Description
This document describes the specifications of ARP assignment, updated after the pandemic emergency.
The new specifications convert the distributed/cooperative model into a single machine implementation. In the next page you can find the new specifications.
The document reports in yellow all the text changes with respect to the original sheet (V1.0).
The problem to be solved has the following specifications.
The goal is to simulate a network of multi-process systems, all identical, each one running on a machine in the same LAN, connected through sockets, exchanging a data token at a specified and measured speed. The number of machines is not given a priori. In figure 1 there are 4 machines as an example. They may be more, or less, or they may even reduce to one single machine for setup and debugging purposes.
Figure 1
In figure 2 the structure of each machine is defined. Note that a black frame encompasses each machine’s multiprocess system, whereas a colored frame includes the system components that must developed by each student. Summing up, each student develops all system components of her/his machine, except for one, which is acquired by her/his partner “right side” student; in her/his turn, the student develops a component that is to be given to, and integrated by the”left side” student (“right” and “left” refer to the schema in figure 2).
Figure 2
| Students will implement on a single machine all processes, simulating a distributed implementation. An artificial | |
| delay (it will be called waiting time) will be used to mimic the network communication delays. | |
Glossary.
- Circles are Posix processes.
| after a given delay | (see | the waiting |
| time |
- Process P receives tokens, computes and sends an updated token below).
- Process L logs data end events in a log file.
- Process G receives tokens and dispatches them to P.
- Process S receives console messages as Posix signals.
- Processes S and G are connected to P through non-deterministic pipes (Posix select).
- A socket connects P to the partner’s G process.
Detailed specifications.
A configuration file in the machine contains in text format the necessary parameters, and is edited by the user before running the multi-process:
- IP address of the machine and port numbers of processes (where relevant) reference frequency (RF) of the generated token wave
| waiting time (in microseconds) applied by process P before sending the updated token (initially: 1,000) |
- other relevant data to be read before starting.
Token is:
- a floating point number between -1. and 1, plus
- a time stamp with the absolute system time (by the operating system) of the instant P writes data on socket.
P does the following computation: new token = received token + DT x (1. – (received token)^2/2) x 2 pi x RF
| system | ). In this way the token series forms a |
| waiting time written in the configuration file. | |
where DT is the time delay between the reception and delivery time instants (that must be measured by the sine wave of frequency RF. Note that DT takes into account the
Log file holds a series of text lines in couples:
<timestamp> <from G | from S > <value>
<timestamp> <sent value> (a sample of the wave)
The log file contents is output anytime G receives a suitable request from the console
S receives signals for carrying out the following actions:
- start (receiving tokens, computing, sending tokens)
- stop ((receiving tokens, computing, sending tokens)
- dump log: outputs the contents of the log file, separating the wave from the actions.
Start phase
P, G, S, L start using the parameters stored in the configuration file.
The process G is given by the “right side” student in executable form, and is run using the exec syscall.
Main syscalls involved:
fork(), exec() pipe() select() read() write() socket() signal() and related ones.
