M362 - Unit 3 Flashcards
What are race conditions?
Race conditions are when the final outcome of a process is dependent on the timing or sequence of other processes.
In a circular buffer, if the data is being overwritten by a producer before it is consumed by the consumer, what problem has occurred?
This is an example of a race between the consumer and the producer processes to get to the shared memory: a race condition has occurred.
In deadlock,a process is blocked while waiting for a resource that is held by another process. In the dining philosophers problem, what would be the resource?
In the dining philosophers problem, forks are the resource.
What is the difference between deadlock and livelock
In deadlock, one or more processes are not able to execute at all. In livelock, a process is executing operations without progressing.
The two-process software solution in Table 2 is deficient and does not satisfy all the properties for correct concurrent programs. Explain which property in particular is not satisfied.
The liveness property is not satisfied, because one process may end up having to wait indefinitely and unnecessarily.
In what ways can semaphores be used?
Semaphores can be used for:
- ensuring mutual exclusion from critical code
- controlling access to single or multiple shared resources
- synchronising cooperating processes.
How can the incorrect use of semaphores lead to problems with safety and liveness?
By forgetting to use semWait it is possible for a resource to become accidentally unprotected (a safety issue). By forgetting to use semSignal it is possible for a resource to become locked indefinitely (a liveness issue)
How can we achieve mutual exclusion in Java?
In Java, mutual exclusion can be achieved by using synchronised methods, using the synchronized keyword. Fully synchronised objects are objects for which all the methods are synchronised, and that is the safest, but possibly not the most efficient, strategy. For some objects, only certain operations or parts of operations need to be carried out under mutual exclusion, so the objects do not need to be fully synchronised.
What is the wait–notify mechanism designed to assist with?
The wait–notify mechanism provides a way to implement condition synchronisation in synchronised methods – a thread can wait on a condition if the condition is not satisfied, and threads may notify other threads if they have altered the state of the shared data.
What is the distinction between the following three thread states: BLOCKED, WAITING and TIMED_WAITING?
A thread ends up in the BLOCKED state if it attempts to execute a synchronised method, but the lock for the object is already held by another thread. Both WAITING and TIMED_WAITING are states that a thread ends up in as a result of the wait method, and hence as a result of a condition not being satisfied. A thread in the TIMED_WAITING state can be woken up as a result of either a notification or the timeslot having expired.
What is the difference between notify and notifyAll, and which of these two methods is considered safer to use?
With notify only one thread gets selected to receive the notification – but we do not know which thread it will be. With notifyAll all threads get notified. On the whole notifyAll is safer to use, because with notify, if the wrong thread receives the notification, the program may end up deadlocked, as the notification cannot be passed on to other threads
In Subsection 2.3 we discussed fairness as a liveness issue. Do you think the algorithm for the writers-preferred policy is fair?
On the face of it, it seems to make sense to give writer processes priority over readers, because the shared resource should always be in the most up-to-date state. This is presumably also in the interest of the reader processes. However, it is possible, in the given scheme, that reader processes never get a chance to run, because writer processes always get priority. This would be unfair not only to the readers, but also it would be unfair to the writers, since they would be wasting their efforts in writing things that never get read!
What are the four conditions for deadlock?
The four conditions for deadlock are: mutual exclusion, hold while waiting, no pre emption and a cycle of processes.
Generally, how can deadlock be prevented?
By arranging for one of the four conditions not to be true (mutual exclusion, hold while waiting, no pre emption and a cycle of processes)
