Layered Queueing Network Documentation
The documentation takes a lot of different forms... we are presently
putting it all into PDF format for easier viewing over the web, and
these are the links in the text... other forms are indicated after
each item, if available.
The original input language is defined informally by executable templates that you can edit, and formally by its grammar
- The LQN basic template is an ASCII input file for basic
lqns, fully commented to explain the syntax. (ASCII) (ASCII OUTPUT)
- The LQN activity template is an ascii input file for lqns
with activities in the tasks, commented to explain the
syntax. (ASCII) (ASCII
- The LQN Input Language defines the
original input file format in BNF.
The XML input language as of 2005 is defined by:
You can convert the old to the new by the filter lqn2xml.
NEW: A version LQNX is being developed for controlling parameter values,
executing multi-run experiments which include computing functions of results, and
outputting formatted results. LQNX includes an experiment language LQX. A guide to using LQX to compute sensitivities of
model outputs to parameters, is given here ,
with the example file and input data file.
- The modeling of activity sequences within a task has been
documented at length in Greg Frank's thesis. An extract of
chapters 8 and 9 covers this topic. (PDF
- Two template files are heavily commented to explain the syntax, on two
- The lqns manual page documents the run-time controls for the LQNS
solver, which is the analytic solver based on mean value analysis
and iteration. (ASCII MAN
- Another substantial documentation of the LQNS solver is found
in Greg Frank's PhD thesis PDF
- The lqsim man page documents the use of the simulation tool lqsim
in place of the analytic solver LQNS. Lqsim employs our simulator parasol
(which is similar to CSIM in many ways), and takes the same input file
and model assumptions as LQNS. It is more accurate for activity
flows which fork and join, expecially if the fork and join are in
different tasks. It should be used if the LQNS iteration scheme
fails to converge. (ASCII MAN PAGE)
- The Java-based editor jlqndef does not have a particular document
at this time... the previous editor lqndef is quite buggy and is not supported.
- Two tools were written to support repetitive sequences of model runs with
different parameter values, including extracting specified results into
a table or a set of graphs. These run with the original modeling language only; for the XML syntax they are replaced by a new feature LQX integrated with the solver (documentation coming soon).
- The first is MultiSRVN, based on gawk. It has been superseded by
SPEX. It utilizes a special language to identify the parameters to be changed.
It does not support models with activities inside tasks, only phases. (POSTSCRIPT
- The new experiment-management tool is the Software Performance Experimenter,
SPEX, which is based on PERL. It represents the parameters to be varied
by character strings in-line in the model. It does support models with
activities. (ASCII DOCUMENTATION)
- The XLQN template is an ascii input file for spex, fully commented
to explain the syntax of spex (ASCII input)
OUTPUT for results template.res)
SPEX also has an iterative version which allows model parameters
to be defined as functions of model outputs. ASCII
DOCUMENTATION PART I PART II
A new component-based system builder is being
- Layered queueing models are built by analyzing the flow of activities along
scenarios. One approach is to create an activity graph which specifies
the resource demands of operations along a scenario. It may be necessary
to reduce this graph by counting loops and probabilities of branching,
and by substituting in for abstract operations, and this can be tedious.
The Activity Graph Reducer AGR does this. POSTSCRIPT
- For local execution, all the LQN tools are found in /home/artt/bin/, which
must be in your path.