Parallelism

Kind: technique

Layer: Performance Core

Record: architecture:parallelism

Severity: contextual

Scope: algorithm, processing, runtime

Canonical: Ontology

A technique for running independent units of work at the same time on several cores or workers.

Listed in Architecture principles, after Concurrency and before Throughput.

Requires

• Independent Work Units

Reinforces

• Throughput
• Performance

Enables

• Multi-Core Utilization

Conflicts with

• Sequential Bottleneck

In tension with

• Coordination Overhead
• Causality
• Causality / Parallelism

Tensions

• Parallelism / Coordination Overhead

Violated by

• Sequential Bottleneck

Severity

• contextual

Category

• Scalability / Performance / Optimization

Enabled by

• Partitioning

Refactors

• Bottleneck Analysis

Reinforced by

• Fan-out/Fan-in

Linked from

• Causality / Ordering / Distributed Time
• Scalability / Performance / Optimization
• Streaming / Pipeline / Dataflow Processing
• Core Vocabulary
• Quality Attributes
• Scalability / Performance / Optimization
• Severity levels
• The resolutions