Granularity refers to how divisible a system is. Fine-grained systems, which have high granularity, are broken down into larger numbers of smaller parts, while a coarse-grained system has a smaller number of larger parts. For example, a measurement of an object's weight in ounces is more granular than a measurement of the same object's weight in pounds, which in turn is more granular than a measurement in tons. The concept is an important one in a number of areas, including science, computer technology, and business.
In parallel computer processing, the term refers to how tasks are divided up. Fine-grained parallel processing divides a task into a large number of smaller tasks, usually of short duration, while coarse-grained parallel processing has larger, longer tasks. Finer granularity increases the amount of work that can be done simultaneously and so is potentially faster, but at the price of requiring more resources for communication between processors.
Granularity is also used to describe the division of data. Data with low granularity is divided into a small number of fields, while data with high granularity is divided into a larger number of more specific fields. For example, a record of a person's physical characteristics with high data might have separate fields for the person's height, weight, age, sex, hair color, eye color, and so on, while a record with low data would record the same information in a smaller number of more general fields, and an even lower record would list all of the information in a single field. Greater granularity makes data more flexible by allowing more specific parts of the data to be processed separately, but requires greater computational resources.
In the physical sciences, the term refers to the level of detail in scientific models. A fine-grained model is highly detailed, while a coarse-grained model averages out low-level details rather than portraying them individually. For example, a fine-grained computer model of interactions between atoms will model them at the subatomic level according to the laws of quantum mechanics, while somewhat coarser models may treat the entire nucleus of an atom as a single-point particle that is then modeled according to classical physics, and still-coarser models treat whole groups of atoms as a single unit. Coarse-grained models are less precise, but require less computing power to model a given system than fine-grained models. They also allow modeling of large-scale systems that would be impractical or impossible to portray with finer-grained models.
This concept is also used in business and finance. In banking, granularity in credit portfolio risk management refers to the diversity of the portfolio. Highly granular portfolios have a larger number of exposures spread across a variety of economic areas, which protects the bank from facing large, sudden losses as a result of a default by a single large debtor or a downturn in a single industry. The term can also refer to a similar principle for reducing risk for investments in equities, bonds, or currencies.