CS6200

Question 1

What is an operating system?

Answer 1

A special piece of software that abstracts and arbitrates the use of a computer system.

It’s a layer of systems software that:
* directly has privileged access to the underlying hardware
* hides the hardware complexity
* manages hardware on behalf of one or more applications according to some predefined policies
* in addition, it ensures that applications are isolated and protected from one another

Question 2

____ operates in kernel mode to access underlying hardware.

Answer 2

The operating system

Question 3

____ operates in user level mode

Answer 3

Application software

Question 4

T/F? kernel-level mode and user-level mode are supported by hardware

Answer 4

True

Question 5

Attempts to directly access hardware by software running in user-level mode can activate a ____ which interrupts the program and hands control to the OS to determine if it should get access

Answer 5

trap

Question 6

Applications can use _____ to interact with underlying hardware

Answer 6

System calls

Question 7

T/F? User/Kernel transitions have a negligent impact on the performance of applications

Answer 7

No, it can have an impact on cache hit frequency after the transition

Question 8

What are some benefits of a monolithic OS?

Answer 8

* Everything is included
* You can use inlining and compile-time optimizations

Question 9

What are some drawbacks of a Monolithic OS?

Answer 9

* customization is harder
* portability is harder
* manageability is harder
* larger memory footprint
* harder to optimize performance for different use cases (e.g. using different file systems for different access patterns)

Question 10

Is linux a monolithic, modular, or microkernel os?

Answer 10

Modular

Question 11

What are benefits of a modular OS?

Answer 11

* Easier to maintain and upgrade
* Smaller footprint
* Less resource needs

Question 12

What are the drawbacks of a modular OS?

Answer 12

* indirection can impact performance
* maintenance can still be an issue, because modules can introduce bugs

Question 13

In a microkernel, system services like the filesystem and disk driver run in ____ level

Answer 13

User level (as opposed to privileged)

Question 14

What are the benefits of a microkernel?

Answer 14

* Small size
* Easy to verify because it’s small

Question 15

What are the downsides of a microkernel?

Answer 15

* Poor portability, typically customized for target hardware
* More complex to develop software for
* Cost of user/kernel transitions

Question 16

What is a process?

Answer 16

An executing program

Question 17

What is the difference between an application and a process?

Answer 17

Application is a program on disk (static entity), whereas a process is a program that has been loaded in memory and is in the state of execution (active entity).

Question 18

What does a process store in memory?

Answer 18

• Stack
• Heap
• Data
• Text

Question 19

What is “address space”?

Answer 19

The “in memory” representation of a process.

Question 20

What are “virtual addresses”?

Answer 20

The abstracted addresses used by a process.

Question 21

What are “physical addresses”?

Answer 21

The actual physical locations in memory.

Question 22

Where is the mapping of virtual addresses to physical addresses stored?

Answer 22

In the Page Table

Question 23

How does an OS keep track of the state of a process?

Answer 23

Via the Process Control Block

Question 24

A Process Control Block contains the state of CPU registers. Since register contents can change much faster than they can be updated in memory (where the PCB is stored), how does the OS make sure those are updated for later use?

Answer 24

By setting the values in the PCB when stopping the process on the CPU.

This is called a context switch.

Question 25

What is a “context switch”?

Answer 25

Switching the CPU from the context of one process to the context of another.

Question 26

Is a context switch computationally expensive?

Answer 26

Yes. It takes CPU cycles to store the old context and load the new one, and caches on the CPU become invalid for the new process.

Question 27

What states can a process be in?

Answer 27

* New * Ready * Running * Waiting * Terminated

Question 28

What state does a process start in?

Answer 28

New

Question 29

What state does a process end in?

Answer 29

Terminated

Question 30

How does a process transition from New to Ready state?

Answer 30

Admitted

Question 31

How does a process transition from Ready to Running state?

Answer 31

A scheduler dispatch

Question 32

How does a process transition from Running to Ready state?

Answer 32

An interrupt

Question 33

How does a process transition from Running to Waiting state?

Answer 33

I/O or event wait

Question 34

How does a process transition from Waiting to Ready state?

Answer 34

I/O or event completion

Question 35

How does a process transition from Running to Terminated state?

Answer 35

It Exits

Question 36

The CPU is able to execute a process when it is in which states?

Answer 36

Running or Ready

Question 37

How are new processes created?

Answer 37

They are spawned by a parent process (via fork or exec)

Question 38

What is the “fork” mechanism for process creation?

Answer 38

Copying the parent Process Control Block into the new child PCB, where the child continues execution at the instruction after the fork.

Question 39

What is the “exec” mechanism for process creation?

Answer 39

Creates a new process control block, loads a new program, and starts from the first instruction.

Question 40

On unix-based OS, which process is “the parent of all processes”?

Answer 40

init

Question 41

On Android, which process is “the parent of all app processes”?

Answer 41

ZYGOTE

Question 42

What is the role of the CPU scheduler?

Answer 42

It determines which one of the currently running processes will be dispatched to the CPU to start running, and how long it should run for.

Question 43

In the context of process scheduling, an OS can “preempt”. What does that mean?

Answer 43

It can interrupt and save current context

Question 44

In the context of process scheduling, an OS can “schedule”. What does that mean?

Answer 44

It can run the scheduler to choose the next process

Question 45

In the context of process scheduling, an OS can “dispatch”. What does that mean?

Answer 45

It can dispatch the process and switch into its context

Question 46

If a process scheduler is run more frequently, ___ goes down.

Answer 46

Useful CPU work Useful CPU work = non-scheduler Processing Time / Total Time

Question 47

In the context of process scheduling, what is “timeslice”?

Answer 47

Time allocated to a process on the CPU

Question 48

What are some ways a process can end up on the ready queue?

Answer 48

* I/O request * Time slice expired * Fork a child * Wait for an interrupt

Question 49

What do Inter-Process Communication (IPC) mechanisms do?

Answer 49

* Transfer data/info between address spaces * Maintain protection and isolation * Provide flexibility and performance

Question 50

What is message-passing IPC (interprocess communication)?

Answer 50

OS manages a messaging channel, which process can write to and read from.

Question 51

What is the shared memory IPC (interprocess communcation) mechanism?

Answer 51

* OS sets up a shared memory channel, and maps it into each process address space

Question 52

What advantage does shared memory IPC have over message passing IPC?

Answer 52

Less overhead, OS is out of the way

Question 53

What is the advantage of message-processing IPC over shared memory IPC?

Answer 53

How to use the shared memory isn't defined by the OS, so it becomes easier to write a bug in code that uses it, and that code might have to be duplicated between different applications.

Question 54

If we want a process to be able to execute on multiple CPUs, we must use \_\_\_\_

Answer 54

Threads

Question 55

Do threads have the same virtual→physical memory mapping? (shared page tables)

Answer 55

Yes

Question 56

What benefits do threads provide?

Answer 56

* Parallelization * Specialization - can dedicate threads to specific tasks, CPU cache can maintain more valid entries in each core, and prioritization can be used

Question 57

Why use threads rather than just having different processes?

Answer 57

Less overhead. We don't have to allocate memory for each thread. Communication between processes is more complicated.

Question 58

Are threads useful on a single CPU?

Answer 58

Yes. It can work on another thread during long I/O operations.

Question 59

When two threads try to access the same memory value at once, it is a ____ \_\_\_\_ problem

Answer 59

data race

Question 60

\_\_\_\_ ____ ensures that only one thread has access to a memory value at a time

Answer 60

mutual exclusion

Question 61

When one thread is waiting on a specific condition to proceed, and will be notified by another thread, it is using a ____ \_\_\_\_

Answer 61

condition variable

Question 62

Condition Variables and Mutex are both ____ mechanisms

Answer 62

synchronization

Question 63

In Birrell's paper, what call is used to create a new thread?

Answer 63

fork(proc, args)

Question 64

In Birrell's paper, what do the params to fork(proc, args) mean?

Answer 64

proc = program counter args = input parameters

Question 65

In Birrell's paper, what call is used to terminate a thread?

Answer 65

join(thread)

Question 66

In Birrell's paper, what does join(c\_thread) do?

Answer 66

It blocks the calling thread until child thread “c\_thread” finishes

Question 67

If a thread cannot acquire a mutex, it is \_\_\_\_

Answer 67

blocked

Question 68

In Birrell's paper, a lock uses a mutex to prevent a ____ \_\_\_\_ from being executed by more than one thread at once.

Answer 68

Critical Section

Question 69

In the context of threads, what is a “spurious wakeup”?

Answer 69

When a thread is woken up, but it still cannot proceed.

Question 70

In the context of threads, what is a deadlock?

Answer 70

Two or more competing threads are waiting on each other to complete, but none of them ever do.

Question 71

In the context of threads, it prevents ____ \_\_\_\_ to maintain the same order in which mutexes are locked.

Answer 71

Dead lock

Question 72

What are the possible ways of dealing with deadlocks?

Answer 72

* Prevention - maintain a lock order - expensive, can limit performance * Rollback - Return to an execution state before the deadlock - not always possible, also can be expensive * Do nothing - just reboot the software if one occurs

Question 73

What are the benefits of the one-to-one model of kernel and user threads?

Answer 73

OS sees/understands threads, synchronization, blocking…

Question 74

What are the drawbacks of the one-to-one model of kernel and user threads?

Answer 74

Must go to kernel space for all operations (which has a cost) OS may have limits on policies and thread count Since thread management happens at the kernel level, it can affect portability of apps between different OS

Question 75

In the many-to-one model of user/kernel threads, which user-level thread is mapped to the kernel level thread is managed at the ____ level

Answer 75

user

Question 76

What are the benefits of the many-to-one model of user/kernel threads?

Answer 76

Easy to port apps – they don't depend on OS specific limits or policies

Question 77

What are the drawbacks of the many-to-one model of user/kernel threads?

Answer 77

OS has no insights into application needs OS may block entire process if one user-level thread blocks on I/O

Question 78

What are bound/unbound threads in the context of the many-to-many model of user/kernel threads?

Answer 78

Bound user threads are associated to just one kernel thread

Question 79

What are the benefits of the many-to-many model of user/kernel threads?

Answer 79

Can avoid problems of one-to-one or many-to-one

Question 80

What are the drawbacks of the many-to-many model of user/kernel threads?

Answer 80

Requires coordination between user and kernel level thread managers

Question 81

What is the boss-workers thread pattern?

Answer 81

Boss assigns work, workers perform the tasks

Question 82

In the boss-worker thread pattern, how does the boss assign tasks to the workers?

Answer 82

Generally with a queue. Although it requires synchronization among the workers to avoid picking up the same task, it has less of a performance impact than using the boss process to keep track of which workers are busy.

Question 83

What are some benefits of specialized workers in the boss-worker thread pattern?

Answer 83

Better locality – more cache hits QoS management is possible

Question 84

What is the pipeline thread pattern?

Answer 84

Threads assigned one subtask Entire tasks are pipelines of subtask threads Multiple tasks can go through the pipeline at different stages

Question 85

What is the layered thread pattern?

Answer 85

Layers are groups of related subtasks end-to-end task must pass through all the layers Like pipeline, but not linear – some threadpools can handle multiple subtasks