8. Threads and Thread Implementations

Question 1

What makes up a thread?

Answer 1

Registers and a stack

Question 2

What are problems with the CPU that the OS tries to correct?

Answer 2

There is only one CPU (or usually fewer than there are processes needing to be run) and it is way faster than the other parts of the system

Question 3

What is batch scheduling?

Answer 3

Running jobs sequentially until completion

Question 4

How does the operating system create the illusion of concurrency?

Answer 4

By rapidly switching the CPU between multiple running tasks

Question 5

How does the OS ensure that it retains control over what threads are using the CPU?

Answer 5

By using a periodic timer to generate hardware interrupts

Question 6

What is the cost of doing a context switch?

Answer 6

Context switches require executing many instructions and saving a lot of state (registers and thread stack). This takes time.

Question 7

How does the cost of context switching affect the rate at which the hardware timer fires to allow the kernel to potentially switch between threads?

Answer 7

The rate of timer firing cannot be too high, otherwise the overhead from the context switches will start to dominate the CPU and drag the system performance down

Question 8

How are registers shared between threads or processes?

Answer 8

They are not. Registers are private to an individual thread. [?]

Question 9

How is the stack shared between threads or processes?

Answer 9

The stack is shared by all the threads of a process. [?] Process’s stack is not shared between processes.

Question 10

How is the address space shared between threads or processes?

Answer 10

The address space is shared by all the threads of a process [?] A process’s address space is not shared with other processes.

Question 11

How is the file descriptor table shared between threads or processes?

Answer 11

The file descriptor table is shared by all the threads of a process. The file descriptor table is copied over if fork is called on a process, but not shared directly with another process.

Question 12

Why do we use threads and not just processes?

Answer 12

Threads can be a good way of thinking about applications that do multiple things “simultaneously”. Threads may help applications hide or parallelize delays caused by slow devices

Question 13

Lay out an analogy for the difference between threads and a process.

Answer 13

Multiple threads within a single process are like multiple cooks trying to prepare the same meal together.

Each one is doing one thing. They are probably doing different things. They all share the same recipe (instructions) but may be looking at different parts of it. They have private state but can communicate with each other easily enough. They must coordinate with shared resources!

Question 14

How does a web server use multithreading?

Answer 14

Web servers use a separate thread to handle each incoming request

Question 15

How do web browsers use multithreading?

Answer 15

Web browsers use separate threads for each open tab.

When loading a page, separate threads are used to request and receive each unique part of the page

Question 16

How does SCIENCE use multithreading?

Answer 16

One example is to execute a divide-and-conquer routine on a highly parallelizable data set (lots of threads computing different parts of the data set)

Question 17

Why threads within a process and not just a bunch of processes?

Answer 17

Threads get to take advantage of lots of shared state to communicate with each other. Interprocess communication (IPC) is more difficult because the kernel tries to protect processes from each other. Additionally, the state associated with processes doesn’t scale well.

Question 18

Describe the M:1 model of thread implementation.

Answer 18

Threads can be implemented in user space by unprivileged libraries such that M user threads are created and look like 1 thread to the OS kernel

Question 19

Describe the 1:1 model of thread implementation

Answer 19

Threads can be implemented by the kernel directly

Question 20

Why is it possible to implement threads in user space?

Answer 20

Creating threads does not involve multiplexing hardware between processes, so no kernel privilege is required.

Question 21

How do we save and restore context when implementing threads in userspace (without the kernel)?

Answer 21

This is just saving and restoring registers. The C library has an implementation called setjmp()/longjmp()

Question 22

How do we preempt other threads from userspace?

Answer 22

We use periodic signals delivered by the OS to activate a user space thread scheduler

Question 23

What are the PROS of the M:1 (userspace) threading model of thread implementation?

Answer 23

Threading operations are much faster because they do not have to cross the user/kernel boundary, AND thread state can be smaller (because you’re only saving register values?)

Question 24

What are the CONS of the M:1 (userspace) threading model of thread implementation?

Answer 24

We can’t use multiple cores, because the process is assigned to a single core (usually).

The OS also may not schedule the application correctly because it doesn’t know about the fact that it contains more than one thread.

Lastly, a single thread may block the entire process in the kernel when there are other threads that could run.

Question 25

What are the PROS of the 1:1 (kernel) threading model of thread implementation?

Answer 25

Scheduling might improve because the kernel can schedule all threads in the process

Question 26

What are the CONS of the 1:1 (kernel) threading model of thread implementation?

Answer 26

Context switch overhead for all threading operations can get costly.