New to the world of HPC and AI?
Get to know our glossary!
This glossary clarifies key terms in order to help you navigate our environment. Whether you are exploring our access models for the first time or deep-diving into AI-optimised supercomputing, these definitions will help you understand how our resources can be harnessed for your innovation and business-critical projects.
| A |
| Access model: One of the available methods for getting access to theLUMI-AI supercomputer for different customer groups such as SMEs, start-ups, and large industries: e.g. national and European calls, grants, challenges, and pay-per-use. |
| Access time: The computing time of a supercomputer that is made available to a user or a group of users to execute their computer programmes. |
| AI Factory: An entity that provides AI supercomputing service infrastructure, introduced in the AI Innovation Package launched by the European Commission. |
| AI Factory Hub: Physical and virtual co-working spaces catering to the diverse needs and goals of various stakeholders, e.g. startups, SMEs, scientific communities, and students. |
| AI/ML inference: Running a trained model to generate predictions on new data, often with low-latency requirements for streaming use cases. |
| AI-optimised supercomputer: A supercomputer that is designed primarily for training large-scale, general purpose AI models and emerging AI applications. |
| Aitta: AI inference service created by CSC. |
| ANN: Artificial neural network |
| API: Application programming interface |
| B |
| Batch job scheduling: A method of managing and executing tasks (jobs) on a computing system without direct user interaction. |
| Batch processing: Processing data in discrete chunks at scheduled times. |
| C |
| CPU: Central processing unit |
| CV: Computer vision; a field of AI that trains computers to interpret the visual world. |
| D |
| DaaS: Dataset-as-a-Service |
| Data Act: European Data Act establishes the rules on access to and use of data, including data generated by connected products and related services. |
| Data governance: Policies and processes ensuring data quality, security, and compliance. |
| Data lake: Centralised storage for raw, structured and unstructured data. |
| Dataset: A structured collection of related data organised for analysis or training AI models. |
| Digital twin: A virtual model designed to accurately reflect a physical object or process. |
| DL: Deep learning |
| E |
| EU AI Act: European Union Artificial Intelligence Act |
| EuroHPC JU: European High-Performance Computing Joint Undertaking |
| Exascale supercomputer: System capable of executing 1018 operations per second (1 Exaflop). |
| F |
| FAIR: Data principles: Findable, Accessible, Interoperable, Reusable. |
| Federated learning: ML technique that trains algorithms across decentralised devices without exchanging local data. |
| Fine-tuning: Adapting a pre-trained model to a specific task using a smaller dataset. |
| Foundation model: Large-scale AI model trained on broad data that can be adapted to various tasks. |
| G |
| GDPR: General Data Protection Regulation |
| GenAI: Generative artificial intelligence; AI that can create new content (e.g. text, images, music). |
| GLM: Generative language model |
| GPU: Graphics processing unit |
| L |
| Large enterprise: Enterprise with more than 250 persons and an annual turnover exceeding 50 million euros. |
| LLM: Large language model |
| LUMI: The current EuroHPC AI supercomputer. |
| LUMI-AI: The upcoming EuroHPC AI supercomputer for next-gen AI workflows. |
| LUMI AI Factory: Pioneering ecosystem integrating computing power, data, and AI talent. |
| LUMI-G: GPU partition of LUMI (AMD MI250X). |
| LUMI-IQ: The upcoming advanced experimental AI-optimised quantum computer. |
| LUMI-K: Kubernetes-based (OpenShift) container cloud service. |
| LUMI-O: Ceph-based object storage. |
| M |
| Metadata: Descriptive information about data (schema, origin). |
| ML: Machine learning |
| Multitenancy: A software architecture where a single instance of a software application serves multiple customers (tenants), with each tenant’s data isolated and remaining invisible to other tenants. |
| N |
| NLG: Natural language generation |
| NLP: Natural language processing |
| P |
| Petascale supercomputer: System capable of 1015 operations per second (1 Petaflop). |
| Pre-exascale supercomputer: System between 100 Petaflops and 1 Exaflop. |
| Q |
| QPU: Quantum processing unit |
| Quantum computer: Device using quantum mechanics to solve complex tasks. |
| R |
| RAG: Retrieval-augmented generation |
| RDI: Research, development, and innovation |
| S |
| Sandbox: Isolated testing environment. |
| SME: Small and medium sized enterprise; enterprises which employ fewer than 250 persons and which have an annual turnover not exceeding 50 million euros, and/or an annual balance sheet total not exceeding 43 million euros. |
| Supercomputing: Computing at performance levels requiring the massive integration of individual computing elements, including quantum components, for solving problems which cannot be handled by standard computing systems. |
| T |
| Trustworthy AI: An approach to artificial intelligence that emphasizes the importance of developing AI systems that are lawful, ethical, and technically robust, ensuring they respect human values and fundamental rights. |
| Try&Buy: A service model that allows organisations to experiment with computing and data resources on a limited scale before committing to full adoption. |