Getting started
Spike is not a self-service cluster where an account is created immediately after a form is submitted. Access starts with a conversation. For TU/e academic research (Spike 1), the Supercomputing Center (SCC) reviews the project and capacity. For commercial and enterprise computing (Spike A), Supercomputing Brabant (SCB) guides the intake and onboarding. In both tracks, an assigned engineer onboards you personally.
Service scope and platform applicability
Section titled “Service scope and platform applicability”Before starting, identify which platform track applies to your project:
| Capability / Service | Spike 1 (TU/e Academic Research) | Spike A (Supercomputing Brabant Commercial) |
|---|---|---|
| Target audience | TU/e researchers, postdocs, PhDs, students | Commercial enterprises, industry R&D teams |
| Compute hardware | Dedicated DGX B200 nodes (8x B200 GPUs) | Dedicated DGX B300 nodes (8x B300 GPUs, 256 cores, 2.3 TB RAM) |
| Parallel storage | 250 TiB DDN project space with S3 option | All-flash VAST Data NVMe parallel cluster (/data or /vast) |
| Engineering support | TU/e Supercomputing Center (SCC) AI Engineers | Supercomputing Brabant (SCB) Engineers |
| Hosted LLM gateway | Available (hosted frontier models) | Not provided (companies deploy custom inference/models) |
| Remote data transfer | TU/e network shares, NetApp, Snellius (SURF), S3 | Corporate on-prem, cloud storage (AWS/Azure/GCP) over VPN |
| Onboarding route | TopDesk Service Desk request | Supercomputing Brabant business development & dual-gate intake |
The route onto Spike
Section titled “The route onto Spike”- Submit an access request. Use the Service Desk form as the first step. If your department already tracks the proposal in the Research Cockpit, include that reference in the request instead of creating an unrelated second record. Describe the research question, the expected GPU workload, your data, and the period in which you need the system.
- Discuss the project with an SCC AI engineer. During the intake, you will look at eligibility, technical fit, data handling, expected scale, and whether Spike offers a real advantage over another computing platform.
- Receive a capacity decision. If the project is a good fit and the cluster has room, onboarding can be scheduled. If the cluster is full, SCC will put the project on the waiting list rather than promise capacity that is not available.
- Onboard together. Set aside part of a day with an AI engineer. You will learn the Spike-specific workflow, including projects, containers, loading data, storage, submitting work, and checking a running workload.
- Start with an engineer nearby. After onboarding, you will remain in close contact with an SCC AI engineer while you bring up the first real workloads, resolve early problems, and learn to use the system efficiently.
- Exploration & use-case intake. Contact Niels te Poel (
ntepoel@bom.nl), Business Developer for Supercomputing Brabant, to explore your AI challenge, define project scope, and establish an NDA where needed. - Technical & data intake. Meet with the Project Manager, Supercomputing Brabant (SCB) Engineer, and Data Steward to evaluate feasibility, Blackwell B300 compute needs, VAST Data storage, security, privacy, IP protection, and finalize the list of authorized users.
- Go / No-Go decision & contracting. Leadership issues formal approval to proceed on Spike A; execute the commercial SLA and Data Processing Agreement (DPA, if required).
- TU/e NOP account registration. Team members submit the User Account Registration Form to obtain external TU/e NOP accounts (Noppers) for secure VPN access.
- Onboarding & project realization. An SCB engineer provisions your Run:AI project and VAST storage, followed by a dedicated session to launch your first container workloads on Blackwell B300 GPUs.
Come prepared
Section titled “Come prepared”The onboarding session teaches you how to work on Spike; it is not an introduction to machine learning, Linux, Python, or your research software. Before applying, have:
- Code that is close to runnable on HPC. It does not need to be packaged perfectly for Spike yet, but it should not still require weeks of development before a meaningful test can run. SCC understands that moving into containers and onto B200 GPUs takes some adjustment.
- An estimate of the dataset size. The data should be ready to transfer, and you should know roughly how much needs to move onto the platform.
- A project name. Use one name consistently in the access request and during onboarding.
- A list of users. Identify everyone who will need access to the project.
- A documented data decision. Use Data classification on Spike to describe the dataset and complete any Data Steward, ethics, privacy, or agreement review before transfer.
You should also understand your own code and be comfortable working from a terminal. Container experience is helpful, but SCC can help translate an otherwise working environment into the Spike workflow.
The Access and intake guide explains what makes a strong request, how SCC assesses fit, and what to prepare for the intake and onboarding sessions.
- Code that is close to runnable on HPC. It does not need to be packaged perfectly for Spike yet, but it should not still require weeks of development before a meaningful test can run. SCB understands that moving into containers and onto Blackwell B300 GPUs takes some adjustment.
- An estimate of the dataset size. The data should be ready to transfer, and you should know roughly how much needs to move onto the platform.
- A project name. Use one name consistently across commercial agreements and technical onboarding.
- Finalized list of authorized users upfront. Identify all company engineers who need cluster access before account onboarding begins so TU/e NOP account registration and security screening can start without delay.
- A documented data decision. Complete data classification, sensitivity review, and IP boundary alignment with the SCB team before transfer.
You should also understand your own code and be comfortable working from a terminal. Container experience is helpful, but SCB engineers can help translate an otherwise working environment into the Spike A workflow.
The Access & intake process guide explains what to prepare for the technical intake and onboarding sessions.
Once you are onboarded
Section titled “Once you are onboarded”Use these guides with your SCC contact while you prepare the first real run: