Computational Task

A computational task is a single unit of work, or an individual step in a larger process, that a computer performs. Tasks are the building blocks of a job: a system breaks a job into tasks, orders them, and processes them to complete the work efficiently. The word "task" is general and may stand in for more specific terms such as thread, process, query, or step depending on context. Almost any operation, retrieving a file, running a calculation, or loading a page, is a task or a series of tasks.

What is a computational task?

In high-performance computing, a task is often a discrete step within a larger workflow run on a set of HPC resources. That step might be a data-processing stage of a genomics pipeline, the force calculation in a computational fluid dynamics simulation, a round of AI training, or a supporting step such as staging directories or aggregating results with a script. Some tasks run alone; many run in parallel. When parallelized, a task may use message passing (MPI) or run as an embarrassingly parallel set of independent units, each approach suited to different applications.

Types of computational tasks

Tasks are commonly grouped by the resource that limits how fast they finish.

• CPU-bound tasks are limited by processor speed and capacity. They are calculation-heavy, the CPU does most of the work while other components stay idle. Mathematical computations and sorting algorithms are typical examples.
• I/O-bound tasks are limited by the speed of input/output operations such as disk read/write or network transfer. They involve little calculation and depend on moving data between systems. File operations and network transfers, including web browsing, streaming, and downloads, are examples.

Why the distinction matters

Knowing whether a task is CPU-bound or I/O-bound guides how to make it faster. CPU-bound work improves when more processing power is available, distributing it across multiple cores or machines through parallel processing raises the total compute that can be applied. I/O-bound work improves through asynchronous programming and non-blocking I/O, which let a system continue working while it waits for data to move. Matching the optimization to the bottleneck is what keeps expensive resources busy rather than stalled.

How tasks fit into jobs and workflows

A scheduler places tasks into a queue and dispatches them to available resources. Grouped and sequenced correctly, tasks form the jobs and pipelines that define an HPC workload. Preprocessing and postprocessing steps, preparing inputs or formatting outputs, are themselves tasks, and depending on how a workflow is organized they may run as separate jobs at different points rather than all at once. This composability is why "task" remains a useful, deliberately general term: it describes the smallest schedulable unit regardless of what that unit actually computes.