0%

​

Share this page with:

Some background on high performance computing

At a high level, all computers consist of three main components: input, output, and the processing unit.The input devices are the mouse and keyboard, the output device is the screen, and the processing unit is everything else:

  • Processing chip: this is the key component performing the calculations needed. The two main architectures are CPUs (Central Processing Unit, used for most tasks) and GPUs (Graphic Processing Unit, historically dedicated to gaming, now commonly used for AI models). The differences are not important for the sake of this course. 
  • Memory: this is the computer’s short-term memory. Data in use is stored here to be quickly accessed by the processor.
  • Storage: this is typical disk storage, either on hard disk (HDD) or solid-state (SDD) drives; it’s the computer’s long-term memory. 

While for most everyday tasks it is enough to have all three in the same place (e.g. a laptop or desktop computer), for intensive computations more processing resources are needed. We can then keep the input and output components on our desk, but move the processing unit elsewhere, where multiple computers are grouped in a room similar to a warehouse: this is what we call a data centre. In this way, the processing unit can be efficiently shared across multiple users, accessing it at their convenience. This is particularly common in science, and referred to by different names such as scientific computing or high performance computing (HPC).

Image illustrating the "local computing" setting (on the left) versus the "interacting with a data centre" setting (on the right). On the local computing setting, all the input (keyboard and mouse) and output (screen) devices are located together with the processing unit, and interact among them. On the data centre setting, the input and output devices interact with processing units (data centre) located in a different place.
In a classical, local computing setting, all hardware components (keyboard and mouse input, processing unit, and screen output) are located in the same place and directly interconnected. However, when more intensive computation is required, using data centres offers an advantageous solution. While we continue working on our personal input and output devices, we can externalise the processing tasks to a data centre. The data centre functions as a shared processing unit for multiple users within a department, university, company, or even globally, depending on the situation, and can be located anywhere in the world. Credits: Flaminia Zane/EMBL-EBI Training.

The ‘cloud’ is no different, as it refers to a data centre operated by a company (a cloud operator) which allows users remote access to shared services. Despite the name, the cloud is very much supported by large scale physical infrastructures. As of 2025, some of the largest ones include Google Cloud, Amazon Web Services and Microsoft Azure.