This guide provides a model for Partners or Service Providers hosting Expertflow CX as a multi-tenant Contact Center as a Service (CCaaS). This is a large, shared infrastructure base and predictable scaling per new tenant.
Deployment Prerequisites (Platform & OS)
See CX Deployment Prerequisites for detailed platform, operating system, and network prerequisites required before deploying Expertflow CX.
Base Infrastructure (Shared by All Tenants)
The hosting environment starts with a Large Enterprise HA/DR deployment, forming the shared core.
|
Component |
Purpose |
Min. Base Instances (Shared) |
Recommended Specs (per VM/Node) |
Notes on Sharing |
|
|
|
|
|
|
|
CX-Core Cluster |
Unified Routing, Agent Management, Digital Channel Connectors |
5+ Nodes (3 Control, 2+ Worker) |
16 vCPU, 32 GB RAM, 500 GB+ HDD/SSD |
All tenants share this core routing capacity. Scale-out must be continuous. |
|
Databases |
Primary Data Persistence (MongoDB, PostgreSQL) |
Dedicated Cluster (5-7 Replicas) |
High-IOPS SSD, dedicated RAM pool |
Databases are shared but logically partitioned (isolated data) for each tenant. Requires aggressive monitoring. |
|
CX Analyzer |
Reporting Interface |
2+ Worker Nodes |
8 vCPU, 16 GB RAM, 1 TB Disk (Shared Reporting DB) |
Supports reporting for all tenants from a single instance, but capacity scales with aggregated report usage. |
|
Data Pipeline Orchestrator |
Shared ETL/Data Integration (Apache Airflow) |
2+ Worker Nodes |
8 vCPU, 16 GB RAM, 250 GB Disk |
Mandatory Component: Shared capacity for all tenants' core ETL jobs. |
Scaling Strategy: Hardware Addition Per New Tenant
When provisioning a new tenant, the primary calculation involves adding capacity for agent-specific and resource-intensive components to the shared pool.
|
Component Type |
Scaling Metric |
Hardware Increment for Capacity |
Notes |
|
CX-Core Worker Capacity |
Every 100-150 new Concurrent Agents (across all tenants) |
+1 Worker Node: 16 vCPU, 32 GB RAM |
Scale-out shared worker nodes horizontally to maintain routing performance. |
|
Media Server Capacity |
Every 100-150 Concurrent Voice Calls |
+1 Media Server: 8 vCPU, 16 GB RAM |
Dedicated resources for voice processing and recording capacity. |
|
Jambonz Capacity |
Every 200-300 Concurrent Voicebot Sessions |
+1 Set of 3 VMs (Feature, RTP, Node) |
Horizontal scaling of the bot gateway is critical for high-volume conversational IVR offerings. |
Dedicated Per-Tenant Add-ons
Components that handle configuration, scheduling, or outbound campaigns often require dedicated instances per tenant for performance isolation and security.
|
Add-on Component |
Deployment Strategy |
Required Specs (Per Tenant Instance) |
Rationale |
|
WFM (Workforce Management) |
Dedicated Instance (Sub-chart) |
4 vCPU, 8 GB RAM, 100 GB Disk |
Ensures scheduling, adherence, and forecasting for Tenant A do not interfere with Tenant B. |
|
Surveys and Campaigns |
Dedicated Instance (Sub-chart) |
4 vCPU, 8 GB RAM, 50 GB Disk |
Campaign dialing velocity is isolated, preventing aggressive outbound activity by one tenant from affecting another. |
|
Third-Party AI/LLM |
Dedicated Cluster/Pool (Consult Vendor) |
Vendor-specific (often GPU) |
Ensures performance isolation and billing accuracy for high-cost ASR/LLM resources per tenant. |
💡 Summary for Presales
The multi-tenant model is a large shared base with incremental hardware additions based on aggregated concurrent usage. Partners must prioritize robust database HA and clear resource isolation for dedicated, performance-sensitive add-ons like WFM and Surveys and Campaigns.