- Storage:
- 128GB SSD/NVMe (system)
- RAID recommended
|
|
Infrastructure Planning Note
All hardware requirements provided on this page are indicative and may vary depending on project-specific requirements, enabled functionality, workload characteristics, integration scope, data volume, and deployment architecture. Final infrastructure sizing and configuration should be validated and adjusted during the solution design and implementation phases.
|
|
Storage Performance Recommendations
To ensure optimal performance of the elDoc AI platform, including OCR, computer vision, image normalization, document classification, indexing, and AI processing workloads, the underlying storage subsystem should meet the minimum performance characteristics outlined below.
High-performance SSD or NVMe storage is strongly recommended for production deployments.
Disk performance validation can be performed using the DiskSpd benchmarking utility provided by Microsoft, available for both Linux and Windows environments.
Sequential Read/Write Performance (1 MB Block Size)
| Metric | Target Performance |
|---|---|
| Sequential Read Speed | ≥ 350 MB/s |
| Sequential Write Speed | ≥ 150 MB/s |
Recommended benchmark command:
diskspd -t1 -o1 -b1M -w25 -d120 -Sh -D -L -h -c5G iotest.dat > diskspd-seq-1m.txt
Random Read/Write Performance (8 KB Block Size)
| Metric | Target Performance |
|---|---|
| Random Read Speed | ≥ 150 MB/s |
| Random Write Speed | ≥ 80 MB/s |
Recommended benchmark command:
diskspd -t4 -o32 -b8K -r -w25 -d120 -Sh -D -L -h -c5G iotest.dat > diskspd-rnd-8K.txt
Additional Recommendations
- SSD or NVMe storage is strongly recommended for all production environments.
- HDD-based storage is not recommended for AI, OCR, vector indexing, or high-throughput document processing workloads.
- Dedicated storage volumes for databases, indexes, and AI model caches are recommended for medium and large-scale deployments.
Last modified: May 18, 2026