Pipe and Filter Architecture

Study Notes

Pipe and Filter Architecture

Allows for the deconstruction of monolithic processes into small, independent components consisting of Pumps, Filters, and a Sink.
Data is passed from Pipes through a variety of Filters, and ultimately ends up in the Sink.

Components and Connectors

Pumps: the initial source of data, which can be streamed or passed into an initial filter.
Filters: independent components that execute a specified task and do not share states with other filters or components.
Filters often have buffers to queue work, but are unaware of their upstream/downstream neighbors.
Sink or consumer: the target where the final output from a series of filters is collected.

Topological Constraints

The design usually must form a uni-directional sequence where the structure lacks cycles.
Ensures a clear defined workflow for data, and that the application isn't cycling back to previous filters.
Multiple pumps are possible, but often the structure of pipe and filter application only has one sink to aggregate processed information.

Applicable Problems

Processing required by an application can be broken down into a set of independent steps.
Allows for modularized content that is not tightly coupled.
Each step can be performed by an application and have different scalability requirements.
Provides flexibility in reordering the processing steps performed by an application, and ease with exchanging components for different use cases.
Enables distributed processing of steps across various servers, and simplifies the development process for developers.

Resilience to Change

Has very low resilience to change as each part of the system is componentized, leading to high cohesion and low coupling.
Easy to create new filters and attach new pipes, without impacting the remaining code base.
Easy to expand a system that adopts this architecture, and make changes to a single component without requiring a full understanding of the entire system.

Negative Behaviours

Requires a large amount of state to be transferred from component to another, as no assumptions can be made about the previous process.
Not suitable for processes that require large amounts of state, as it becomes inefficient.
Not recommended for developing highly dynamic, interactive systems.
Linearized implementation of pipe and filters can give rise to delays in program execution.
Cannot be used if a program requires a set of steps that cannot be executed independently or concurrently.

Description

Learn about the Pipe and Filter architecture, a component-based style that breaks down monolithic processes into smaller, independent components. Understand the roles of Pumps, Filters, and Sink in data processing.