Process Management Concepts

Process Management

• Process management involves memory management and processor sharing, where multiple programs compete for limited resources.
• Deadlocks occur when two or more processes are blocked, waiting for each other to release resources, preventing any of them from continuing execution.
• A deadlock is often described as a "deadly embrace."
• Jobs in a deadlock are placed in a HOLD state, waiting for vital resources that are unavailable.
• A system in deadlock state is completely unresponsive and frozen.
• External intervention may be necessary to resolve a deadlock.
• Starvation occurs when a job is indefinitely postponed due to being consistently blocked from resources.
• A race condition happens when multiple processes access and modify shared resources concurrently, potentially leading to unexpected results.
• The difference between deadlock, starvation and race conditions will be needed to solve conflicts during execution efficiently.

Learning Objectives

• Students learn the causes of system deadlocks and livelock.
• Learners will understand the differences between preventing and avoiding deadlocks.
• The ability to detect and recover from deadlocks will be taught.
• Students learn about process starvation and detection/recovery methods. Concepts of a race condition and prevention strategies are included.
• Learners know the difference between deadlock, starvation and race.

Deadlock

• Many programs have to compete for limited resources like memory and processor.
• Deadlock leads to "deadly embrace" situation where two or more jobs are in HOLD state, waiting for a vital resource unavailable.

Case 1: Deadlocks on File Requests

• Jobs request and hold files for the duration of their execution.
• A scenario is illustrated with two programs (P1, P2) competing for two files (F1, F2).
• Deadlock happens when, for example, P1 has access to F1 and F2 and requires F2 and F1, which is impossible.

Case 2: Deadlocks in Databases

• Database locking can cause conflicts where two or more processes access and lock records.
• Locking strategies can either lock the entire database or subsections.

Case 3: Deadlocks in Dedicated Device Allocation

• A scenario is given where two programs need two tape drives, but only two are available in the system.
• One program requests a tape drive and gets it. Then, the second program also requests a tape drive and gets it.
• The first program then requests the second tape drive, which becomes blocked, while the second program similarly requests the first tape drive, creating a deadlock situation.

Case 4: Deadlocks in Multiple Device Allocation

• Multiple processes might request and hold dedicated devices (tape drive, printer, plotter).
• A scenario of three programs competing for three devices is presented.
• Processes can become blocked waiting for resources held by another process.

Case 5: Deadlocks in Spooling

• A dedicated device like a printer is made sharable using a high-speed disk.
• The spooling mechanism acts as temporary storage for output.

Case 6: Deadlocks in a Network

• No network protocols control message flow between network computers.
• The sequence is given where several computers exchange messages with each other.
• The available buffer space fills up, causing a deadlock.

Case 7: Deadlocks in Disk Sharing

• Competing processes send conflicting commands for disk access.
• Processes get blocked while waiting for different parts of disk access which leads to deadlock.
• A two-process scenario is used to demonstrate these issues with respect to disk accesses.

Conditions for Deadlock

• Four conditions must occur simultaneously prior to deadlock or livelock: mutual exclusion, resource holding, no preemption, and circular wait.
• Operating systems need to recognize simultaneous occurrence of these conditions to deal with these conflicts. Deadlocks can only be resolved if one condition is removed.

Mutual Exclusion

• Only one process can access a given resource at a time.

Resource Holding

• A process holds a resource and requests another resource while waiting for another process to release it.

No Preemption

• Resources previously granted to a process cannot be forcibly taken away. These have to be explicitly released by the process holding them.

Circular Wait

• A circular chain of two or more processes exists, where each process waits for a resource held by the next process in the chain.

Modeling Deadlocks

• Diagrams show processes and resources in scenarios and the possible events that can happen when more than one process is requesting resources that are already held by another, leading to a deadlock.

Modeling Avoidance

• System proactively prevents deadlocks by knowing the sequence of requests in advance, allowing for possible delays.

Banker's Algorithm

• A resource allocation algorithm use to solve deadlock avoidance. The algorithm regulates resources allocation for a system, to prevent deadlock from occurring.

Safe State

• The system is in a safe state if there is a sequence of resource processes in the system.

Unsafe State

• The system is in an unsafe state if there's no such sequence.

Avoidance Strategy

• The system keeps track of resources to only allow requests for processes in a safe state.

Detection Strategy

• Check for cycles in the resource graph to identify a deadlock.

Recovery Strategy

• When deadlock is detected, the system must recover by removing deadlocked processes in a safe manner. This might involve restarting processes, or reclaiming resources.
• Recovery methods may result in job loss, or system downtime.

Starvation

• Situations where certain processes are indefinitely postponed due to a constant refusal of resource access. This may result from conservative resource allocation.
• "The Dining Philosophers" problem exemplifies the challenges of avoiding starvation in system.
• Starvation avoidance algorithms implement an aging mechanism to track how long a job waits, and allow prioritization if necessary.

Summary

• The operating system dynamically allocates resources.
• Avoiding deadlock or starvation requires management of resource requests.
• Four methods for deadlock management include prevention, avoidance, detection, and recovery.
• Prevention removes the precondition for deadlock, whereas avoidance proactively avoids deadlock if the precondition is possible.
• Detection identifies deadlock and recovery resolves the deadlock situation, potentially with some process loss
• Prevention algorithms may lead to inefficient resource utilization.
• Deadlock prevention mechanisms might lead to delays or reduce system responsiveness.