Deploy on a Kubernetes Environment

Use this deployment model when you want to run KloudMate Infinity on Kubernetes-managed infrastructure.

Infrastructure Prerequisites

• A running Kubernetes cluster, whether self-hosted or managed through services such as EKS, AKS, or GKE
• Kubernetes version >= 1.22
• x86-64/amd64 workloads only
• Helm version >= 3.14.3
• A default StorageClass that supports dynamic provisioning of PersistentVolumeClaims, with allowVolumeExpansion: true and volumeBindingMode: WaitForFirstConsumer
• A high-performance SSD-backed StorageClass for data plane workloads — ReadWriteOnce block storage on NVMe or equivalent, with a sustained target of at least 3,000 IOPS and 125 MB/s throughput per volume (for example, AWS gp3/io2, Azure Premium SSD v2, or GCP pd-ssd/pd-extreme)
• A ReadWriteMany (RWX) shared-filesystem StorageClass for shared blob storage — required because multiple workloads mount the same PVC from different nodes to read and write shared data files (for example, AWS EFS, Azure Files, GCP Filestore, NFS, or CephFS)
• A Kubernetes LoadBalancer or ingress controller for exposing services externally
• A domain name with a wildcard SSL certificate
• SMTP credentials for sending email notifications
• A Stripe account if online billing is required
• kubectl access to the cluster

Note

All storage for KloudMate Infinity is provisioned through PersistentVolumeClaims, not raw disks attached to nodes. Plan capacity at the StorageClass and PVC level so volumes can be rescheduled with their pods. Workloads use different access modes — ReadWriteOnce block volumes for the data plane, and a ReadWriteMany shared filesystem for blob storage shared across the control plane.

Minimum Sizing

Compute sizing for the cluster. PVC storage is sized separately under Persistent Volume Requirements — PVCs are cluster-scoped and provisioned through your StorageClass, not attached to specific nodes.

Resource	Minimum / POC	Minimum Production
Nodes	3	5
CPU (total)	8 cores	16 cores
Memory (total)	16 GB	32 GB

Moderate Workload Cluster

This setup is suitable for approximately 10 TB of monthly ingestion.

Tier	Node Count	vCPUs Per Node	RAM Per Node
Control plane	3	4-8	8-16 GB
Data plane	2	8-16	32-64 GB

High Workload Cluster

This setup is intended for very high ingestion rates and larger production deployments.

Tier	Node Count	vCPUs Per Node	RAM Per Node
Control plane	3	4	8-16 GB
Data plane	4	16-32	64-128 GB

Persistent Volume Requirements

PVCs are provisioned dynamically through StorageClasses and bound to pods, not nodes. Plan the following PVC capacity independently of node sizing.

Purpose	Access Mode	Storage Class	Moderate Workload	High Workload
Data plane (analytics / telemetry storage)	ReadWriteOnce	High-IOPS NVMe SSD	1-2 TB per volume	2-5 TB per volume
Control plane (databases, queues, cache)	ReadWriteOnce	General-purpose SSD	50 GB per volume	50 GB per volume
Shared blob storage	ReadWriteMany	Shared filesystem	50-100 GB total	50-100 GB total

• Data plane PVCs must target at least 3,000 IOPS and 125 MB/s throughput per volume for moderate workloads, and at least 16,000 IOPS and 500 MB/s for high workloads.
• Shared blob storage must be ReadWriteMany because multiple workloads mount the same PVC from different nodes.

After completing the prerequisites, contact KloudMate to proceed with deployment.