2019-12-09 Pedro Canadilla
* Solved class(x) == "something" with inherits(x, "something") as suggested by CRAN team.
* General Birth and Death Process model
2019-05-04 Pedro Canadilla
* Removed unit test with testthat package.
* General Birth and Death Process model
2017-11-13 Pedro Canadilla
* Included unit test with testthat package.
2017-05-29 Pedro Canadilla
* For the M/M/1, M/M/c and M/M/Infinte models, the parameter has been updated to the following description: "number of customers in the system from which you want to obtain its probabilities. Put n=0 for a idle probability (no customer present in the system or system idle). With n=-1, no probabilities are computed". No changes has been done in source code.
2017-05-13 Pedro Canadilla
* The big source code queueing.R has been divided into small model.R files.
2017-03-20 Pedro Canadilla
* Model MCMN had a bug. Solved.
* New summary functions following R FAQ. summary now returns an object of class summary.o_MM1, and then new print functions prints in a table format (using data.frame) the main results of the model.
2016-12-25 Pedro Canadilla
* old summary functions renamed to Report.
2016-10-16 Pedro Canadilla
* Three bugs detected and fixed in the Multiple Class Open Network. CheckInput was checking capacity constraints for delay nodes; the checking of values was wrong, getting array index out of limits error, and finally, some W was wrong calculated in the model when delay centers was provided in a model with routing probabilities.
* In the same way, in the Multiple Class Closed Network the checking of values was wrong, getting array index out of limits errors.
* Thank you very much for the contribution of Alexandre Martins to bring my attention a function developed by Tom Potter at , which makes the B_erlang function atonishing fast.
2016-08-13 Pedro Canadilla
* Fixed Engset calculator.
2015-11-08 Pedro Canadilla
* Fixed NOTE as CRAN checks has detected in the new version of R.
2015-05-15 Pedro Canadilla
* Fixed a bug in the CJN, Wk of MMInf server in the ExactMethod. It was written mu instead of 1/mu. Detected and fixed by Duncan Garmonsway (nacnudus at gmail.com). Thank you very much Duncan.
2015-04-11 Pedro Canadilla
* Some minor changes requested by CRAN.
2014-08-20 Pedro Canadilla
* Fixed an error on pdf explaining Wqq and Lqq.
* Fixed notes as cran reported it.
* Included a new method to get the M/M/c/c probabilities. It has been taken from [Kobayashi2012], pag. 709. It gets more probabilities that the exact method, and in lesser time than the approximate method. For large sites, it is the recomended method. All examples of M/M/c/c model has been changed to this way.
* Variances has been included in the basic models. For models M/M/c/K, M/M/c/K/K and M/M/c/K/m, variances respect the time in system are not inluded. Nor the variance respect the waiting time has been calculated for M/M/c/K/m model.
* To obtain the pn in the M/M/1/K model for high values of k and in case that lambda < mu, dgeom and pgeom has been used to speedup the process. In the case of lambda > mu, it persists the numerical overflow for high values of k.
* FW for models M/M/c/K, M/M/c/K/m and M/M/c/K has been removed. FWq also has been removed for M/M/c/K/m.
2013-11-02 Pedro Canadilla
* Included a new method to get the M/M/1/K/K probabilities. It has been taken from [Kobayashi2012], pag. 709. It gets more probabilities that the exact method, and in lesser time than the approximate method. For large sites, it is the recomended method. All examples of M/M/1/K/K model has been changed to this way.
* Fixed typos in the documentation (including the paper). Word "client" has been replaced by "customer".
2013-02-24 Pedro Canadilla
* Included a new method to get the M/M/1/K/K and M/M/C/K/K probabilities. It has been taken from [Jain2007], pag. 26. It gets more probabilities that the exact method, but in more time than the approximate method. In my computer, for the M/M/1/K/K, the value 336 is the absolute value for the k parameter in which the new method and the approximate method gets values. Profiling the code, maybe I can give more performance to Jain's Method using other techniques.
* Fixed typos and suggestions given to me by See Chuen-Teck in the documentation pdf explaining the Wqq and Lqq formulae included in the package. Also I thank Alvaro Mayol and Alfonso Mateos for reading the paper.
2013-01-27 Pedro Canadilla
* Fixed typos in the SP help pages.
* Included a documentation pdf explaining the lastest formulae included in the package.
2012-12-24 Pedro Canadilla
* Included the package description help file.
2012-12-09 Pedro Canadilla
* Fixed the Wqq and Lqq functions in M/M/c/K/m model.
* Fixed doc typos.
* Consolidation of the summary function for Open and Closed Jackson networks, to get inconsistencies solved.
* Added new ways to define the inputs of the Open and Closed Jackson Networks, more convenient from an operational point of view. Also, included a new way to pass a list of nodes built before instead of use the dots parameter. More convenient for automatization or scripting.
2012-11-18 Pedro Canadilla
* Fixed a bug in the models M/M/1/K, MMC, M/M/c/K, M/M/1/K/K, M/M/c/K/K and M/M/c/K/m. FWq and FW were calculating the wrong values.
* Added Qn probabilities to the o_MM1K object to get the probabilities that an arrival that enters the systems see n clients in it. Also it has been added for the rest of infinite capacity objects to homogeneization of all the objects.
* Review of the doc. Minor changes to maintain consistency.
* Calculated the Wqq values for all models except networks. Some values corrected due a bug. Thanks to Alfonso Mateos for help me.
* Calculated the Lqq for all models except networks. Thanks to See Chuen-Teck for giving me valuable suggestions to get it, including a formulae and with the corresponding Litle Law's version.
* Reorganization of the code in general. Summary function for the markovian models is now centralized and the code not repeated again and again.
* Fixed the function for comparing models to include Lqq and P0.
* Removed the file news, I hope that no more link in CRAN to a empty NEWS.
* Fixed RO value for the model M/M/Infinite/K/K.
2012-11-09 Pedro Canadilla
* Fixed a bug in model M/M/C. The names of FWq and FW components of the output was interchanged respectively.
2012-11-07 Pedro Canadilla
* Fixed another bug in the MMCKM model. The summary function was returnin a wrong value for Ws. Thanks to See Chuen-Teck (see_chuenteck at yahoo.com.sg) for detect it.
2012-11-04 Pedro Canadilla
* Fixed a bug in the MMCKM model. The RO functions were returning values greather than zero. Thanks to See Chuen-Teck (see_chuenteck at yahoo.com.sg) for detect it.
* Changed the names of the WWq and LLq functions to Wqq and Lqq as suggested for See Chuen-Teck, because there are more natural the new names than previous.
* Implemented functions to compare basic queueing models. No networks models can be compared yet. Thanks to See Chuen-Teck for the suggestion.
2012-10-21 Pedro Canadilla
* New version 0.1.6. Added the Mixed queueing network model.
2012-10-13 Pedro Canadilla
* New version 0.1.5. Added the aproximation method for the MCCN.
* Checked for anomalous for the tol and method parameters in Checkparams method of the CJN model.
2012-10-10 Pedro Canadilla
* Fixed doc of the MCCN functions. The descriptions were wrong.
* Changed the vMu parameters to be vService, more natural in the models MCCN and MCON.
* Fixed a fatal error that makes the function not work for vNumber with elements greater than one.
* Changed some checks with no sense in CheckParams for the MCCN model.
2012-08-04 Pedro Canadilla
* Fixed doc of the NewInput functions. The usage section was wrong and included a S3 method section not needed and confusing.
* Included the Engset calculator
2011-11-23 Pedro Canadilla
* Removed the floor conditions used to check that the number of servers c has to be an integer number. Now used is.wholenumber as recommended in the doc.
* Added the check that the the c value has to be an integer in the M/M/c, M/M/c/K/M, M/M/c/K/K, M/M/c/K, M/M/c/c models.
* Added the CheckInput to every node in an Open JackSon Network.
* Implemented the Bard-Schweitzer approximation algorithm in the Closed Jackson Network.
2011-11-22 Pedro Canadilla
* The WWq value of the M/M/1 model has been fixed to the correct value (W). Thanks to Juan Antonio Fernandez
(jafernandez at fi.upm.es) for detect it.
2011-11-02 Pedro Canadilla
* The B_erlang function has been modified to be faster. High order functional functions like Reduce has been tried but finally a simplified -for- winned the system.time simple benchmark.
2011-10-30 Pedro Canadilla
* The RO funtion in the M/M/Infinite model has been established to the L value, because that has been seen in more than one book of the literature.
* In consecuence, the Open Jackson Network inherits this change and also the Closed Jackson Network has been updated.
2011-10-18 Pedro Canadilla
* Fixed the summary function of the Open Jakson Network, Closed Jakson Network, Multi Class Open Network and
Multi Class Closed Network. The print function has been replaced by the cat function.
* Fixed the bug with the RO and W values of the Multi Class Closed Network.
* The Multi Class Closed Network QueueingModel returns a vector Throughputcn with the Throughput of each class and of
each n population passed as Input parameter. The reason is performance to compute multi class load dependent models.
2011-09-24 Pedro Canadilla
* The Closed Jackson Network QueueingModel returns a vector Throughputn with the Throughput of each n population
passed as Input parameter. The reason is performance to compute load dependent models.
2011-09-23 Pedro Canadilla
* Introduced the Closed Multi Classes model.
* Nomenclature more consistent. Wi, Li, Throughputi has been substituted by Wk, Lk, Throughputk respectively.
* Several minor errors in the documentation has been fixed.
* Included a parameter operational in the closed jackson network model to be able to use visit ratio with a stacionary
probability meaning or with an operational point of view.
2011-08-20 Pedro Canadilla
* Fixed values for M/M/c/K/K and M/M/c/K/m models in case of RO = 1.
* logFact function used has been sustituted by built-in function lfactorial, more precise with small values of n.
* CheckInput for multiple Class Open Network has been revisited to detect saturated nodes.
* The variable classes has been introduced to NewInput.MCON model input. The reason is performance.
2011-08-12 Pedro Canadilla
* Fixed B-erlang and C-erlang functions to have defaults values and to check values.
* Fixed extreme value of Pn in M/M/1/K/K model.
* More detailed documentation has been created for the saturation point in the M/M/1/K/K model. The optimal number of
clients (or saturation) point coincides with the inverse of the serialization parameter of Amdalh's Law. That is, the value
which converges the speedup function k/(1 + ser(k-1)). It makes sense, because the saturation point is the maximun value in
which no syncronization happens.
* Fixed in the M/M/1/K/K model the Pn function. It was returning RO instead of Pn.
* Fixed and improved the model M/M/c. The parameter n was not working as needed.
* Included a new parameter method which controls if the probabilities are calculated exactly or if an aproximation is
used in models M/M/1/K/K, M/M/C/K/K, M/M/C/K/M, M/M/C. The aprox method are useful to get the model work with higher
parameter values.
* Fixed also the parameter n in models M/M/1, M/M/C and M/M/Inf. With n=-1, no probabilities were calculated.
* Documented the resolution of a Open Jackson Network with visit ratios instead of probabilities.
* Added the node of type M/M/Inf (or delay node in the literature) in the Open Jackson Network.
* Added the model M/M/Inf/K/K.
* Added more examples in the documentation of Open Jackson Models and Closed Jackson Model to use visit ratios.
* Fixed a bug in Closed Jackson Network with several nodes of class M/M/C
* Added the Multi Class Open Network model.
* Fixed some doc, W in the Open Jackson Network.
* Top level Copying file deleted.
* removed is.nan(x) just because of CRAN checks.