This shows you the differences between two versions of the page.
Both sides previous revision Previous revision Next revision | Previous revision | ||
hpc:hpc_glossary [2023/10/04 09:15] Yann Sagon |
hpc:hpc_glossary [2023/10/04 14:33] (current) Yann Sagon [Glossary] |
||
---|---|---|---|
Line 2: | Line 2: | ||
====== Glossary ====== | ====== Glossary ====== | ||
- | * **HPC** High Performance Computing. You can do that on an HPC cluster. | + | |
- | * **Baobab** This is the name of our HPC service, as well as the name of our first HPC cluster. | + | * **Baobab**: This is the name of our HPC service, as well as the name of our first HPC cluster. |
- | * **HPC cluster** Our compute infrastructure, composed by a bunch of compute node, some storage, at least one head node, all that part of a network | + | * **Batch**: When you want to launch software with a program (e.g., Stata with a script), you do it in batch mode. |
- | * **CPU** Central processing unit. It' | + | * **CPU**: Central processing unit, the main computing unit of every computer |
- | * **Core** Modern CPU contains many cores. You can see each core as a computation unit. | + | * **Core**: Modern CPU contains many cores, each of which can be seen as a computation unit. |
- | * **GPU** Graphics Processing Unit. Co-processors originally designed for graphic rendering. They are now widely used to do general purpose computing. | + | * **FLOPS**: Floating-point Operations Per Second. |
- | * **RAM** Random Access Memory. This is a fast memory that is used to store programs and related data while they are executed. | + | * **GPU**: Graphics Processing Unit, co-processors originally designed for graphic rendering, now widely used for general-purpose computing. |
- | * **Network interface**. The piece of hardware that allows each computer (or node) to exchange data with each other through a network. | + | * **Head node**: The head node, or login node is the entry point for the users. You connect to the head node to have access to the cluster. |
- | * **Batch** When you want to launch a software with a program (Stata with a script), you do it in batch mode. | + | * **HPC**: High Performance Computing, and also the name of your HPC service. |
- | * **Head node** The head node, or login node is the entry point for the users. You connect to the head node to have access to the cluster. | + | * **HPC cluster**: Our compute infrastructure, |
- | * **Interactive** When you want to launch | + | * **Interactive**: When you want to launch software like you would do on your own computer and be able to interact with it. It's not good to do that on clusters because as you are working on your software, there are many times when the compute node is just waiting |
- | * **Job** | + | * **Job**: Your program executed against your data on a resource. |
- | * **Node (compute)** A server of the cluster, where the actual computation is made. | + | * **Network interface**: |
- | * **Partition** In Slurm, the compute nodes are grouped by partitions. Each partition can have different characteristics. | + | * **Node (compute)**: A server of the cluster where the actual computation is made. |
- | * **Resources** | + | * **Partition**: In Slurm, the compute nodes are grouped by partitions. Each partition can have different characteristics. |
- | * **Scheduler** As the cluster is shared between users, when you want to use a resource on the cluster, the scheduler is responsible | + | * **Primary Investigator or PI**: Your boss. |
- | * **FLOPS** FLoating-point Operations Per Second | + | * **RAM**: Random Access Memory, fast memory used to store programs and related data while they are executed. |
- | * **Scratch space** | + | * **Resources**: For example, the CPUs, memory, disks. |
+ | * **Scheduler**: As the cluster is shared between users, when you want to use a resource on the cluster, the scheduler is responsible | ||
+ | * **Scratch space**: A space on a hard disk drive that is dedicated for the storage of temporary user data. It is unreliable by intention and has no backup. |