Promethee workbench was initially designed to provide an easy-to-use graphical front-end for engineers. The main feature was just the possibility to parameterize input files and remotely launch (or distribute) calculations at the same time, over all computing resources.
From this basic usage, we developed a full featured workbench for extended engineering issues, based on “Design of Computer Experiments” (DoCE) framework.
So, it now provides optimization algorithms, inversion, propagation of uncertainties, for any code integrated in the Promethee environment. The new degrees of freedom offered by this generic workbench lets the user test any algorithm with any computer code. We are often surprised of the new engineering opportunities emerging form this simple fact.

Yes or no.
Yes: if you just need to distribute calculations over may CPUs: Promethee can do this, with less constraints and more efficiency in general (say merging heterogeneous hardware and software for instance).
No: if you need High Performance Computing infrastructure (HPC) for parallelization of simulations over thousands of CPU, Promethee is NOT the solution to replace a true Unix queueing system. Nevertheless, you can also merge Promethee in an existing queueing environment, where Promethee just stands for another user of the queue.

An “algorithm” in our lingo, means a design policy which leads the calculations to be launch. In other words, when you have variables in your data sets, you can either define their values by yourself, or let an “algorithm” do it for you, following a goal you define (like an optimization).

For now, Promethee is free as in “free beer”. We are already working on the “true free” (as in “free software” defined by GNU) next generation version, which will provide more flexibility and extensibility.
For short term extensibility issues, an operating license is nevertheless available (see http://promethee.irsn.fr/doku.php?id=support).

This version has a limit of 4 calculations at the same time. Yes, 4 is not so much, but in most cases, it is already a disruptive change in terms of engineering user time.
If you need to run more calculations in parallel (in the sense of launching many independent calculations at the same time), you can get industrial (not free) support: http://promethee.irsn.fr/doku.php?id=support

The following “engineering” projects are close in terms of features, and may partially replace Promethee as an engineering software:

SUNSET: https://gforge.irsn.fr/gf/project/sunset/
OpenTURNS: http://www.openturns.org/
URANIE: http://www.gdr-mascotnum.fr/media/presuranie20081119esreda.pdf
DAKOTA: http://dakota.sandia.gov/

For alternative versions of “algorithms”, we recommend to work with the [R] environment, which provides lot of “Computer Experiments” packages (see http://cran.r-project.org/web/views/ExperimentalDesign.html). Some of these toolboxes are also used as a basis for some Promethee algorithms plugin.

A demo version which just provides one code wrapper (for [R]) is available for this purpose. It should be tested on a standard, no-admin-rights, Windows desktop computer: [XXXX].

Even if a Windows desktop network is not a dedicated environment to provide significant computing power, many operational entities may benefit to deploy a grid on their desktop computers:

total cost ownership (TCO) of servers fleet is too hight to be covered by numerical engineering gains,
high constraints of server fleet operation
- reduced disk space or lack of computing power,
- computer codes portability,
- inadequate network architecture (latency or communication tools),
- administration lack of flexibility (outsourcing),
need for punctual autonomy from server fleet:
- network instability or unavailability,
- office laptop disconnection.

At the same time, raw computing power of a grid is nowadays close to office network one:

raising of unit power and decreasing of load average,
multiplication of CPU/core inside each computer,
high and continuous hardware turnover.

However if this potential power can theoretically compete with a fleet of servers, the constraints for using an office network as a grid are:

Desktop ergonomic impact: when a calculation is started, the user of the machine hosting the calculation should not be penalized in his office use,
Portability of computer codes on Windows: most computer codes are compiled only for Linux or Unix architectures, Windows porting is not available,
Hardware heterogeneity of office desktops: need for dynamic and efficient load balancing,
Instability of configuration: disconnection and dynamic unplug of desktop or laptops.

PWinG : Promethee Windows Grid

The PWinG subproject is designed to solve the previous technical issues in order to build a computing grid from an office desktop environment.
Office desktop usage is then seamlessly integrated inside the computing power offered by the traditional dedicated servers calculations (if available).

In common use, the Promethee environment is installed as a daemon on servers dedicated to calculation and provides a graphical interface on the workstations of users.
In a PWinG grid, in addition to the graphical user interface, the daemon is also running on desktop computers to integrate their power into the grid. In such a case, the Promethee grid may include only desktop computers or merge both office desktop and servers.
The previous constraints and issues are solved as follows:

Desktop ergonomy impact:
- day-parting control of Promethee calculation service,
- selection of authorized computing tasks,
- definition of number of CPU used,
Portability of computer codes on Windows: ideally, computing software may be available on Windows platform.
However, this portage on Windows x86 architecture is not possible for many codes, for technical reasons (sources are not portable) or commercial (source code not available). As part of the project, a virtualization solution is used to allow execution of code compiled for Linux on x86 Windows: a coLinux virtual machine is packaged to execute the Promethee daemon installed in the directory of the Windows host machine.
Management of computing fleet: heterogeneity and dynamic load balancing are handled by Promethee grid, designed to support such constraints.

Administration & deployment guide

Overview of Promethee environment

Requirements

Typical deployment configurations

FAQ

PWinG : Promethee Windows Grid

Submit your own question