Presentation of practical exercise 3

1
Presentation of practical exercise 3

• Queuing networks
• Process scheduling
• Event-driven simulation

2
A basic computer system

• Consists of a CPU, a memory unit and I/O devices,
  connected by buses.
• The operating system must control and facilitate
  the use of these resources.
• The OSs clients are processes executing
  application programs.
• The OS controls the processes use of and access
  to the hardware resources.

3
Queuing networks

• Queuing networks are used to study the
  performance of computer systems.
• A queuing network abstracts a system into a
  network of interconnected resources with
  associated queues of requests.
• We will consider a QN of our basic computer
  system, and use it to study process scheduling.

4
Our queuing network
Processes (requests) in queue
Resources with queues
Processes (requests) currently being processed
5
How the QN works

• Processes enter the system by entering the memory
  queue.
• When a process has been allocated the required
  amount of memory, it is ready to start execution,
  so it enters the CPU queue.
• When a process is done executing, it leaves the
  CPU and the system.
• While processing, a process may need to perform
  I/O. In that case, it leaves the CPU and enters
  the I/O queue. When the I/O is done, it reenters
  the CPU queue to continue processing.

6
Allowing for multiprogramming

• The system is able to serve multiple processes
  simultaneously.
• This is done by queuing processes, and
  time-slicing their access to the CPU, using the
  Round Robin algorithm.

• When a process has used up its time slice it is
  removed from the CPU and inserted at the end of
  the CPU queue to enable waiting processes to use
  the CPU.
• This introduces another path in the QN

Time slice exhausted
7
The exercise

• The QN has been partially implemented in Java.
  You must complete it.
• The GUI, the Queue class and the simulation
  framework has been implemented.
• The memory unit has been implemented, you need to
  implement the CPU and the I/O device.
• The CPU part is the most interesting as it
  implements the Round Robin process scheduling
  algorithm.

8
Simulation parameters

• These parameters are inputted by the user at the
  start of the simulation
• Available memory.
• The length of the Round Robin time slice.
• Average duration of an I/O operation.
• Average time between the arrivals of new
  processes.
• Total simulation time.

9
Process characteristics

• Each process that arrives has some properties
  that are randomly generated
• Memory required (100 KB ? 25 of total memory)
• Runtime or CPU-time (100ms ? 10000ms)
• Average time between I/O operations
  (1 ? 25 of runtime)
• NB Runtime means time spent in the CPU,
  excluding time spent in I/O

Lifetime
Runtime
10
The simulation framework

• The simulator uses a queue of events to simulate
  the workings of the queuing network.
• An event occurs every time a process is moved
  from one resource/queue to another in the QN.
• The event queue holds future events. The events
  are kept chronologically sorted. The main loop
  repeatedly removes the first event from the event
  queue and processes it.
• This means that the simulation proceeds in
  discrete time steps.
• The processing of an event can lead to processes
  being moved around in the QN, and it can lead to
  new events being added to the event queue.

11
Event processing

• Example The arrival of a new process.
• This event is of type NEW_PROCESS defined in the
  Constants interface.
• When a NEW_PROCESS event is removed from the
  event queue, a new Process object is created and
  added to the memory queue.
• If there is enough memory to allocate to the
  newly created process, it is immediately moved on
  to the CPU queue.
• If the CPU queue is empty, the process
  immediately enters the CPU, and a new event will
  be added to the event queue.

12
Event processing

• Depending on random variables and the simulation
  parameters, the event added to the event queue is
  one of the following
• SWITCH_PROCESS if the process is allowed to
  exhaust its time slice and then must yield the
  CPU to waiting processes.
• IO_REQUEST if the process will need to perform
  some I/O operation before the time slice is
  exhausted.
• END_PROCESS if the process will complete its
  execution before the time slice is exhausted.
• When the NEW_PROCESS event has been processed,
  the time until the next process arrival is
  determined, and a new NEW_PROCESS event for that
  arrival is created and added to the event queue.

13
Event processing

• Processing routines must be written for the
  SWITCH_PROCESS, END_PROCESS and IO_REQUEST event
  types as well.
• The fifth and final event type is END_IO, which
  marks the completion of an I/O operation.
• This event is added to the event queue when a
  IO_REQUEST event is processed.
• The processing of this event involves moving a
  process from the I/O device to the CPU queue, and
  starting a new I/O operation if there are any
  processes waiting in the I/O queue.
• If the CPU queue was originally empty, the
  process that entered the queue immediately enters
  the CPU, as described earlier.
• The event processing routines drive the
  simulation forward !

14
Your tasks

• Complete the partial solution. This involves
  completing the classes Process, Simulator, Cpu
  and Io.
• Use your solution to explore the effects of
  varying the different simulation parameters.
• Hand in a report including your source code and
  the results of your experiments.
• The program must be demonstrated to one of the
  assistants.

15
Hints

• Study the partial solution to understand how the
  program works.
• The Cpu and Io classes will resemble the Memory
  class, but are a bit more complicated.
• Take advantage of the notes from this lecture
  when writing code for processing events.
• Use the methods in the Gui interface to update
  the GUI during the simulation. This will enable
  you to see how your QN works.
• You can run the handed out code by running the
  P3.bat batch file. In the partial solution,
  processes completely bypass the CPU and IO
  resources, as these parts of the system are
  missing.