Minimum resource requirements

Hitachi iQ Studio Installation Guide

Version
1.0.x
Audience
anonymous
Part Number
MK-26HIQS000-00
ft:lastEdition
2026-06-05

The resource requirements listed below are for all iQ Studio application services that run on worker nodes. These are not per-node requirements unless otherwise specified. No workloads are deployed on master/control-plane nodes. The application is designed as a multi-namespace deployment for separation of environments.

Resource Minimum requirement Notes / purpose
Control plane master node
CPU 32 cores Control plane, inference services, system components, and supporting workloads.
Memory 128 GB Model loading, inference execution, caching, and runtime operations.
Ephemeral Storage 300 GB Temporary storage space
GPU Worker node
CPU 128 cores Control plane, inference services, system components, and supporting workloads.
Memory 1 TB Model loading, inference execution, caching, and runtime operations.
Ephemeral Storage 1 TB Temporary storage space
GPU NVIDIA GPUs with >= 160GB memory

Worker nodes must have NVIDIA GPUs for model inferencing and document embeddings, and GPU requirements vary based on your workloads, model size, and inference patterns; for basic workloads, 160 GB or more of GPU memory is recommended, and installation prerequisites are met when the GPUs are registered and available for scheduling in Kubernetes, for example, exposed as nvidia.com/gpu resources.

Storage
Primary shared storage 2 TB (HCSF, Hammerspace or equivalent)

Shared centralized storage is required for models, artifacts, shared data, and the vector and transactional databases in iQ Studio, as well as metrics, logs, and traces in the observability component. This storage must support RWX (ReadWriteMany) PersistentVolumeClaims so multiple pods can access the same persistent data. The cluster must use HCSF, HammerSpace, or an equivalent shared storage platform (such as Rook CEPH) that provides Kubernetes CSI integration, dynamic provisioning, and consistent multi‑node access. For reference, a 1024‑dimension embedding model requires 5–6 times the size of the source text; for example, embeddings for a 1‑MB document require 5–6 MB of storage.

Internal container registry 2 TB A container registry is required for deployment.
Local registry space 75 GB of free local disk space

Make sure that you have the required local disk space to successfully complete extraction and all bundle‑related operations.

Object storage 2 TB An S3-compatible object storage server is required for ingesting documents into knowledge collections. For example, Hitachi Vantara’s or Hitachi Content Platform (HCP). If it is not available, the deployment includes a small object server that you can deploy on the cluster.