10 - Concurrency / Parallelism

Question 1

What is Parallelism and what are its main purposes?

Answer 1

• Parallelism involves doing many tasks simultaneously instead of sequentially.
• Its main purposes are to improve throughput and decrease the runtime of programs by utilizing multiple cores or computers.

Question 2

What are the different types of Parallelism?

Answer 2

• Data Parallelism: Performing the same computation on a lot of data.
• Task Parallelism: Different computations/programs running at the same time.
• Pipeline Parallelism: Similar to an assembly line process.

Question 3

What is Concurrency and what is its purpose?

Answer 3

• Concurrency is the execution of two or more tasks overlapping in time.
• Its purpose is to decrease response time by switching between different threads of control and working on one thread when it can make useful progress.

Question 4

How is concurrency relevant in the real world and in computing?

Answer 4

• In the real world, different activities often proceed simultaneously and are inter-related.
• In computing, concurrency is evident with multiple resources like processors, disks, network interfaces, and multiple CPUs/Cores handling multiple users/applications simultaneously.

Question 5

Why is concurrency/parallelism important for programmers?

Answer 5

• Network services must handle multiple requests from clients simultaneously.
• Modern applications often have user interfaces and execute application logic simultaneously.
• Parallel programs need to efficiently map work onto multiple processors.
• There is a need to mask the latency of disk and network operations

Question 6

What are the different levels of concurrency?

Answer 6

• Instruction Level: Execution of two or more machine instructions simultaneously.
• Statement Level: Execution of two or more statements at the same time.
• Unit Level: Execution of two or more subprogram units simultaneously.
• Program Level: Execution of two or more programs simultaneously

Question 7

Can there be concurrency without parallelism and parallelism without concurrency?

Answer 7

• Concurrency without parallelism involves tasks interacting with different subsystems while other tasks are scheduled when the first waits on the subsystem’s response.
• Parallelism without concurrency involves executing code across multiple CPUs or cores, multiple PCs, or alongside GPU and I/O processors​

Question 8

How do single-threading and multithreading differ in processes and threads?

Answer 8

• Single Threading (Uniprogrammed): Involves a single CPU or multiple CPUs executing one thread at a time.
• Multithreading: Allows simultaneous execution of multiple threads, either in a single or multiple CPU setup.
• Examples include MSDOS (single-threaded, uniprogrammed) and Linux/Windows (multithreaded, multiprogrammed)

Question 9

What are the key concepts in scheduling within operating systems?

Answer 9

• Scheduling involves queues to store information about tasks/processes, including the Job Queue (processes awaiting admission), Ready Queue (processes in main memory ready to execute), and Wait Queue(s) (processes waiting for an IO device or other processes).
• The Job scheduler selects processes to put onto the ready queue.
• The CPU scheduler selects the next process to execute and allocates CPU resources
• Guarantees
  
  • Fairness - every process gets its fair share
  • Efficiency - keep CPU busy 100%
  • Response time - for interactive processes
  • Turn-around - for batch processes
  • Throughput - maximize results

Question 10

What is the difference between preemptive and non-preemptive scheduling?

Answer 10

• Non-preemptive scheduling allows a process to run until it blocks or exits, which is simple to implement but open to denial-of-service as a malicious process can refuse to yield.
• Preemptive scheduling, implemented in modern OS, solves denial-of-service by allowing the OS to preempt long-running processes. However, it is more complex due to timer management and concurrency issues

Question 11

What are some exaples of scheduling algorithms?

Answer 11

• Round Robin
• Priority (static and dynamic)
• Shortest job first
• Guaranteed Scheduling
• Earliest deadline first