Tentative Project Information

General issues:

• The project will count as a double homework assignment.
• The goal is to produce something generally useful to the SimPy user community. If your project is done well, it will be submitted to the SimPy project team for the "Contributed" portion of their Web site, and possible inclusion into SimPy itself.
• Group sizes must be either two or three; you cannot work alone on this one..
• Projects must work under SimPy version 1.9 (which I have in ~matloff/Pub/SimPy on CSIF).
• You will submit a written report, in L^AT_EX  , first in draft form for my comments, then in final form. The report will clearly explain the project goals, directions for users and an overview of implementation.
• You will present your project to the class, in a 15-20 minute talk. Desired length is at least 15-20 minutes, not including time for questions.
• The project descriptions below are meant to be minimal, and you are urged to be creative in terms of the details. You will need to keep me posted on this, and of course are welcome to seek advice from me.
• You may also propose your own project. To be approved, it must satisfy the "generally useful" criterion above, and should be of an appropriate size and scope.

Key dates:

• February 18 (11:59 p.m.): Ranked project preferences and group compositions due. Let me know by e-mail. In your preference rankings, let me know whether your preferences are really strong or not, and if so, why. I believe everyone can get their first or second preference.
• February 19 (in class): Project assignments announced, pep talk given. :-)
• March 5 (11:59 p.m.): Rough drafts of written reports due.
• March 6: Presentations begin.
• March 17 (11:59 p.m.): Final written reports due.

Project choices:

• SimPy debugging/code verification tool
  • supplement to, not replacement for, existing debugging tools
  • each thread has its own window, displaying its information: next line to be executed; next event time if any, passive/action status; resources/waitevents queued for, if any, etc.
  • user hooks
  • see SimulationStep
  • need very strong knowledge of SimPy internals and Python foundations
  • SimPy has a Tcl/Tk capability, but you may prefer wxPython or ...

• general SimPy framework for simulating networks of queues
• have many servers, output from one server routed into another
• jobs can also come in from outside, and can exit the system
• applications: manufacturing, communication networks, computer I/O systems
• examples:
  • Jackson network, on SimPy Web page (this is actually a less general version of our project)
  • purchase order chain
  • our ALOHA example, with more than one message allowed at a node at one time
  • Stallings CS examples--packet-switching network, database server, multiprocessor (starting on p.8)

• project involves:
  • you write some SimPy classes (as subclasses of Process, Resource etc.) that enable users to simulate any kind of network of queues as described above
  • your code also reads in specifications supplied by the user in an input file, including nodes, routing, buffer sizes, service times etc.

• computation of regenerative confidence intervals
  • a regeneration point is a point in time at which the system being simulated in which "time starts over"; see our PLN unit on inference on simulation output
  • we want our simulation output to say, for instance, "We are 95% sure that the long-run average queue size is between 12.2 and 15.8," instead of "We estimate the long-run average queue size to be 14.0"
  • no statistical knowledge needed
  • project involves:
    • adding a new member method to the Monitor class, implementing the formula given in the PLN
    • devising several additional examples beyond the two in the PLN; this can probably be done on existing programs, e.g. in the PLNs or from your homework
    • confirming that the examples work, by doing a "simulation of simulations" (i.e. embedding the SimPy code within an outer loop)
    • writing a clear report so that users know what the method does and how to apply it

• other queuing disciplines
• addition of new member methods to the Resource class
• a queuing discipline is a priority policy
• examples:
  • Adan-Resing online notes, Chapter 9 (ignore the math)

• implement a few of the most famous policies, again with good examples needed to teach users

• implement ns-2 in SimPy
• ns-2 is a popular network simulator
• home page
• tutorials etc.:
  • Greis tutorial
  • ns by Example
  • NS Simulator for Beginners

• see this slide presentation and this for quick introductions
• go here for a large collection of example scripts

• e.g. implement node type in ns-2 as a subclass of Process or Resource in SimPy
• implement enough features to demonstrate that the project is viable
• need to know or learn a bit of Tcl (like Python and Perl)