System and Sizing Requirements for Skylar AI

Download this manual as a PDF file

The installation and configuration of Skylar AI and Skylar Advisor uses Harbor for the registry and Helm for deployment. The Skylar AI platform is deployed as a single Helm umbrella chart containing over 20 microservices, databases, and supporting infrastructure components.

This chapter details the different cluster requirements for Skylar Analytics and Skylar Advisor, and it also includes requirements for only Skylar Analytics if you are currently not using Skylar Advisor. This chapter also includes the required and recommended infrastructure dependencies for a Skylar AI deployment.

Cluster Requirements for Skylar Analytics and Skylar Advisor

Before beginning the installation process, verify that your environment has the required versions of Kubernetes and the container runtime (where relevant) listed in the following sections.

Cluster Requirements for Skylar Analytics

  • Kubernetes: Version 1.32 or later (Skylar Analytics only).

    You can run the following command to verify your environment:

    kubectl version

    Kubernetes upgrades must be performed one minor version at a time. If you are planning a cluster upgrade, see the Kubernetes upgrade documentation.

  • Storage: Configured EBS StorageClass capable of dynamic PV/PVC provisioning. Storage cannot be NFS-backed, as this is not supported by ScienceLogic databases. Also, you must have one storage class marked as default.

  • Attached Storage Capacity: Recommended 1 TB or more total storage capacity. This requirement varies by tenant datapoints per minute (DPM) requirements. For more sizing information, see Cluster Sizing Guidelines for Skylar Analytics.

  • Node Storage Capacity: Around 300 GB per node of internal storage for images and Kubernetes-related functions. Ensure that the root directory has most of this space provisioned. For more sizing information, see Cluster Sizing Guidelines for Skylar Analytics.

  • Networking: Currently, only IPv4 is supported. IPv6 is not supported at this time.

  • External Networking:

    • The Skylar AI Kubernetes cluster will need Internet access to https://registry.scilo.tools/. This connection can be through a proxy if needed (additional configuration required).
    • 443 connectivity from the Skylar One instance to the Skylar AI Kubernetes cluster.

  • Deployment Options:

    • Self-hosted Kubernetes clusters, such as bare metal or VMware
    • Cloud-managed Kubernetes services (EKS, GKE, AKS)

Additional Requirements for Skylar Advisor

If you are installing only Skylar Analytics, you can skip this topic.

  • Storage: For Skylar Advisor, ScienceLogic requires that you use RWX Storage configurations only for Skylar Advisor services. This is in addition to the RW EBS storage class required for Skylar Analytics.

  • Kubernetes: Version 1.35 with ContainerD 2.1 or cri-o 1.32 or later runtime.

    You can run the following commands to verify your environment:

    kubectl version

    containerd --version

  • GPU resources provisioning: You need to dynamically allocate GPU resources that can be correctly provisioned to Kubernetes nodes under the Resources tab. Examples would be NVIDIA time-slicing/MIG or VMware vGPU.

  • GPU Driver. Version 570.195.03 with CUDA 12.8.

For more sizing information, see Cluster Sizing Guidelines for Skylar Analytics.

Third-Party Chart Dependencies for Skylar Analytics and Skylar Advisor

Skylar AI includes several third-party Helm charts from Bitnami, Chainguard, and other providers:

  • Bitnami Charts: ClickHouse, PostgreSQL HA, and associated components
  • Maintenance Notice: ScienceLogic validates and updates third-party chart versions with each Skylar AI release. However, customers are responsible for:
  • Security patching of third-party components between Skylar AI releases
  • Vulnerability management for non-Skylar AI components
  • Understanding the security posture of included third-party dependencies

For additional Skylar Advisor configuration information, see Cluster Sizing Guidelines for Skylar Advisor.

Cluster Sizing Guidelines for Skylar Analytics

This section lists the recommended configuration for your Skylar Analytics environment based on your expected volume of data, measured in datapoints per minute (DPM).

The following sizings are only for the Skylar Analytics application. You will need to consider additional resource requirements for system and cluster level services. In other words, do not size a node to just the requirements listed below.

Also, these sizing are for a single tenant cluster. Skylar AI can be multi-tenant, which will change the sizing recommendations as more Data Visualization tenants are added.

Micro Deployment for Customer Proof Of Concept (POC)

  • DPM Range: 0 - 5,000 DPM (datapoints per minute)
  • Target Use Case: POC environments
  • Data Visualization: 5 CPU and 32 GB RAM

Estimated Total: 18 cores, 80 Gi RAM

  • About 150 Gi free space for every Kubernetes node for images, logs, etc.
  • About 250 Gi attached storage for database mounts.

Small Deployment

  • DPM Range: 5,000 - 30,000 DPM
  • Target Use Case: Development, QA, and penetration-testing environments
  • Data Visualization: 5 CPU and 32 GB RAM

Estimated Total: 25 cores, 70 Gi RAM

  • About 150 Gi free space for every Kubernetes node for images, logs, etc.
  • About 512 Gi attached storage for database mounts.

Medium Deployment

  • DPM Range: 30,000 - 215,000 DPM
  • Target Use Case: Small production environments
  • Data Visualization: 5 CPU and 64 GB RAM

Estimated Total: 34 cores, 225 Gi RAM

  • About 150 Gi free space for every Kubernetes node for images, logs, etc.
  • About 1 Ti Storage attached storage for database mounts.

Large Deployment

  • DPM Range: 215,000 - 300,000 DPM
  • Target Use Case: Production environments with moderate to high load
  • Data Visualization: 5 cpu and 64gb RAM

Estimated Total: 53 cores,432 Gi RAM

  • About 150 Gi free space for every Kubernetes node for images, logs, etc.
  • About 2 Ti Storage attached storage for database mounts.

Extra Large Deployment

  • DPM Range: 300,000 - 900,000 DPM
  • Target Use Case: High-scale production environments with heavy workloads
  • Maximum Single Instance: 977,000 DPM
  • Data Visualization: 8 CPU and 128 GB RAM

Estimated Total: 93 cores, 700 Gi RAM

  • About 150 Gi free space for every Kubernetes node for images, logs, etc.
  • About 4 Ti Storage attached storage for database mounts.

Cluster Sizing Guidelines for Skylar Advisor

The sizing guidelines for Skylar Advisor are independent of Skylar Analytics. For example, you could have a medium-sized deployment of Skylar Analytics with a small-sized deployment of Skylar Advisor. These requirements are in addition to the requirements of Skylar Advisor.

The deployment size is based on the number of expected active concurrent users and the size of the Corpus in Skylar Advisor.

Small Deployment

  • GPU: 4 NVIDIA RTX 6000 GPUs or 4 NVIDIA l40s
  • CPU: 80 cores
  • Memory: 450 GB memory
  • Storage: 250 GB of node memory
  • Attached Storage: Depends on the expected Corpus size
  • Number of active concurrent users: 1-15

Medium Deployment

  • GPU: 4 NVIDIA H100 or H200s
  • CPU: 90 cores
  • Storage: 650 GB of node memory
  • Memory: 250 GB memory
  • Attached Storage: Depends on the expected Corpus size
  • Number of active concurrent users: 16-30

Large Deployments

  • GPU: 8 NVIDIA H100 or H200 GPUs
  • CPU: 170 cores
  • Memory: 1 Tb
  • Storage: 250 GB of node memory
  • Attached Storage: Depends on the expected Corpus size
  • Number of active concurrent users: 31-120

Required Infrastructure Dependencies

The information in this section is relevant for both Skylar Analytics and Skylar Advisor deployments.

Ingress Controller

HTTP/HTTPS traffic routing and SSL termination:

  • Required: An ingress controller must be installed and configured
  • Recommended: ingress-nginx controller
  • Alternatives: AWS Load Balancer Controller, GKE Ingress, Azure Application Gateway, Traefik, or HAProxy Ingress

OpenTelemetry Operator

Install the OpenTelemetry Operator with custom image:

# Install OpenTelemetry Operator with custom image
helm registry login 
helm install opentelemetry-operator oci://registry.scilo.tools/skylar/opentelemetry-helm-charts/opentelemetry-operator \
  --version 0.114.0 \
  --namespace opentelemetry-operator \
  --create-namespace \
  --set "manager.collectorImage.repository=registry.scilo.tools/skylar/sl-otelcol" \
  --set "manager.collectorImage.tag=0.16" \
  --set admissionWebhooks.certManager.enabled=false \
  --set admissionWebhooks.autoGenerateCert.enabled=true

If you are using Cert-Manager for your certificate management, you can remove the following lines:

--set admissionWebhooks.certManager.enabled=false \
--set admissionWebhooks.autoGenerateCert.enabled=true

Recommended Infrastructure Components

The information in this section is relevant for both Skylar Analytics and Skylar Advisor deployments.

Load Balancing

A load balancer solution is needed to distribute traffic across worker nodes:

  • Self-hosted: F5 BIG-IP, HAProxy with floating IP, or MetalLB
  • Cloud: Application Load Balancer (ALB) or Network Load Balancer (NLB)

DNS and TLS Requirements

Required:

  • FQDN: A fully qualified domain name pointing to the load balancer
  • TLS Certificate: Valid TLS certificate for the FQDN, provided as a Kubernetes secret

Optional Monitoring Integration

Prometheus-based Monitoring

  • Supported: Integration with existing Prometheus deployments
  • Benefits: Custom metrics export from Skylar AI services