General and Port Requirements

CTERA Content Services includes the following functionality:

• CTERA Search
• CTERA Classify
• CTERA Expert

The CTERA Content Services configuration includes the following components:

• An Ingestor
  Takes information, breaks it down, and prepares it for the LLM to process.
• A Collector
  Gathers and organizes the data required to train or operate AI models (like LLMs), which require high-quality datasets to learn.
• An Embedder
  A specialized machine learning algorithm that translates human-readable data—like text, images, or audio—into a machine-readable vectors.

These components can be deployed on a single VM. The Ingestor can also be installed on a VM and the Collector and Embedder can then be installed on a second VM.

For optimum performance the collector and embedder VM should be deployed close to the storage bucket where the raw data is collected, to reduce latency. The Ingestor component should be deployed close to the CTERA Portal.

Separating the ingestor component from the collector and embedder components separates the deployments into a data plane consisting of the collector and embedder components and a control plane consisting of the ingestor component.

Installing all the components on a single VM can be used for testing and proof of concept deployments, but for a production environment the deployment of the data and control planes should be on different VMs.

Software Requirements

The following software requirements are for the configuration when all three components are installed on the same VM. When an external edge collector is used, CTERA recommends that it includes a minimum of 2 NVIDIA L40S GPUs. The resources for an external edge collector VM are dependent on the LLLM models that are used by CTERA Content Services. Contact CTERA Support for help with sizing for this VM.

• A virtual machine running Ubuntu 64-bit 24.04.3 LTS (Noble Numbat) or later. By default, this VM will host both CTERA Content Services and as an edge collector.

  	Minimum	Recommended
  CPU	8	–
  RAM	32GB	64GB
  Disk	500GB	1TB
  NVIDIA L40S GPU	2	–

  The VM must have a fixed IP and the DNS configured for remote access:
  For example:

  A            <base_URL>.com                 <base_URL_IP>
  A            *.<base_URL>.com               <base_URL_IP>
  

  Where <base_URL> is the DNS suffix and base_URL_IP is the IP of the Content Services VM.

• When one VM will host CTERA Content Services and anotherVM will host an edge collector.

  • CTERA Content Services VM

    	Minimum	Recommended
    CPU	8	–
    RAM	32GB	64GB
    Disk	500GB	1TB

  • Collector VM
    GPU with 48GB VRAM and 1TB Disk

General Requirements

• A browser with access to the Internet. You can use any of the latest two releases of Google Chrome, Apple Safari, Microsoft Edge, and Mozilla Firefox.
• Microsoft Entra ID P1. For details, see Setting Up SSO in Azure Entra ID.
  Groups of end users require the role ctera-ai-user assigned to the OPENID_USER_ROLE_ID role ID in Entra ID.
  Groups of administrators require the role ctera-ai-admin assigned to the OPENID_ADMIN_ROLE_ID role ID in Entra ID.
• The latest Content Services OVA file, provided by CTERA Support.
  Note

  On platforms other than ESXi, a token is required from CTERA Support to access the CTERA docker repository in Azure Container Registry (ACR) to retrieve the latest CTERA Content Services dockers.

Inbound Ports

Outbound Ports

Internal Ports (not requiring Internet access)

Other ports can be optionally opened for maintenance. For the full list, contact CTERA Support.