AI data lake: petabytes within German jurisdiction
Training data grows faster than any plan – object storage grows with it, without booking volume in advance. As a data lake host in Germany, centron offers the technological basis to advance your data analysis securely, flexibly scalably and within German jurisdiction – whether high-resolution image files for computer vision, log files, audio recordings or other unstructured data.
- Scaling into the petabyte range – store large data volumes without complex partitioning or storage bottlenecks.
- In Hallstadt by default – your data sits in our own data centre near Bamberg, within German jurisdiction.
- S3-compatible object storage – high-performance object storage for demanding deep learning scenarios.
- Evidence in the Trust Center – ISO 27001 based on IT-Grundschutz, BSI C5 attestation, DPA and TOM documentation.
Why object storage is the foundation
Classic storage systems hit limits with AI data volumes – not in capacity, but in structure.
Scales into petabytes
Large data volumes without complex partitioning or capacity bottlenecks – storage grows with the training data, not the other way round.
Storage and compute separated
Compute scales independently of storage. GPU capacity runs only during training while the data stays put – separating the cost curves.
S3 compatible
Standard tools, SDKs and frameworks speak S3. Existing pipelines can be connected without rewriting them.
In Hallstadt by default
Data resides in our own data centre near Bamberg, with no transfer to third countries – a European alternative to US hyperscalers.
Why centron is the right infrastructure for your data lake in the cloud
The smooth flow of data between data sources, storage and ML models requires a coordinated system landscape: data sources load into S3 storage via standard protocols (data streaming as the upstream stage), analysis frameworks such as Apache Spark, Presto or Databricks access the raw data directly, GPU clusters and MLOps pipelines in PyTorch, TensorFlow or MLflow load the training sets straight from storage – for parallel training runs see HPC servers.
Smart factory: predictive maintenance
In Industry 4.0, streams of machine data flow in continuously. By storing IoT sensor data in the S3 data lake, you lay the foundation for predictive maintenance and automated quality control.
E-commerce and retail: customer analytics
Link click paths, historical purchase data, support interactions and unstructured customer reviews in a central data lake. On this data basis, recommendation systems can be trained that strengthen revenue and customer loyalty – infrastructure for e-commerce.
Healthcare and MedTech: medical data
Whether MRI images, lab results or patient histories: processing health data via cloud services is subject to additional requirements under Section 393 SGB V regarding location, establishment and C5 attestation. The evidence centron provides can be found in the Trust Center – details on cloud for healthcare.
Fintech and banking: analyses on financial data
Use centron as storage for transaction patterns, account movements and historical market data to train fraud detection algorithms. Outsourcing to IT service providers is subject to the requirements of BAIT and MaRisk.
From data lake to lakehouse
The data lake stores raw data, the data warehouse structured analytics – the lakehouse combines both through table formats such as Delta Lake or Apache Iceberg on the same object storage. For the infrastructure little changes: you need S3-compatible storage that supports the metadata operations of these formats and grows with the data. The analytics layer on top is yours to choose.
- Store raw – unstructured, no schema up front
- S3 compatible – standard tools just work
- Scale separately – storage and compute
- Governance – data sovereignty under German law
The data lake foundation: plan options compared
The S3-compatible object storage is based on a redundant storage architecture and grows with your requirements – you pay for what you actually use: calculable costs from €0.02 per gigabyte per month, no traffic costs for inbound and outbound transfer. The complement is ccloud³ GPU VMs as AI computing power: thanks to the separation of storage and compute, your data sits permanently and cheaply in S3 storage while you train models on virtual GPU servers – with a direct connection between bucket and VM and flexible runtimes. Once the model is trained, the requirement shifts from training to LLM inference in ongoing operation; all terms on S3 pricing.
| Plan option | S3 Storage Cloud S | S3 Storage Cloud M | S3 Storage pay-as-you-go |
|---|---|---|---|
| Storage included | 250 GB | 1,000 GB | by usage |
| Monthly price | €5.00 flat | €20.00 flat | €0.02 per GB/month |
| Egress traffic | free | free | free |
| Special features | API access, S3-compatible | API access, S3-compatible | scalability into the petabyte range |
The right centron products
Clinics and medical institutions typically implement IT security with these modules – combinable and extensible at any time.
- €0.02 per GB per month
- Outbound traffic free
- S3-compatible API
- NVIDIA GPUs
- Scales separately from storage
- Hourly billing available
- Fixed monthly plans
- Standard control plane free
- AutoScaler & traffic included
What does an AI data lake need as its foundation?
An AI data lake needs object storage that ingests unstructured data in raw form and scales into the petabyte range without coupling storage to compute. The core building block is S3 Object Storage from €5.00 per month at €0.02 per gigabyte, complemented by Cloud GPU for training and Managed Kubernetes for pipelines. The infrastructure runs in our own German data centres, certified to ISO 27001 based on IT-Grundschutz (BSI certificate BSI-IGZ-0773) and with an unqualified BSI C5:2020 Type 1 attestation. New accounts receive €200 in starting credit.
| Module | Price |
|---|---|
| S3 Object Storage | from €5.00 / month |
| Cloud GPU | from €92.59 / month |
| Managed Kubernetes | from €29.99 / month |
Frequently asked questions
What is the advantage of separating storage and compute?
How fast is access to centron S3 storage during AI training?
Where is my data located and which law applies to it?
What distinguishes a data lake from a data warehouse?
Do common tools work with centron S3 Object Storage?
What does outbound traffic cost?
More on data & AI
Get started for free
Sign up and receive €200 credit at centron within your first 60 days.
This promotional offer applies to new accounts only. Available exclusively to businesses.