Chapter 3 Multiprocessing

Question 1

Process

Answer 1

Process API takes a function to execute and runs it in a new process.
Process(target=process_task, name=”process-{0}”.format(i))

Question 2

Fork

Answer 2

Process can be created by fork, spawn and forkserver. This is set in the library using multiprocessing.set_start_method(‘fork’)

Question 3

Fork disadvantages

Answer 3

For copies almost everything from the current process, this could be problem with lock states, open file descriptors etc

Question 4

Spawn

Answer 4

Spawn uses exec to replace the state. Spawn is set using multiprocessing.set_start_method(‘spawn’)

Question 5

Forkserver

Answer 5

Need to study

Question 6

Queues and pipes

Answer 6

Queues take items, pipes are like files.

Question 7

Shared memory

Answer 7

created using Value

Question 8

Locks and reentrant locks

Answer 8

as usual

Question 9

Barrier, semaphore and condition variable

Answer 9

same as threads

Question 10

Pool

Answer 10

Creates a pool of processes which execute multiple tasks and do not die immediately

Question 11

Maps

Answer 11

map api are available to distribute work to a pool

Question 12

Managers

Answer 12

Managers are a way to share data and co-ordinate between processes running on different machines.

Question 13

Creating Manager

Answer 13

BaseManager creates a manager and binds it to a port. To connect, we create a BaseManager and call connect. Then we can call the api to get work.

Question 14

Types of Managers

Answer 14

1. BaseManager

2. CustomManager, subclassing from BaseManager

Question 15

SyncManager

Answer 15

SyncManager synchronizes between server and client.

Question 16

SyncManager Synchronization

Answer 16

SyncManager provides synchronization premitives to other processes like semaphore, mutex, cv etc.