CIQ partner

HPC clusters on Rocky Linux: every node the same, always

We deploy your cluster with Warewulf Pro, which boots every node over the network from the same image, and run your jobs in Apptainer containers. We set it up with Slurm and look after it with our own team.

Talk about your clusterMessage us on WhatsApp

For x86_64 and ARM, with NVIDIA or AMD GPUs.

Why clusters drift

A cluster starts with every node the same. Months later, each one has its own library version, its own patch and its own manual fix. The day a job fails on node 47 but not on node 12, nobody knows why, and reinstalling by hand does not scale.

With Warewulf, nodes have nothing installed on disk: they boot over the network from an image, and when the image changes, they all change. On top of that, Apptainer packs each application and its dependencies into a container, so what works on one node works on all of them.

What the CIQ HPC stack includes
Warewulf ProStateless network boot for the nodes (iPXE and HTTP), with images ready for OpenHPC, Slurm and OpenPBS.
ManagementWeb interface for the images, an overlay catalogue and guided major version upgrades.
Apptainer ProLong-term support releases, GPU, CUDA, ROCm and MPI checks before launch, and a record of every run.
ContainersSIF, OCI and Docker, rootless, with Open MPI, MPICH, Intel MPI and PMIx.
OfflineLocal image mirrors and signed packages for air-gapped networks.
FoundationRocky Linux or RLC Pro on x86_64 and ARM.

Which piece you need

Warewulf Pro

If you are building a new cluster or the one you have has become unmanageable. Everything else sits on it.

Apptainer Pro

If your users already work with containers, you are moving from Singularity or you need to know which image ran on which node.

Fuzzball

If you want researchers to submit jobs from a browser, without the command line, on top of your Slurm or PBS.

Ascender Pro

If you also need to patch and audit lots of servers with Ansible, inside and outside the cluster.

For GPU nodes, the OS that fits is RLC Pro AI. Every Rocky edition is on our Rocky Linux support page.

What we bring

CIQ makes the software; designing the cluster, deploying it and getting your team comfortable running it is our job.

  1. Design

    Nodes, network and storage

    With Supermicro servers if needed, and the storage your workload calls for: Ceph, Lustre or IBM Storage Scale.

  2. Deployment

    Warewulf and Slurm

    Node images, overlays and the scheduler, tested with your real jobs.

  3. Containers

    Apptainer for your applications

    We move your Singularity containers across and build the images your team uses.

  4. Training

    Your team in charge

    With our Slurm course and the rest of our HPC training.

  5. Day to day

    Support in English, Spanish and French

    Updates, incidents and cluster growth, with CIQ behind us for Warewulf and Apptainer issues.

Frequently asked questions

What is Warewulf?

An open source system for deploying and managing HPC clusters. Nodes boot over the network from a shared image and keep nothing on disk, so they are all identical, and changing the OS on a hundred nodes means changing one image. CIQ is the main force behind the project.

What does Warewulf Pro add to community Warewulf?

Node images built and maintained for OpenHPC, Slurm and OpenPBS, a catalogue of configuration overlays, a web interface to manage the images, Warewulf and Apptainer packages kept current, and guided upgrades between major versions.

Are Apptainer and Singularity the same thing?

Apptainer is the Singularity project under a new name since it moved to the Linux Foundation in 2021. SIF containers keep working, and it runs without root privileges, which is what makes it suitable for shared clusters.

Can I buy Apptainer Pro yet?

It is in early access, aimed mainly at teams moving from Singularity. We will tell you where it stands when we talk about your cluster.

Does it work with Slurm or PBS?

Yes. Warewulf Pro ships images ready for Slurm, OpenPBS and OpenHPC, and Apptainer Pro works with Slurm and Fuzzball. You keep your scheduler.

Tell us what you will run

With the number of nodes, the kind of jobs and the storage you need, we will tell you how to build it and which pieces you need.

Talk about your clusterMessage us on WhatsApp