Lecture_18-CSCI_U511_01-Jahangir_Majumder-Spring_2025

These flashcards cover key concepts and definitions from the lecture on parallel and distributed systems in the context of operating systems.

What is an algorithm?

A detailed sequence of actions to perform to accomplish some task.

How does a parallel algorithm operate?

It performs tasks efficiently on multiple processors, often starting with a serial algorithm.

What is partitioning in the context of parallel algorithm design?

Deciding how workers talk to each other by cutting a project into pieces.

What are the two types of decomposition in partitioning?

Functional decomposition and domain decomposition.

What does agglomeration refer to in parallel computing?

Combining tasks and revisiting initial partitioning decisions.

What are the goals of performance optimization in parallel systems?

Shorter overall execution time, better speedup, and better scalability.

What is Amdahl's Law?

A formula that describes the potential speedup of a task using parallel processing, indicating that speedup is limited by serial portions of the task.

What does communication cost consist of?

The cost associated with sending messages and the size of those messages.

What is the definition of scalability in parallel systems?

The ability of the system to increase speedup proportionally with the number of processing elements.

What does 'load balancing' ensure in parallel computing?

It ensures that the workload is evenly distributed among processors to optimize performance.

Define performance metrics in the context of parallel systems.

Criteria used to measure the effectiveness and efficiency of parallel operations, such as execution time, speedup, efficiency, cost, and scalability.

What is super-linear speedup?

A performance improvement greater than linear, often due to better utilization of resources like cache and memory.

What is the significance of communication design in parallel systems?

It addresses how well the communication between processors is structured to minimize latency and optimize performance.

What does increase in granularity aim to improve in parallel computing?

It aims to reduce communication costs and enhance computation efficiency.

What is emphasized in the Communication Design Checklist?

Ensuring load balancing, restricted group communications, and concurrent operations among communication and computation.