Deadlock Avoidance and SJF/LLF Scheduling
In this assignment, you will implement a deadlock avoidance algorithm as part of the process manager to avoid deadlocks in a Unix/Linux system. Part of the assignment requires the manipulation of Unix/Linux processes and part of it consists of simulation.
Both the deadlock-handling process and the processes requesting resources are real Unix/Linux processes created using fork(). However, you do not need to actually allocate any resource. The main process executes the Banker's algorithm. The resource-requesting processes make requests by communicating with the deadlock-handling process with either Unix/Linux pipes or shared memorycontrolled by semaphores (your choice, but pipes are easier to use).
The deadlock-handling process chooses the next process with the shortest computation time (SJF) to be serviced. After having one request serviced, a process has to allow another process to make a request before making another one, that is, another process with the shortest computation time is chosen for service. Ties are broken arbitrarily. A process can also release resources during its execution, and releases all resources held when it terminates.
Associated with each process is also a relative deadline (that is, the deadline from the current time instant). One time unit is needed for servicing each request (whether the resource is allocated or not), so the relative deadline for each process is decreased by 1 whether it is serviced or not. If the resource is allocated for a request, then the computation time of this process decreases by 1; otherwise, the computation time remains the same. If a process misses its deadline, keep servicing its requests but indicate the deadline miss in the output. A release instruction also needs 1 unit of computation time just like a request.
The input format is as follows:
m /* number of resources */
n /* number of processes */
available[1] = number of instances of resource 1
:
available[m] = number of instances of resource m
max[1,1] = maximum demand for resource 1 by process 1
:
max[n,m] = maximum demand for resource m by process n
process_1:
deadline_1 /* an integer, must be >= computation time */
computation_time_1 /* an integer, equal to number of requests and releases */
:
request(0,1,0,...,2) /* request vector, m integers */
:
release(0,1,0,...,1) /* release vector, m integers */:
request(1,0,3,...,1) /* request vector, m integers */
:
end.
:
process_n:
deadline_n /* an integer */
computation_time_n /* an integer, equal to number of requests and releases */
:
request(0,2,0,...,2) /* request vector, m integers */
:
release(0,1,0,...,2) /* release vector, m integers */
:
request(1,0,3,...,1) /* request vector, m integers */
:
end.
For output, print the state of the system after servicing each request: the arrays available, allocation, need, and deadline misses, if any.
Next, let's try LLF (least laxity rst scheduling) instead of SJF. The deadlock-handling process chooses the next process with the least laxity (slack) to be serviced. Laxity (l) is de ned as deadline (d) minus remaining computation time (c) at a speci c time (t), that is, l(t) = d(t) - c(t). Which scheduling technique yields fewer deadline misses?