Ciw 3.2.6

A discrete event simulation library for open queueing networks

A discrete event simulation library for queueing networks

Ciw is a discrete event simulation library for open queueing networks. It’s core features include the capability to simulate networks of queues, multiple customer classes, and implementation of Type I blocking for restricted networks.

• Read the documentation

• Contribution guidelines

• Our great contributors

Install with pip install ciw.

Current supported version of Python:

• Python 3.8

• Python 3.9

• Python 3.10

• Python 3.11

• Python 3.12

Usage

Import Ciw:

>>> import ciw

To define an M/M/3 queue, with λ = 0.2 and μ = 0.1:

>>> N = ciw.create_network(
...     arrival_distributions=[ciw.dists.Exponential(rate=0.2)],
...     service_distributions=[ciw.dists.Exponential(rate=0.1)],
...     number_of_servers=[3]
... )

Now set a seed, create a Simulation object, and simulate for 1440 time units:

>>> ciw.seed(1)
>>> Q = ciw.Simulation(N)
>>> Q.simulate_until_max_time(1440)

Collect results:

>>> recs = Q.get_all_records()

Manipulate results to get useful statistics, e.g. average waiting time:

>>> waits = [r.waiting_time for r in recs]
>>> sum(waits) / len(waits)
4.2305...

Features

A number of other features are also implemented, including:

• Type I blocking

• A large range of sampling distributions

• Phase-Type distributions

• Time-dependent and state-dependent distributions

• Batch arrivals

• Baulking customers

• Reneging customers

• Processor sharing

• Multiple customer classes

• Priorities

• Server priorities

• Service disciplines

• Customers changing classes while queueing

• Customers changing classes after service

• Server schedules

• Slotted services

• State tracking

• Stopping the simulation after a certain amount of customers

• Process-based routing

• Logical routing

• Deadlock detection

History

• 3.2.6 (2025-06-13)

  • Add the include_incomplete keyword tot he get_all_records method to allow for collection of incomplete data records.

• 3.2.5 (2025-03-25)

  • Fixes bug in PoissonIntervals distribution where if the last rate is zero or so small there is no samples, the cycle begins again too early.

• 3.2.4 (2024-12-04)

  • Adds a new method for the simulate_until_max_customers: “Complete” simulates until a specific number of completed customer journeys; while “Finish” simulates until a specific number of customers have reached the exit node (through bailking or reneging).

• 3.2.3 (2024-10-15)

  • Allow some numerical imprecision in the PMF probability sums. This allows for very large arrays of probabilities and use of Pandas and Numpy to define probabilities.

• 3.2.2 (2024-07-05)

  • Allow custom exit node classes.

• 3.2.1 (2024-06-19)

  • Remove reneging_destinations, move to a jockeying routing option within the routing objects.

• 3.2.0 (2024-05-07)

  • System capacity implemented.

  • Server schedules objects now take numbers_of_servers and shift_end_dates keywards, instead of schedule.

  • Server schedules and slotted services can be offset.

  • Added a GroupedNodePopulation tracker.

  • Restructures documentation, adds more guides, and updates reference pages.

  • Routing objects implemented: TransitionMatrix object, ProcessBased object, FlexibleProcessBased object, NetworkRouting object, Direct object, Leave object, Probabilistic object, JoinShortestQueue object, LoadBalancing object

  • Added reroute as an option for interrupted preemptive services.

• 3.1.4 (2024-04-09)

  • Fix bug where Individual’s server isn’t reset after a slotted service.

• 3.1.3 (2024-04-08)

  • Allows class methods as generator functions for process-based routing.

• 3.1.2 (2024-04-08)

  • Fix bug when using Mixture distribution.

• 3.1.1 (2024-04-04)

  • Add a MixtureDistrubution that probabilistically chooses from a number of other distributions to sample from.

  • Baulking functions now take the simulation, current node, and current individual.

  • Service disciplines now take the current time.

  • Service disciplines now called when a customer arrives (allowing for lingering customers).

  • Distributions have parameters in their reprs.

  • Adding type hints and better doctrings.

• v3.1.0 (2023-12-03)

  • Server schedules now defined with objects.

  • Slotted services feature added, with capacitated and non-capacitated slots, and pre-emption options.

  • Rename ‘continue’ pre-emption option to ‘resume’.

  • Allows simultaneous renegeing customers.

  • Internal refactoring.

• v3.0.2 (2023-11-14)

  • Fix bug where simulation errors if fist event of a node is a shift change.

  • Minor docstring and documentation changes.

  • Default statetracker is now None.

• v3.0.1 (2023-10-31)

  • Fix bug where class change individual not reset after preemption.

• v3.0.0 (2023-08-22)

  • Adds service disciplines - FIFO, LIFO, SIRO and ability for custom disciplines.

  • Removes baulking and rejection dictionaries, these are recorded as DataRecords.

  • Adds an optional filter to Q.get_all_records to filter only records of given types.

  • Removes ciw.dists.NoArrivals, can just use None.

  • Removes ability to write to file and read from file.

  • Allows any string to represent customer classes.

  • Api changes to class_change_matrices and class_change_distributions

  • ciw.trackers.NodeClassMatrix now required a class_ordering keyword to ordet the arbitrary customer class string names.

  • Performance improvements.

• v2.3.7 (2023-04-06)

  • Adds Poisson, Geometric, and Binomial distributions for batching.

• v2.3.6 (2023-02-16)

  • Writes a data record for interrupted services caused by server schedules.

  • Raises an error when an inconsistant number of custom node classes are used.

• v2.3.5 (2023-02-07)

  • Fixes bug caused when implementing preemptive priority classes and server schedules that take all servers off duty.

• v2.3.4 (2022-12-07)

  • PoissonIntervals distribution now allows rates of zero

• v2.3.2 (2022-10-07)

  • Add PoissonIntervals distribution

  • Add numpy random generator, ciw.seed now also creates new random generator

  • Documentation on parallelising trials

  • Remove support for Python 3.6

• v2.3.1 (2022-07-25)

  • Fixes bug where blocked customers were candidates for finish_service when more than one customer finishes service simultaneously

  • Correctly writes csv’s in Windows

  • Support for Python 3.9 by updating tqdm requirement

• v2.3.0 (2022-05-10)

  • Reneging customers implemented

  • Customer class changes while waiting implemented

  • Preemptive interruption options implemented

  • New record_type field added to DataRecords

• v2.2.4 (2022-03-02)

  • Improve docs on pausing simulations and server priorities

  • Record server ID in the DataRecords

  • Move CI to GitHub Actions

• v2.2.3 (2022-01-27)

  • Server priority functions implemented.

• v2.2.2 (2021-12-17)

  • State trackers now take objects not indices

  • Servers are attached to individuals before sampling service times

  • Docs on sever-dependant distributions

  • Docs on DES+SD hybrid simulations

• v2.2.1 (2021-11-04)

  • PhaseType distributions implemented

  • Classes for specific PhaseType distributions: Erlang, HyperExponential, HyperErlang, and Coxian

• v2.2.0 (2021-07-22)

  • Processor sharing implemented (limited and capacitated)

  • Ability to use a different node_class per node of the network

  • State tracking now works with simulate_until_max_customers

  • Remove testing on Python 3.5

• v2.1.3 (2020-10-06)

  • Small refactor to Node adding new servers, and to Individuals receiving the Simulation object.

  • Add a library of custom behaviour to docs

  • Support Python 3.8, update hypothesis

• v2.1.2 (2020-09-26)

  • Ability to incorporate customer behaviour Server and Individual classes.

• v2.1.1 (2020-05-27)

  • State Trackers slightly more efficent, they do not record any state changes that result in the same state as before.

  • Add the NodePopulationSubset tracker.

  • Distribution objects can now see the Simulation object, for true state dependent distributions.

• v2.1.0 (2020-04-23)

  • State Trackers now track history

  • State Trackers give state probabilities

  • A number of performance improvements

  • Fix some documentation

  • Test on PyPy3.6 and Python 3.7 too

• v2.0.1 (2019-07-17)

  • setup.py now finds packages to fix pip install bug

• v2.0.0 (2019-07-10)

  • Large refactor:

  • Drop support for Python 2.7, Python 3.4.

  • Update networkx and pyyaml requirements.

  • Refactor time so that Simulation has current_time attribute.

  • Change Transition_matrices keyword to routing.

  • routing can take a process-based routing function.

  • Refactor distributions to be objects: [‘Exponential’, 0.5] -> ciw.dists.Exponential(0.5).

  • Distribution objects can be manipulated with +, -, * and /.

  • All keywords lower case to conform to Pep8.

  • deadlock_detector keyword takes object, not string.

  • tracker keyword takes object, not string.

  • Add tests and docs to show how objects can be used for state-dependent distributions.

  • All user facing api now takes float(‘inf’) not ‘Inf’, expect for .yml files.

  • Reference Ciw paper in docs.

  • Add AUTHORS.rst to docs.

• v1.1.6 (2018-10-22)

  • Fixed bug in which preemptively iterrupted individuals remained blocked once service resampled.

  • Fixed bug in which interrupted individuals not removed from interrupted list when restarting service.

  • Some performance improvements.

  • Improve deadlock detection to check for knots less often.

• v1.1.5 (2018-01-11)

  • Fixed bug calculating the utilisation of servers.

• v1.1.4 (2017-12-12)

  • Time dependent batching distributions

  • Hard pin requirements versions

• v1.1.3 (2017-08-18)

  • Replace DataRecord object with namedtuple.

  • Number of minor tweaks for speed improvements.

• v1.1.2 (2017-07-05)

  • Batch arrivals.

• v1.1.1 (2017-06-23)

  • Server utilisation & overtime.

  • Small fixes to docs.

  • Testing on Python 3.6.

• v1.1.0 (2017-04-26)

  • Replace kwargs with actual keyword arguments in ciw.create_network.

  • Refactor server schedule inputs (schedules placed inside Number_of_servers instead of as their own keyword).

• v1.0.0 (2017-04-04)

  • ciw.create_network takes in kwargs, not dictionary.

  • Add Sequential distribution.

  • Add truncated Normal distribution.

  • Refactor inputs for custom PDF.

  • Refactor inputs for server schedules.

  • Transition matrix now optional for 1 node networks.

  • Overhaul of documentation.

  • Add CONTRIBUTING.rst.

  • Slight improvement of ciw.random_choice.

• v0.2.11 (2017-03-13)

  • Add ability to simulate until max number of customers have passed arrived/been accepted/passed through the system.

• v0.2.10 (2017-03-10)

  • Performance improvements.

  • Drop dependency on numpy.

• v0.2.9 (2017-02-24)

  • Allow zero servers.

• v0.2.8 (2016-11-10)

  • Add option for time dependent distributions.

• v0.2.7 (2016-10-26)

  • Run tests on Appveyor.

  • Check docs build and pip installable on Travis.

  • Remove hypothesis cache.

• v0.2.6 (2016-10-17)

  • Add AUTHORS.rst.

  • Add progress bar option.

• v0.2.5 (2016-10-06)

  • Fix bug that didn’t include .rst files in MANIFEST.in.

• v0.2.4 (2016-09-27)

  • Fixed bug in which priority classes and dynamic classes didn’t work together.

  • New feature: preemptive interruptions for server schedules.

• v0.2.3 (2016-07-27)

  • Ability to set seed. More docs. Fixes to tests.

• v0.2.2 (2016-07-06)

  • Baulking implemented, and minor fixes to order of unblocking.

• v0.2.1 (2016-06-29)

  • Priority classes implemented.

• v0.2.0 (2016-06-20)

  • Python 3.4 and 3.5 compatible along with 2.7.

  • Data records now kept in list.

• v0.1.1 (2016-06-06)

  • Ability to incorporate behaviour nodes.

  • Data records are now named tuples.

• v0.1.0 (2016-04-25)

  • Re-factor inputs.

  • Simulation takes in a Network object.

  • Helper functions to import yml and dictionary to a Network object.

  • Simulation object takes optional arguments: deadlock_detector, exact, tracker.

  • simulate_until_max_time() takes argument max_simulation_time.

• v0.0.6 (2016-04-04)

  • Exactness implemented.

  • Restructure some features e.g. times_to_deadlock.

  • Custom simulation names.

• v0.0.5 (2016-03-18)

  • State space tracker plug-and-playable.

  • Add rejection dictionary.

• v0.0.4 (2016-02-20)

  • Empirical and UserDefined distributions added.

  • Tidy ups.

• v0.0.3 (2016-02-09)

  • Arrival distributions.

  • MMC options removed.

  • Fix server schedule bugs.

• v0.0.2 (2016-01-06)

  • Some kwargs optional.

  • Hypothesis tests.

  • Minor enhancements.

• v0.0.1 (2015-12-14)

  • Initial release.

• v0.0.1dev (2015-12-14)

  • Initial release (dev).

Contributors

Ciw has been developed by the following contributors:

• Geraint Palmer

• Vincent Knight

• Paul Harper

• Lieke Hölscher

• Sam Luen-English

• Alex Carney

• Adam Johnson

• Nikoleta Glynatsi

• caipirginka

• Emma Aspland

• Henry Wilde

• timlathy

• Michalis Panayides

• Jorge Martín Pérez

• KernelA

• Matthew Howells

• Galen Seilis