In contrast to the familiar situation where a user runs, for example, Word on his/her personal laptop, a supercomputer may serve hundreds or thousands of users (as an example, CSC currently has over two thousand users) who connect to the system through the internet. Also, where most laptop users utilize a graphical user interface (i.e., start applications by double-clicking an icon with a mouse and generally use a mouse extensively), the most common way to use supercomputers is via command line. So, instead of using a mouse, one works by typing commands with a keyboard.
Even though the command line interface might at first feel arcane to those who have only used graphical desktop applications, it is actually a very powerful and flexible way to use the computer. After all, supercomputers are to normal computers what a Formula One car is to a common station wagon, and great power may come with more difficult usage. However, while very few people have a chance even to try to drive a Formula One car, anybody can learn the basics of using supercomputers and the command line in a single day.
Supercomputers are not made more challenging to use on purpose, and supercomputing centers are constantly looking for ways to make them more accessible. Nowadays, some supercomputing centers also provide web browser-based access, allowing users to utilize supercomputers with more graphical interfaces using a web browser (Firefox, Chrome, Safari, Edge, etc.) running on their local laptop or desktop computer.
As a supercomputer may have thousands of users, there are not necessarily enough resources (e.g., CPUs, GPUs, and memory) for everyone to run their jobs at the same time. Thus, users typically cannot run their heavy computations directly whenever they want interactively but instead need to submit their jobs through a batch queueing system. The batch system then takes care of scheduling the distribution of the resources fairly. If the supercomputer has enough free resources, the job can start immediately, but otherwise, it may take hours or even days for the computing to begin.
A batch system also ensures that single users cannot monopolize the system, as there are typically limits on how much resources a single user can utilize at a time. Batch systems make the utilization of supercomputers more efficient. With only interactive use, there would be load mainly during normal working hours, but with the help of a batch system, jobs can be run on supercomputers 24/7 throughout the year.
With a batch system, the user describes, in most cases in a special file called a batch script, what kind of an executable is to be run, how many nodes, cores, and possibly GPUs are needed, how much memory is to be used, and finally, how much time is to be reserved for the job. The batch queueing system then uses a complicated algorithm to decide when and in which subset of the nodes the job will run. Typically, the more nodes and other resources and/or time a job requires, the longer is the wait in the queue.
Link to optional material (not included in the exam): Using command line in CSC supercomputers
Most supercomputing centers all over the world have similar policies for accessing supercomputers. Supercomputing centers typically have national public funding, and people performing academic research in the country hosting the supercomputers can use them for free. For example, CSC provides free access to researchers and sometimes also for students with an affiliation to Finnish higher education institutions (Universities or Universities of Applied Sciences) or state research institutes if research results are made publicly available. Supercomputers are typically available for commercial use, but it usually is separately invoiced.
Many supercomputing centers participate in transnational access programs in which users outside the hosting country are also provided resources in the supercomputer based on special calls. There may also be scientific criteria for approving applications.
The forthcoming LUMI supercomputer is funded by a consortium of ten European countries (Finland, Belgium, Czechia, Denmark, Estonia, Iceland, Norway, Poland, Sweden, and Switzerland) and the European Union. Half of the resources of LUMI are allocated to consortium members, while the remainder is also available for researchers and industries outside the consortium via a special application process.