Concurrency and Deadlocks PDF (STI IT2105)

IT2105 Concurrency and Deadlocks 1. Multiple applications The core of operating system designs focuses on the management of 2. Structured applications processes and threads, such as the following (Stallings, 2018):...

IT2105 Concurrency and Deadlocks 1. Multiple applications The core of operating system designs focuses on the management of 2. Structured applications processes and threads, such as the following (Stallings, 2018): 3. Operating system structure Multiprogramming – It is the management of multiple processes within a uniprocessor system. Below are some important terminologies related to concurrency: Multiprocessing - It is the management of multiple processes within a Atomic operation – It is a function or an action implemented as a multiprocessor. sequence of one (1) or more instructions that appear to be indivisible, Distributed processing – It is the management of multiple processes wherein no other process can see an intermediate state or interrupt the executing on multiple distributed computer systems. operation. In this operation, the sequence of instruction is guaranteed to execute as a group, or not execute at all, having no visible effect on the The basic requirement to support concurrent processes is the ability of a system state. Note that the concept of atomicity guarantees isolation from program to enforce mutual exclusion, which pertains to the ability to execute concurrent processes. all other processes from a course of action while one (1) process is granted Critical section – It is a section of code within a process that requires that ability. Any facility or capability that is to provide support for mutual access to shared resources, and must not be executed while other exclusion should meet the following requirements (Stallings, 2018): process is in a corresponding section of code. 1. Only one process at a time is allowed into its critical section, among Race condition – It is a situation in which multiple threads or processes all processes that have critical sections for the same resource or read and write a shared data item, and the final result depends on the shared object. relative timing of their execution. 2. A process that halts in its noncritical section must do so without Starvation – It is a situation in which a runnable process is overlooked interfering with other processes. indefinitely by the scheduler. Although the process is able to proceed, it is 3. No deadlock of starvation. Processes requiring access to a critical never chosen. section must not be delayed indefinitely. 4. When no process is in a critical section, any process that requests In a single-processor multiprogramming system, processes are interleaved in entry to its critical section must be permitted to enter without delay. time to yield the appearance of simultaneous execution. Even actual parallel 5. No assumptions must be made about the relative process speeds or processing is not achieved, and there is an overhead in switching back and number of processors. forth between processes, interleaved execution provides a major benefit in 6. A process shall remain inside the critical section for a finite time only. processing efficiency and program structuring. Principles of Concurrency It is possible to classify how processes interact based on the degree of their Concurrency is an application performance technique that encompasses the awareness about the other existing processes (Stallings, 2018). ability to load and execute multiple runnable programs. Each of these Competition: Processes unaware of each other – This comprises programs may be an application process that does not necessarily execute on independent processes that are not intended to work together, which can the central processing unit (CPU) at the same instance; even their runtimes either be batch jobs, interactive sessions, or a mixture of both. Some of may overlap (Gregg, 2021). the potential control problems in this process interaction set-up are mutual exclusion, deadlock, and starvation. The best example of this situation is A concurrent system supports more than one (1) task by allowing all the tasks the multiprogramming of multiple independent processes. to make progress. Concurrency involves an array of design issues, including Cooperation by sharing: Processes indirectly aware of each other – process communication, accessing and sharing of resources, synchronization This involves processes that are not necessarily aware of each other by of activities of multiple processes, and allocation of processor time to different their respective process identifications but shares access to some objects. processes. Concurrency can be viewed based on the following contexts: The result of one (1) process may depend on the information obtained 04 Handout 1 *Property of STI  [email protected] Page 1 of 4  IT2105 from the other processes. Some of the potential control problems in this locks the mutex must be the one to unlock it, and only the holder of the process interaction set-up are mutual exclusion, deadlock, starvation, and lock can operate. data coherence. Condition Variable – This is a data type that is used to block a process Cooperation by communication: Processes directly aware of each or a thread until a specific condition is true. other – This encompasses processes that are able to communicate with Monitor – This is a programming construct that encapsulates variables, each other by process identification and that are designed to work jointly access procedures, and initialization code within an abstract data type. It on some activity. Some of the potential control problems in this process is easier to control and has been implemented in a number of interaction set-up are deadlock and starvation. programming languages such as Java and Pascal-Plus. The monitor's variable may only be accessed via its access procedures, and that only Below are some operating system (OS) concerns that are raised by the one (1) process can actively access the monitor at any given time. existence of concurrency (Stallings, 2018): Event Flag – It is a memory word used as a synchronization mechanism. 1. The OS must be able to keep track of the various processes through A specific application code is associated with each bit in a flag. A thread the utilization of process control blocks. can wait for either a single event or a combination of events by checking 2. The OS must allocate and deallocate different resources for each one (1) or multiple bits in the corresponding flag. active process. Mailbox or Message Passing – This mechanism is considered as a 3. The OS must protect the data and physical resources of each process means for two (2) processes to exchange information, and that may be against unintended interference by other processes. used for process synchronization. 4. The operation of a process, and the output it produces, must be Spinlock – This is a mechanism in which a process executes in an infinite independent of the speed at which its execution is carried out relative loop waiting for the value of a lock variable to indicate availability. to the speed of other concurrent processes. Principles of Deadlocks (Stallings, 2018) Common Concurrency Mechanisms (Stallings, 2018) Deadlocks can be defined as permanent blocking of a set of processes that The first major advancement in dealing with concurrency-related problems either compete for system resources or communicate with each other. A set of came in around 1965 with Dijkstra's written formal works. Dijkstra's work processes is deadlocked when each process in the set is blocked waiting for focuses on operating system (OS) design as a collection of cooperating an event, typically freeing up some requested resource that can only be sequential processes, with the development of an efficient and reliable triggered by another blocked process in the set. Note that all deadlocks involve mechanism for supporting cooperation. conflicting needs for resources by two (2) or more processes. Below are the common OS and programming language mechanisms that are The following are the general resource categories: used to provide or support concurrency: Reusable resources – These resources can be used by only one process Counting Semaphore – This involves an integer value that is used for at a time and are not depleted by usage. As an example of a deadlock signaling processes. Only three (3) operations may be performed on a involving reusable resources, consider two (2) programs that compete for semaphore, all of which are atomic: initialize, decrement, and increment. exclusive access to a disk file D and a tape drive T. Deadlock occurs if The decrement operation may result in the blocking of a process, and the each process holds one (1) resource and requests the other. increment operation may result in the unblocking of a process. This Consumable resources – These are resources that can be created concurrency mechanism is also known as the general semaphore. (produced) and destroyed (consumed). An unblocked producing process Binary Semaphore – This is a semaphore that only takes the values zero may create any number of resources. Then, when a resource is acquired (0) and one (1). by a consuming process, the resource ceases to exist. As an example of Mutual Exclusion (Mutex) Lock – This mechanism is similar to a binary a deadlock involving consumable resources, consider a pair of processes semaphore. The key difference between the two is that the process that in which each process attempts to receive a message from the other 04 Handout 1 *Property of STI  [email protected] Page 2 of 4  IT2105 process, then sends a message to the other. Deadlock occurs if the receiving process is blocked until the message is received. Figure 2. Examples of resource allocation graph. Source: Operating Systems: Internal and Design Principles (9th ed.), 2018 p. 297 Figure 1. Example of a deadlock. Deadlock Prevention, Avoidance, and Detection Source: Operating Systems: Internal and Design Principles (9th ed.), 2018 p. 291 The following conditions must be present for a deadlock to occur (Silberschatz, Galvin & Gagne, 2018): The resource allocation graph, which was introduced by Richard Holt, is a 1. Mutual exclusion: At least one resource must be held in a non- useful tool in characterizing the allocation of resources to processes. It is a sharable mode. This means that only one (1) thread at a time can use directed graph that depicts the state of system resource processes, wherein the resource. If another thread requests that specific resource, the processes and resources are represented by nodes connected by edges. A requesting thread must be delayed until the resource has been resource allocation graph can be described as follows: released. o An edge directed from a process to a resource indicates a resource 2. Hold and wait: A thread must be holding at least one (1) resource and that has been requested by the process but not yet granted. waiting to acquire additional resources that are currently being held by o A dot within the resource node represents an instance of a resource. other threads. o An edge directed from a reusable resource node dot to a process 3. No preemption: Resources cannot be preempted. This means that a indicates a request that has been granted, and one (1) unit of the resource can only be released voluntarily by the thread holding it after resource has been assigned to the process. that thread has completed its task. o An edge directed from a consumable resource node dot to a process 4. Circular wait: A closed chain of threads exists, such that each thread indicates the process is the procedure of that resource. holds at least one resource needed by the next thread in the chain. The first three (3) conditions are necessary, but not sufficient, for a deadlock to occur. The fourth condition is required for a deadlock to actually take place. Technically, the fourth condition is the potential consequence of the first three conditions. Deadlock can also be described as an unresolved circular wait. 04 Handout 1 *Property of STI  [email protected] Page 3 of 4  IT2105 Deadlock Prevention: Disallows one of the four conditions for deadlock The processes under consideration must be unconstrained by any occurrence – This strategy of involves the designing of a system in such a synchronization requirements; way that the possibility of deadlock is excluded (Stallings, 2018). There must be a fixed number of resources to allocate; and A. Indirect method of deadlock prevention (preventing the first three No process may exit while holding resources. conditions) o Mutual exclusion: In general, this condition cannot be disallowed. If Deadlock Detection: Grant resource requests when possible, but access to a resource requires mutual execution, then mutual exclusion periodically check for deadlock and act to recover – This strategy does not must be supported by the operating system. limit resource access or restricts process executions. Requested resources o Hold and wait: This condition can be prevented by requiring a process are granted to processes whenever possible. Then, the operating system to request all of its required resources at once and blocking the periodically executes an algorithm that detects the presence of a circular wait process until all resources can be granted simultaneously. condition. Once the deadlock has been detected, any of the following recovery o No preemption: This condition can be prevented through the following methods can be performed, whichever is more appropriate for the program ways: (Stallings, 2018):  If a process is denied of further request, that process must release a. Aborting all deadlock processes is the most common solution in operating the resources that it is currently holding; systems.  If necessary, request the process again with the additional b. Back up each deadlocked process to some previously defined checkpoint, resources; and and restart all processes. This requires that the rollback and restart  Let go of other resources in order to proceed with other process mechanisms are built into the system. However, the original deadlock may execution. recur. B. Direct method of deadlock prevention (preventing the occurrence of the c. Successively abort deadlocked processes until deadlock no longer exists. fourth condition) After each abortion, the detection algorithm must be reinvoked to see o Circular wait: This condition can be prevented by defining a linear whether a deadlock still exists. ordering of resource types. d. Successively preempt resources until deadlock no longer exists. A process that encompasses preempted resource must be rolled back to the Deadlock Avoidance: Do not grant a resource request if the allocation point prior to its resource acquisition. might lead to a deadlock condition – This strategy allows the three necessary conditions but makes judicious choices to assure that the deadlock The selection criteria in successively aborting deadlocked processes or point is never reached. Thus, allowing more concurrency. With deadlock preempt resources could be one of the following: avoidance, decisions are made dynamically. Hence, knowledge of future Process with the least amount of processor time consumed so far process resource requests must be known. Two (2) approaches of deadlock Process with least amount of output produced so far avoidance are as follows (Stallings, 2018): Process with the most estimated remaining time A. Process initiation denial: Do not start a process if its demands might Process with the least total of resources allocated so far lead to a deadlock; and Process with the lowest priority B. Resource allocation denial: Do not grant an incremental resource request to a process if this allocation might lead to a deadlock. References: Gregg, B. (2021). System performance: Enterprise and cloud (2nd ed.). Pearson Education, Inc. Below are some restrictions in implementing the deadlock avoidance strategy: Silberschatz, A., Galvin, P. & Gagne, G. (2018). Operating systems concepts (10th ed.). John Wiley & Sons, Inc. Stallings, W. (2018). Operating systems: Internal and design principles (9th ed.). Pearson Education Limited The maximum resource requirement for each process must be stated in advance; 04 Handout 1 *Property of STI  [email protected] Page 4 of 4