Glossary
Term | Description |
Access to the queues | Ability to run a job. |
ARCHIE-WeSt | Academic and Research Computer Hosting Industry and Enterprise in the West of Scotland. |
AU | Allocation Unit. |
“Back-filling” | A method to use the computational resources in efficient way. Short parallel jobs might use (for short period of time) the nodes which are already reserved for other, large parallel job. Back filling is possible only when the queuing system managed to reserve all the required nodes yet. |
Core | Refers to the part of processor which does calculations (in modern multi-core processors). |
Core hour | Computing time on one core. |
CPU | Central processing Unit. Traditionally CPU referred to the processor, nowadays to the processor’s core. |
dm1 | Data mover 1; fast connection to ARCHIE-WeSt (10Gb/s) which should be used for data transfer. |
Environmental Module | The way how the software “installation” is handled. to use particular software the user needs to load appropriate module. |
GPU node | The part of HPC made from Graphic Processing Unit |
qpu.q | Gpu queue. The queue dedicated for calculations ran on graphic card(s). |
GUI | Graphical User Interface. |
Hard quota | The disk allocation which can;t be exceeded. |
HPC | High Performance Computing, e.g. the practice of aggregating computing power in a way that delivers much higher performance than one could get out of a typical desktop computer or workstation in order to solve large problems in science, engineering, or business. High Performance Computer can be treated as cluster of computers. |
Interactive jobs | Jobs ran on ARCHIE-WeSt with usage of software GUI. Not efficient way of using the HPC facility. |
Job script | A job submission script that tells the Sun Grid Engine (SGE) system what compute nodes should be used (normal, SMP, GPU), is the job serial or parallel and what is the project ID. |
Login node | Part of the HPC system which copes with login operations and sends the job to the particular compute node. Not dedicated to run any calculations. |
Lustre | High performance storage. The drive from which all jobs should be launched (submitted). |
Node | If HPC is treated as a cluster of computers, the node is one computer in the cluster. Each core might have several processors and multiple cores on it. |
Parallel.q | Parallel queue. The queue dedicated for multi-core (parallel) jobs which will run on normal compute nodes. |
Project allocation | CPU allocation per project. |
Project ID | Project identifier, needs to be specified in all job scripts, otherwise the job will not run. Used to calculate the CPU usage per project (for monitoring he machine usage and reporting). |
Putty | Software to connect to any remote Linux computer firm Windows Desktop. |
Quota | Disk quota: disk space allocation per user. |
Serial.q | Serial queue. The queue dedicated for single-core jobs. |
SGE | Sun Grid Engine – the queuing system. |
SL1 | High priority access (via project). |
SL2 | Normal priority access (via project). |
SMP node | Large memory node, called also fat node. |
Smp.q | Smp queue. The queue dedicated for both serial and parallel jobs which require large amount or RAM per core. |
Soft quota | The disk allocation which might be exceeded up to hard limit and for not longer than certain time (for example 7 days). |
Submit a job | Launch a job. Send the job to the queuing system to run (once there will be enough resources). |
Subscription fee | Annual payment £500 per user to allow the access to the queues (e.g. run jobs). |
ThinLinc | Remote desktop software to connect to ARCHIE-WeSt. |
Vizualization Server | Part of the system dedicated for extensive pre/post processing data which needs to be done graphically. The data does not need to be transferred to the desktop computer for analysis. |
WinSCP | Software used to transfer the data between any Linux and Windows Computers. |
X | Graphical support (for ssh). |