Hardware Requirements for Hybrid CCaaS: Customer Hosted Data and AI Services
This guide outlines the hardware requirements for the Hybrid CCaaS Model, where Expertflow hosts the cloud applications, but the customer retains control by hosting critical data components (databases, ETL, file store, and AI services) on-premise or in a regional cloud.
Deployment Prerequisites (Platform & OS)
The components hosted by the customer must still meet the following environment requirements:
Component | Requirement | Notes |
Operating System (OS) | Linux Distribution (e.g., Debian, CentOS) | Required for hosting data components and AI services. |
Container Platform | Kubernetes (RKE2, K3s, or other) | Recommended for hosting scalable data components and ETL services. |
Networking | Secure, Low-Latency Connection | Mandatory for the cloud-hosted applications (CX-Core, Agents) to reach the regional databases. |
2. Customer Hosted Data Components (Mandatory)
The following components must be provisioned and maintained by the customer/partner to ensure data residency and compliance. Specs are per individual component instance.
Component | Purpose | Min. Recommended Specs | Storage Details | HA/DR Notes |
MongoDB Cluster | Storing component configurations, customer, agent, and bot activities. | 0.5 vCPU, 1.5 GiB RAM (Per Node) | 8 GiB vDisk (System) | Requires a Replica Set configuration (3+ nodes) for production HA. |
ETL Pipeline (Airflow) | Data transformation and loading for historical reports. | 2 vCPU, 4 GiB RAM | 200 GiB (System) + 500 GiB (Data) | Must be sized for transactional load; redundancy is highly recommended. |
File Store (MinIO) | Storage for media files exchanged in conversations. | 0.5 vCPU, 128 MiB RAM | 8 GiB vDisk (System) + Large S3 compatible storage pool. | Must use S3-compatible object storage; size based on media retention policy. |
3. Customer Hosted AI Services (Optional)
Hosting open-source LLM/AI services (Ollama, Whisper, Mistral) requires specialized hardware, often leveraging GPUs. This is a high-cost component that scales significantly with concurrent usage.
Concurrent Users | Minimum Hardware Requirements | Key Requirement |
5 Users | GPU: 1 x RTX 3090 (24GB) / RAM: 32GB | GPU is mandatory for real-time LLM inference. |
10 Users | GPU: 1 x A100 (40GB) / RAM: 64GB | Server-grade CPUs required for high concurrency. |
100 Users | GPU: 2 x A100 (80GB) / RAM: 256GB | Requires high-speed network (10Gbps) and Load Balancer. |
💡 Summary of Customer Responsibility
The customer is responsible for the provision, installation, configuration, management, high availability, backups, and disaster recovery for all listed components. All application components (Agent Desktop, Routing, Connectors) are hosted and managed by Expertflow Cloud.